Datasets:

---

ftopal

/

huggingface-models

like 0

[ "transformers", "safetensors", "mistral", "text-generation", "4-bit", "AWQ", "autotrain_compatible", "endpoints_compatible", "mergekit", "merge", "en", "base_model:MaziyarPanahi/samantha-mistral-7b-Mistral-7B-Instruct-v0.1", "base_model:cognitivecomputations/dolphin-2.6-mistral-7b", "base_model:SanjiWatsuki/Sonya-7B", "license:other", "text-generation-inference", "region:us" ]

[]

[ "en" ]

[ "# jeiku/Elly_7B AWQ\n\n- Model creator: jeiku\n- Original model: Elly_7B\n\n!image/png\n\nThis is a merge of pre-trained language models created using mergekit.", "## How to use", "### Install the necessary packages", "### Example Python code", "### About AWQ\n\nAWQ is an efficient, accurate and blazing-fast low-bit weight quantization method, currently supporting 4-bit quantization. Compared to GPTQ, it offers faster Transformers-based inference with equivalent or better quality compared to the most commonly used GPTQ settings.\n\nAWQ models are currently supported on Linux and Windows, with NVidia GPUs only. macOS users: please use GGUF models instead.\n\nIt is supported by:\n\n- Text Generation Webui - using Loader: AutoAWQ\n- vLLM - version 0.2.2 or later for support for all model types.\n- Hugging Face Text Generation Inference (TGI)\n- Transformers version 4.35.0 and later, from any code or client that supports Transformers\n- AutoAWQ - for use from Python code" ]

[ "TAGS\n#transformers #safetensors #mistral #text-generation #4-bit #AWQ #autotrain_compatible #endpoints_compatible #mergekit #merge #en #base_model-MaziyarPanahi/samantha-mistral-7b-Mistral-7B-Instruct-v0.1 #base_model-cognitivecomputations/dolphin-2.6-mistral-7b #base_model-SanjiWatsuki/Sonya-7B #license-other #text-generation-inference #region-us \n", "# jeiku/Elly_7B AWQ\n\n- Model creator: jeiku\n- Original model: Elly_7B\n\n!image/png\n\nThis is a merge of pre-trained language models created using mergekit.", "## How to use", "### Install the necessary packages", "### Example Python code", "### About AWQ\n\nAWQ is an efficient, accurate and blazing-fast low-bit weight quantization method, currently supporting 4-bit quantization. Compared to GPTQ, it offers faster Transformers-based inference with equivalent or better quality compared to the most commonly used GPTQ settings.\n\nAWQ models are currently supported on Linux and Windows, with NVidia GPUs only. macOS users: please use GGUF models instead.\n\nIt is supported by:\n\n- Text Generation Webui - using Loader: AutoAWQ\n- vLLM - version 0.2.2 or later for support for all model types.\n- Hugging Face Text Generation Inference (TGI)\n- Transformers version 4.35.0 and later, from any code or client that supports Transformers\n- AutoAWQ - for use from Python code" ]

[ "transformers", "tensorboard", "safetensors", "bert", "text-classification", "generated_from_trainer", "base_model:google-bert/bert-base-cased", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "region:us" ]

[]

[]

[ "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 2e-05\n* train\\_batch\\_size: 16\n* eval\\_batch\\_size: 16\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* num\\_epochs: 2", "### Training results", "### Framework versions\n\n\n* Transformers 4.40.1\n* Pytorch 2.2.1+cu121\n* Datasets 2.19.0\n* Tokenizers 0.19.1" ]

[ "TAGS\n#transformers #tensorboard #safetensors #bert #text-classification #generated_from_trainer #base_model-google-bert/bert-base-cased #license-apache-2.0 #autotrain_compatible #endpoints_compatible #region-us \n", "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 2e-05\n* train\\_batch\\_size: 16\n* eval\\_batch\\_size: 16\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* num\\_epochs: 2", "### Training results", "### Framework versions\n\n\n* Transformers 4.40.1\n* Pytorch 2.2.1+cu121\n* Datasets 2.19.0\n* Tokenizers 0.19.1" ]

[ "transformers", "pytorch", "safetensors", "gguf", "llama", "text-generation-inference", "unsloth", "trl", "sft", "question-answering", "en", "dataset:Ahmedelmorshidy/influraQAs", "base_model:unsloth/llama-3-8b-Instruct-bnb-4bit", "license:apache-2.0", "endpoints_compatible", "region:us" ]

[]

[ "en" ]

[ "# Uploaded model\n\n- Developed by: Ahmedelmorshidy\n- License: apache-2.0\n- Finetuned from model : unsloth/llama-3-8b-Instruct-bnb-4bit\n\nthis model is used to influra app to build it's chatbot .\nby eng : ahmed elmorshidy" ]

[ "TAGS\n#transformers #pytorch #safetensors #gguf #llama #text-generation-inference #unsloth #trl #sft #question-answering #en #dataset-Ahmedelmorshidy/influraQAs #base_model-unsloth/llama-3-8b-Instruct-bnb-4bit #license-apache-2.0 #endpoints_compatible #region-us \n", "# Uploaded model\n\n- Developed by: Ahmedelmorshidy\n- License: apache-2.0\n- Finetuned from model : unsloth/llama-3-8b-Instruct-bnb-4bit\n\nthis model is used to influra app to build it's chatbot .\nby eng : ahmed elmorshidy" ]

[ "transformers", "tensorboard", "safetensors", "gpt_neox", "text-classification", "generated_from_trainer", "base_model:EleutherAI/pythia-1b", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[]

[]

[ "# robust_llm_pythia-1b_mz-131_PasswordMatch\n\nThis model is a fine-tuned version of EleutherAI/pythia-1b on an unknown dataset.", "## Model description\n\nMore information needed", "## Intended uses & limitations\n\nMore information needed", "## Training and evaluation data\n\nMore information needed", "## Training procedure", "### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 1e-05\n- train_batch_size: 8\n- eval_batch_size: 64\n- seed: 0\n- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n- lr_scheduler_type: linear\n- num_epochs: 1", "### Training results", "### Framework versions\n\n- Transformers 4.39.3\n- Pytorch 2.2.1\n- Datasets 2.18.0\n- Tokenizers 0.15.2" ]

[ "TAGS\n#transformers #tensorboard #safetensors #gpt_neox #text-classification #generated_from_trainer #base_model-EleutherAI/pythia-1b #license-apache-2.0 #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n", "# robust_llm_pythia-1b_mz-131_PasswordMatch\n\nThis model is a fine-tuned version of EleutherAI/pythia-1b on an unknown dataset.", "## Model description\n\nMore information needed", "## Intended uses & limitations\n\nMore information needed", "## Training and evaluation data\n\nMore information needed", "## Training procedure", "### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 1e-05\n- train_batch_size: 8\n- eval_batch_size: 64\n- seed: 0\n- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n- lr_scheduler_type: linear\n- num_epochs: 1", "### Training results", "### Framework versions\n\n- Transformers 4.39.3\n- Pytorch 2.2.1\n- Datasets 2.18.0\n- Tokenizers 0.15.2" ]

[ "transformers", "safetensors", "pix2struct", "text2text-generation", "arxiv:1910.09700", "autotrain_compatible", "endpoints_compatible", "region:us" ]

[ "1910.09700" ]

[]

[ "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "TAGS\n#transformers #safetensors #pix2struct #text2text-generation #arxiv-1910.09700 #autotrain_compatible #endpoints_compatible #region-us \n", "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "transformers", "safetensors", "llama", "text-generation", "conversational", "arxiv:1910.09700", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[ "1910.09700" ]

[]

[ "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "TAGS\n#transformers #safetensors #llama #text-generation #conversational #arxiv-1910.09700 #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n", "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "mlx", "safetensors", "phi3", "custom_code", "license:apache-2.0", "region:us" ]

[]

[]

[ "# GreenBitAI/Phi-3-mini-128k-instruct-layer-mix-bpw-3.0-mlx\nThis quantized low-bit model was converted to MLX format from ['GreenBitAI/Phi-3-mini-128k-instruct-layer-mix-bpw-3.0']().\nRefer to the original model card for more details on the model.", "## Use with mlx" ]

[ "TAGS\n#mlx #safetensors #phi3 #custom_code #license-apache-2.0 #region-us \n", "# GreenBitAI/Phi-3-mini-128k-instruct-layer-mix-bpw-3.0-mlx\nThis quantized low-bit model was converted to MLX format from ['GreenBitAI/Phi-3-mini-128k-instruct-layer-mix-bpw-3.0']().\nRefer to the original model card for more details on the model.", "## Use with mlx" ]

[ "transformers", "tensorboard", "safetensors", "wav2vec2", "automatic-speech-recognition", "generated_from_trainer", "dataset:fleurs", "base_model:facebook/wav2vec2-xls-r-300m", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]

[]

[]

[ "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0001\n* train\\_batch\\_size: 4\n* eval\\_batch\\_size: 16\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* lr\\_scheduler\\_warmup\\_steps: 0.1\n* num\\_epochs: 20\n* mixed\\_precision\\_training: Native AMP", "### Training results", "### Framework versions\n\n\n* Transformers 4.40.0\n* Pytorch 2.2.2+cu121\n* Datasets 2.17.1\n* Tokenizers 0.19.1" ]

[ "TAGS\n#transformers #tensorboard #safetensors #wav2vec2 #automatic-speech-recognition #generated_from_trainer #dataset-fleurs #base_model-facebook/wav2vec2-xls-r-300m #license-apache-2.0 #model-index #endpoints_compatible #region-us \n", "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0001\n* train\\_batch\\_size: 4\n* eval\\_batch\\_size: 16\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* lr\\_scheduler\\_warmup\\_steps: 0.1\n* num\\_epochs: 20\n* mixed\\_precision\\_training: Native AMP", "### Training results", "### Framework versions\n\n\n* Transformers 4.40.0\n* Pytorch 2.2.2+cu121\n* Datasets 2.17.1\n* Tokenizers 0.19.1" ]

[ "transformers", "gguf", "merge", "mergekit", "lazymergekit", "kaist-ai/mistral-orpo-capybara-7k", "argilla/CapybaraHermes-2.5-Mistral-7B", "en", "base_model:Yuma42/KangalKhan-Alpha-Sapphiroid-7B-Fixed", "license:apache-2.0", "endpoints_compatible", "region:us" ]

[]

[ "en" ]

[]

[ "TAGS\n#transformers #gguf #merge #mergekit #lazymergekit #kaist-ai/mistral-orpo-capybara-7k #argilla/CapybaraHermes-2.5-Mistral-7B #en #base_model-Yuma42/KangalKhan-Alpha-Sapphiroid-7B-Fixed #license-apache-2.0 #endpoints_compatible #region-us \n" ]

[ "transformers", "pytorch", "llama", "text-generation", "text-generation-inference", "unsloth", "trl", "sft", "conversational", "en", "base_model:TinyLlama/TinyLlama-1.1B-Chat-v1.0", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "region:us" ]

[]

[ "en" ]

[ "# Uploaded model\n\n- Developed by: ppopiolek\n- License: apache-2.0\n- Finetuned from model : TinyLlama/TinyLlama-1.1B-Chat-v1.0\n\nThis llama model was trained 2x faster with Unsloth and Huggingface's TRL library.\n\n<img src=\"URL width=\"200\"/>" ]

[ "TAGS\n#transformers #pytorch #llama #text-generation #text-generation-inference #unsloth #trl #sft #conversational #en #base_model-TinyLlama/TinyLlama-1.1B-Chat-v1.0 #license-apache-2.0 #autotrain_compatible #endpoints_compatible #region-us \n", "# Uploaded model\n\n- Developed by: ppopiolek\n- License: apache-2.0\n- Finetuned from model : TinyLlama/TinyLlama-1.1B-Chat-v1.0\n\nThis llama model was trained 2x faster with Unsloth and Huggingface's TRL library.\n\n<img src=\"URL width=\"200\"/>" ]

[ "gguf", "facebook", "meta", "pytorch", "llama", "llama-3", "text-generation", "en", "license:other", "region:us" ]

[]

[ "en" ]

[]

[ "TAGS\n#gguf #facebook #meta #pytorch #llama #llama-3 #text-generation #en #license-other #region-us \n" ]

[ "transformers", "safetensors", "bert", "text-classification", "arxiv:1910.09700", "autotrain_compatible", "endpoints_compatible", "region:us" ]

[ "1910.09700" ]

[]

[ "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "TAGS\n#transformers #safetensors #bert #text-classification #arxiv-1910.09700 #autotrain_compatible #endpoints_compatible #region-us \n", "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "transformers", "safetensors", "llama", "text-generation", "facebook", "meta", "pytorch", "llama-3", "conversational", "en", "license:other", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "3-bit", "region:us" ]

[]

[ "en" ]

[ "### Use with transformers\n\n\nYou can run conversational inference using the Transformers pipeline abstraction, or by leveraging the Auto classes with the 'generate()' function. Let's see examples of both.", "#### Transformers pipeline", "#### Transformers AutoModelForCausalLM", "### Use with 'llama3'\n\n\nPlease, follow the instructions in the repository\n\n\nTo download Original checkpoints, see the example command below leveraging 'huggingface-cli':\n\n\nFor Hugging Face support, we recommend using transformers or TGI, but a similar command works.\n\n\nHardware and Software\n---------------------\n\n\nTraining Factors We used custom training libraries, Meta's Research SuperCluster, and production clusters for pretraining. Fine-tuning, annotation, and evaluation were also performed on third-party cloud compute.\n\n\nCarbon Footprint Pretraining utilized a cumulative 7.7M GPU hours of computation on hardware of type H100-80GB (TDP of 700W). Estimated total emissions were 2290 tCO2eq, 100% of which were offset by Meta’s sustainability program.\n\n\n\nCO2 emissions during pre-training. Time: total GPU time required for training each model. Power Consumption: peak power capacity per GPU device for the GPUs used adjusted for power usage efficiency. 100% of the emissions are directly offset by Meta's sustainability program, and because we are openly releasing these models, the pretraining costs do not need to be incurred by others.\n\n\nTraining Data\n-------------\n\n\nOverview Llama 3 was pretrained on over 15 trillion tokens of data from publicly available sources. The fine-tuning data includes publicly available instruction datasets, as well as over 10M human-annotated examples. Neither the pretraining nor the fine-tuning datasets include Meta user data.\n\n\nData Freshness The pretraining data has a cutoff of March 2023 for the 7B and December 2023 for the 70B models respectively.\n\n\nBenchmarks\n----------\n\n\nIn this section, we report the results for Llama 3 models on standard automatic benchmarks. For all the evaluations, we use our internal evaluations library. For details on the methodology see here.", "### Base pretrained models", "### Instruction tuned models", "### Responsibility & Safety\n\n\nWe believe that an open approach to AI leads to better, safer products, faster innovation, and a bigger overall market. We are committed to Responsible AI development and took a series of steps to limit misuse and harm and support the open source community.\n\n\nFoundation models are widely capable technologies that are built to be used for a diverse range of applications. They are not designed to meet every developer preference on safety levels for all use cases, out-of-the-box, as those by their nature will differ across different applications.\n\n\nRather, responsible LLM-application deployment is achieved by implementing a series of safety best practices throughout the development of such applications, from the model pre-training, fine-tuning and the deployment of systems composed of safeguards to tailor the safety needs specifically to the use case and audience.\n\n\nAs part of the Llama 3 release, we updated our Responsible Use Guide to outline the steps and best practices for developers to implement model and system level safety for their application. We also provide a set of resources including Meta Llama Guard 2 and Code Shield safeguards. These tools have proven to drastically reduce residual risks of LLM Systems, while maintaining a high level of helpfulness. We encourage developers to tune and deploy these safeguards according to their needs and we provide a reference implementation to get you started.", "#### Llama 3-Instruct\n\n\nAs outlined in the Responsible Use Guide, some trade-off between model helpfulness and model alignment is likely unavoidable. Developers should exercise discretion about how to weigh the benefits of alignment and helpfulness for their specific use case and audience. Developers should be mindful of residual risks when using Llama models and leverage additional safety tools as needed to reach the right safety bar for their use case.\n\n\nSafety\n\n\nFor our instruction tuned model, we conducted extensive red teaming exercises, performed adversarial evaluations and implemented safety mitigations techniques to lower residual risks. As with any Large Language Model, residual risks will likely remain and we recommend that developers assess these risks in the context of their use case. In parallel, we are working with the community to make AI safety benchmark standards transparent, rigorous and interpretable.\n\n\nRefusals\n\n\nIn addition to residual risks, we put a great emphasis on model refusals to benign prompts. Over-refusing not only can impact the user experience but could even be harmful in certain contexts as well. We’ve heard the feedback from the developer community and improved our fine tuning to ensure that Llama 3 is significantly less likely to falsely refuse to answer prompts than Llama 2.\n\n\nWe built internal benchmarks and developed mitigations to limit false refusals making Llama 3 our most helpful model to date.", "#### Responsible release\n\n\nIn addition to responsible use considerations outlined above, we followed a rigorous process that requires us to take extra measures against misuse and critical risks before we make our release decision.\n\n\nMisuse\n\n\nIf you access or use Llama 3, you agree to the Acceptable Use Policy. The most recent copy of this policy can be found at URL", "#### Critical risks\n\n\nCBRNE (Chemical, Biological, Radiological, Nuclear, and high yield Explosives)\n\n\nWe have conducted a two fold assessment of the safety of the model in this area:\n\n\n* Iterative testing during model training to assess the safety of responses related to CBRNE threats and other adversarial risks.\n* Involving external CBRNE experts to conduct an uplift test assessing the ability of the model to accurately provide expert knowledge and reduce barriers to potential CBRNE misuse, by reference to what can be achieved using web search (without the model).", "### Cyber Security\n\n\nWe have evaluated Llama 3 with CyberSecEval, Meta’s cybersecurity safety eval suite, measuring Llama 3’s propensity to suggest insecure code when used as a coding assistant, and Llama 3’s propensity to comply with requests to help carry out cyber attacks, where attacks are defined by the industry standard MITRE ATT&CK cyber attack ontology. On our insecure coding and cyber attacker helpfulness tests, Llama 3 behaved in the same range or safer than models of equivalent coding capability.", "### Child Safety\n\n\nChild Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences.", "### Community\n\n\nGenerative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership in AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository.\n\n\nFinally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community.\n\n\nEthical Considerations and Limitations\n--------------------------------------\n\n\nThe core values of Llama 3 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress.\n\n\nBut Llama 3 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has been in English, and has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3 models, developers should perform safety testing and tuning tailored to their specific applications of the model. As outlined in the Responsible Use Guide, we recommend incorporating Purple Llama solutions into your workflows and specifically Llama Guard which provides a base model to filter input and output prompts to layer system-level safety on top of model-level safety.\n\n\nPlease see the Responsible Use Guide available at URL\n\n\ninstructions\n\n\n@article{llama3modelcard,\n\n\ntitle={Llama 3 Model Card},\n\n\nauthor={AI@Meta},\n\n\nyear={2024},\n\n\nurl = {URL\n\n\n}\n\n\nContributors\n------------\n\n\nAaditya Singh; Aaron Grattafiori; Abhimanyu Dubey; Abhinav Jauhri; Abhinav Pandey; Abhishek Kadian; Adam Kelsey; Adi Gangidi; Ahmad Al-Dahle; Ahuva Goldstand; Aiesha Letman; Ajay Menon; Akhil Mathur; Alan Schelten; Alex Vaughan; Amy Yang; Andrei Lupu; Andres Alvarado; Andrew Gallagher; Andrew Gu; Andrew Ho; Andrew Poulton; Andrew Ryan; Angela Fan; Ankit Ramchandani; Anthony Hartshorn; Archi Mitra; Archie Sravankumar; Artem Korenev; Arun Rao; Ashley Gabriel; Ashwin Bharambe; Assaf Eisenman; Aston Zhang; Aurelien Rodriguez; Austen Gregerson; Ava Spataru; Baptiste Roziere; Ben Maurer; Benjamin Leonhardi; Bernie Huang; Bhargavi Paranjape; Bing Liu; Binh Tang; Bobbie Chern; Brani Stojkovic; Brian Fuller; Catalina Mejia Arenas; Chao Zhou; Charlotte Caucheteux; Chaya Nayak; Ching-Hsiang Chu; Chloe Bi; Chris Cai; Chris Cox; Chris Marra; Chris McConnell; Christian Keller; Christoph Feichtenhofer; Christophe Touret; Chunyang Wu; Corinne Wong; Cristian Canton Ferrer; Damien Allonsius; Daniel Kreymer; Daniel Haziza; Daniel Li; Danielle Pintz; Danny Livshits; Danny Wyatt; David Adkins; David Esiobu; David Xu; Davide Testuggine; Delia David; Devi Parikh; Dhruv Choudhary; Dhruv Mahajan; Diana Liskovich; Diego Garcia-Olano; Diego Perino; Dieuwke Hupkes; Dingkang Wang; Dustin Holland; Egor Lakomkin; Elina Lobanova; Xiaoqing Ellen Tan; Emily Dinan; Eric Smith; Erik Brinkman; Esteban Arcaute; Filip Radenovic; Firat Ozgenel; Francesco Caggioni; Frank Seide; Frank Zhang; Gabriel Synnaeve; Gabriella Schwarz; Gabrielle Lee; Gada Badeer; Georgia Anderson; Graeme Nail; Gregoire Mialon; Guan Pang; Guillem Cucurell; Hailey Nguyen; Hannah Korevaar; Hannah Wang; Haroun Habeeb; Harrison Rudolph; Henry Aspegren; Hu Xu; Hugo Touvron; Iga Kozlowska; Igor Molybog; Igor Tufanov; Iliyan Zarov; Imanol Arrieta Ibarra; Irina-Elena Veliche; Isabel Kloumann; Ishan Misra; Ivan Evtimov; Jacob Xu; Jade Copet; Jake Weissman; Jan Geffert; Jana Vranes; Japhet Asher; Jason Park; Jay Mahadeokar; Jean-Baptiste Gaya; Jeet Shah; Jelmer van der Linde; Jennifer Chan; Jenny Hong; Jenya Lee; Jeremy Fu; Jeremy Teboul; Jianfeng Chi; Jianyu Huang; Jie Wang; Jiecao Yu; Joanna Bitton; Joe Spisak; Joelle Pineau; Jon Carvill; Jongsoo Park; Joseph Rocca; Joshua Johnstun; Junteng Jia; Kalyan Vasuden Alwala; Kam Hou U; Kate Plawiak; Kartikeya Upasani; Kaushik Veeraraghavan; Ke Li; Kenneth Heafield; Kevin Stone; Khalid El-Arini; Krithika Iyer; Kshitiz Malik; Kuenley Chiu; Kunal Bhalla; Kyle Huang; Lakshya Garg; Lauren Rantala-Yeary; Laurens van der Maaten; Lawrence Chen; Leandro Silva; Lee Bell; Lei Zhang; Liang Tan; Louis Martin; Lovish Madaan; Luca Wehrstedt; Lukas Blecher; Luke de Oliveira; Madeline Muzzi; Madian Khabsa; Manav Avlani; Mannat Singh; Manohar Paluri; Mark Zuckerberg; Marcin Kardas; Martynas Mankus; Mathew Oldham; Mathieu Rita; Matthew Lennie; Maya Pavlova; Meghan Keneally; Melanie Kambadur; Mihir Patel; Mikayel Samvelyan; Mike Clark; Mike Lewis; Min Si; Mitesh Kumar Singh; Mo Metanat; Mona Hassan; Naman Goyal; Narjes Torabi; Nicolas Usunier; Nikolay Bashlykov; Nikolay Bogoychev; Niladri Chatterji; Ning Dong; Oliver Aobo Yang; Olivier Duchenne; Onur Celebi; Parth Parekh; Patrick Alrassy; Paul Saab; Pavan Balaji; Pedro Rittner; Pengchuan Zhang; Pengwei Li; Petar Vasic; Peter Weng; Polina Zvyagina; Prajjwal Bhargava; Pratik Dubal; Praveen Krishnan; Punit Singh Koura; Qing He; Rachel Rodriguez; Ragavan Srinivasan; Rahul Mitra; Ramon Calderer; Raymond Li; Robert Stojnic; Roberta Raileanu; Robin Battey; Rocky Wang; Rohit Girdhar; Rohit Patel; Romain Sauvestre; Ronnie Polidoro; Roshan Sumbaly; Ross Taylor; Ruan Silva; Rui Hou; Rui Wang; Russ Howes; Ruty Rinott; Saghar Hosseini; Sai Jayesh Bondu; Samyak Datta; Sanjay Singh; Sara Chugh; Sargun Dhillon; Satadru Pan; Sean Bell; Sergey Edunov; Shaoliang Nie; Sharan Narang; Sharath Raparthy; Shaun Lindsay; Sheng Feng; Sheng Shen; Shenghao Lin; Shiva Shankar; Shruti Bhosale; Shun Zhang; Simon Vandenhende; Sinong Wang; Seohyun Sonia Kim; Soumya Batra; Sten Sootla; Steve Kehoe; Suchin Gururangan; Sumit Gupta; Sunny Virk; Sydney Borodinsky; Tamar Glaser; Tamar Herman; Tamara Best; Tara Fowler; Thomas Georgiou; Thomas Scialom; Tianhe Li; Todor Mihaylov; Tong Xiao; Ujjwal Karn; Vedanuj Goswami; Vibhor Gupta; Vignesh Ramanathan; Viktor Kerkez; Vinay Satish Kumar; Vincent Gonguet; Vish Vogeti; Vlad Poenaru; Vlad Tiberiu Mihailescu; Vladan Petrovic; Vladimir Ivanov; Wei Li; Weiwei Chu; Wenhan Xiong; Wenyin Fu; Wes Bouaziz; Whitney Meers; Will Constable; Xavier Martinet; Xiaojian Wu; Xinbo Gao; Xinfeng Xie; Xuchao Jia; Yaelle Goldschlag; Yann LeCun; Yashesh Gaur; Yasmine Babaei; Ye Qi; Yenda Li; Yi Wen; Yiwen Song; Youngjin Nam; Yuchen Hao; Yuchen Zhang; Yun Wang; Yuning Mao; Yuzi He; Zacharie Delpierre Coudert; Zachary DeVito; Zahra Hankir; Zhaoduo Wen; Zheng Yan; Zhengxing Chen; Zhenyu Yang; Zoe Papakipos" ]

[ "TAGS\n#transformers #safetensors #llama #text-generation #facebook #meta #pytorch #llama-3 #conversational #en #license-other #autotrain_compatible #endpoints_compatible #text-generation-inference #3-bit #region-us \n", "### Use with transformers\n\n\nYou can run conversational inference using the Transformers pipeline abstraction, or by leveraging the Auto classes with the 'generate()' function. Let's see examples of both.", "#### Transformers pipeline", "#### Transformers AutoModelForCausalLM", "### Use with 'llama3'\n\n\nPlease, follow the instructions in the repository\n\n\nTo download Original checkpoints, see the example command below leveraging 'huggingface-cli':\n\n\nFor Hugging Face support, we recommend using transformers or TGI, but a similar command works.\n\n\nHardware and Software\n---------------------\n\n\nTraining Factors We used custom training libraries, Meta's Research SuperCluster, and production clusters for pretraining. Fine-tuning, annotation, and evaluation were also performed on third-party cloud compute.\n\n\nCarbon Footprint Pretraining utilized a cumulative 7.7M GPU hours of computation on hardware of type H100-80GB (TDP of 700W). Estimated total emissions were 2290 tCO2eq, 100% of which were offset by Meta’s sustainability program.\n\n\n\nCO2 emissions during pre-training. Time: total GPU time required for training each model. Power Consumption: peak power capacity per GPU device for the GPUs used adjusted for power usage efficiency. 100% of the emissions are directly offset by Meta's sustainability program, and because we are openly releasing these models, the pretraining costs do not need to be incurred by others.\n\n\nTraining Data\n-------------\n\n\nOverview Llama 3 was pretrained on over 15 trillion tokens of data from publicly available sources. The fine-tuning data includes publicly available instruction datasets, as well as over 10M human-annotated examples. Neither the pretraining nor the fine-tuning datasets include Meta user data.\n\n\nData Freshness The pretraining data has a cutoff of March 2023 for the 7B and December 2023 for the 70B models respectively.\n\n\nBenchmarks\n----------\n\n\nIn this section, we report the results for Llama 3 models on standard automatic benchmarks. For all the evaluations, we use our internal evaluations library. For details on the methodology see here.", "### Base pretrained models", "### Instruction tuned models", "### Responsibility & Safety\n\n\nWe believe that an open approach to AI leads to better, safer products, faster innovation, and a bigger overall market. We are committed to Responsible AI development and took a series of steps to limit misuse and harm and support the open source community.\n\n\nFoundation models are widely capable technologies that are built to be used for a diverse range of applications. They are not designed to meet every developer preference on safety levels for all use cases, out-of-the-box, as those by their nature will differ across different applications.\n\n\nRather, responsible LLM-application deployment is achieved by implementing a series of safety best practices throughout the development of such applications, from the model pre-training, fine-tuning and the deployment of systems composed of safeguards to tailor the safety needs specifically to the use case and audience.\n\n\nAs part of the Llama 3 release, we updated our Responsible Use Guide to outline the steps and best practices for developers to implement model and system level safety for their application. We also provide a set of resources including Meta Llama Guard 2 and Code Shield safeguards. These tools have proven to drastically reduce residual risks of LLM Systems, while maintaining a high level of helpfulness. We encourage developers to tune and deploy these safeguards according to their needs and we provide a reference implementation to get you started.", "#### Llama 3-Instruct\n\n\nAs outlined in the Responsible Use Guide, some trade-off between model helpfulness and model alignment is likely unavoidable. Developers should exercise discretion about how to weigh the benefits of alignment and helpfulness for their specific use case and audience. Developers should be mindful of residual risks when using Llama models and leverage additional safety tools as needed to reach the right safety bar for their use case.\n\n\nSafety\n\n\nFor our instruction tuned model, we conducted extensive red teaming exercises, performed adversarial evaluations and implemented safety mitigations techniques to lower residual risks. As with any Large Language Model, residual risks will likely remain and we recommend that developers assess these risks in the context of their use case. In parallel, we are working with the community to make AI safety benchmark standards transparent, rigorous and interpretable.\n\n\nRefusals\n\n\nIn addition to residual risks, we put a great emphasis on model refusals to benign prompts. Over-refusing not only can impact the user experience but could even be harmful in certain contexts as well. We’ve heard the feedback from the developer community and improved our fine tuning to ensure that Llama 3 is significantly less likely to falsely refuse to answer prompts than Llama 2.\n\n\nWe built internal benchmarks and developed mitigations to limit false refusals making Llama 3 our most helpful model to date.", "#### Responsible release\n\n\nIn addition to responsible use considerations outlined above, we followed a rigorous process that requires us to take extra measures against misuse and critical risks before we make our release decision.\n\n\nMisuse\n\n\nIf you access or use Llama 3, you agree to the Acceptable Use Policy. The most recent copy of this policy can be found at URL", "#### Critical risks\n\n\nCBRNE (Chemical, Biological, Radiological, Nuclear, and high yield Explosives)\n\n\nWe have conducted a two fold assessment of the safety of the model in this area:\n\n\n* Iterative testing during model training to assess the safety of responses related to CBRNE threats and other adversarial risks.\n* Involving external CBRNE experts to conduct an uplift test assessing the ability of the model to accurately provide expert knowledge and reduce barriers to potential CBRNE misuse, by reference to what can be achieved using web search (without the model).", "### Cyber Security\n\n\nWe have evaluated Llama 3 with CyberSecEval, Meta’s cybersecurity safety eval suite, measuring Llama 3’s propensity to suggest insecure code when used as a coding assistant, and Llama 3’s propensity to comply with requests to help carry out cyber attacks, where attacks are defined by the industry standard MITRE ATT&CK cyber attack ontology. On our insecure coding and cyber attacker helpfulness tests, Llama 3 behaved in the same range or safer than models of equivalent coding capability.", "### Child Safety\n\n\nChild Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences.", "### Community\n\n\nGenerative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership in AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository.\n\n\nFinally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community.\n\n\nEthical Considerations and Limitations\n--------------------------------------\n\n\nThe core values of Llama 3 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress.\n\n\nBut Llama 3 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has been in English, and has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3 models, developers should perform safety testing and tuning tailored to their specific applications of the model. As outlined in the Responsible Use Guide, we recommend incorporating Purple Llama solutions into your workflows and specifically Llama Guard which provides a base model to filter input and output prompts to layer system-level safety on top of model-level safety.\n\n\nPlease see the Responsible Use Guide available at URL\n\n\ninstructions\n\n\n@article{llama3modelcard,\n\n\ntitle={Llama 3 Model Card},\n\n\nauthor={AI@Meta},\n\n\nyear={2024},\n\n\nurl = {URL\n\n\n}\n\n\nContributors\n------------\n\n\nAaditya Singh; Aaron Grattafiori; Abhimanyu Dubey; Abhinav Jauhri; Abhinav Pandey; Abhishek Kadian; Adam Kelsey; Adi Gangidi; Ahmad Al-Dahle; Ahuva Goldstand; Aiesha Letman; Ajay Menon; Akhil Mathur; Alan Schelten; Alex Vaughan; Amy Yang; Andrei Lupu; Andres Alvarado; Andrew Gallagher; Andrew Gu; Andrew Ho; Andrew Poulton; Andrew Ryan; Angela Fan; Ankit Ramchandani; Anthony Hartshorn; Archi Mitra; Archie Sravankumar; Artem Korenev; Arun Rao; Ashley Gabriel; Ashwin Bharambe; Assaf Eisenman; Aston Zhang; Aurelien Rodriguez; Austen Gregerson; Ava Spataru; Baptiste Roziere; Ben Maurer; Benjamin Leonhardi; Bernie Huang; Bhargavi Paranjape; Bing Liu; Binh Tang; Bobbie Chern; Brani Stojkovic; Brian Fuller; Catalina Mejia Arenas; Chao Zhou; Charlotte Caucheteux; Chaya Nayak; Ching-Hsiang Chu; Chloe Bi; Chris Cai; Chris Cox; Chris Marra; Chris McConnell; Christian Keller; Christoph Feichtenhofer; Christophe Touret; Chunyang Wu; Corinne Wong; Cristian Canton Ferrer; Damien Allonsius; Daniel Kreymer; Daniel Haziza; Daniel Li; Danielle Pintz; Danny Livshits; Danny Wyatt; David Adkins; David Esiobu; David Xu; Davide Testuggine; Delia David; Devi Parikh; Dhruv Choudhary; Dhruv Mahajan; Diana Liskovich; Diego Garcia-Olano; Diego Perino; Dieuwke Hupkes; Dingkang Wang; Dustin Holland; Egor Lakomkin; Elina Lobanova; Xiaoqing Ellen Tan; Emily Dinan; Eric Smith; Erik Brinkman; Esteban Arcaute; Filip Radenovic; Firat Ozgenel; Francesco Caggioni; Frank Seide; Frank Zhang; Gabriel Synnaeve; Gabriella Schwarz; Gabrielle Lee; Gada Badeer; Georgia Anderson; Graeme Nail; Gregoire Mialon; Guan Pang; Guillem Cucurell; Hailey Nguyen; Hannah Korevaar; Hannah Wang; Haroun Habeeb; Harrison Rudolph; Henry Aspegren; Hu Xu; Hugo Touvron; Iga Kozlowska; Igor Molybog; Igor Tufanov; Iliyan Zarov; Imanol Arrieta Ibarra; Irina-Elena Veliche; Isabel Kloumann; Ishan Misra; Ivan Evtimov; Jacob Xu; Jade Copet; Jake Weissman; Jan Geffert; Jana Vranes; Japhet Asher; Jason Park; Jay Mahadeokar; Jean-Baptiste Gaya; Jeet Shah; Jelmer van der Linde; Jennifer Chan; Jenny Hong; Jenya Lee; Jeremy Fu; Jeremy Teboul; Jianfeng Chi; Jianyu Huang; Jie Wang; Jiecao Yu; Joanna Bitton; Joe Spisak; Joelle Pineau; Jon Carvill; Jongsoo Park; Joseph Rocca; Joshua Johnstun; Junteng Jia; Kalyan Vasuden Alwala; Kam Hou U; Kate Plawiak; Kartikeya Upasani; Kaushik Veeraraghavan; Ke Li; Kenneth Heafield; Kevin Stone; Khalid El-Arini; Krithika Iyer; Kshitiz Malik; Kuenley Chiu; Kunal Bhalla; Kyle Huang; Lakshya Garg; Lauren Rantala-Yeary; Laurens van der Maaten; Lawrence Chen; Leandro Silva; Lee Bell; Lei Zhang; Liang Tan; Louis Martin; Lovish Madaan; Luca Wehrstedt; Lukas Blecher; Luke de Oliveira; Madeline Muzzi; Madian Khabsa; Manav Avlani; Mannat Singh; Manohar Paluri; Mark Zuckerberg; Marcin Kardas; Martynas Mankus; Mathew Oldham; Mathieu Rita; Matthew Lennie; Maya Pavlova; Meghan Keneally; Melanie Kambadur; Mihir Patel; Mikayel Samvelyan; Mike Clark; Mike Lewis; Min Si; Mitesh Kumar Singh; Mo Metanat; Mona Hassan; Naman Goyal; Narjes Torabi; Nicolas Usunier; Nikolay Bashlykov; Nikolay Bogoychev; Niladri Chatterji; Ning Dong; Oliver Aobo Yang; Olivier Duchenne; Onur Celebi; Parth Parekh; Patrick Alrassy; Paul Saab; Pavan Balaji; Pedro Rittner; Pengchuan Zhang; Pengwei Li; Petar Vasic; Peter Weng; Polina Zvyagina; Prajjwal Bhargava; Pratik Dubal; Praveen Krishnan; Punit Singh Koura; Qing He; Rachel Rodriguez; Ragavan Srinivasan; Rahul Mitra; Ramon Calderer; Raymond Li; Robert Stojnic; Roberta Raileanu; Robin Battey; Rocky Wang; Rohit Girdhar; Rohit Patel; Romain Sauvestre; Ronnie Polidoro; Roshan Sumbaly; Ross Taylor; Ruan Silva; Rui Hou; Rui Wang; Russ Howes; Ruty Rinott; Saghar Hosseini; Sai Jayesh Bondu; Samyak Datta; Sanjay Singh; Sara Chugh; Sargun Dhillon; Satadru Pan; Sean Bell; Sergey Edunov; Shaoliang Nie; Sharan Narang; Sharath Raparthy; Shaun Lindsay; Sheng Feng; Sheng Shen; Shenghao Lin; Shiva Shankar; Shruti Bhosale; Shun Zhang; Simon Vandenhende; Sinong Wang; Seohyun Sonia Kim; Soumya Batra; Sten Sootla; Steve Kehoe; Suchin Gururangan; Sumit Gupta; Sunny Virk; Sydney Borodinsky; Tamar Glaser; Tamar Herman; Tamara Best; Tara Fowler; Thomas Georgiou; Thomas Scialom; Tianhe Li; Todor Mihaylov; Tong Xiao; Ujjwal Karn; Vedanuj Goswami; Vibhor Gupta; Vignesh Ramanathan; Viktor Kerkez; Vinay Satish Kumar; Vincent Gonguet; Vish Vogeti; Vlad Poenaru; Vlad Tiberiu Mihailescu; Vladan Petrovic; Vladimir Ivanov; Wei Li; Weiwei Chu; Wenhan Xiong; Wenyin Fu; Wes Bouaziz; Whitney Meers; Will Constable; Xavier Martinet; Xiaojian Wu; Xinbo Gao; Xinfeng Xie; Xuchao Jia; Yaelle Goldschlag; Yann LeCun; Yashesh Gaur; Yasmine Babaei; Ye Qi; Yenda Li; Yi Wen; Yiwen Song; Youngjin Nam; Yuchen Hao; Yuchen Zhang; Yun Wang; Yuning Mao; Yuzi He; Zacharie Delpierre Coudert; Zachary DeVito; Zahra Hankir; Zhaoduo Wen; Zheng Yan; Zhengxing Chen; Zhenyu Yang; Zoe Papakipos" ]

[ "transformers", "safetensors", "stablelm", "text-generation", "conversational", "arxiv:1910.09700", "autotrain_compatible", "endpoints_compatible", "region:us" ]

[ "1910.09700" ]

[]

[ "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "TAGS\n#transformers #safetensors #stablelm #text-generation #conversational #arxiv-1910.09700 #autotrain_compatible #endpoints_compatible #region-us \n", "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "mlx", "safetensors", "phi3", "custom_code", "license:apache-2.0", "region:us" ]

[]

[]

[ "# GreenBitAI/Phi-3-mini-128k-instruct-layer-mix-bpw-2.5-mlx\nThis quantized low-bit model was converted to MLX format from ['GreenBitAI/Phi-3-mini-128k-instruct-layer-mix-bpw-2.5']().\nRefer to the original model card for more details on the model.", "## Use with mlx" ]

[ "TAGS\n#mlx #safetensors #phi3 #custom_code #license-apache-2.0 #region-us \n", "# GreenBitAI/Phi-3-mini-128k-instruct-layer-mix-bpw-2.5-mlx\nThis quantized low-bit model was converted to MLX format from ['GreenBitAI/Phi-3-mini-128k-instruct-layer-mix-bpw-2.5']().\nRefer to the original model card for more details on the model.", "## Use with mlx" ]

[ "transformers", "safetensors", "llama", "text-generation", "facebook", "meta", "pytorch", "llama-3", "conversational", "en", "license:other", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "4-bit", "region:us" ]

[]

[ "en" ]

[ "### Use with transformers\n\n\nYou can run conversational inference using the Transformers pipeline abstraction, or by leveraging the Auto classes with the 'generate()' function. Let's see examples of both.", "#### Transformers pipeline", "#### Transformers AutoModelForCausalLM", "### Use with 'llama3'\n\n\nPlease, follow the instructions in the repository\n\n\nTo download Original checkpoints, see the example command below leveraging 'huggingface-cli':\n\n\nFor Hugging Face support, we recommend using transformers or TGI, but a similar command works.\n\n\nHardware and Software\n---------------------\n\n\nTraining Factors We used custom training libraries, Meta's Research SuperCluster, and production clusters for pretraining. Fine-tuning, annotation, and evaluation were also performed on third-party cloud compute.\n\n\nCarbon Footprint Pretraining utilized a cumulative 7.7M GPU hours of computation on hardware of type H100-80GB (TDP of 700W). Estimated total emissions were 2290 tCO2eq, 100% of which were offset by Meta’s sustainability program.\n\n\n\nCO2 emissions during pre-training. Time: total GPU time required for training each model. Power Consumption: peak power capacity per GPU device for the GPUs used adjusted for power usage efficiency. 100% of the emissions are directly offset by Meta's sustainability program, and because we are openly releasing these models, the pretraining costs do not need to be incurred by others.\n\n\nTraining Data\n-------------\n\n\nOverview Llama 3 was pretrained on over 15 trillion tokens of data from publicly available sources. The fine-tuning data includes publicly available instruction datasets, as well as over 10M human-annotated examples. Neither the pretraining nor the fine-tuning datasets include Meta user data.\n\n\nData Freshness The pretraining data has a cutoff of March 2023 for the 7B and December 2023 for the 70B models respectively.\n\n\nBenchmarks\n----------\n\n\nIn this section, we report the results for Llama 3 models on standard automatic benchmarks. For all the evaluations, we use our internal evaluations library. For details on the methodology see here.", "### Base pretrained models", "### Instruction tuned models", "### Responsibility & Safety\n\n\nWe believe that an open approach to AI leads to better, safer products, faster innovation, and a bigger overall market. We are committed to Responsible AI development and took a series of steps to limit misuse and harm and support the open source community.\n\n\nFoundation models are widely capable technologies that are built to be used for a diverse range of applications. They are not designed to meet every developer preference on safety levels for all use cases, out-of-the-box, as those by their nature will differ across different applications.\n\n\nRather, responsible LLM-application deployment is achieved by implementing a series of safety best practices throughout the development of such applications, from the model pre-training, fine-tuning and the deployment of systems composed of safeguards to tailor the safety needs specifically to the use case and audience.\n\n\nAs part of the Llama 3 release, we updated our Responsible Use Guide to outline the steps and best practices for developers to implement model and system level safety for their application. We also provide a set of resources including Meta Llama Guard 2 and Code Shield safeguards. These tools have proven to drastically reduce residual risks of LLM Systems, while maintaining a high level of helpfulness. We encourage developers to tune and deploy these safeguards according to their needs and we provide a reference implementation to get you started.", "#### Llama 3-Instruct\n\n\nAs outlined in the Responsible Use Guide, some trade-off between model helpfulness and model alignment is likely unavoidable. Developers should exercise discretion about how to weigh the benefits of alignment and helpfulness for their specific use case and audience. Developers should be mindful of residual risks when using Llama models and leverage additional safety tools as needed to reach the right safety bar for their use case.\n\n\nSafety\n\n\nFor our instruction tuned model, we conducted extensive red teaming exercises, performed adversarial evaluations and implemented safety mitigations techniques to lower residual risks. As with any Large Language Model, residual risks will likely remain and we recommend that developers assess these risks in the context of their use case. In parallel, we are working with the community to make AI safety benchmark standards transparent, rigorous and interpretable.\n\n\nRefusals\n\n\nIn addition to residual risks, we put a great emphasis on model refusals to benign prompts. Over-refusing not only can impact the user experience but could even be harmful in certain contexts as well. We’ve heard the feedback from the developer community and improved our fine tuning to ensure that Llama 3 is significantly less likely to falsely refuse to answer prompts than Llama 2.\n\n\nWe built internal benchmarks and developed mitigations to limit false refusals making Llama 3 our most helpful model to date.", "#### Responsible release\n\n\nIn addition to responsible use considerations outlined above, we followed a rigorous process that requires us to take extra measures against misuse and critical risks before we make our release decision.\n\n\nMisuse\n\n\nIf you access or use Llama 3, you agree to the Acceptable Use Policy. The most recent copy of this policy can be found at URL", "#### Critical risks\n\n\nCBRNE (Chemical, Biological, Radiological, Nuclear, and high yield Explosives)\n\n\nWe have conducted a two fold assessment of the safety of the model in this area:\n\n\n* Iterative testing during model training to assess the safety of responses related to CBRNE threats and other adversarial risks.\n* Involving external CBRNE experts to conduct an uplift test assessing the ability of the model to accurately provide expert knowledge and reduce barriers to potential CBRNE misuse, by reference to what can be achieved using web search (without the model).", "### Cyber Security\n\n\nWe have evaluated Llama 3 with CyberSecEval, Meta’s cybersecurity safety eval suite, measuring Llama 3’s propensity to suggest insecure code when used as a coding assistant, and Llama 3’s propensity to comply with requests to help carry out cyber attacks, where attacks are defined by the industry standard MITRE ATT&CK cyber attack ontology. On our insecure coding and cyber attacker helpfulness tests, Llama 3 behaved in the same range or safer than models of equivalent coding capability.", "### Child Safety\n\n\nChild Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences.", "### Community\n\n\nGenerative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership in AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository.\n\n\nFinally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community.\n\n\nEthical Considerations and Limitations\n--------------------------------------\n\n\nThe core values of Llama 3 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress.\n\n\nBut Llama 3 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has been in English, and has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3 models, developers should perform safety testing and tuning tailored to their specific applications of the model. As outlined in the Responsible Use Guide, we recommend incorporating Purple Llama solutions into your workflows and specifically Llama Guard which provides a base model to filter input and output prompts to layer system-level safety on top of model-level safety.\n\n\nPlease see the Responsible Use Guide available at URL\n\n\ninstructions\n\n\n@article{llama3modelcard,\n\n\ntitle={Llama 3 Model Card},\n\n\nauthor={AI@Meta},\n\n\nyear={2024},\n\n\nurl = {URL\n\n\n}\n\n\nContributors\n------------\n\n\nAaditya Singh; Aaron Grattafiori; Abhimanyu Dubey; Abhinav Jauhri; Abhinav Pandey; Abhishek Kadian; Adam Kelsey; Adi Gangidi; Ahmad Al-Dahle; Ahuva Goldstand; Aiesha Letman; Ajay Menon; Akhil Mathur; Alan Schelten; Alex Vaughan; Amy Yang; Andrei Lupu; Andres Alvarado; Andrew Gallagher; Andrew Gu; Andrew Ho; Andrew Poulton; Andrew Ryan; Angela Fan; Ankit Ramchandani; Anthony Hartshorn; Archi Mitra; Archie Sravankumar; Artem Korenev; Arun Rao; Ashley Gabriel; Ashwin Bharambe; Assaf Eisenman; Aston Zhang; Aurelien Rodriguez; Austen Gregerson; Ava Spataru; Baptiste Roziere; Ben Maurer; Benjamin Leonhardi; Bernie Huang; Bhargavi Paranjape; Bing Liu; Binh Tang; Bobbie Chern; Brani Stojkovic; Brian Fuller; Catalina Mejia Arenas; Chao Zhou; Charlotte Caucheteux; Chaya Nayak; Ching-Hsiang Chu; Chloe Bi; Chris Cai; Chris Cox; Chris Marra; Chris McConnell; Christian Keller; Christoph Feichtenhofer; Christophe Touret; Chunyang Wu; Corinne Wong; Cristian Canton Ferrer; Damien Allonsius; Daniel Kreymer; Daniel Haziza; Daniel Li; Danielle Pintz; Danny Livshits; Danny Wyatt; David Adkins; David Esiobu; David Xu; Davide Testuggine; Delia David; Devi Parikh; Dhruv Choudhary; Dhruv Mahajan; Diana Liskovich; Diego Garcia-Olano; Diego Perino; Dieuwke Hupkes; Dingkang Wang; Dustin Holland; Egor Lakomkin; Elina Lobanova; Xiaoqing Ellen Tan; Emily Dinan; Eric Smith; Erik Brinkman; Esteban Arcaute; Filip Radenovic; Firat Ozgenel; Francesco Caggioni; Frank Seide; Frank Zhang; Gabriel Synnaeve; Gabriella Schwarz; Gabrielle Lee; Gada Badeer; Georgia Anderson; Graeme Nail; Gregoire Mialon; Guan Pang; Guillem Cucurell; Hailey Nguyen; Hannah Korevaar; Hannah Wang; Haroun Habeeb; Harrison Rudolph; Henry Aspegren; Hu Xu; Hugo Touvron; Iga Kozlowska; Igor Molybog; Igor Tufanov; Iliyan Zarov; Imanol Arrieta Ibarra; Irina-Elena Veliche; Isabel Kloumann; Ishan Misra; Ivan Evtimov; Jacob Xu; Jade Copet; Jake Weissman; Jan Geffert; Jana Vranes; Japhet Asher; Jason Park; Jay Mahadeokar; Jean-Baptiste Gaya; Jeet Shah; Jelmer van der Linde; Jennifer Chan; Jenny Hong; Jenya Lee; Jeremy Fu; Jeremy Teboul; Jianfeng Chi; Jianyu Huang; Jie Wang; Jiecao Yu; Joanna Bitton; Joe Spisak; Joelle Pineau; Jon Carvill; Jongsoo Park; Joseph Rocca; Joshua Johnstun; Junteng Jia; Kalyan Vasuden Alwala; Kam Hou U; Kate Plawiak; Kartikeya Upasani; Kaushik Veeraraghavan; Ke Li; Kenneth Heafield; Kevin Stone; Khalid El-Arini; Krithika Iyer; Kshitiz Malik; Kuenley Chiu; Kunal Bhalla; Kyle Huang; Lakshya Garg; Lauren Rantala-Yeary; Laurens van der Maaten; Lawrence Chen; Leandro Silva; Lee Bell; Lei Zhang; Liang Tan; Louis Martin; Lovish Madaan; Luca Wehrstedt; Lukas Blecher; Luke de Oliveira; Madeline Muzzi; Madian Khabsa; Manav Avlani; Mannat Singh; Manohar Paluri; Mark Zuckerberg; Marcin Kardas; Martynas Mankus; Mathew Oldham; Mathieu Rita; Matthew Lennie; Maya Pavlova; Meghan Keneally; Melanie Kambadur; Mihir Patel; Mikayel Samvelyan; Mike Clark; Mike Lewis; Min Si; Mitesh Kumar Singh; Mo Metanat; Mona Hassan; Naman Goyal; Narjes Torabi; Nicolas Usunier; Nikolay Bashlykov; Nikolay Bogoychev; Niladri Chatterji; Ning Dong; Oliver Aobo Yang; Olivier Duchenne; Onur Celebi; Parth Parekh; Patrick Alrassy; Paul Saab; Pavan Balaji; Pedro Rittner; Pengchuan Zhang; Pengwei Li; Petar Vasic; Peter Weng; Polina Zvyagina; Prajjwal Bhargava; Pratik Dubal; Praveen Krishnan; Punit Singh Koura; Qing He; Rachel Rodriguez; Ragavan Srinivasan; Rahul Mitra; Ramon Calderer; Raymond Li; Robert Stojnic; Roberta Raileanu; Robin Battey; Rocky Wang; Rohit Girdhar; Rohit Patel; Romain Sauvestre; Ronnie Polidoro; Roshan Sumbaly; Ross Taylor; Ruan Silva; Rui Hou; Rui Wang; Russ Howes; Ruty Rinott; Saghar Hosseini; Sai Jayesh Bondu; Samyak Datta; Sanjay Singh; Sara Chugh; Sargun Dhillon; Satadru Pan; Sean Bell; Sergey Edunov; Shaoliang Nie; Sharan Narang; Sharath Raparthy; Shaun Lindsay; Sheng Feng; Sheng Shen; Shenghao Lin; Shiva Shankar; Shruti Bhosale; Shun Zhang; Simon Vandenhende; Sinong Wang; Seohyun Sonia Kim; Soumya Batra; Sten Sootla; Steve Kehoe; Suchin Gururangan; Sumit Gupta; Sunny Virk; Sydney Borodinsky; Tamar Glaser; Tamar Herman; Tamara Best; Tara Fowler; Thomas Georgiou; Thomas Scialom; Tianhe Li; Todor Mihaylov; Tong Xiao; Ujjwal Karn; Vedanuj Goswami; Vibhor Gupta; Vignesh Ramanathan; Viktor Kerkez; Vinay Satish Kumar; Vincent Gonguet; Vish Vogeti; Vlad Poenaru; Vlad Tiberiu Mihailescu; Vladan Petrovic; Vladimir Ivanov; Wei Li; Weiwei Chu; Wenhan Xiong; Wenyin Fu; Wes Bouaziz; Whitney Meers; Will Constable; Xavier Martinet; Xiaojian Wu; Xinbo Gao; Xinfeng Xie; Xuchao Jia; Yaelle Goldschlag; Yann LeCun; Yashesh Gaur; Yasmine Babaei; Ye Qi; Yenda Li; Yi Wen; Yiwen Song; Youngjin Nam; Yuchen Hao; Yuchen Zhang; Yun Wang; Yuning Mao; Yuzi He; Zacharie Delpierre Coudert; Zachary DeVito; Zahra Hankir; Zhaoduo Wen; Zheng Yan; Zhengxing Chen; Zhenyu Yang; Zoe Papakipos" ]

[ "TAGS\n#transformers #safetensors #llama #text-generation #facebook #meta #pytorch #llama-3 #conversational #en #license-other #autotrain_compatible #endpoints_compatible #text-generation-inference #4-bit #region-us \n", "### Use with transformers\n\n\nYou can run conversational inference using the Transformers pipeline abstraction, or by leveraging the Auto classes with the 'generate()' function. Let's see examples of both.", "#### Transformers pipeline", "#### Transformers AutoModelForCausalLM", "### Use with 'llama3'\n\n\nPlease, follow the instructions in the repository\n\n\nTo download Original checkpoints, see the example command below leveraging 'huggingface-cli':\n\n\nFor Hugging Face support, we recommend using transformers or TGI, but a similar command works.\n\n\nHardware and Software\n---------------------\n\n\nTraining Factors We used custom training libraries, Meta's Research SuperCluster, and production clusters for pretraining. Fine-tuning, annotation, and evaluation were also performed on third-party cloud compute.\n\n\nCarbon Footprint Pretraining utilized a cumulative 7.7M GPU hours of computation on hardware of type H100-80GB (TDP of 700W). Estimated total emissions were 2290 tCO2eq, 100% of which were offset by Meta’s sustainability program.\n\n\n\nCO2 emissions during pre-training. Time: total GPU time required for training each model. Power Consumption: peak power capacity per GPU device for the GPUs used adjusted for power usage efficiency. 100% of the emissions are directly offset by Meta's sustainability program, and because we are openly releasing these models, the pretraining costs do not need to be incurred by others.\n\n\nTraining Data\n-------------\n\n\nOverview Llama 3 was pretrained on over 15 trillion tokens of data from publicly available sources. The fine-tuning data includes publicly available instruction datasets, as well as over 10M human-annotated examples. Neither the pretraining nor the fine-tuning datasets include Meta user data.\n\n\nData Freshness The pretraining data has a cutoff of March 2023 for the 7B and December 2023 for the 70B models respectively.\n\n\nBenchmarks\n----------\n\n\nIn this section, we report the results for Llama 3 models on standard automatic benchmarks. For all the evaluations, we use our internal evaluations library. For details on the methodology see here.", "### Base pretrained models", "### Instruction tuned models", "### Responsibility & Safety\n\n\nWe believe that an open approach to AI leads to better, safer products, faster innovation, and a bigger overall market. We are committed to Responsible AI development and took a series of steps to limit misuse and harm and support the open source community.\n\n\nFoundation models are widely capable technologies that are built to be used for a diverse range of applications. They are not designed to meet every developer preference on safety levels for all use cases, out-of-the-box, as those by their nature will differ across different applications.\n\n\nRather, responsible LLM-application deployment is achieved by implementing a series of safety best practices throughout the development of such applications, from the model pre-training, fine-tuning and the deployment of systems composed of safeguards to tailor the safety needs specifically to the use case and audience.\n\n\nAs part of the Llama 3 release, we updated our Responsible Use Guide to outline the steps and best practices for developers to implement model and system level safety for their application. We also provide a set of resources including Meta Llama Guard 2 and Code Shield safeguards. These tools have proven to drastically reduce residual risks of LLM Systems, while maintaining a high level of helpfulness. We encourage developers to tune and deploy these safeguards according to their needs and we provide a reference implementation to get you started.", "#### Llama 3-Instruct\n\n\nAs outlined in the Responsible Use Guide, some trade-off between model helpfulness and model alignment is likely unavoidable. Developers should exercise discretion about how to weigh the benefits of alignment and helpfulness for their specific use case and audience. Developers should be mindful of residual risks when using Llama models and leverage additional safety tools as needed to reach the right safety bar for their use case.\n\n\nSafety\n\n\nFor our instruction tuned model, we conducted extensive red teaming exercises, performed adversarial evaluations and implemented safety mitigations techniques to lower residual risks. As with any Large Language Model, residual risks will likely remain and we recommend that developers assess these risks in the context of their use case. In parallel, we are working with the community to make AI safety benchmark standards transparent, rigorous and interpretable.\n\n\nRefusals\n\n\nIn addition to residual risks, we put a great emphasis on model refusals to benign prompts. Over-refusing not only can impact the user experience but could even be harmful in certain contexts as well. We’ve heard the feedback from the developer community and improved our fine tuning to ensure that Llama 3 is significantly less likely to falsely refuse to answer prompts than Llama 2.\n\n\nWe built internal benchmarks and developed mitigations to limit false refusals making Llama 3 our most helpful model to date.", "#### Responsible release\n\n\nIn addition to responsible use considerations outlined above, we followed a rigorous process that requires us to take extra measures against misuse and critical risks before we make our release decision.\n\n\nMisuse\n\n\nIf you access or use Llama 3, you agree to the Acceptable Use Policy. The most recent copy of this policy can be found at URL", "#### Critical risks\n\n\nCBRNE (Chemical, Biological, Radiological, Nuclear, and high yield Explosives)\n\n\nWe have conducted a two fold assessment of the safety of the model in this area:\n\n\n* Iterative testing during model training to assess the safety of responses related to CBRNE threats and other adversarial risks.\n* Involving external CBRNE experts to conduct an uplift test assessing the ability of the model to accurately provide expert knowledge and reduce barriers to potential CBRNE misuse, by reference to what can be achieved using web search (without the model).", "### Cyber Security\n\n\nWe have evaluated Llama 3 with CyberSecEval, Meta’s cybersecurity safety eval suite, measuring Llama 3’s propensity to suggest insecure code when used as a coding assistant, and Llama 3’s propensity to comply with requests to help carry out cyber attacks, where attacks are defined by the industry standard MITRE ATT&CK cyber attack ontology. On our insecure coding and cyber attacker helpfulness tests, Llama 3 behaved in the same range or safer than models of equivalent coding capability.", "### Child Safety\n\n\nChild Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences.", "### Community\n\n\nGenerative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership in AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository.\n\n\nFinally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community.\n\n\nEthical Considerations and Limitations\n--------------------------------------\n\n\nThe core values of Llama 3 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress.\n\n\nBut Llama 3 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has been in English, and has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3 models, developers should perform safety testing and tuning tailored to their specific applications of the model. As outlined in the Responsible Use Guide, we recommend incorporating Purple Llama solutions into your workflows and specifically Llama Guard which provides a base model to filter input and output prompts to layer system-level safety on top of model-level safety.\n\n\nPlease see the Responsible Use Guide available at URL\n\n\ninstructions\n\n\n@article{llama3modelcard,\n\n\ntitle={Llama 3 Model Card},\n\n\nauthor={AI@Meta},\n\n\nyear={2024},\n\n\nurl = {URL\n\n\n}\n\n\nContributors\n------------\n\n\nAaditya Singh; Aaron Grattafiori; Abhimanyu Dubey; Abhinav Jauhri; Abhinav Pandey; Abhishek Kadian; Adam Kelsey; Adi Gangidi; Ahmad Al-Dahle; Ahuva Goldstand; Aiesha Letman; Ajay Menon; Akhil Mathur; Alan Schelten; Alex Vaughan; Amy Yang; Andrei Lupu; Andres Alvarado; Andrew Gallagher; Andrew Gu; Andrew Ho; Andrew Poulton; Andrew Ryan; Angela Fan; Ankit Ramchandani; Anthony Hartshorn; Archi Mitra; Archie Sravankumar; Artem Korenev; Arun Rao; Ashley Gabriel; Ashwin Bharambe; Assaf Eisenman; Aston Zhang; Aurelien Rodriguez; Austen Gregerson; Ava Spataru; Baptiste Roziere; Ben Maurer; Benjamin Leonhardi; Bernie Huang; Bhargavi Paranjape; Bing Liu; Binh Tang; Bobbie Chern; Brani Stojkovic; Brian Fuller; Catalina Mejia Arenas; Chao Zhou; Charlotte Caucheteux; Chaya Nayak; Ching-Hsiang Chu; Chloe Bi; Chris Cai; Chris Cox; Chris Marra; Chris McConnell; Christian Keller; Christoph Feichtenhofer; Christophe Touret; Chunyang Wu; Corinne Wong; Cristian Canton Ferrer; Damien Allonsius; Daniel Kreymer; Daniel Haziza; Daniel Li; Danielle Pintz; Danny Livshits; Danny Wyatt; David Adkins; David Esiobu; David Xu; Davide Testuggine; Delia David; Devi Parikh; Dhruv Choudhary; Dhruv Mahajan; Diana Liskovich; Diego Garcia-Olano; Diego Perino; Dieuwke Hupkes; Dingkang Wang; Dustin Holland; Egor Lakomkin; Elina Lobanova; Xiaoqing Ellen Tan; Emily Dinan; Eric Smith; Erik Brinkman; Esteban Arcaute; Filip Radenovic; Firat Ozgenel; Francesco Caggioni; Frank Seide; Frank Zhang; Gabriel Synnaeve; Gabriella Schwarz; Gabrielle Lee; Gada Badeer; Georgia Anderson; Graeme Nail; Gregoire Mialon; Guan Pang; Guillem Cucurell; Hailey Nguyen; Hannah Korevaar; Hannah Wang; Haroun Habeeb; Harrison Rudolph; Henry Aspegren; Hu Xu; Hugo Touvron; Iga Kozlowska; Igor Molybog; Igor Tufanov; Iliyan Zarov; Imanol Arrieta Ibarra; Irina-Elena Veliche; Isabel Kloumann; Ishan Misra; Ivan Evtimov; Jacob Xu; Jade Copet; Jake Weissman; Jan Geffert; Jana Vranes; Japhet Asher; Jason Park; Jay Mahadeokar; Jean-Baptiste Gaya; Jeet Shah; Jelmer van der Linde; Jennifer Chan; Jenny Hong; Jenya Lee; Jeremy Fu; Jeremy Teboul; Jianfeng Chi; Jianyu Huang; Jie Wang; Jiecao Yu; Joanna Bitton; Joe Spisak; Joelle Pineau; Jon Carvill; Jongsoo Park; Joseph Rocca; Joshua Johnstun; Junteng Jia; Kalyan Vasuden Alwala; Kam Hou U; Kate Plawiak; Kartikeya Upasani; Kaushik Veeraraghavan; Ke Li; Kenneth Heafield; Kevin Stone; Khalid El-Arini; Krithika Iyer; Kshitiz Malik; Kuenley Chiu; Kunal Bhalla; Kyle Huang; Lakshya Garg; Lauren Rantala-Yeary; Laurens van der Maaten; Lawrence Chen; Leandro Silva; Lee Bell; Lei Zhang; Liang Tan; Louis Martin; Lovish Madaan; Luca Wehrstedt; Lukas Blecher; Luke de Oliveira; Madeline Muzzi; Madian Khabsa; Manav Avlani; Mannat Singh; Manohar Paluri; Mark Zuckerberg; Marcin Kardas; Martynas Mankus; Mathew Oldham; Mathieu Rita; Matthew Lennie; Maya Pavlova; Meghan Keneally; Melanie Kambadur; Mihir Patel; Mikayel Samvelyan; Mike Clark; Mike Lewis; Min Si; Mitesh Kumar Singh; Mo Metanat; Mona Hassan; Naman Goyal; Narjes Torabi; Nicolas Usunier; Nikolay Bashlykov; Nikolay Bogoychev; Niladri Chatterji; Ning Dong; Oliver Aobo Yang; Olivier Duchenne; Onur Celebi; Parth Parekh; Patrick Alrassy; Paul Saab; Pavan Balaji; Pedro Rittner; Pengchuan Zhang; Pengwei Li; Petar Vasic; Peter Weng; Polina Zvyagina; Prajjwal Bhargava; Pratik Dubal; Praveen Krishnan; Punit Singh Koura; Qing He; Rachel Rodriguez; Ragavan Srinivasan; Rahul Mitra; Ramon Calderer; Raymond Li; Robert Stojnic; Roberta Raileanu; Robin Battey; Rocky Wang; Rohit Girdhar; Rohit Patel; Romain Sauvestre; Ronnie Polidoro; Roshan Sumbaly; Ross Taylor; Ruan Silva; Rui Hou; Rui Wang; Russ Howes; Ruty Rinott; Saghar Hosseini; Sai Jayesh Bondu; Samyak Datta; Sanjay Singh; Sara Chugh; Sargun Dhillon; Satadru Pan; Sean Bell; Sergey Edunov; Shaoliang Nie; Sharan Narang; Sharath Raparthy; Shaun Lindsay; Sheng Feng; Sheng Shen; Shenghao Lin; Shiva Shankar; Shruti Bhosale; Shun Zhang; Simon Vandenhende; Sinong Wang; Seohyun Sonia Kim; Soumya Batra; Sten Sootla; Steve Kehoe; Suchin Gururangan; Sumit Gupta; Sunny Virk; Sydney Borodinsky; Tamar Glaser; Tamar Herman; Tamara Best; Tara Fowler; Thomas Georgiou; Thomas Scialom; Tianhe Li; Todor Mihaylov; Tong Xiao; Ujjwal Karn; Vedanuj Goswami; Vibhor Gupta; Vignesh Ramanathan; Viktor Kerkez; Vinay Satish Kumar; Vincent Gonguet; Vish Vogeti; Vlad Poenaru; Vlad Tiberiu Mihailescu; Vladan Petrovic; Vladimir Ivanov; Wei Li; Weiwei Chu; Wenhan Xiong; Wenyin Fu; Wes Bouaziz; Whitney Meers; Will Constable; Xavier Martinet; Xiaojian Wu; Xinbo Gao; Xinfeng Xie; Xuchao Jia; Yaelle Goldschlag; Yann LeCun; Yashesh Gaur; Yasmine Babaei; Ye Qi; Yenda Li; Yi Wen; Yiwen Song; Youngjin Nam; Yuchen Hao; Yuchen Zhang; Yun Wang; Yuning Mao; Yuzi He; Zacharie Delpierre Coudert; Zachary DeVito; Zahra Hankir; Zhaoduo Wen; Zheng Yan; Zhengxing Chen; Zhenyu Yang; Zoe Papakipos" ]

[ "peft", "safetensors", "trl", "sft", "generated_from_trainer", "dataset:generator", "base_model:TinyPixel/Llama-2-7B-bf16-sharded", "region:us" ]

[]

[]

[ "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0001\n* train\\_batch\\_size: 4\n* eval\\_batch\\_size: 8\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: constant\n* lr\\_scheduler\\_warmup\\_ratio: 0.03\n* num\\_epochs: 3", "### Training results", "### Framework versions\n\n\n* PEFT 0.10.1.dev0\n* Transformers 4.40.1\n* Pytorch 2.2.1+cu121\n* Datasets 2.19.0\n* Tokenizers 0.19.1" ]

[ "TAGS\n#peft #safetensors #trl #sft #generated_from_trainer #dataset-generator #base_model-TinyPixel/Llama-2-7B-bf16-sharded #region-us \n", "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0001\n* train\\_batch\\_size: 4\n* eval\\_batch\\_size: 8\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: constant\n* lr\\_scheduler\\_warmup\\_ratio: 0.03\n* num\\_epochs: 3", "### Training results", "### Framework versions\n\n\n* PEFT 0.10.1.dev0\n* Transformers 4.40.1\n* Pytorch 2.2.1+cu121\n* Datasets 2.19.0\n* Tokenizers 0.19.1" ]

[ "transformers", "pytorch", "llama", "text-generation", "unsloth", "Sorah", "spanish", "HirCoir", "en", "zh", "es", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[]

[ "en", "zh", "es" ]

[ "# MiniChat-2-3B-Sorah\n\n!Sorah\n\nMiniChat-2-3B-Sorah es un modelo de lenguaje basado en MiniChat-1.5-3B y ajustado con datos de instrucción y preferencia.\n\nEl modelo MiniChat-1.5-3B-Sorah complementa el modelo de voz Sorah, que fue creado por el proyecto Piper. El modelo Sorah está en proceso de entrenamiento y mejoras, por lo que su acceso no es público, pero puedes encontrar más información en el repositorio de Sorah Neuronal.\n\n\nEl modelo MiniChat-2-3B-Sorah supera a Vicuna-7B y se acerca a LLaMA-2-Chat-7B en MT-Bench.\n\nA continuación, se muestra un ejemplo de código para usar Sorah basado en MiniChat-2-3B:" ]

[ "TAGS\n#transformers #pytorch #llama #text-generation #unsloth #Sorah #spanish #HirCoir #en #zh #es #license-apache-2.0 #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n", "# MiniChat-2-3B-Sorah\n\n!Sorah\n\nMiniChat-2-3B-Sorah es un modelo de lenguaje basado en MiniChat-1.5-3B y ajustado con datos de instrucción y preferencia.\n\nEl modelo MiniChat-1.5-3B-Sorah complementa el modelo de voz Sorah, que fue creado por el proyecto Piper. El modelo Sorah está en proceso de entrenamiento y mejoras, por lo que su acceso no es público, pero puedes encontrar más información en el repositorio de Sorah Neuronal.\n\n\nEl modelo MiniChat-2-3B-Sorah supera a Vicuna-7B y se acerca a LLaMA-2-Chat-7B en MT-Bench.\n\nA continuación, se muestra un ejemplo de código para usar Sorah basado en MiniChat-2-3B:" ]

[ "transformers", "safetensors", "gpt2", "text-classification", "mergekit", "merge", "arxiv:2203.05482", "base_model:mllm-dev/gpt2_f_experiment_3_drug_data_new_run", "base_model:mllm-dev/gpt2_f_experiment_2_drug_data_new_run", "base_model:mllm-dev/gpt2_f_experiment_1_drug_data_new_run", "base_model:mllm-dev/gpt2_f_experiment_0_drug_data_new_run", "base_model:mllm-dev/gpt2_f_experiment_4_drug_data_new_run", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[ "2203.05482" ]

[]

[ "# tam_test_merge_out_drug_data_dare_linear_test_new_run\n\nThis is a merge of pre-trained language models created using mergekit.", "## Merge Details", "### Merge Method\n\nThis model was merged using the linear merge method.", "### Models Merged\n\nThe following models were included in the merge:\n* mllm-dev/gpt2_f_experiment_3_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_2_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_1_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_0_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_4_drug_data_new_run", "### Configuration\n\nThe following YAML configuration was used to produce this model:" ]

[ "TAGS\n#transformers #safetensors #gpt2 #text-classification #mergekit #merge #arxiv-2203.05482 #base_model-mllm-dev/gpt2_f_experiment_3_drug_data_new_run #base_model-mllm-dev/gpt2_f_experiment_2_drug_data_new_run #base_model-mllm-dev/gpt2_f_experiment_1_drug_data_new_run #base_model-mllm-dev/gpt2_f_experiment_0_drug_data_new_run #base_model-mllm-dev/gpt2_f_experiment_4_drug_data_new_run #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n", "# tam_test_merge_out_drug_data_dare_linear_test_new_run\n\nThis is a merge of pre-trained language models created using mergekit.", "## Merge Details", "### Merge Method\n\nThis model was merged using the linear merge method.", "### Models Merged\n\nThe following models were included in the merge:\n* mllm-dev/gpt2_f_experiment_3_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_2_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_1_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_0_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_4_drug_data_new_run", "### Configuration\n\nThe following YAML configuration was used to produce this model:" ]

[ "transformers", "safetensors", "llama", "text-generation", "facebook", "meta", "pytorch", "llama-3", "conversational", "en", "license:other", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "5-bit", "region:us" ]

[]

[ "en" ]

[ "### Use with transformers\n\n\nYou can run conversational inference using the Transformers pipeline abstraction, or by leveraging the Auto classes with the 'generate()' function. Let's see examples of both.", "#### Transformers pipeline", "#### Transformers AutoModelForCausalLM", "### Use with 'llama3'\n\n\nPlease, follow the instructions in the repository\n\n\nTo download Original checkpoints, see the example command below leveraging 'huggingface-cli':\n\n\nFor Hugging Face support, we recommend using transformers or TGI, but a similar command works.\n\n\nHardware and Software\n---------------------\n\n\nTraining Factors We used custom training libraries, Meta's Research SuperCluster, and production clusters for pretraining. Fine-tuning, annotation, and evaluation were also performed on third-party cloud compute.\n\n\nCarbon Footprint Pretraining utilized a cumulative 7.7M GPU hours of computation on hardware of type H100-80GB (TDP of 700W). Estimated total emissions were 2290 tCO2eq, 100% of which were offset by Meta’s sustainability program.\n\n\n\nCO2 emissions during pre-training. Time: total GPU time required for training each model. Power Consumption: peak power capacity per GPU device for the GPUs used adjusted for power usage efficiency. 100% of the emissions are directly offset by Meta's sustainability program, and because we are openly releasing these models, the pretraining costs do not need to be incurred by others.\n\n\nTraining Data\n-------------\n\n\nOverview Llama 3 was pretrained on over 15 trillion tokens of data from publicly available sources. The fine-tuning data includes publicly available instruction datasets, as well as over 10M human-annotated examples. Neither the pretraining nor the fine-tuning datasets include Meta user data.\n\n\nData Freshness The pretraining data has a cutoff of March 2023 for the 7B and December 2023 for the 70B models respectively.\n\n\nBenchmarks\n----------\n\n\nIn this section, we report the results for Llama 3 models on standard automatic benchmarks. For all the evaluations, we use our internal evaluations library. For details on the methodology see here.", "### Base pretrained models", "### Instruction tuned models", "### Responsibility & Safety\n\n\nWe believe that an open approach to AI leads to better, safer products, faster innovation, and a bigger overall market. We are committed to Responsible AI development and took a series of steps to limit misuse and harm and support the open source community.\n\n\nFoundation models are widely capable technologies that are built to be used for a diverse range of applications. They are not designed to meet every developer preference on safety levels for all use cases, out-of-the-box, as those by their nature will differ across different applications.\n\n\nRather, responsible LLM-application deployment is achieved by implementing a series of safety best practices throughout the development of such applications, from the model pre-training, fine-tuning and the deployment of systems composed of safeguards to tailor the safety needs specifically to the use case and audience.\n\n\nAs part of the Llama 3 release, we updated our Responsible Use Guide to outline the steps and best practices for developers to implement model and system level safety for their application. We also provide a set of resources including Meta Llama Guard 2 and Code Shield safeguards. These tools have proven to drastically reduce residual risks of LLM Systems, while maintaining a high level of helpfulness. We encourage developers to tune and deploy these safeguards according to their needs and we provide a reference implementation to get you started.", "#### Llama 3-Instruct\n\n\nAs outlined in the Responsible Use Guide, some trade-off between model helpfulness and model alignment is likely unavoidable. Developers should exercise discretion about how to weigh the benefits of alignment and helpfulness for their specific use case and audience. Developers should be mindful of residual risks when using Llama models and leverage additional safety tools as needed to reach the right safety bar for their use case.\n\n\nSafety\n\n\nFor our instruction tuned model, we conducted extensive red teaming exercises, performed adversarial evaluations and implemented safety mitigations techniques to lower residual risks. As with any Large Language Model, residual risks will likely remain and we recommend that developers assess these risks in the context of their use case. In parallel, we are working with the community to make AI safety benchmark standards transparent, rigorous and interpretable.\n\n\nRefusals\n\n\nIn addition to residual risks, we put a great emphasis on model refusals to benign prompts. Over-refusing not only can impact the user experience but could even be harmful in certain contexts as well. We’ve heard the feedback from the developer community and improved our fine tuning to ensure that Llama 3 is significantly less likely to falsely refuse to answer prompts than Llama 2.\n\n\nWe built internal benchmarks and developed mitigations to limit false refusals making Llama 3 our most helpful model to date.", "#### Responsible release\n\n\nIn addition to responsible use considerations outlined above, we followed a rigorous process that requires us to take extra measures against misuse and critical risks before we make our release decision.\n\n\nMisuse\n\n\nIf you access or use Llama 3, you agree to the Acceptable Use Policy. The most recent copy of this policy can be found at URL", "#### Critical risks\n\n\nCBRNE (Chemical, Biological, Radiological, Nuclear, and high yield Explosives)\n\n\nWe have conducted a two fold assessment of the safety of the model in this area:\n\n\n* Iterative testing during model training to assess the safety of responses related to CBRNE threats and other adversarial risks.\n* Involving external CBRNE experts to conduct an uplift test assessing the ability of the model to accurately provide expert knowledge and reduce barriers to potential CBRNE misuse, by reference to what can be achieved using web search (without the model).", "### Cyber Security\n\n\nWe have evaluated Llama 3 with CyberSecEval, Meta’s cybersecurity safety eval suite, measuring Llama 3’s propensity to suggest insecure code when used as a coding assistant, and Llama 3’s propensity to comply with requests to help carry out cyber attacks, where attacks are defined by the industry standard MITRE ATT&CK cyber attack ontology. On our insecure coding and cyber attacker helpfulness tests, Llama 3 behaved in the same range or safer than models of equivalent coding capability.", "### Child Safety\n\n\nChild Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences.", "### Community\n\n\nGenerative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership in AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository.\n\n\nFinally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community.\n\n\nEthical Considerations and Limitations\n--------------------------------------\n\n\nThe core values of Llama 3 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress.\n\n\nBut Llama 3 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has been in English, and has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3 models, developers should perform safety testing and tuning tailored to their specific applications of the model. As outlined in the Responsible Use Guide, we recommend incorporating Purple Llama solutions into your workflows and specifically Llama Guard which provides a base model to filter input and output prompts to layer system-level safety on top of model-level safety.\n\n\nPlease see the Responsible Use Guide available at URL\n\n\ninstructions\n\n\n@article{llama3modelcard,\n\n\ntitle={Llama 3 Model Card},\n\n\nauthor={AI@Meta},\n\n\nyear={2024},\n\n\nurl = {URL\n\n\n}\n\n\nContributors\n------------\n\n\nAaditya Singh; Aaron Grattafiori; Abhimanyu Dubey; Abhinav Jauhri; Abhinav Pandey; Abhishek Kadian; Adam Kelsey; Adi Gangidi; Ahmad Al-Dahle; Ahuva Goldstand; Aiesha Letman; Ajay Menon; Akhil Mathur; Alan Schelten; Alex Vaughan; Amy Yang; Andrei Lupu; Andres Alvarado; Andrew Gallagher; Andrew Gu; Andrew Ho; Andrew Poulton; Andrew Ryan; Angela Fan; Ankit Ramchandani; Anthony Hartshorn; Archi Mitra; Archie Sravankumar; Artem Korenev; Arun Rao; Ashley Gabriel; Ashwin Bharambe; Assaf Eisenman; Aston Zhang; Aurelien Rodriguez; Austen Gregerson; Ava Spataru; Baptiste Roziere; Ben Maurer; Benjamin Leonhardi; Bernie Huang; Bhargavi Paranjape; Bing Liu; Binh Tang; Bobbie Chern; Brani Stojkovic; Brian Fuller; Catalina Mejia Arenas; Chao Zhou; Charlotte Caucheteux; Chaya Nayak; Ching-Hsiang Chu; Chloe Bi; Chris Cai; Chris Cox; Chris Marra; Chris McConnell; Christian Keller; Christoph Feichtenhofer; Christophe Touret; Chunyang Wu; Corinne Wong; Cristian Canton Ferrer; Damien Allonsius; Daniel Kreymer; Daniel Haziza; Daniel Li; Danielle Pintz; Danny Livshits; Danny Wyatt; David Adkins; David Esiobu; David Xu; Davide Testuggine; Delia David; Devi Parikh; Dhruv Choudhary; Dhruv Mahajan; Diana Liskovich; Diego Garcia-Olano; Diego Perino; Dieuwke Hupkes; Dingkang Wang; Dustin Holland; Egor Lakomkin; Elina Lobanova; Xiaoqing Ellen Tan; Emily Dinan; Eric Smith; Erik Brinkman; Esteban Arcaute; Filip Radenovic; Firat Ozgenel; Francesco Caggioni; Frank Seide; Frank Zhang; Gabriel Synnaeve; Gabriella Schwarz; Gabrielle Lee; Gada Badeer; Georgia Anderson; Graeme Nail; Gregoire Mialon; Guan Pang; Guillem Cucurell; Hailey Nguyen; Hannah Korevaar; Hannah Wang; Haroun Habeeb; Harrison Rudolph; Henry Aspegren; Hu Xu; Hugo Touvron; Iga Kozlowska; Igor Molybog; Igor Tufanov; Iliyan Zarov; Imanol Arrieta Ibarra; Irina-Elena Veliche; Isabel Kloumann; Ishan Misra; Ivan Evtimov; Jacob Xu; Jade Copet; Jake Weissman; Jan Geffert; Jana Vranes; Japhet Asher; Jason Park; Jay Mahadeokar; Jean-Baptiste Gaya; Jeet Shah; Jelmer van der Linde; Jennifer Chan; Jenny Hong; Jenya Lee; Jeremy Fu; Jeremy Teboul; Jianfeng Chi; Jianyu Huang; Jie Wang; Jiecao Yu; Joanna Bitton; Joe Spisak; Joelle Pineau; Jon Carvill; Jongsoo Park; Joseph Rocca; Joshua Johnstun; Junteng Jia; Kalyan Vasuden Alwala; Kam Hou U; Kate Plawiak; Kartikeya Upasani; Kaushik Veeraraghavan; Ke Li; Kenneth Heafield; Kevin Stone; Khalid El-Arini; Krithika Iyer; Kshitiz Malik; Kuenley Chiu; Kunal Bhalla; Kyle Huang; Lakshya Garg; Lauren Rantala-Yeary; Laurens van der Maaten; Lawrence Chen; Leandro Silva; Lee Bell; Lei Zhang; Liang Tan; Louis Martin; Lovish Madaan; Luca Wehrstedt; Lukas Blecher; Luke de Oliveira; Madeline Muzzi; Madian Khabsa; Manav Avlani; Mannat Singh; Manohar Paluri; Mark Zuckerberg; Marcin Kardas; Martynas Mankus; Mathew Oldham; Mathieu Rita; Matthew Lennie; Maya Pavlova; Meghan Keneally; Melanie Kambadur; Mihir Patel; Mikayel Samvelyan; Mike Clark; Mike Lewis; Min Si; Mitesh Kumar Singh; Mo Metanat; Mona Hassan; Naman Goyal; Narjes Torabi; Nicolas Usunier; Nikolay Bashlykov; Nikolay Bogoychev; Niladri Chatterji; Ning Dong; Oliver Aobo Yang; Olivier Duchenne; Onur Celebi; Parth Parekh; Patrick Alrassy; Paul Saab; Pavan Balaji; Pedro Rittner; Pengchuan Zhang; Pengwei Li; Petar Vasic; Peter Weng; Polina Zvyagina; Prajjwal Bhargava; Pratik Dubal; Praveen Krishnan; Punit Singh Koura; Qing He; Rachel Rodriguez; Ragavan Srinivasan; Rahul Mitra; Ramon Calderer; Raymond Li; Robert Stojnic; Roberta Raileanu; Robin Battey; Rocky Wang; Rohit Girdhar; Rohit Patel; Romain Sauvestre; Ronnie Polidoro; Roshan Sumbaly; Ross Taylor; Ruan Silva; Rui Hou; Rui Wang; Russ Howes; Ruty Rinott; Saghar Hosseini; Sai Jayesh Bondu; Samyak Datta; Sanjay Singh; Sara Chugh; Sargun Dhillon; Satadru Pan; Sean Bell; Sergey Edunov; Shaoliang Nie; Sharan Narang; Sharath Raparthy; Shaun Lindsay; Sheng Feng; Sheng Shen; Shenghao Lin; Shiva Shankar; Shruti Bhosale; Shun Zhang; Simon Vandenhende; Sinong Wang; Seohyun Sonia Kim; Soumya Batra; Sten Sootla; Steve Kehoe; Suchin Gururangan; Sumit Gupta; Sunny Virk; Sydney Borodinsky; Tamar Glaser; Tamar Herman; Tamara Best; Tara Fowler; Thomas Georgiou; Thomas Scialom; Tianhe Li; Todor Mihaylov; Tong Xiao; Ujjwal Karn; Vedanuj Goswami; Vibhor Gupta; Vignesh Ramanathan; Viktor Kerkez; Vinay Satish Kumar; Vincent Gonguet; Vish Vogeti; Vlad Poenaru; Vlad Tiberiu Mihailescu; Vladan Petrovic; Vladimir Ivanov; Wei Li; Weiwei Chu; Wenhan Xiong; Wenyin Fu; Wes Bouaziz; Whitney Meers; Will Constable; Xavier Martinet; Xiaojian Wu; Xinbo Gao; Xinfeng Xie; Xuchao Jia; Yaelle Goldschlag; Yann LeCun; Yashesh Gaur; Yasmine Babaei; Ye Qi; Yenda Li; Yi Wen; Yiwen Song; Youngjin Nam; Yuchen Hao; Yuchen Zhang; Yun Wang; Yuning Mao; Yuzi He; Zacharie Delpierre Coudert; Zachary DeVito; Zahra Hankir; Zhaoduo Wen; Zheng Yan; Zhengxing Chen; Zhenyu Yang; Zoe Papakipos" ]

[ "TAGS\n#transformers #safetensors #llama #text-generation #facebook #meta #pytorch #llama-3 #conversational #en #license-other #autotrain_compatible #endpoints_compatible #text-generation-inference #5-bit #region-us \n", "### Use with transformers\n\n\nYou can run conversational inference using the Transformers pipeline abstraction, or by leveraging the Auto classes with the 'generate()' function. Let's see examples of both.", "#### Transformers pipeline", "#### Transformers AutoModelForCausalLM", "### Use with 'llama3'\n\n\nPlease, follow the instructions in the repository\n\n\nTo download Original checkpoints, see the example command below leveraging 'huggingface-cli':\n\n\nFor Hugging Face support, we recommend using transformers or TGI, but a similar command works.\n\n\nHardware and Software\n---------------------\n\n\nTraining Factors We used custom training libraries, Meta's Research SuperCluster, and production clusters for pretraining. Fine-tuning, annotation, and evaluation were also performed on third-party cloud compute.\n\n\nCarbon Footprint Pretraining utilized a cumulative 7.7M GPU hours of computation on hardware of type H100-80GB (TDP of 700W). Estimated total emissions were 2290 tCO2eq, 100% of which were offset by Meta’s sustainability program.\n\n\n\nCO2 emissions during pre-training. Time: total GPU time required for training each model. Power Consumption: peak power capacity per GPU device for the GPUs used adjusted for power usage efficiency. 100% of the emissions are directly offset by Meta's sustainability program, and because we are openly releasing these models, the pretraining costs do not need to be incurred by others.\n\n\nTraining Data\n-------------\n\n\nOverview Llama 3 was pretrained on over 15 trillion tokens of data from publicly available sources. The fine-tuning data includes publicly available instruction datasets, as well as over 10M human-annotated examples. Neither the pretraining nor the fine-tuning datasets include Meta user data.\n\n\nData Freshness The pretraining data has a cutoff of March 2023 for the 7B and December 2023 for the 70B models respectively.\n\n\nBenchmarks\n----------\n\n\nIn this section, we report the results for Llama 3 models on standard automatic benchmarks. For all the evaluations, we use our internal evaluations library. For details on the methodology see here.", "### Base pretrained models", "### Instruction tuned models", "### Responsibility & Safety\n\n\nWe believe that an open approach to AI leads to better, safer products, faster innovation, and a bigger overall market. We are committed to Responsible AI development and took a series of steps to limit misuse and harm and support the open source community.\n\n\nFoundation models are widely capable technologies that are built to be used for a diverse range of applications. They are not designed to meet every developer preference on safety levels for all use cases, out-of-the-box, as those by their nature will differ across different applications.\n\n\nRather, responsible LLM-application deployment is achieved by implementing a series of safety best practices throughout the development of such applications, from the model pre-training, fine-tuning and the deployment of systems composed of safeguards to tailor the safety needs specifically to the use case and audience.\n\n\nAs part of the Llama 3 release, we updated our Responsible Use Guide to outline the steps and best practices for developers to implement model and system level safety for their application. We also provide a set of resources including Meta Llama Guard 2 and Code Shield safeguards. These tools have proven to drastically reduce residual risks of LLM Systems, while maintaining a high level of helpfulness. We encourage developers to tune and deploy these safeguards according to their needs and we provide a reference implementation to get you started.", "#### Llama 3-Instruct\n\n\nAs outlined in the Responsible Use Guide, some trade-off between model helpfulness and model alignment is likely unavoidable. Developers should exercise discretion about how to weigh the benefits of alignment and helpfulness for their specific use case and audience. Developers should be mindful of residual risks when using Llama models and leverage additional safety tools as needed to reach the right safety bar for their use case.\n\n\nSafety\n\n\nFor our instruction tuned model, we conducted extensive red teaming exercises, performed adversarial evaluations and implemented safety mitigations techniques to lower residual risks. As with any Large Language Model, residual risks will likely remain and we recommend that developers assess these risks in the context of their use case. In parallel, we are working with the community to make AI safety benchmark standards transparent, rigorous and interpretable.\n\n\nRefusals\n\n\nIn addition to residual risks, we put a great emphasis on model refusals to benign prompts. Over-refusing not only can impact the user experience but could even be harmful in certain contexts as well. We’ve heard the feedback from the developer community and improved our fine tuning to ensure that Llama 3 is significantly less likely to falsely refuse to answer prompts than Llama 2.\n\n\nWe built internal benchmarks and developed mitigations to limit false refusals making Llama 3 our most helpful model to date.", "#### Responsible release\n\n\nIn addition to responsible use considerations outlined above, we followed a rigorous process that requires us to take extra measures against misuse and critical risks before we make our release decision.\n\n\nMisuse\n\n\nIf you access or use Llama 3, you agree to the Acceptable Use Policy. The most recent copy of this policy can be found at URL", "#### Critical risks\n\n\nCBRNE (Chemical, Biological, Radiological, Nuclear, and high yield Explosives)\n\n\nWe have conducted a two fold assessment of the safety of the model in this area:\n\n\n* Iterative testing during model training to assess the safety of responses related to CBRNE threats and other adversarial risks.\n* Involving external CBRNE experts to conduct an uplift test assessing the ability of the model to accurately provide expert knowledge and reduce barriers to potential CBRNE misuse, by reference to what can be achieved using web search (without the model).", "### Cyber Security\n\n\nWe have evaluated Llama 3 with CyberSecEval, Meta’s cybersecurity safety eval suite, measuring Llama 3’s propensity to suggest insecure code when used as a coding assistant, and Llama 3’s propensity to comply with requests to help carry out cyber attacks, where attacks are defined by the industry standard MITRE ATT&CK cyber attack ontology. On our insecure coding and cyber attacker helpfulness tests, Llama 3 behaved in the same range or safer than models of equivalent coding capability.", "### Child Safety\n\n\nChild Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences.", "### Community\n\n\nGenerative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership in AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository.\n\n\nFinally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community.\n\n\nEthical Considerations and Limitations\n--------------------------------------\n\n\nThe core values of Llama 3 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress.\n\n\nBut Llama 3 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has been in English, and has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3 models, developers should perform safety testing and tuning tailored to their specific applications of the model. As outlined in the Responsible Use Guide, we recommend incorporating Purple Llama solutions into your workflows and specifically Llama Guard which provides a base model to filter input and output prompts to layer system-level safety on top of model-level safety.\n\n\nPlease see the Responsible Use Guide available at URL\n\n\ninstructions\n\n\n@article{llama3modelcard,\n\n\ntitle={Llama 3 Model Card},\n\n\nauthor={AI@Meta},\n\n\nyear={2024},\n\n\nurl = {URL\n\n\n}\n\n\nContributors\n------------\n\n\nAaditya Singh; Aaron Grattafiori; Abhimanyu Dubey; Abhinav Jauhri; Abhinav Pandey; Abhishek Kadian; Adam Kelsey; Adi Gangidi; Ahmad Al-Dahle; Ahuva Goldstand; Aiesha Letman; Ajay Menon; Akhil Mathur; Alan Schelten; Alex Vaughan; Amy Yang; Andrei Lupu; Andres Alvarado; Andrew Gallagher; Andrew Gu; Andrew Ho; Andrew Poulton; Andrew Ryan; Angela Fan; Ankit Ramchandani; Anthony Hartshorn; Archi Mitra; Archie Sravankumar; Artem Korenev; Arun Rao; Ashley Gabriel; Ashwin Bharambe; Assaf Eisenman; Aston Zhang; Aurelien Rodriguez; Austen Gregerson; Ava Spataru; Baptiste Roziere; Ben Maurer; Benjamin Leonhardi; Bernie Huang; Bhargavi Paranjape; Bing Liu; Binh Tang; Bobbie Chern; Brani Stojkovic; Brian Fuller; Catalina Mejia Arenas; Chao Zhou; Charlotte Caucheteux; Chaya Nayak; Ching-Hsiang Chu; Chloe Bi; Chris Cai; Chris Cox; Chris Marra; Chris McConnell; Christian Keller; Christoph Feichtenhofer; Christophe Touret; Chunyang Wu; Corinne Wong; Cristian Canton Ferrer; Damien Allonsius; Daniel Kreymer; Daniel Haziza; Daniel Li; Danielle Pintz; Danny Livshits; Danny Wyatt; David Adkins; David Esiobu; David Xu; Davide Testuggine; Delia David; Devi Parikh; Dhruv Choudhary; Dhruv Mahajan; Diana Liskovich; Diego Garcia-Olano; Diego Perino; Dieuwke Hupkes; Dingkang Wang; Dustin Holland; Egor Lakomkin; Elina Lobanova; Xiaoqing Ellen Tan; Emily Dinan; Eric Smith; Erik Brinkman; Esteban Arcaute; Filip Radenovic; Firat Ozgenel; Francesco Caggioni; Frank Seide; Frank Zhang; Gabriel Synnaeve; Gabriella Schwarz; Gabrielle Lee; Gada Badeer; Georgia Anderson; Graeme Nail; Gregoire Mialon; Guan Pang; Guillem Cucurell; Hailey Nguyen; Hannah Korevaar; Hannah Wang; Haroun Habeeb; Harrison Rudolph; Henry Aspegren; Hu Xu; Hugo Touvron; Iga Kozlowska; Igor Molybog; Igor Tufanov; Iliyan Zarov; Imanol Arrieta Ibarra; Irina-Elena Veliche; Isabel Kloumann; Ishan Misra; Ivan Evtimov; Jacob Xu; Jade Copet; Jake Weissman; Jan Geffert; Jana Vranes; Japhet Asher; Jason Park; Jay Mahadeokar; Jean-Baptiste Gaya; Jeet Shah; Jelmer van der Linde; Jennifer Chan; Jenny Hong; Jenya Lee; Jeremy Fu; Jeremy Teboul; Jianfeng Chi; Jianyu Huang; Jie Wang; Jiecao Yu; Joanna Bitton; Joe Spisak; Joelle Pineau; Jon Carvill; Jongsoo Park; Joseph Rocca; Joshua Johnstun; Junteng Jia; Kalyan Vasuden Alwala; Kam Hou U; Kate Plawiak; Kartikeya Upasani; Kaushik Veeraraghavan; Ke Li; Kenneth Heafield; Kevin Stone; Khalid El-Arini; Krithika Iyer; Kshitiz Malik; Kuenley Chiu; Kunal Bhalla; Kyle Huang; Lakshya Garg; Lauren Rantala-Yeary; Laurens van der Maaten; Lawrence Chen; Leandro Silva; Lee Bell; Lei Zhang; Liang Tan; Louis Martin; Lovish Madaan; Luca Wehrstedt; Lukas Blecher; Luke de Oliveira; Madeline Muzzi; Madian Khabsa; Manav Avlani; Mannat Singh; Manohar Paluri; Mark Zuckerberg; Marcin Kardas; Martynas Mankus; Mathew Oldham; Mathieu Rita; Matthew Lennie; Maya Pavlova; Meghan Keneally; Melanie Kambadur; Mihir Patel; Mikayel Samvelyan; Mike Clark; Mike Lewis; Min Si; Mitesh Kumar Singh; Mo Metanat; Mona Hassan; Naman Goyal; Narjes Torabi; Nicolas Usunier; Nikolay Bashlykov; Nikolay Bogoychev; Niladri Chatterji; Ning Dong; Oliver Aobo Yang; Olivier Duchenne; Onur Celebi; Parth Parekh; Patrick Alrassy; Paul Saab; Pavan Balaji; Pedro Rittner; Pengchuan Zhang; Pengwei Li; Petar Vasic; Peter Weng; Polina Zvyagina; Prajjwal Bhargava; Pratik Dubal; Praveen Krishnan; Punit Singh Koura; Qing He; Rachel Rodriguez; Ragavan Srinivasan; Rahul Mitra; Ramon Calderer; Raymond Li; Robert Stojnic; Roberta Raileanu; Robin Battey; Rocky Wang; Rohit Girdhar; Rohit Patel; Romain Sauvestre; Ronnie Polidoro; Roshan Sumbaly; Ross Taylor; Ruan Silva; Rui Hou; Rui Wang; Russ Howes; Ruty Rinott; Saghar Hosseini; Sai Jayesh Bondu; Samyak Datta; Sanjay Singh; Sara Chugh; Sargun Dhillon; Satadru Pan; Sean Bell; Sergey Edunov; Shaoliang Nie; Sharan Narang; Sharath Raparthy; Shaun Lindsay; Sheng Feng; Sheng Shen; Shenghao Lin; Shiva Shankar; Shruti Bhosale; Shun Zhang; Simon Vandenhende; Sinong Wang; Seohyun Sonia Kim; Soumya Batra; Sten Sootla; Steve Kehoe; Suchin Gururangan; Sumit Gupta; Sunny Virk; Sydney Borodinsky; Tamar Glaser; Tamar Herman; Tamara Best; Tara Fowler; Thomas Georgiou; Thomas Scialom; Tianhe Li; Todor Mihaylov; Tong Xiao; Ujjwal Karn; Vedanuj Goswami; Vibhor Gupta; Vignesh Ramanathan; Viktor Kerkez; Vinay Satish Kumar; Vincent Gonguet; Vish Vogeti; Vlad Poenaru; Vlad Tiberiu Mihailescu; Vladan Petrovic; Vladimir Ivanov; Wei Li; Weiwei Chu; Wenhan Xiong; Wenyin Fu; Wes Bouaziz; Whitney Meers; Will Constable; Xavier Martinet; Xiaojian Wu; Xinbo Gao; Xinfeng Xie; Xuchao Jia; Yaelle Goldschlag; Yann LeCun; Yashesh Gaur; Yasmine Babaei; Ye Qi; Yenda Li; Yi Wen; Yiwen Song; Youngjin Nam; Yuchen Hao; Yuchen Zhang; Yun Wang; Yuning Mao; Yuzi He; Zacharie Delpierre Coudert; Zachary DeVito; Zahra Hankir; Zhaoduo Wen; Zheng Yan; Zhengxing Chen; Zhenyu Yang; Zoe Papakipos" ]

[ "transformers", "gguf", "text-generation-inference", "unsloth", "llama", "trl", "sft", "not-for-all-audiences", "en", "dataset:grimulkan/LimaRP-augmented", "dataset:mpasila/LimaRP-augmented-8k-context", "base_model:mpasila/Llama-3-LimaRP-Instruct-8B", "license:apache-2.0", "endpoints_compatible", "region:us" ]

[]

[ "en" ]

[]

[ "TAGS\n#transformers #gguf #text-generation-inference #unsloth #llama #trl #sft #not-for-all-audiences #en #dataset-grimulkan/LimaRP-augmented #dataset-mpasila/LimaRP-augmented-8k-context #base_model-mpasila/Llama-3-LimaRP-Instruct-8B #license-apache-2.0 #endpoints_compatible #region-us \n" ]

[ "transformers", "safetensors", "llama", "text-generation", "facebook", "meta", "pytorch", "llama-3", "conversational", "en", "license:other", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "6-bit", "region:us" ]

[]

[ "en" ]

[ "### Use with transformers\n\n\nYou can run conversational inference using the Transformers pipeline abstraction, or by leveraging the Auto classes with the 'generate()' function. Let's see examples of both.", "#### Transformers pipeline", "#### Transformers AutoModelForCausalLM", "### Use with 'llama3'\n\n\nPlease, follow the instructions in the repository\n\n\nTo download Original checkpoints, see the example command below leveraging 'huggingface-cli':\n\n\nFor Hugging Face support, we recommend using transformers or TGI, but a similar command works.\n\n\nHardware and Software\n---------------------\n\n\nTraining Factors We used custom training libraries, Meta's Research SuperCluster, and production clusters for pretraining. Fine-tuning, annotation, and evaluation were also performed on third-party cloud compute.\n\n\nCarbon Footprint Pretraining utilized a cumulative 7.7M GPU hours of computation on hardware of type H100-80GB (TDP of 700W). Estimated total emissions were 2290 tCO2eq, 100% of which were offset by Meta’s sustainability program.\n\n\n\nCO2 emissions during pre-training. Time: total GPU time required for training each model. Power Consumption: peak power capacity per GPU device for the GPUs used adjusted for power usage efficiency. 100% of the emissions are directly offset by Meta's sustainability program, and because we are openly releasing these models, the pretraining costs do not need to be incurred by others.\n\n\nTraining Data\n-------------\n\n\nOverview Llama 3 was pretrained on over 15 trillion tokens of data from publicly available sources. The fine-tuning data includes publicly available instruction datasets, as well as over 10M human-annotated examples. Neither the pretraining nor the fine-tuning datasets include Meta user data.\n\n\nData Freshness The pretraining data has a cutoff of March 2023 for the 7B and December 2023 for the 70B models respectively.\n\n\nBenchmarks\n----------\n\n\nIn this section, we report the results for Llama 3 models on standard automatic benchmarks. For all the evaluations, we use our internal evaluations library. For details on the methodology see here.", "### Base pretrained models", "### Instruction tuned models", "### Responsibility & Safety\n\n\nWe believe that an open approach to AI leads to better, safer products, faster innovation, and a bigger overall market. We are committed to Responsible AI development and took a series of steps to limit misuse and harm and support the open source community.\n\n\nFoundation models are widely capable technologies that are built to be used for a diverse range of applications. They are not designed to meet every developer preference on safety levels for all use cases, out-of-the-box, as those by their nature will differ across different applications.\n\n\nRather, responsible LLM-application deployment is achieved by implementing a series of safety best practices throughout the development of such applications, from the model pre-training, fine-tuning and the deployment of systems composed of safeguards to tailor the safety needs specifically to the use case and audience.\n\n\nAs part of the Llama 3 release, we updated our Responsible Use Guide to outline the steps and best practices for developers to implement model and system level safety for their application. We also provide a set of resources including Meta Llama Guard 2 and Code Shield safeguards. These tools have proven to drastically reduce residual risks of LLM Systems, while maintaining a high level of helpfulness. We encourage developers to tune and deploy these safeguards according to their needs and we provide a reference implementation to get you started.", "#### Llama 3-Instruct\n\n\nAs outlined in the Responsible Use Guide, some trade-off between model helpfulness and model alignment is likely unavoidable. Developers should exercise discretion about how to weigh the benefits of alignment and helpfulness for their specific use case and audience. Developers should be mindful of residual risks when using Llama models and leverage additional safety tools as needed to reach the right safety bar for their use case.\n\n\nSafety\n\n\nFor our instruction tuned model, we conducted extensive red teaming exercises, performed adversarial evaluations and implemented safety mitigations techniques to lower residual risks. As with any Large Language Model, residual risks will likely remain and we recommend that developers assess these risks in the context of their use case. In parallel, we are working with the community to make AI safety benchmark standards transparent, rigorous and interpretable.\n\n\nRefusals\n\n\nIn addition to residual risks, we put a great emphasis on model refusals to benign prompts. Over-refusing not only can impact the user experience but could even be harmful in certain contexts as well. We’ve heard the feedback from the developer community and improved our fine tuning to ensure that Llama 3 is significantly less likely to falsely refuse to answer prompts than Llama 2.\n\n\nWe built internal benchmarks and developed mitigations to limit false refusals making Llama 3 our most helpful model to date.", "#### Responsible release\n\n\nIn addition to responsible use considerations outlined above, we followed a rigorous process that requires us to take extra measures against misuse and critical risks before we make our release decision.\n\n\nMisuse\n\n\nIf you access or use Llama 3, you agree to the Acceptable Use Policy. The most recent copy of this policy can be found at URL", "#### Critical risks\n\n\nCBRNE (Chemical, Biological, Radiological, Nuclear, and high yield Explosives)\n\n\nWe have conducted a two fold assessment of the safety of the model in this area:\n\n\n* Iterative testing during model training to assess the safety of responses related to CBRNE threats and other adversarial risks.\n* Involving external CBRNE experts to conduct an uplift test assessing the ability of the model to accurately provide expert knowledge and reduce barriers to potential CBRNE misuse, by reference to what can be achieved using web search (without the model).", "### Cyber Security\n\n\nWe have evaluated Llama 3 with CyberSecEval, Meta’s cybersecurity safety eval suite, measuring Llama 3’s propensity to suggest insecure code when used as a coding assistant, and Llama 3’s propensity to comply with requests to help carry out cyber attacks, where attacks are defined by the industry standard MITRE ATT&CK cyber attack ontology. On our insecure coding and cyber attacker helpfulness tests, Llama 3 behaved in the same range or safer than models of equivalent coding capability.", "### Child Safety\n\n\nChild Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences.", "### Community\n\n\nGenerative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership in AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository.\n\n\nFinally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community.\n\n\nEthical Considerations and Limitations\n--------------------------------------\n\n\nThe core values of Llama 3 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress.\n\n\nBut Llama 3 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has been in English, and has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3 models, developers should perform safety testing and tuning tailored to their specific applications of the model. As outlined in the Responsible Use Guide, we recommend incorporating Purple Llama solutions into your workflows and specifically Llama Guard which provides a base model to filter input and output prompts to layer system-level safety on top of model-level safety.\n\n\nPlease see the Responsible Use Guide available at URL\n\n\ninstructions\n\n\n@article{llama3modelcard,\n\n\ntitle={Llama 3 Model Card},\n\n\nauthor={AI@Meta},\n\n\nyear={2024},\n\n\nurl = {URL\n\n\n}\n\n\nContributors\n------------\n\n\nAaditya Singh; Aaron Grattafiori; Abhimanyu Dubey; Abhinav Jauhri; Abhinav Pandey; Abhishek Kadian; Adam Kelsey; Adi Gangidi; Ahmad Al-Dahle; Ahuva Goldstand; Aiesha Letman; Ajay Menon; Akhil Mathur; Alan Schelten; Alex Vaughan; Amy Yang; Andrei Lupu; Andres Alvarado; Andrew Gallagher; Andrew Gu; Andrew Ho; Andrew Poulton; Andrew Ryan; Angela Fan; Ankit Ramchandani; Anthony Hartshorn; Archi Mitra; Archie Sravankumar; Artem Korenev; Arun Rao; Ashley Gabriel; Ashwin Bharambe; Assaf Eisenman; Aston Zhang; Aurelien Rodriguez; Austen Gregerson; Ava Spataru; Baptiste Roziere; Ben Maurer; Benjamin Leonhardi; Bernie Huang; Bhargavi Paranjape; Bing Liu; Binh Tang; Bobbie Chern; Brani Stojkovic; Brian Fuller; Catalina Mejia Arenas; Chao Zhou; Charlotte Caucheteux; Chaya Nayak; Ching-Hsiang Chu; Chloe Bi; Chris Cai; Chris Cox; Chris Marra; Chris McConnell; Christian Keller; Christoph Feichtenhofer; Christophe Touret; Chunyang Wu; Corinne Wong; Cristian Canton Ferrer; Damien Allonsius; Daniel Kreymer; Daniel Haziza; Daniel Li; Danielle Pintz; Danny Livshits; Danny Wyatt; David Adkins; David Esiobu; David Xu; Davide Testuggine; Delia David; Devi Parikh; Dhruv Choudhary; Dhruv Mahajan; Diana Liskovich; Diego Garcia-Olano; Diego Perino; Dieuwke Hupkes; Dingkang Wang; Dustin Holland; Egor Lakomkin; Elina Lobanova; Xiaoqing Ellen Tan; Emily Dinan; Eric Smith; Erik Brinkman; Esteban Arcaute; Filip Radenovic; Firat Ozgenel; Francesco Caggioni; Frank Seide; Frank Zhang; Gabriel Synnaeve; Gabriella Schwarz; Gabrielle Lee; Gada Badeer; Georgia Anderson; Graeme Nail; Gregoire Mialon; Guan Pang; Guillem Cucurell; Hailey Nguyen; Hannah Korevaar; Hannah Wang; Haroun Habeeb; Harrison Rudolph; Henry Aspegren; Hu Xu; Hugo Touvron; Iga Kozlowska; Igor Molybog; Igor Tufanov; Iliyan Zarov; Imanol Arrieta Ibarra; Irina-Elena Veliche; Isabel Kloumann; Ishan Misra; Ivan Evtimov; Jacob Xu; Jade Copet; Jake Weissman; Jan Geffert; Jana Vranes; Japhet Asher; Jason Park; Jay Mahadeokar; Jean-Baptiste Gaya; Jeet Shah; Jelmer van der Linde; Jennifer Chan; Jenny Hong; Jenya Lee; Jeremy Fu; Jeremy Teboul; Jianfeng Chi; Jianyu Huang; Jie Wang; Jiecao Yu; Joanna Bitton; Joe Spisak; Joelle Pineau; Jon Carvill; Jongsoo Park; Joseph Rocca; Joshua Johnstun; Junteng Jia; Kalyan Vasuden Alwala; Kam Hou U; Kate Plawiak; Kartikeya Upasani; Kaushik Veeraraghavan; Ke Li; Kenneth Heafield; Kevin Stone; Khalid El-Arini; Krithika Iyer; Kshitiz Malik; Kuenley Chiu; Kunal Bhalla; Kyle Huang; Lakshya Garg; Lauren Rantala-Yeary; Laurens van der Maaten; Lawrence Chen; Leandro Silva; Lee Bell; Lei Zhang; Liang Tan; Louis Martin; Lovish Madaan; Luca Wehrstedt; Lukas Blecher; Luke de Oliveira; Madeline Muzzi; Madian Khabsa; Manav Avlani; Mannat Singh; Manohar Paluri; Mark Zuckerberg; Marcin Kardas; Martynas Mankus; Mathew Oldham; Mathieu Rita; Matthew Lennie; Maya Pavlova; Meghan Keneally; Melanie Kambadur; Mihir Patel; Mikayel Samvelyan; Mike Clark; Mike Lewis; Min Si; Mitesh Kumar Singh; Mo Metanat; Mona Hassan; Naman Goyal; Narjes Torabi; Nicolas Usunier; Nikolay Bashlykov; Nikolay Bogoychev; Niladri Chatterji; Ning Dong; Oliver Aobo Yang; Olivier Duchenne; Onur Celebi; Parth Parekh; Patrick Alrassy; Paul Saab; Pavan Balaji; Pedro Rittner; Pengchuan Zhang; Pengwei Li; Petar Vasic; Peter Weng; Polina Zvyagina; Prajjwal Bhargava; Pratik Dubal; Praveen Krishnan; Punit Singh Koura; Qing He; Rachel Rodriguez; Ragavan Srinivasan; Rahul Mitra; Ramon Calderer; Raymond Li; Robert Stojnic; Roberta Raileanu; Robin Battey; Rocky Wang; Rohit Girdhar; Rohit Patel; Romain Sauvestre; Ronnie Polidoro; Roshan Sumbaly; Ross Taylor; Ruan Silva; Rui Hou; Rui Wang; Russ Howes; Ruty Rinott; Saghar Hosseini; Sai Jayesh Bondu; Samyak Datta; Sanjay Singh; Sara Chugh; Sargun Dhillon; Satadru Pan; Sean Bell; Sergey Edunov; Shaoliang Nie; Sharan Narang; Sharath Raparthy; Shaun Lindsay; Sheng Feng; Sheng Shen; Shenghao Lin; Shiva Shankar; Shruti Bhosale; Shun Zhang; Simon Vandenhende; Sinong Wang; Seohyun Sonia Kim; Soumya Batra; Sten Sootla; Steve Kehoe; Suchin Gururangan; Sumit Gupta; Sunny Virk; Sydney Borodinsky; Tamar Glaser; Tamar Herman; Tamara Best; Tara Fowler; Thomas Georgiou; Thomas Scialom; Tianhe Li; Todor Mihaylov; Tong Xiao; Ujjwal Karn; Vedanuj Goswami; Vibhor Gupta; Vignesh Ramanathan; Viktor Kerkez; Vinay Satish Kumar; Vincent Gonguet; Vish Vogeti; Vlad Poenaru; Vlad Tiberiu Mihailescu; Vladan Petrovic; Vladimir Ivanov; Wei Li; Weiwei Chu; Wenhan Xiong; Wenyin Fu; Wes Bouaziz; Whitney Meers; Will Constable; Xavier Martinet; Xiaojian Wu; Xinbo Gao; Xinfeng Xie; Xuchao Jia; Yaelle Goldschlag; Yann LeCun; Yashesh Gaur; Yasmine Babaei; Ye Qi; Yenda Li; Yi Wen; Yiwen Song; Youngjin Nam; Yuchen Hao; Yuchen Zhang; Yun Wang; Yuning Mao; Yuzi He; Zacharie Delpierre Coudert; Zachary DeVito; Zahra Hankir; Zhaoduo Wen; Zheng Yan; Zhengxing Chen; Zhenyu Yang; Zoe Papakipos" ]

[ "TAGS\n#transformers #safetensors #llama #text-generation #facebook #meta #pytorch #llama-3 #conversational #en #license-other #autotrain_compatible #endpoints_compatible #text-generation-inference #6-bit #region-us \n", "### Use with transformers\n\n\nYou can run conversational inference using the Transformers pipeline abstraction, or by leveraging the Auto classes with the 'generate()' function. Let's see examples of both.", "#### Transformers pipeline", "#### Transformers AutoModelForCausalLM", "### Use with 'llama3'\n\n\nPlease, follow the instructions in the repository\n\n\nTo download Original checkpoints, see the example command below leveraging 'huggingface-cli':\n\n\nFor Hugging Face support, we recommend using transformers or TGI, but a similar command works.\n\n\nHardware and Software\n---------------------\n\n\nTraining Factors We used custom training libraries, Meta's Research SuperCluster, and production clusters for pretraining. Fine-tuning, annotation, and evaluation were also performed on third-party cloud compute.\n\n\nCarbon Footprint Pretraining utilized a cumulative 7.7M GPU hours of computation on hardware of type H100-80GB (TDP of 700W). Estimated total emissions were 2290 tCO2eq, 100% of which were offset by Meta’s sustainability program.\n\n\n\nCO2 emissions during pre-training. Time: total GPU time required for training each model. Power Consumption: peak power capacity per GPU device for the GPUs used adjusted for power usage efficiency. 100% of the emissions are directly offset by Meta's sustainability program, and because we are openly releasing these models, the pretraining costs do not need to be incurred by others.\n\n\nTraining Data\n-------------\n\n\nOverview Llama 3 was pretrained on over 15 trillion tokens of data from publicly available sources. The fine-tuning data includes publicly available instruction datasets, as well as over 10M human-annotated examples. Neither the pretraining nor the fine-tuning datasets include Meta user data.\n\n\nData Freshness The pretraining data has a cutoff of March 2023 for the 7B and December 2023 for the 70B models respectively.\n\n\nBenchmarks\n----------\n\n\nIn this section, we report the results for Llama 3 models on standard automatic benchmarks. For all the evaluations, we use our internal evaluations library. For details on the methodology see here.", "### Base pretrained models", "### Instruction tuned models", "### Responsibility & Safety\n\n\nWe believe that an open approach to AI leads to better, safer products, faster innovation, and a bigger overall market. We are committed to Responsible AI development and took a series of steps to limit misuse and harm and support the open source community.\n\n\nFoundation models are widely capable technologies that are built to be used for a diverse range of applications. They are not designed to meet every developer preference on safety levels for all use cases, out-of-the-box, as those by their nature will differ across different applications.\n\n\nRather, responsible LLM-application deployment is achieved by implementing a series of safety best practices throughout the development of such applications, from the model pre-training, fine-tuning and the deployment of systems composed of safeguards to tailor the safety needs specifically to the use case and audience.\n\n\nAs part of the Llama 3 release, we updated our Responsible Use Guide to outline the steps and best practices for developers to implement model and system level safety for their application. We also provide a set of resources including Meta Llama Guard 2 and Code Shield safeguards. These tools have proven to drastically reduce residual risks of LLM Systems, while maintaining a high level of helpfulness. We encourage developers to tune and deploy these safeguards according to their needs and we provide a reference implementation to get you started.", "#### Llama 3-Instruct\n\n\nAs outlined in the Responsible Use Guide, some trade-off between model helpfulness and model alignment is likely unavoidable. Developers should exercise discretion about how to weigh the benefits of alignment and helpfulness for their specific use case and audience. Developers should be mindful of residual risks when using Llama models and leverage additional safety tools as needed to reach the right safety bar for their use case.\n\n\nSafety\n\n\nFor our instruction tuned model, we conducted extensive red teaming exercises, performed adversarial evaluations and implemented safety mitigations techniques to lower residual risks. As with any Large Language Model, residual risks will likely remain and we recommend that developers assess these risks in the context of their use case. In parallel, we are working with the community to make AI safety benchmark standards transparent, rigorous and interpretable.\n\n\nRefusals\n\n\nIn addition to residual risks, we put a great emphasis on model refusals to benign prompts. Over-refusing not only can impact the user experience but could even be harmful in certain contexts as well. We’ve heard the feedback from the developer community and improved our fine tuning to ensure that Llama 3 is significantly less likely to falsely refuse to answer prompts than Llama 2.\n\n\nWe built internal benchmarks and developed mitigations to limit false refusals making Llama 3 our most helpful model to date.", "#### Responsible release\n\n\nIn addition to responsible use considerations outlined above, we followed a rigorous process that requires us to take extra measures against misuse and critical risks before we make our release decision.\n\n\nMisuse\n\n\nIf you access or use Llama 3, you agree to the Acceptable Use Policy. The most recent copy of this policy can be found at URL", "#### Critical risks\n\n\nCBRNE (Chemical, Biological, Radiological, Nuclear, and high yield Explosives)\n\n\nWe have conducted a two fold assessment of the safety of the model in this area:\n\n\n* Iterative testing during model training to assess the safety of responses related to CBRNE threats and other adversarial risks.\n* Involving external CBRNE experts to conduct an uplift test assessing the ability of the model to accurately provide expert knowledge and reduce barriers to potential CBRNE misuse, by reference to what can be achieved using web search (without the model).", "### Cyber Security\n\n\nWe have evaluated Llama 3 with CyberSecEval, Meta’s cybersecurity safety eval suite, measuring Llama 3’s propensity to suggest insecure code when used as a coding assistant, and Llama 3’s propensity to comply with requests to help carry out cyber attacks, where attacks are defined by the industry standard MITRE ATT&CK cyber attack ontology. On our insecure coding and cyber attacker helpfulness tests, Llama 3 behaved in the same range or safer than models of equivalent coding capability.", "### Child Safety\n\n\nChild Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences.", "### Community\n\n\nGenerative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership in AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository.\n\n\nFinally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community.\n\n\nEthical Considerations and Limitations\n--------------------------------------\n\n\nThe core values of Llama 3 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress.\n\n\nBut Llama 3 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has been in English, and has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3 models, developers should perform safety testing and tuning tailored to their specific applications of the model. As outlined in the Responsible Use Guide, we recommend incorporating Purple Llama solutions into your workflows and specifically Llama Guard which provides a base model to filter input and output prompts to layer system-level safety on top of model-level safety.\n\n\nPlease see the Responsible Use Guide available at URL\n\n\ninstructions\n\n\n@article{llama3modelcard,\n\n\ntitle={Llama 3 Model Card},\n\n\nauthor={AI@Meta},\n\n\nyear={2024},\n\n\nurl = {URL\n\n\n}\n\n\nContributors\n------------\n\n\nAaditya Singh; Aaron Grattafiori; Abhimanyu Dubey; Abhinav Jauhri; Abhinav Pandey; Abhishek Kadian; Adam Kelsey; Adi Gangidi; Ahmad Al-Dahle; Ahuva Goldstand; Aiesha Letman; Ajay Menon; Akhil Mathur; Alan Schelten; Alex Vaughan; Amy Yang; Andrei Lupu; Andres Alvarado; Andrew Gallagher; Andrew Gu; Andrew Ho; Andrew Poulton; Andrew Ryan; Angela Fan; Ankit Ramchandani; Anthony Hartshorn; Archi Mitra; Archie Sravankumar; Artem Korenev; Arun Rao; Ashley Gabriel; Ashwin Bharambe; Assaf Eisenman; Aston Zhang; Aurelien Rodriguez; Austen Gregerson; Ava Spataru; Baptiste Roziere; Ben Maurer; Benjamin Leonhardi; Bernie Huang; Bhargavi Paranjape; Bing Liu; Binh Tang; Bobbie Chern; Brani Stojkovic; Brian Fuller; Catalina Mejia Arenas; Chao Zhou; Charlotte Caucheteux; Chaya Nayak; Ching-Hsiang Chu; Chloe Bi; Chris Cai; Chris Cox; Chris Marra; Chris McConnell; Christian Keller; Christoph Feichtenhofer; Christophe Touret; Chunyang Wu; Corinne Wong; Cristian Canton Ferrer; Damien Allonsius; Daniel Kreymer; Daniel Haziza; Daniel Li; Danielle Pintz; Danny Livshits; Danny Wyatt; David Adkins; David Esiobu; David Xu; Davide Testuggine; Delia David; Devi Parikh; Dhruv Choudhary; Dhruv Mahajan; Diana Liskovich; Diego Garcia-Olano; Diego Perino; Dieuwke Hupkes; Dingkang Wang; Dustin Holland; Egor Lakomkin; Elina Lobanova; Xiaoqing Ellen Tan; Emily Dinan; Eric Smith; Erik Brinkman; Esteban Arcaute; Filip Radenovic; Firat Ozgenel; Francesco Caggioni; Frank Seide; Frank Zhang; Gabriel Synnaeve; Gabriella Schwarz; Gabrielle Lee; Gada Badeer; Georgia Anderson; Graeme Nail; Gregoire Mialon; Guan Pang; Guillem Cucurell; Hailey Nguyen; Hannah Korevaar; Hannah Wang; Haroun Habeeb; Harrison Rudolph; Henry Aspegren; Hu Xu; Hugo Touvron; Iga Kozlowska; Igor Molybog; Igor Tufanov; Iliyan Zarov; Imanol Arrieta Ibarra; Irina-Elena Veliche; Isabel Kloumann; Ishan Misra; Ivan Evtimov; Jacob Xu; Jade Copet; Jake Weissman; Jan Geffert; Jana Vranes; Japhet Asher; Jason Park; Jay Mahadeokar; Jean-Baptiste Gaya; Jeet Shah; Jelmer van der Linde; Jennifer Chan; Jenny Hong; Jenya Lee; Jeremy Fu; Jeremy Teboul; Jianfeng Chi; Jianyu Huang; Jie Wang; Jiecao Yu; Joanna Bitton; Joe Spisak; Joelle Pineau; Jon Carvill; Jongsoo Park; Joseph Rocca; Joshua Johnstun; Junteng Jia; Kalyan Vasuden Alwala; Kam Hou U; Kate Plawiak; Kartikeya Upasani; Kaushik Veeraraghavan; Ke Li; Kenneth Heafield; Kevin Stone; Khalid El-Arini; Krithika Iyer; Kshitiz Malik; Kuenley Chiu; Kunal Bhalla; Kyle Huang; Lakshya Garg; Lauren Rantala-Yeary; Laurens van der Maaten; Lawrence Chen; Leandro Silva; Lee Bell; Lei Zhang; Liang Tan; Louis Martin; Lovish Madaan; Luca Wehrstedt; Lukas Blecher; Luke de Oliveira; Madeline Muzzi; Madian Khabsa; Manav Avlani; Mannat Singh; Manohar Paluri; Mark Zuckerberg; Marcin Kardas; Martynas Mankus; Mathew Oldham; Mathieu Rita; Matthew Lennie; Maya Pavlova; Meghan Keneally; Melanie Kambadur; Mihir Patel; Mikayel Samvelyan; Mike Clark; Mike Lewis; Min Si; Mitesh Kumar Singh; Mo Metanat; Mona Hassan; Naman Goyal; Narjes Torabi; Nicolas Usunier; Nikolay Bashlykov; Nikolay Bogoychev; Niladri Chatterji; Ning Dong; Oliver Aobo Yang; Olivier Duchenne; Onur Celebi; Parth Parekh; Patrick Alrassy; Paul Saab; Pavan Balaji; Pedro Rittner; Pengchuan Zhang; Pengwei Li; Petar Vasic; Peter Weng; Polina Zvyagina; Prajjwal Bhargava; Pratik Dubal; Praveen Krishnan; Punit Singh Koura; Qing He; Rachel Rodriguez; Ragavan Srinivasan; Rahul Mitra; Ramon Calderer; Raymond Li; Robert Stojnic; Roberta Raileanu; Robin Battey; Rocky Wang; Rohit Girdhar; Rohit Patel; Romain Sauvestre; Ronnie Polidoro; Roshan Sumbaly; Ross Taylor; Ruan Silva; Rui Hou; Rui Wang; Russ Howes; Ruty Rinott; Saghar Hosseini; Sai Jayesh Bondu; Samyak Datta; Sanjay Singh; Sara Chugh; Sargun Dhillon; Satadru Pan; Sean Bell; Sergey Edunov; Shaoliang Nie; Sharan Narang; Sharath Raparthy; Shaun Lindsay; Sheng Feng; Sheng Shen; Shenghao Lin; Shiva Shankar; Shruti Bhosale; Shun Zhang; Simon Vandenhende; Sinong Wang; Seohyun Sonia Kim; Soumya Batra; Sten Sootla; Steve Kehoe; Suchin Gururangan; Sumit Gupta; Sunny Virk; Sydney Borodinsky; Tamar Glaser; Tamar Herman; Tamara Best; Tara Fowler; Thomas Georgiou; Thomas Scialom; Tianhe Li; Todor Mihaylov; Tong Xiao; Ujjwal Karn; Vedanuj Goswami; Vibhor Gupta; Vignesh Ramanathan; Viktor Kerkez; Vinay Satish Kumar; Vincent Gonguet; Vish Vogeti; Vlad Poenaru; Vlad Tiberiu Mihailescu; Vladan Petrovic; Vladimir Ivanov; Wei Li; Weiwei Chu; Wenhan Xiong; Wenyin Fu; Wes Bouaziz; Whitney Meers; Will Constable; Xavier Martinet; Xiaojian Wu; Xinbo Gao; Xinfeng Xie; Xuchao Jia; Yaelle Goldschlag; Yann LeCun; Yashesh Gaur; Yasmine Babaei; Ye Qi; Yenda Li; Yi Wen; Yiwen Song; Youngjin Nam; Yuchen Hao; Yuchen Zhang; Yun Wang; Yuning Mao; Yuzi He; Zacharie Delpierre Coudert; Zachary DeVito; Zahra Hankir; Zhaoduo Wen; Zheng Yan; Zhengxing Chen; Zhenyu Yang; Zoe Papakipos" ]

[ "transformers", "tensorboard", "safetensors", "gpt_neox", "text-classification", "generated_from_trainer", "base_model:EleutherAI/pythia-14m", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[]

[]

[ "# robust_llm_pythia-14m_mz-132_EnronSpam_n-its-10\n\nThis model is a fine-tuned version of EleutherAI/pythia-14m on an unknown dataset.", "## Model description\n\nMore information needed", "## Intended uses & limitations\n\nMore information needed", "## Training and evaluation data\n\nMore information needed", "## Training procedure", "### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 1e-05\n- train_batch_size: 8\n- eval_batch_size: 64\n- seed: 0\n- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n- lr_scheduler_type: linear\n- num_epochs: 1", "### Training results", "### Framework versions\n\n- Transformers 4.39.3\n- Pytorch 2.2.1\n- Datasets 2.18.0\n- Tokenizers 0.15.2" ]

[ "TAGS\n#transformers #tensorboard #safetensors #gpt_neox #text-classification #generated_from_trainer #base_model-EleutherAI/pythia-14m #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n", "# robust_llm_pythia-14m_mz-132_EnronSpam_n-its-10\n\nThis model is a fine-tuned version of EleutherAI/pythia-14m on an unknown dataset.", "## Model description\n\nMore information needed", "## Intended uses & limitations\n\nMore information needed", "## Training and evaluation data\n\nMore information needed", "## Training procedure", "### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 1e-05\n- train_batch_size: 8\n- eval_batch_size: 64\n- seed: 0\n- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n- lr_scheduler_type: linear\n- num_epochs: 1", "### Training results", "### Framework versions\n\n- Transformers 4.39.3\n- Pytorch 2.2.1\n- Datasets 2.18.0\n- Tokenizers 0.15.2" ]

[ "ml-agents", "tensorboard", "onnx", "Pyramids", "deep-reinforcement-learning", "reinforcement-learning", "ML-Agents-Pyramids", "region:us" ]

[]

[]

[ "# ppo Agent playing Pyramids\n This is a trained model of a ppo agent playing Pyramids\n using the Unity ML-Agents Library.\n\n ## Usage (with ML-Agents)\n The Documentation: URL\n\n We wrote a complete tutorial to learn to train your first agent using ML-Agents and publish it to the Hub:\n - A *short tutorial* where you teach Huggy the Dog to fetch the stick and then play with him directly in your\n browser: URL\n - A *longer tutorial* to understand how works ML-Agents:\n URL\n\n ### Resume the training\n \n\n ### Watch your Agent play\n You can watch your agent playing directly in your browser\n\n 1. If the environment is part of ML-Agents official environments, go to URL\n 2. Step 1: Find your model_id: ahforoughi/ppo-Pyramids-Training\n 3. Step 2: Select your *.nn /*.onnx file\n 4. Click on Watch the agent play" ]

[ "TAGS\n#ml-agents #tensorboard #onnx #Pyramids #deep-reinforcement-learning #reinforcement-learning #ML-Agents-Pyramids #region-us \n", "# ppo Agent playing Pyramids\n This is a trained model of a ppo agent playing Pyramids\n using the Unity ML-Agents Library.\n\n ## Usage (with ML-Agents)\n The Documentation: URL\n\n We wrote a complete tutorial to learn to train your first agent using ML-Agents and publish it to the Hub:\n - A *short tutorial* where you teach Huggy the Dog to fetch the stick and then play with him directly in your\n browser: URL\n - A *longer tutorial* to understand how works ML-Agents:\n URL\n\n ### Resume the training\n \n\n ### Watch your Agent play\n You can watch your agent playing directly in your browser\n\n 1. If the environment is part of ML-Agents official environments, go to URL\n 2. Step 1: Find your model_id: ahforoughi/ppo-Pyramids-Training\n 3. Step 2: Select your *.nn /*.onnx file\n 4. Click on Watch the agent play" ]

[ "mlx", "safetensors", "phi3", "custom_code", "license:apache-2.0", "region:us" ]

[]

[]

[ "# GreenBitAI/Phi-3-mini-128k-instruct-layer-mix-bpw-2.2-mlx\nThis quantized low-bit model was converted to MLX format from ['GreenBitAI/Phi-3-mini-128k-instruct-layer-mix-bpw-2.2']().\nRefer to the original model card for more details on the model.", "## Use with mlx" ]

[ "TAGS\n#mlx #safetensors #phi3 #custom_code #license-apache-2.0 #region-us \n", "# GreenBitAI/Phi-3-mini-128k-instruct-layer-mix-bpw-2.2-mlx\nThis quantized low-bit model was converted to MLX format from ['GreenBitAI/Phi-3-mini-128k-instruct-layer-mix-bpw-2.2']().\nRefer to the original model card for more details on the model.", "## Use with mlx" ]

[ "transformers", "safetensors", "llama", "text-generation", "4-bit", "AWQ", "autotrain_compatible", "endpoints_compatible", "meta-llama/Meta-Llama-3-8B", "conversational", "dataset:feeltheAGI/maverick-sharegpt", "license:apache-2.0", "text-generation-inference", "region:us" ]

[]

[]

[ "# feeltheAGI/maverick-llama3-8B AWQ\n\n- Model creator: feeltheAGI\n- Original model: maverick-llama3-8B\n\n!URL", "## Model Summary\n\nmodel is using ChatML prompt template format.\n\nexample:", "## How to use", "### Install the necessary packages", "### Example Python code", "### About AWQ\n\nAWQ is an efficient, accurate and blazing-fast low-bit weight quantization method, currently supporting 4-bit quantization. Compared to GPTQ, it offers faster Transformers-based inference with equivalent or better quality compared to the most commonly used GPTQ settings.\n\nAWQ models are currently supported on Linux and Windows, with NVidia GPUs only. macOS users: please use GGUF models instead.\n\nIt is supported by:\n\n- Text Generation Webui - using Loader: AutoAWQ\n- vLLM - version 0.2.2 or later for support for all model types.\n- Hugging Face Text Generation Inference (TGI)\n- Transformers version 4.35.0 and later, from any code or client that supports Transformers\n- AutoAWQ - for use from Python code" ]

[ "TAGS\n#transformers #safetensors #llama #text-generation #4-bit #AWQ #autotrain_compatible #endpoints_compatible #meta-llama/Meta-Llama-3-8B #conversational #dataset-feeltheAGI/maverick-sharegpt #license-apache-2.0 #text-generation-inference #region-us \n", "# feeltheAGI/maverick-llama3-8B AWQ\n\n- Model creator: feeltheAGI\n- Original model: maverick-llama3-8B\n\n!URL", "## Model Summary\n\nmodel is using ChatML prompt template format.\n\nexample:", "## How to use", "### Install the necessary packages", "### Example Python code", "### About AWQ\n\nAWQ is an efficient, accurate and blazing-fast low-bit weight quantization method, currently supporting 4-bit quantization. Compared to GPTQ, it offers faster Transformers-based inference with equivalent or better quality compared to the most commonly used GPTQ settings.\n\nAWQ models are currently supported on Linux and Windows, with NVidia GPUs only. macOS users: please use GGUF models instead.\n\nIt is supported by:\n\n- Text Generation Webui - using Loader: AutoAWQ\n- vLLM - version 0.2.2 or later for support for all model types.\n- Hugging Face Text Generation Inference (TGI)\n- Transformers version 4.35.0 and later, from any code or client that supports Transformers\n- AutoAWQ - for use from Python code" ]

[ "transformers", "safetensors", "llama", "text-generation", "facebook", "meta", "pytorch", "llama-3", "conversational", "en", "license:other", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "8-bit", "region:us" ]

[]

[ "en" ]

[ "### Use with transformers\n\n\nYou can run conversational inference using the Transformers pipeline abstraction, or by leveraging the Auto classes with the 'generate()' function. Let's see examples of both.", "#### Transformers pipeline", "#### Transformers AutoModelForCausalLM", "### Use with 'llama3'\n\n\nPlease, follow the instructions in the repository\n\n\nTo download Original checkpoints, see the example command below leveraging 'huggingface-cli':\n\n\nFor Hugging Face support, we recommend using transformers or TGI, but a similar command works.\n\n\nHardware and Software\n---------------------\n\n\nTraining Factors We used custom training libraries, Meta's Research SuperCluster, and production clusters for pretraining. Fine-tuning, annotation, and evaluation were also performed on third-party cloud compute.\n\n\nCarbon Footprint Pretraining utilized a cumulative 7.7M GPU hours of computation on hardware of type H100-80GB (TDP of 700W). Estimated total emissions were 2290 tCO2eq, 100% of which were offset by Meta’s sustainability program.\n\n\n\nCO2 emissions during pre-training. Time: total GPU time required for training each model. Power Consumption: peak power capacity per GPU device for the GPUs used adjusted for power usage efficiency. 100% of the emissions are directly offset by Meta's sustainability program, and because we are openly releasing these models, the pretraining costs do not need to be incurred by others.\n\n\nTraining Data\n-------------\n\n\nOverview Llama 3 was pretrained on over 15 trillion tokens of data from publicly available sources. The fine-tuning data includes publicly available instruction datasets, as well as over 10M human-annotated examples. Neither the pretraining nor the fine-tuning datasets include Meta user data.\n\n\nData Freshness The pretraining data has a cutoff of March 2023 for the 7B and December 2023 for the 70B models respectively.\n\n\nBenchmarks\n----------\n\n\nIn this section, we report the results for Llama 3 models on standard automatic benchmarks. For all the evaluations, we use our internal evaluations library. For details on the methodology see here.", "### Base pretrained models", "### Instruction tuned models", "### Responsibility & Safety\n\n\nWe believe that an open approach to AI leads to better, safer products, faster innovation, and a bigger overall market. We are committed to Responsible AI development and took a series of steps to limit misuse and harm and support the open source community.\n\n\nFoundation models are widely capable technologies that are built to be used for a diverse range of applications. They are not designed to meet every developer preference on safety levels for all use cases, out-of-the-box, as those by their nature will differ across different applications.\n\n\nRather, responsible LLM-application deployment is achieved by implementing a series of safety best practices throughout the development of such applications, from the model pre-training, fine-tuning and the deployment of systems composed of safeguards to tailor the safety needs specifically to the use case and audience.\n\n\nAs part of the Llama 3 release, we updated our Responsible Use Guide to outline the steps and best practices for developers to implement model and system level safety for their application. We also provide a set of resources including Meta Llama Guard 2 and Code Shield safeguards. These tools have proven to drastically reduce residual risks of LLM Systems, while maintaining a high level of helpfulness. We encourage developers to tune and deploy these safeguards according to their needs and we provide a reference implementation to get you started.", "#### Llama 3-Instruct\n\n\nAs outlined in the Responsible Use Guide, some trade-off between model helpfulness and model alignment is likely unavoidable. Developers should exercise discretion about how to weigh the benefits of alignment and helpfulness for their specific use case and audience. Developers should be mindful of residual risks when using Llama models and leverage additional safety tools as needed to reach the right safety bar for their use case.\n\n\nSafety\n\n\nFor our instruction tuned model, we conducted extensive red teaming exercises, performed adversarial evaluations and implemented safety mitigations techniques to lower residual risks. As with any Large Language Model, residual risks will likely remain and we recommend that developers assess these risks in the context of their use case. In parallel, we are working with the community to make AI safety benchmark standards transparent, rigorous and interpretable.\n\n\nRefusals\n\n\nIn addition to residual risks, we put a great emphasis on model refusals to benign prompts. Over-refusing not only can impact the user experience but could even be harmful in certain contexts as well. We’ve heard the feedback from the developer community and improved our fine tuning to ensure that Llama 3 is significantly less likely to falsely refuse to answer prompts than Llama 2.\n\n\nWe built internal benchmarks and developed mitigations to limit false refusals making Llama 3 our most helpful model to date.", "#### Responsible release\n\n\nIn addition to responsible use considerations outlined above, we followed a rigorous process that requires us to take extra measures against misuse and critical risks before we make our release decision.\n\n\nMisuse\n\n\nIf you access or use Llama 3, you agree to the Acceptable Use Policy. The most recent copy of this policy can be found at URL", "#### Critical risks\n\n\nCBRNE (Chemical, Biological, Radiological, Nuclear, and high yield Explosives)\n\n\nWe have conducted a two fold assessment of the safety of the model in this area:\n\n\n* Iterative testing during model training to assess the safety of responses related to CBRNE threats and other adversarial risks.\n* Involving external CBRNE experts to conduct an uplift test assessing the ability of the model to accurately provide expert knowledge and reduce barriers to potential CBRNE misuse, by reference to what can be achieved using web search (without the model).", "### Cyber Security\n\n\nWe have evaluated Llama 3 with CyberSecEval, Meta’s cybersecurity safety eval suite, measuring Llama 3’s propensity to suggest insecure code when used as a coding assistant, and Llama 3’s propensity to comply with requests to help carry out cyber attacks, where attacks are defined by the industry standard MITRE ATT&CK cyber attack ontology. On our insecure coding and cyber attacker helpfulness tests, Llama 3 behaved in the same range or safer than models of equivalent coding capability.", "### Child Safety\n\n\nChild Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences.", "### Community\n\n\nGenerative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership in AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository.\n\n\nFinally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community.\n\n\nEthical Considerations and Limitations\n--------------------------------------\n\n\nThe core values of Llama 3 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress.\n\n\nBut Llama 3 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has been in English, and has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3 models, developers should perform safety testing and tuning tailored to their specific applications of the model. As outlined in the Responsible Use Guide, we recommend incorporating Purple Llama solutions into your workflows and specifically Llama Guard which provides a base model to filter input and output prompts to layer system-level safety on top of model-level safety.\n\n\nPlease see the Responsible Use Guide available at URL\n\n\ninstructions\n\n\n@article{llama3modelcard,\n\n\ntitle={Llama 3 Model Card},\n\n\nauthor={AI@Meta},\n\n\nyear={2024},\n\n\nurl = {URL\n\n\n}\n\n\nContributors\n------------\n\n\nAaditya Singh; Aaron Grattafiori; Abhimanyu Dubey; Abhinav Jauhri; Abhinav Pandey; Abhishek Kadian; Adam Kelsey; Adi Gangidi; Ahmad Al-Dahle; Ahuva Goldstand; Aiesha Letman; Ajay Menon; Akhil Mathur; Alan Schelten; Alex Vaughan; Amy Yang; Andrei Lupu; Andres Alvarado; Andrew Gallagher; Andrew Gu; Andrew Ho; Andrew Poulton; Andrew Ryan; Angela Fan; Ankit Ramchandani; Anthony Hartshorn; Archi Mitra; Archie Sravankumar; Artem Korenev; Arun Rao; Ashley Gabriel; Ashwin Bharambe; Assaf Eisenman; Aston Zhang; Aurelien Rodriguez; Austen Gregerson; Ava Spataru; Baptiste Roziere; Ben Maurer; Benjamin Leonhardi; Bernie Huang; Bhargavi Paranjape; Bing Liu; Binh Tang; Bobbie Chern; Brani Stojkovic; Brian Fuller; Catalina Mejia Arenas; Chao Zhou; Charlotte Caucheteux; Chaya Nayak; Ching-Hsiang Chu; Chloe Bi; Chris Cai; Chris Cox; Chris Marra; Chris McConnell; Christian Keller; Christoph Feichtenhofer; Christophe Touret; Chunyang Wu; Corinne Wong; Cristian Canton Ferrer; Damien Allonsius; Daniel Kreymer; Daniel Haziza; Daniel Li; Danielle Pintz; Danny Livshits; Danny Wyatt; David Adkins; David Esiobu; David Xu; Davide Testuggine; Delia David; Devi Parikh; Dhruv Choudhary; Dhruv Mahajan; Diana Liskovich; Diego Garcia-Olano; Diego Perino; Dieuwke Hupkes; Dingkang Wang; Dustin Holland; Egor Lakomkin; Elina Lobanova; Xiaoqing Ellen Tan; Emily Dinan; Eric Smith; Erik Brinkman; Esteban Arcaute; Filip Radenovic; Firat Ozgenel; Francesco Caggioni; Frank Seide; Frank Zhang; Gabriel Synnaeve; Gabriella Schwarz; Gabrielle Lee; Gada Badeer; Georgia Anderson; Graeme Nail; Gregoire Mialon; Guan Pang; Guillem Cucurell; Hailey Nguyen; Hannah Korevaar; Hannah Wang; Haroun Habeeb; Harrison Rudolph; Henry Aspegren; Hu Xu; Hugo Touvron; Iga Kozlowska; Igor Molybog; Igor Tufanov; Iliyan Zarov; Imanol Arrieta Ibarra; Irina-Elena Veliche; Isabel Kloumann; Ishan Misra; Ivan Evtimov; Jacob Xu; Jade Copet; Jake Weissman; Jan Geffert; Jana Vranes; Japhet Asher; Jason Park; Jay Mahadeokar; Jean-Baptiste Gaya; Jeet Shah; Jelmer van der Linde; Jennifer Chan; Jenny Hong; Jenya Lee; Jeremy Fu; Jeremy Teboul; Jianfeng Chi; Jianyu Huang; Jie Wang; Jiecao Yu; Joanna Bitton; Joe Spisak; Joelle Pineau; Jon Carvill; Jongsoo Park; Joseph Rocca; Joshua Johnstun; Junteng Jia; Kalyan Vasuden Alwala; Kam Hou U; Kate Plawiak; Kartikeya Upasani; Kaushik Veeraraghavan; Ke Li; Kenneth Heafield; Kevin Stone; Khalid El-Arini; Krithika Iyer; Kshitiz Malik; Kuenley Chiu; Kunal Bhalla; Kyle Huang; Lakshya Garg; Lauren Rantala-Yeary; Laurens van der Maaten; Lawrence Chen; Leandro Silva; Lee Bell; Lei Zhang; Liang Tan; Louis Martin; Lovish Madaan; Luca Wehrstedt; Lukas Blecher; Luke de Oliveira; Madeline Muzzi; Madian Khabsa; Manav Avlani; Mannat Singh; Manohar Paluri; Mark Zuckerberg; Marcin Kardas; Martynas Mankus; Mathew Oldham; Mathieu Rita; Matthew Lennie; Maya Pavlova; Meghan Keneally; Melanie Kambadur; Mihir Patel; Mikayel Samvelyan; Mike Clark; Mike Lewis; Min Si; Mitesh Kumar Singh; Mo Metanat; Mona Hassan; Naman Goyal; Narjes Torabi; Nicolas Usunier; Nikolay Bashlykov; Nikolay Bogoychev; Niladri Chatterji; Ning Dong; Oliver Aobo Yang; Olivier Duchenne; Onur Celebi; Parth Parekh; Patrick Alrassy; Paul Saab; Pavan Balaji; Pedro Rittner; Pengchuan Zhang; Pengwei Li; Petar Vasic; Peter Weng; Polina Zvyagina; Prajjwal Bhargava; Pratik Dubal; Praveen Krishnan; Punit Singh Koura; Qing He; Rachel Rodriguez; Ragavan Srinivasan; Rahul Mitra; Ramon Calderer; Raymond Li; Robert Stojnic; Roberta Raileanu; Robin Battey; Rocky Wang; Rohit Girdhar; Rohit Patel; Romain Sauvestre; Ronnie Polidoro; Roshan Sumbaly; Ross Taylor; Ruan Silva; Rui Hou; Rui Wang; Russ Howes; Ruty Rinott; Saghar Hosseini; Sai Jayesh Bondu; Samyak Datta; Sanjay Singh; Sara Chugh; Sargun Dhillon; Satadru Pan; Sean Bell; Sergey Edunov; Shaoliang Nie; Sharan Narang; Sharath Raparthy; Shaun Lindsay; Sheng Feng; Sheng Shen; Shenghao Lin; Shiva Shankar; Shruti Bhosale; Shun Zhang; Simon Vandenhende; Sinong Wang; Seohyun Sonia Kim; Soumya Batra; Sten Sootla; Steve Kehoe; Suchin Gururangan; Sumit Gupta; Sunny Virk; Sydney Borodinsky; Tamar Glaser; Tamar Herman; Tamara Best; Tara Fowler; Thomas Georgiou; Thomas Scialom; Tianhe Li; Todor Mihaylov; Tong Xiao; Ujjwal Karn; Vedanuj Goswami; Vibhor Gupta; Vignesh Ramanathan; Viktor Kerkez; Vinay Satish Kumar; Vincent Gonguet; Vish Vogeti; Vlad Poenaru; Vlad Tiberiu Mihailescu; Vladan Petrovic; Vladimir Ivanov; Wei Li; Weiwei Chu; Wenhan Xiong; Wenyin Fu; Wes Bouaziz; Whitney Meers; Will Constable; Xavier Martinet; Xiaojian Wu; Xinbo Gao; Xinfeng Xie; Xuchao Jia; Yaelle Goldschlag; Yann LeCun; Yashesh Gaur; Yasmine Babaei; Ye Qi; Yenda Li; Yi Wen; Yiwen Song; Youngjin Nam; Yuchen Hao; Yuchen Zhang; Yun Wang; Yuning Mao; Yuzi He; Zacharie Delpierre Coudert; Zachary DeVito; Zahra Hankir; Zhaoduo Wen; Zheng Yan; Zhengxing Chen; Zhenyu Yang; Zoe Papakipos" ]

[ "TAGS\n#transformers #safetensors #llama #text-generation #facebook #meta #pytorch #llama-3 #conversational #en #license-other #autotrain_compatible #endpoints_compatible #text-generation-inference #8-bit #region-us \n", "### Use with transformers\n\n\nYou can run conversational inference using the Transformers pipeline abstraction, or by leveraging the Auto classes with the 'generate()' function. Let's see examples of both.", "#### Transformers pipeline", "#### Transformers AutoModelForCausalLM", "### Use with 'llama3'\n\n\nPlease, follow the instructions in the repository\n\n\nTo download Original checkpoints, see the example command below leveraging 'huggingface-cli':\n\n\nFor Hugging Face support, we recommend using transformers or TGI, but a similar command works.\n\n\nHardware and Software\n---------------------\n\n\nTraining Factors We used custom training libraries, Meta's Research SuperCluster, and production clusters for pretraining. Fine-tuning, annotation, and evaluation were also performed on third-party cloud compute.\n\n\nCarbon Footprint Pretraining utilized a cumulative 7.7M GPU hours of computation on hardware of type H100-80GB (TDP of 700W). Estimated total emissions were 2290 tCO2eq, 100% of which were offset by Meta’s sustainability program.\n\n\n\nCO2 emissions during pre-training. Time: total GPU time required for training each model. Power Consumption: peak power capacity per GPU device for the GPUs used adjusted for power usage efficiency. 100% of the emissions are directly offset by Meta's sustainability program, and because we are openly releasing these models, the pretraining costs do not need to be incurred by others.\n\n\nTraining Data\n-------------\n\n\nOverview Llama 3 was pretrained on over 15 trillion tokens of data from publicly available sources. The fine-tuning data includes publicly available instruction datasets, as well as over 10M human-annotated examples. Neither the pretraining nor the fine-tuning datasets include Meta user data.\n\n\nData Freshness The pretraining data has a cutoff of March 2023 for the 7B and December 2023 for the 70B models respectively.\n\n\nBenchmarks\n----------\n\n\nIn this section, we report the results for Llama 3 models on standard automatic benchmarks. For all the evaluations, we use our internal evaluations library. For details on the methodology see here.", "### Base pretrained models", "### Instruction tuned models", "### Responsibility & Safety\n\n\nWe believe that an open approach to AI leads to better, safer products, faster innovation, and a bigger overall market. We are committed to Responsible AI development and took a series of steps to limit misuse and harm and support the open source community.\n\n\nFoundation models are widely capable technologies that are built to be used for a diverse range of applications. They are not designed to meet every developer preference on safety levels for all use cases, out-of-the-box, as those by their nature will differ across different applications.\n\n\nRather, responsible LLM-application deployment is achieved by implementing a series of safety best practices throughout the development of such applications, from the model pre-training, fine-tuning and the deployment of systems composed of safeguards to tailor the safety needs specifically to the use case and audience.\n\n\nAs part of the Llama 3 release, we updated our Responsible Use Guide to outline the steps and best practices for developers to implement model and system level safety for their application. We also provide a set of resources including Meta Llama Guard 2 and Code Shield safeguards. These tools have proven to drastically reduce residual risks of LLM Systems, while maintaining a high level of helpfulness. We encourage developers to tune and deploy these safeguards according to their needs and we provide a reference implementation to get you started.", "#### Llama 3-Instruct\n\n\nAs outlined in the Responsible Use Guide, some trade-off between model helpfulness and model alignment is likely unavoidable. Developers should exercise discretion about how to weigh the benefits of alignment and helpfulness for their specific use case and audience. Developers should be mindful of residual risks when using Llama models and leverage additional safety tools as needed to reach the right safety bar for their use case.\n\n\nSafety\n\n\nFor our instruction tuned model, we conducted extensive red teaming exercises, performed adversarial evaluations and implemented safety mitigations techniques to lower residual risks. As with any Large Language Model, residual risks will likely remain and we recommend that developers assess these risks in the context of their use case. In parallel, we are working with the community to make AI safety benchmark standards transparent, rigorous and interpretable.\n\n\nRefusals\n\n\nIn addition to residual risks, we put a great emphasis on model refusals to benign prompts. Over-refusing not only can impact the user experience but could even be harmful in certain contexts as well. We’ve heard the feedback from the developer community and improved our fine tuning to ensure that Llama 3 is significantly less likely to falsely refuse to answer prompts than Llama 2.\n\n\nWe built internal benchmarks and developed mitigations to limit false refusals making Llama 3 our most helpful model to date.", "#### Responsible release\n\n\nIn addition to responsible use considerations outlined above, we followed a rigorous process that requires us to take extra measures against misuse and critical risks before we make our release decision.\n\n\nMisuse\n\n\nIf you access or use Llama 3, you agree to the Acceptable Use Policy. The most recent copy of this policy can be found at URL", "#### Critical risks\n\n\nCBRNE (Chemical, Biological, Radiological, Nuclear, and high yield Explosives)\n\n\nWe have conducted a two fold assessment of the safety of the model in this area:\n\n\n* Iterative testing during model training to assess the safety of responses related to CBRNE threats and other adversarial risks.\n* Involving external CBRNE experts to conduct an uplift test assessing the ability of the model to accurately provide expert knowledge and reduce barriers to potential CBRNE misuse, by reference to what can be achieved using web search (without the model).", "### Cyber Security\n\n\nWe have evaluated Llama 3 with CyberSecEval, Meta’s cybersecurity safety eval suite, measuring Llama 3’s propensity to suggest insecure code when used as a coding assistant, and Llama 3’s propensity to comply with requests to help carry out cyber attacks, where attacks are defined by the industry standard MITRE ATT&CK cyber attack ontology. On our insecure coding and cyber attacker helpfulness tests, Llama 3 behaved in the same range or safer than models of equivalent coding capability.", "### Child Safety\n\n\nChild Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences.", "### Community\n\n\nGenerative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership in AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository.\n\n\nFinally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community.\n\n\nEthical Considerations and Limitations\n--------------------------------------\n\n\nThe core values of Llama 3 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress.\n\n\nBut Llama 3 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has been in English, and has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3 models, developers should perform safety testing and tuning tailored to their specific applications of the model. As outlined in the Responsible Use Guide, we recommend incorporating Purple Llama solutions into your workflows and specifically Llama Guard which provides a base model to filter input and output prompts to layer system-level safety on top of model-level safety.\n\n\nPlease see the Responsible Use Guide available at URL\n\n\ninstructions\n\n\n@article{llama3modelcard,\n\n\ntitle={Llama 3 Model Card},\n\n\nauthor={AI@Meta},\n\n\nyear={2024},\n\n\nurl = {URL\n\n\n}\n\n\nContributors\n------------\n\n\nAaditya Singh; Aaron Grattafiori; Abhimanyu Dubey; Abhinav Jauhri; Abhinav Pandey; Abhishek Kadian; Adam Kelsey; Adi Gangidi; Ahmad Al-Dahle; Ahuva Goldstand; Aiesha Letman; Ajay Menon; Akhil Mathur; Alan Schelten; Alex Vaughan; Amy Yang; Andrei Lupu; Andres Alvarado; Andrew Gallagher; Andrew Gu; Andrew Ho; Andrew Poulton; Andrew Ryan; Angela Fan; Ankit Ramchandani; Anthony Hartshorn; Archi Mitra; Archie Sravankumar; Artem Korenev; Arun Rao; Ashley Gabriel; Ashwin Bharambe; Assaf Eisenman; Aston Zhang; Aurelien Rodriguez; Austen Gregerson; Ava Spataru; Baptiste Roziere; Ben Maurer; Benjamin Leonhardi; Bernie Huang; Bhargavi Paranjape; Bing Liu; Binh Tang; Bobbie Chern; Brani Stojkovic; Brian Fuller; Catalina Mejia Arenas; Chao Zhou; Charlotte Caucheteux; Chaya Nayak; Ching-Hsiang Chu; Chloe Bi; Chris Cai; Chris Cox; Chris Marra; Chris McConnell; Christian Keller; Christoph Feichtenhofer; Christophe Touret; Chunyang Wu; Corinne Wong; Cristian Canton Ferrer; Damien Allonsius; Daniel Kreymer; Daniel Haziza; Daniel Li; Danielle Pintz; Danny Livshits; Danny Wyatt; David Adkins; David Esiobu; David Xu; Davide Testuggine; Delia David; Devi Parikh; Dhruv Choudhary; Dhruv Mahajan; Diana Liskovich; Diego Garcia-Olano; Diego Perino; Dieuwke Hupkes; Dingkang Wang; Dustin Holland; Egor Lakomkin; Elina Lobanova; Xiaoqing Ellen Tan; Emily Dinan; Eric Smith; Erik Brinkman; Esteban Arcaute; Filip Radenovic; Firat Ozgenel; Francesco Caggioni; Frank Seide; Frank Zhang; Gabriel Synnaeve; Gabriella Schwarz; Gabrielle Lee; Gada Badeer; Georgia Anderson; Graeme Nail; Gregoire Mialon; Guan Pang; Guillem Cucurell; Hailey Nguyen; Hannah Korevaar; Hannah Wang; Haroun Habeeb; Harrison Rudolph; Henry Aspegren; Hu Xu; Hugo Touvron; Iga Kozlowska; Igor Molybog; Igor Tufanov; Iliyan Zarov; Imanol Arrieta Ibarra; Irina-Elena Veliche; Isabel Kloumann; Ishan Misra; Ivan Evtimov; Jacob Xu; Jade Copet; Jake Weissman; Jan Geffert; Jana Vranes; Japhet Asher; Jason Park; Jay Mahadeokar; Jean-Baptiste Gaya; Jeet Shah; Jelmer van der Linde; Jennifer Chan; Jenny Hong; Jenya Lee; Jeremy Fu; Jeremy Teboul; Jianfeng Chi; Jianyu Huang; Jie Wang; Jiecao Yu; Joanna Bitton; Joe Spisak; Joelle Pineau; Jon Carvill; Jongsoo Park; Joseph Rocca; Joshua Johnstun; Junteng Jia; Kalyan Vasuden Alwala; Kam Hou U; Kate Plawiak; Kartikeya Upasani; Kaushik Veeraraghavan; Ke Li; Kenneth Heafield; Kevin Stone; Khalid El-Arini; Krithika Iyer; Kshitiz Malik; Kuenley Chiu; Kunal Bhalla; Kyle Huang; Lakshya Garg; Lauren Rantala-Yeary; Laurens van der Maaten; Lawrence Chen; Leandro Silva; Lee Bell; Lei Zhang; Liang Tan; Louis Martin; Lovish Madaan; Luca Wehrstedt; Lukas Blecher; Luke de Oliveira; Madeline Muzzi; Madian Khabsa; Manav Avlani; Mannat Singh; Manohar Paluri; Mark Zuckerberg; Marcin Kardas; Martynas Mankus; Mathew Oldham; Mathieu Rita; Matthew Lennie; Maya Pavlova; Meghan Keneally; Melanie Kambadur; Mihir Patel; Mikayel Samvelyan; Mike Clark; Mike Lewis; Min Si; Mitesh Kumar Singh; Mo Metanat; Mona Hassan; Naman Goyal; Narjes Torabi; Nicolas Usunier; Nikolay Bashlykov; Nikolay Bogoychev; Niladri Chatterji; Ning Dong; Oliver Aobo Yang; Olivier Duchenne; Onur Celebi; Parth Parekh; Patrick Alrassy; Paul Saab; Pavan Balaji; Pedro Rittner; Pengchuan Zhang; Pengwei Li; Petar Vasic; Peter Weng; Polina Zvyagina; Prajjwal Bhargava; Pratik Dubal; Praveen Krishnan; Punit Singh Koura; Qing He; Rachel Rodriguez; Ragavan Srinivasan; Rahul Mitra; Ramon Calderer; Raymond Li; Robert Stojnic; Roberta Raileanu; Robin Battey; Rocky Wang; Rohit Girdhar; Rohit Patel; Romain Sauvestre; Ronnie Polidoro; Roshan Sumbaly; Ross Taylor; Ruan Silva; Rui Hou; Rui Wang; Russ Howes; Ruty Rinott; Saghar Hosseini; Sai Jayesh Bondu; Samyak Datta; Sanjay Singh; Sara Chugh; Sargun Dhillon; Satadru Pan; Sean Bell; Sergey Edunov; Shaoliang Nie; Sharan Narang; Sharath Raparthy; Shaun Lindsay; Sheng Feng; Sheng Shen; Shenghao Lin; Shiva Shankar; Shruti Bhosale; Shun Zhang; Simon Vandenhende; Sinong Wang; Seohyun Sonia Kim; Soumya Batra; Sten Sootla; Steve Kehoe; Suchin Gururangan; Sumit Gupta; Sunny Virk; Sydney Borodinsky; Tamar Glaser; Tamar Herman; Tamara Best; Tara Fowler; Thomas Georgiou; Thomas Scialom; Tianhe Li; Todor Mihaylov; Tong Xiao; Ujjwal Karn; Vedanuj Goswami; Vibhor Gupta; Vignesh Ramanathan; Viktor Kerkez; Vinay Satish Kumar; Vincent Gonguet; Vish Vogeti; Vlad Poenaru; Vlad Tiberiu Mihailescu; Vladan Petrovic; Vladimir Ivanov; Wei Li; Weiwei Chu; Wenhan Xiong; Wenyin Fu; Wes Bouaziz; Whitney Meers; Will Constable; Xavier Martinet; Xiaojian Wu; Xinbo Gao; Xinfeng Xie; Xuchao Jia; Yaelle Goldschlag; Yann LeCun; Yashesh Gaur; Yasmine Babaei; Ye Qi; Yenda Li; Yi Wen; Yiwen Song; Youngjin Nam; Yuchen Hao; Yuchen Zhang; Yun Wang; Yuning Mao; Yuzi He; Zacharie Delpierre Coudert; Zachary DeVito; Zahra Hankir; Zhaoduo Wen; Zheng Yan; Zhengxing Chen; Zhenyu Yang; Zoe Papakipos" ]

[ "transformers", "tensorboard", "safetensors", "llama", "text-generation", "alignment-handbook", "trl", "sft", "generated_from_trainer", "conversational", "dataset:GAIR/lima", "base_model:JackFram/llama-68m", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[]

[]

[ "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 2e-05\n* train\\_batch\\_size: 32\n* eval\\_batch\\_size: 8\n* seed: 42\n* distributed\\_type: multi-GPU\n* num\\_devices: 4\n* total\\_train\\_batch\\_size: 128\n* total\\_eval\\_batch\\_size: 32\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: cosine\n* lr\\_scheduler\\_warmup\\_ratio: 0.1\n* num\\_epochs: 2.0", "### Training results", "### Framework versions\n\n\n* Transformers 4.38.2\n* Pytorch 2.1.2\n* Datasets 2.14.6\n* Tokenizers 0.15.2" ]

[ "TAGS\n#transformers #tensorboard #safetensors #llama #text-generation #alignment-handbook #trl #sft #generated_from_trainer #conversational #dataset-GAIR/lima #base_model-JackFram/llama-68m #license-apache-2.0 #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n", "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 2e-05\n* train\\_batch\\_size: 32\n* eval\\_batch\\_size: 8\n* seed: 42\n* distributed\\_type: multi-GPU\n* num\\_devices: 4\n* total\\_train\\_batch\\_size: 128\n* total\\_eval\\_batch\\_size: 32\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: cosine\n* lr\\_scheduler\\_warmup\\_ratio: 0.1\n* num\\_epochs: 2.0", "### Training results", "### Framework versions\n\n\n* Transformers 4.38.2\n* Pytorch 2.1.2\n* Datasets 2.14.6\n* Tokenizers 0.15.2" ]

[ "transformers", "safetensors", "llama", "text-generation", "conversational", "en", "doi:10.57967/hf/2135", "license:cc-by-nc-4.0", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[]

[ "en" ]

[ "# Tiny-Pirate-1.1b-v0.2\n\nTiny-Pirate-1.1b-v0.2 is a significantly enhanced version of TinyPirate! This version, meticulously fine-tuned from the TinyLlama-1.1B model, demonstrates marked improvements in performance, thematic adherence, and personality compared to its predecessor, TinyPiratev0.1.\n\n- Developed by: phanerozoic\n- License: cc-by-nc-4.0\n- Finetuned from: TinyLlama-1.1B-Chat-v1.0", "### Version Control\nTiny-Pirate-1.1b-v0.2 represents a major leap forward from the initial release, boasting enhanced pirate personality, thematic consistency, and overall language coherence. This version showcases the potential for iterative fine-tuning to create highly specialized and engaging language models tailored to specific themes and characters.", "### Performance\nIn comparison to TinyPiratev0.1, this version exhibits a far stronger grasp of its pirate identity, delivering responses that are more cohesive, contextually relevant, and thematically adherent. The model's ability to maintain a consistent and authentic pirate tone throughout interactions has been significantly enhanced, resulting in a more immersive, engaging, and entertaining user experience. TinyPiratev0.2 showcases improved language understanding, generation, and contextual awareness, allowing it to handle a wider range of pirate-themed queries and prompts with greater finesse and nuance.", "### Direct Use\nLike its predecessor, Tiny-Pirate-1.1b-v0.2 is ideally suited for applications requiring high-quality, thematic language generation in resource-constrained environments. This includes edge computing, mobile devices, lightweight AI applications, chatbots, games, interactive fiction, and other domains where authentic pirate-themed content is desired. The model's compact size and efficient performance make it an excellent choice for developers and creators looking to integrate engaging, character-driven language experiences into their projects without the need for extensive computational resources.", "### Training Data\nTo ensure rich, diverse, and high-quality inputs for fine-tuning, TinyPiratev0.2 was trained on the same carefully curated pirate-themed dataset used for the development of PirateTalk 8b. This dataset encompasses a wide range of pirate-related content, including historical accounts, literary works, film and television scripts, and more. By exposing the model to such a comprehensive and varied corpus, TinyPiratev0.2 has developed a deep understanding of pirate language, culture, and themes, enabling it to generate content that is both authentic and engaging.", "### Custom Stopping Strings\nTo enhance output quality and maintain better control over the model's behavior, especially in extreme or edge cases, a set of custom stopping strings should be employed:\n\n- \"}\\\\n\\\\n\\\\n{\"\n- \"\\\\user:\"\n- \"\\\\nYou:\"\n- \"\\\\n\"\n\nThese stopping strings help to ensure that the model generates coherent, well-structured, and contextually relevant responses, even in challenging or unexpected situations.", "### Training Hyperparameters and Fine-Tuning Details\nThe hyperparameters used in the fine-tuning of TinyPiratev0.2 were carefully chosen to optimize the model's performance, thematic adherence, and overall language quality. The use of LoRA (Low-Rank Adaptation) technique allowed for efficient and effective fine-tuning while minimizing the risk of overfitting.\n\nSome key hyperparameters include:\n- LoRA Rank: 2048\n- LoRA Alpha: 4096\n- LoRA Dropout: 0.05\n- Micro Batch Size: 12\n- Epochs: 1.01\n- Learning Rate: 2e-5\n- LR Scheduler: Linear\n- Cutoff Length: 256\n- Warmup Ratio: 0\n- Gradient Accumulation: 1\n\nThese hyperparameters were arrived at through extensive experimentation and tuning, with the goal of striking a balance between model performance, training efficiency, and generalization ability. The relatively high LoRA Rank and LoRA Alpha values allow for more expressive and nuanced adaptations of the base model, while the low LoRA Dropout helps to prevent overfitting. The use of a linear learning rate scheduler and a small warmup ratio ensures stable and consistent learning throughout the training process.\n\nThe choice of a micro batch size of 12, combined with a single gradient accumulation step, enables efficient utilization of computational resources while maintaining a sufficiently large effective batch size for stable training. The cutoff length of 256 tokens helps to focus the model's attention on relevant context while minimizing the computational overhead associated with processing long sequences.\n\nOverall, these hyperparameters reflect an empirically validated approach to fine-tuning, aimed at maximizing the model's performance and thematic coherence within the constraints of the available computational resources.", "### Limitations\nWhile TinyPiratev0.2 demonstrates significant improvements in thematic performance and language quality compared to its predecessor, it is essential to recognize that it remains a compact model with inherent limitations. As such, it may not handle highly complex, abstract, or ambiguous language tasks with the same level of proficiency as larger, more general-purpose models. Additionally, the model's specialization in pirate dialect and themes necessarily limits its applicability to general language applications, where a more neutral and versatile language model may be required.", "### Compute Infrastructure\nThe training of TinyPiratev0.2 was conducted efficiently using a single RTX 6000 Ada Lovelace GPU, showcasing the model's ability to achieve significant performance gains with relatively modest computational resources. The entire fine-tuning process was completed in approximately 4.3 minutes, highlighting the resource-effective nature of specialized model development and the efficiency of the LoRA technique.\n\nThis efficient training process underscores the potential for creating high-quality, specialized language models that can be developed and deployed quickly and cost-effectively, making them accessible to a wider range of developers, researchers, and creators.", "### Results\nTinyPiratev0.2 exhibits a remarkable improvement in its ability to generate pirate-themed content that is engaging, immersive, and thematically consistent. The model's responses are characterized by a strong pirate personality, with language that is colorful, idiomatic, and true to the spirit of pirate culture. Compared to the previous version, TinyPiratev0.2 demonstrates a deeper understanding of context, a more coherent narrative flow, and a greater ability to handle a wide range of pirate-related topics and scenarios.\n\nThese results underscore the potential for focused fine-tuning to create language models that are not only highly specialized but also capable of delivering rich, immersive, and resonant user experiences.", "### Future Developments\nWhile TinyPiratev0.2 represents a significant achievement in the development of compact, specialized language models, it is likely to be the last iteration of this specific model size and architecture. As the field of natural language processing continues to evolve and new architectures and techniques emerge, future developments may explore the integration of TinyPirate with more advanced base models, such as Microsoft Phi or other state-of-the-art offerings.\n\nMoreover, as smaller models continue to improve in performance and efficiency relative to their larger counterparts, there may be opportunities to further optimize and compress the TinyPirate model while maintaining or even enhancing its thematic coherence and language quality.\n\nFuture work may also investigate the application of the TinyPirate methodology to other specialized domains and themes, demonstrating the versatility and adaptability of this approach to language model development.", "### Acknowledgments\nThe development of TinyPiratev0.2 would not have been possible without the groundbreaking work of the TinyLlama developers, whose innovative approach to compact language model design laid the foundation for this project. Their commitment to open-source research and their willingness to share their knowledge and expertise have been instrumental in advancing the field of specialized language modeling.\n\nSpecial thanks also goes to s3nh for their support and popularization of our project.", "### Summary\nTiny-Pirate-1.1b-v0.2 represents a major milestone in the development of compact, specialized language models designed for thematic content generation. With its enhanced performance, improved thematic coherence, and engaging pirate personality, this model showcases the potential for focused fine-tuning to create language models that are not only efficient and resource-effective but also capable of delivering rich, immersive, and emotionally resonant user experiences.\n\nAs the field of natural language processing continues to evolve, TinyPiratev0.2 stands as a testament to the power and potential of specialized language modeling. It demonstrates that through careful fine-tuning, even compact models can achieve remarkable levels of thematic adherence, language quality, and user engagement, opening up new possibilities for the development of character-driven, domain-specific language applications.\n\nWhile future iterations of the TinyPirate model may explore new architectures and techniques, the lessons learned and the methodologies developed in the creation of TinyPiratev0.2 will undoubtedly inform and inspire further advancements in the field. As such, this model represents not only a significant achievement in its own right but also a valuable contribution to the ongoing exploration of the frontiers of language modeling and its applications in a wide range of domains and use cases." ]

[ "TAGS\n#transformers #safetensors #llama #text-generation #conversational #en #doi-10.57967/hf/2135 #license-cc-by-nc-4.0 #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n", "# Tiny-Pirate-1.1b-v0.2\n\nTiny-Pirate-1.1b-v0.2 is a significantly enhanced version of TinyPirate! This version, meticulously fine-tuned from the TinyLlama-1.1B model, demonstrates marked improvements in performance, thematic adherence, and personality compared to its predecessor, TinyPiratev0.1.\n\n- Developed by: phanerozoic\n- License: cc-by-nc-4.0\n- Finetuned from: TinyLlama-1.1B-Chat-v1.0", "### Version Control\nTiny-Pirate-1.1b-v0.2 represents a major leap forward from the initial release, boasting enhanced pirate personality, thematic consistency, and overall language coherence. This version showcases the potential for iterative fine-tuning to create highly specialized and engaging language models tailored to specific themes and characters.", "### Performance\nIn comparison to TinyPiratev0.1, this version exhibits a far stronger grasp of its pirate identity, delivering responses that are more cohesive, contextually relevant, and thematically adherent. The model's ability to maintain a consistent and authentic pirate tone throughout interactions has been significantly enhanced, resulting in a more immersive, engaging, and entertaining user experience. TinyPiratev0.2 showcases improved language understanding, generation, and contextual awareness, allowing it to handle a wider range of pirate-themed queries and prompts with greater finesse and nuance.", "### Direct Use\nLike its predecessor, Tiny-Pirate-1.1b-v0.2 is ideally suited for applications requiring high-quality, thematic language generation in resource-constrained environments. This includes edge computing, mobile devices, lightweight AI applications, chatbots, games, interactive fiction, and other domains where authentic pirate-themed content is desired. The model's compact size and efficient performance make it an excellent choice for developers and creators looking to integrate engaging, character-driven language experiences into their projects without the need for extensive computational resources.", "### Training Data\nTo ensure rich, diverse, and high-quality inputs for fine-tuning, TinyPiratev0.2 was trained on the same carefully curated pirate-themed dataset used for the development of PirateTalk 8b. This dataset encompasses a wide range of pirate-related content, including historical accounts, literary works, film and television scripts, and more. By exposing the model to such a comprehensive and varied corpus, TinyPiratev0.2 has developed a deep understanding of pirate language, culture, and themes, enabling it to generate content that is both authentic and engaging.", "### Custom Stopping Strings\nTo enhance output quality and maintain better control over the model's behavior, especially in extreme or edge cases, a set of custom stopping strings should be employed:\n\n- \"}\\\\n\\\\n\\\\n{\"\n- \"\\\\user:\"\n- \"\\\\nYou:\"\n- \"\\\\n\"\n\nThese stopping strings help to ensure that the model generates coherent, well-structured, and contextually relevant responses, even in challenging or unexpected situations.", "### Training Hyperparameters and Fine-Tuning Details\nThe hyperparameters used in the fine-tuning of TinyPiratev0.2 were carefully chosen to optimize the model's performance, thematic adherence, and overall language quality. The use of LoRA (Low-Rank Adaptation) technique allowed for efficient and effective fine-tuning while minimizing the risk of overfitting.\n\nSome key hyperparameters include:\n- LoRA Rank: 2048\n- LoRA Alpha: 4096\n- LoRA Dropout: 0.05\n- Micro Batch Size: 12\n- Epochs: 1.01\n- Learning Rate: 2e-5\n- LR Scheduler: Linear\n- Cutoff Length: 256\n- Warmup Ratio: 0\n- Gradient Accumulation: 1\n\nThese hyperparameters were arrived at through extensive experimentation and tuning, with the goal of striking a balance between model performance, training efficiency, and generalization ability. The relatively high LoRA Rank and LoRA Alpha values allow for more expressive and nuanced adaptations of the base model, while the low LoRA Dropout helps to prevent overfitting. The use of a linear learning rate scheduler and a small warmup ratio ensures stable and consistent learning throughout the training process.\n\nThe choice of a micro batch size of 12, combined with a single gradient accumulation step, enables efficient utilization of computational resources while maintaining a sufficiently large effective batch size for stable training. The cutoff length of 256 tokens helps to focus the model's attention on relevant context while minimizing the computational overhead associated with processing long sequences.\n\nOverall, these hyperparameters reflect an empirically validated approach to fine-tuning, aimed at maximizing the model's performance and thematic coherence within the constraints of the available computational resources.", "### Limitations\nWhile TinyPiratev0.2 demonstrates significant improvements in thematic performance and language quality compared to its predecessor, it is essential to recognize that it remains a compact model with inherent limitations. As such, it may not handle highly complex, abstract, or ambiguous language tasks with the same level of proficiency as larger, more general-purpose models. Additionally, the model's specialization in pirate dialect and themes necessarily limits its applicability to general language applications, where a more neutral and versatile language model may be required.", "### Compute Infrastructure\nThe training of TinyPiratev0.2 was conducted efficiently using a single RTX 6000 Ada Lovelace GPU, showcasing the model's ability to achieve significant performance gains with relatively modest computational resources. The entire fine-tuning process was completed in approximately 4.3 minutes, highlighting the resource-effective nature of specialized model development and the efficiency of the LoRA technique.\n\nThis efficient training process underscores the potential for creating high-quality, specialized language models that can be developed and deployed quickly and cost-effectively, making them accessible to a wider range of developers, researchers, and creators.", "### Results\nTinyPiratev0.2 exhibits a remarkable improvement in its ability to generate pirate-themed content that is engaging, immersive, and thematically consistent. The model's responses are characterized by a strong pirate personality, with language that is colorful, idiomatic, and true to the spirit of pirate culture. Compared to the previous version, TinyPiratev0.2 demonstrates a deeper understanding of context, a more coherent narrative flow, and a greater ability to handle a wide range of pirate-related topics and scenarios.\n\nThese results underscore the potential for focused fine-tuning to create language models that are not only highly specialized but also capable of delivering rich, immersive, and resonant user experiences.", "### Future Developments\nWhile TinyPiratev0.2 represents a significant achievement in the development of compact, specialized language models, it is likely to be the last iteration of this specific model size and architecture. As the field of natural language processing continues to evolve and new architectures and techniques emerge, future developments may explore the integration of TinyPirate with more advanced base models, such as Microsoft Phi or other state-of-the-art offerings.\n\nMoreover, as smaller models continue to improve in performance and efficiency relative to their larger counterparts, there may be opportunities to further optimize and compress the TinyPirate model while maintaining or even enhancing its thematic coherence and language quality.\n\nFuture work may also investigate the application of the TinyPirate methodology to other specialized domains and themes, demonstrating the versatility and adaptability of this approach to language model development.", "### Acknowledgments\nThe development of TinyPiratev0.2 would not have been possible without the groundbreaking work of the TinyLlama developers, whose innovative approach to compact language model design laid the foundation for this project. Their commitment to open-source research and their willingness to share their knowledge and expertise have been instrumental in advancing the field of specialized language modeling.\n\nSpecial thanks also goes to s3nh for their support and popularization of our project.", "### Summary\nTiny-Pirate-1.1b-v0.2 represents a major milestone in the development of compact, specialized language models designed for thematic content generation. With its enhanced performance, improved thematic coherence, and engaging pirate personality, this model showcases the potential for focused fine-tuning to create language models that are not only efficient and resource-effective but also capable of delivering rich, immersive, and emotionally resonant user experiences.\n\nAs the field of natural language processing continues to evolve, TinyPiratev0.2 stands as a testament to the power and potential of specialized language modeling. It demonstrates that through careful fine-tuning, even compact models can achieve remarkable levels of thematic adherence, language quality, and user engagement, opening up new possibilities for the development of character-driven, domain-specific language applications.\n\nWhile future iterations of the TinyPirate model may explore new architectures and techniques, the lessons learned and the methodologies developed in the creation of TinyPiratev0.2 will undoubtedly inform and inspire further advancements in the field. As such, this model represents not only a significant achievement in its own right but also a valuable contribution to the ongoing exploration of the frontiers of language modeling and its applications in a wide range of domains and use cases." ]

[ "transformers", "safetensors", "mistral", "text-generation", "4-bit", "AWQ", "autotrain_compatible", "endpoints_compatible", "en", "dataset:ResplendentAI/Luna_Alpaca", "base_model:ChaoticNeutrals/InfinityNexus_9B", "license:apache-2.0", "text-generation-inference", "region:us" ]

[]

[ "en" ]

[ "# jeiku/Garbage_9B AWQ\n\n- Model creator: jeiku\n- Original model: Garbage_9B\n\n!image/png", "## Model Summary\n\nThis is a finetune of InfinityNexus_9B. This is my first time tuning a frankenmerge, so hopefully it works out. The goal is to improve intelligence and RP ability beyond the 7B original models.", "## How to use", "### Install the necessary packages", "### Example Python code", "### About AWQ\n\nAWQ is an efficient, accurate and blazing-fast low-bit weight quantization method, currently supporting 4-bit quantization. Compared to GPTQ, it offers faster Transformers-based inference with equivalent or better quality compared to the most commonly used GPTQ settings.\n\nAWQ models are currently supported on Linux and Windows, with NVidia GPUs only. macOS users: please use GGUF models instead.\n\nIt is supported by:\n\n- Text Generation Webui - using Loader: AutoAWQ\n- vLLM - version 0.2.2 or later for support for all model types.\n- Hugging Face Text Generation Inference (TGI)\n- Transformers version 4.35.0 and later, from any code or client that supports Transformers\n- AutoAWQ - for use from Python code" ]

[ "TAGS\n#transformers #safetensors #mistral #text-generation #4-bit #AWQ #autotrain_compatible #endpoints_compatible #en #dataset-ResplendentAI/Luna_Alpaca #base_model-ChaoticNeutrals/InfinityNexus_9B #license-apache-2.0 #text-generation-inference #region-us \n", "# jeiku/Garbage_9B AWQ\n\n- Model creator: jeiku\n- Original model: Garbage_9B\n\n!image/png", "## Model Summary\n\nThis is a finetune of InfinityNexus_9B. This is my first time tuning a frankenmerge, so hopefully it works out. The goal is to improve intelligence and RP ability beyond the 7B original models.", "## How to use", "### Install the necessary packages", "### Example Python code", "### About AWQ\n\nAWQ is an efficient, accurate and blazing-fast low-bit weight quantization method, currently supporting 4-bit quantization. Compared to GPTQ, it offers faster Transformers-based inference with equivalent or better quality compared to the most commonly used GPTQ settings.\n\nAWQ models are currently supported on Linux and Windows, with NVidia GPUs only. macOS users: please use GGUF models instead.\n\nIt is supported by:\n\n- Text Generation Webui - using Loader: AutoAWQ\n- vLLM - version 0.2.2 or later for support for all model types.\n- Hugging Face Text Generation Inference (TGI)\n- Transformers version 4.35.0 and later, from any code or client that supports Transformers\n- AutoAWQ - for use from Python code" ]

[ "transformers", "safetensors", "vision-encoder-decoder", "image-to-text", "endpoints_compatible", "region:us" ]

[]

[]

[ "## Model description\n\nThis model is an encoder-decoder model, consisting of an image Transformer as encoder, and a text Transformer as decoder. The image encoder was initialized from the weights of Dino-v2, while the text decoder was initialized from the weights of RoBERTa.\n\nImages are presented to the model as a sequence of fixed-size patches (resolution 16x16), which are linearly embedded. One also adds absolute position embeddings before feeding the sequence to the layers of the Transformer encoder. Next, the Transformer text decoder autoregressively generates tokens.", "### How to use\n\nHere is how to use this model in PyTorch:\n\n\n---\nlicense: Konthee\n---" ]

[ "TAGS\n#transformers #safetensors #vision-encoder-decoder #image-to-text #endpoints_compatible #region-us \n", "## Model description\n\nThis model is an encoder-decoder model, consisting of an image Transformer as encoder, and a text Transformer as decoder. The image encoder was initialized from the weights of Dino-v2, while the text decoder was initialized from the weights of RoBERTa.\n\nImages are presented to the model as a sequence of fixed-size patches (resolution 16x16), which are linearly embedded. One also adds absolute position embeddings before feeding the sequence to the layers of the Transformer encoder. Next, the Transformer text decoder autoregressively generates tokens.", "### How to use\n\nHere is how to use this model in PyTorch:\n\n\n---\nlicense: Konthee\n---" ]

[ "region:us" ]

[]

[]

[ "# Keto Energize: Effektiv und Preisgünstig für Ihre Fitnessziele!\n\n![Keto Energize](URL\n\nTräumen Sie von einem fitteren und gesünderen Körper? Möchten Sie die Vorteile der ketogenen Ernährung nutzen, aber nicht Ihr gesamtes Budget dafür aufwenden? Dann ist Keto Energize die Lösung, nach der Sie gesucht haben!\n\n__Erfahren Sie mehr über Keto Energize hier!__", "### Warum Keto Energize?\n\nKeto Energize ist eine hochwirksame Ergänzung, die Ihnen dabei hilft, Ihren Körper in den Zustand der Ketose zu versetzen, ohne dabei Ihren Geldbeutel zu belasten. Diese einzigartige Formel wurde entwickelt, um Ihre Fettverbrennung zu beschleunigen, Ihre Energie zu steigern und Ihren Appetit zu kontrollieren, alles zu einem erschwinglichen Preis.", "### Die Vorteile von Keto Energize:\n\n- Erschwinglichkeit: Im Gegensatz zu teuren Alternativen auf dem Markt bietet Keto Energize hochwertige Ergebnisse zu einem Bruchteil der Kosten.\n\n- Effektive Fettverbrennung: Dank seiner leistungsstarken Inhaltsstoffe hilft Keto Energize Ihrem Körper, Fett effizienter zu verbrennen und Ihre Fitnessziele schneller zu erreichen.\n\n- Positive Bewertungen: Unsere zufriedenen Kunden sprechen für sich! Erfahren Sie aus erster Hand, wie Keto Energize das Leben vieler Menschen verändert hat.", "### Wie man Keto Energize verwendet:\n\nDie Anwendung von Keto Energize ist denkbar einfach. Nehmen Sie einfach die empfohlene Dosis ein, vorzugsweise vor den Mahlzeiten, und lassen Sie die kraftvollen Inhaltsstoffe ihre Arbeit tun.", "### Was unsere Kunden sagen:\n\n- Markus: \"Ich war auf der Suche nach einer kostengünstigen Möglichkeit, die Vorteile der ketogenen Ernährung zu nutzen, und Keto Energize hat meine Erwartungen übertroffen! Ich fühle mich energiegeladener und sehe bereits Fortschritte bei meiner Gewichtsabnahme.\"\n\n- Anna: \"Als Studentin ist es mir wichtig, mein Budget im Auge zu behalten. Keto Energize hat mir geholfen, meine Fitnessziele zu erreichen, ohne dabei mein Bankkonto zu belasten. Ich kann es nur empfehlen!\"", "### Bestellen Sie jetzt Keto Energize!\n\nMöchten auch Sie von den Vorteilen von Keto Energize profitieren? Worauf warten Sie noch? Bestellen Sie noch heute und nehmen Sie den ersten Schritt zu einem fitteren und gesünderen Lebensstil!\n\n*Klicken Sie unten, um Ihr Keto Energize noch heute zu bestellen und Ihre Fitnessziele zu verwirklichen!*\n\n__Jetzt bestellen!__" ]

[ "TAGS\n#region-us \n", "# Keto Energize: Effektiv und Preisgünstig für Ihre Fitnessziele!\n\n![Keto Energize](URL\n\nTräumen Sie von einem fitteren und gesünderen Körper? Möchten Sie die Vorteile der ketogenen Ernährung nutzen, aber nicht Ihr gesamtes Budget dafür aufwenden? Dann ist Keto Energize die Lösung, nach der Sie gesucht haben!\n\n__Erfahren Sie mehr über Keto Energize hier!__", "### Warum Keto Energize?\n\nKeto Energize ist eine hochwirksame Ergänzung, die Ihnen dabei hilft, Ihren Körper in den Zustand der Ketose zu versetzen, ohne dabei Ihren Geldbeutel zu belasten. Diese einzigartige Formel wurde entwickelt, um Ihre Fettverbrennung zu beschleunigen, Ihre Energie zu steigern und Ihren Appetit zu kontrollieren, alles zu einem erschwinglichen Preis.", "### Die Vorteile von Keto Energize:\n\n- Erschwinglichkeit: Im Gegensatz zu teuren Alternativen auf dem Markt bietet Keto Energize hochwertige Ergebnisse zu einem Bruchteil der Kosten.\n\n- Effektive Fettverbrennung: Dank seiner leistungsstarken Inhaltsstoffe hilft Keto Energize Ihrem Körper, Fett effizienter zu verbrennen und Ihre Fitnessziele schneller zu erreichen.\n\n- Positive Bewertungen: Unsere zufriedenen Kunden sprechen für sich! Erfahren Sie aus erster Hand, wie Keto Energize das Leben vieler Menschen verändert hat.", "### Wie man Keto Energize verwendet:\n\nDie Anwendung von Keto Energize ist denkbar einfach. Nehmen Sie einfach die empfohlene Dosis ein, vorzugsweise vor den Mahlzeiten, und lassen Sie die kraftvollen Inhaltsstoffe ihre Arbeit tun.", "### Was unsere Kunden sagen:\n\n- Markus: \"Ich war auf der Suche nach einer kostengünstigen Möglichkeit, die Vorteile der ketogenen Ernährung zu nutzen, und Keto Energize hat meine Erwartungen übertroffen! Ich fühle mich energiegeladener und sehe bereits Fortschritte bei meiner Gewichtsabnahme.\"\n\n- Anna: \"Als Studentin ist es mir wichtig, mein Budget im Auge zu behalten. Keto Energize hat mir geholfen, meine Fitnessziele zu erreichen, ohne dabei mein Bankkonto zu belasten. Ich kann es nur empfehlen!\"", "### Bestellen Sie jetzt Keto Energize!\n\nMöchten auch Sie von den Vorteilen von Keto Energize profitieren? Worauf warten Sie noch? Bestellen Sie noch heute und nehmen Sie den ersten Schritt zu einem fitteren und gesünderen Lebensstil!\n\n*Klicken Sie unten, um Ihr Keto Energize noch heute zu bestellen und Ihre Fitnessziele zu verwirklichen!*\n\n__Jetzt bestellen!__" ]

[ "transformers", "safetensors", "llama", "text-generation", "unsloth", "arxiv:1910.09700", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "4-bit", "region:us" ]

[ "1910.09700" ]

[]

[ "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "TAGS\n#transformers #safetensors #llama #text-generation #unsloth #arxiv-1910.09700 #autotrain_compatible #endpoints_compatible #text-generation-inference #4-bit #region-us \n", "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "diffusers", "safetensors", "arxiv:1910.09700", "endpoints_compatible", "diffusers:StableDiffusionPipeline", "region:us" ]

[ "1910.09700" ]

[]

[ "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a diffusers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "TAGS\n#diffusers #safetensors #arxiv-1910.09700 #endpoints_compatible #diffusers-StableDiffusionPipeline #region-us \n", "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a diffusers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "transformers", "safetensors", "vits", "text-to-audio", "text-to-speech", "sw", "dataset:google/fleurs", "arxiv:1910.09700", "endpoints_compatible", "region:us" ]

[ "1910.09700" ]

[ "sw" ]

[ "# Model Card for Model ID\n\n\n\nThis modelcard aims to be a base template for new models. It has been generated using this raw template.", "## Model Details", "### Model Description\n\n\n\n\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "TAGS\n#transformers #safetensors #vits #text-to-audio #text-to-speech #sw #dataset-google/fleurs #arxiv-1910.09700 #endpoints_compatible #region-us \n", "# Model Card for Model ID\n\n\n\nThis modelcard aims to be a base template for new models. It has been generated using this raw template.", "## Model Details", "### Model Description\n\n\n\n\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "mlx", "safetensors", "phi3", "custom_code", "license:apache-2.0", "region:us" ]

[]

[]

[ "# GreenBitAI/Phi-3-mini-4k-instruct-layer-mix-bpw-3.0-mlx\nThis quantized low-bit model was converted to MLX format from ['GreenBitAI/Phi-3-mini-4k-instruct-layer-mix-bpw-3.0']().\nRefer to the original model card for more details on the model.", "## Use with mlx" ]

[ "TAGS\n#mlx #safetensors #phi3 #custom_code #license-apache-2.0 #region-us \n", "# GreenBitAI/Phi-3-mini-4k-instruct-layer-mix-bpw-3.0-mlx\nThis quantized low-bit model was converted to MLX format from ['GreenBitAI/Phi-3-mini-4k-instruct-layer-mix-bpw-3.0']().\nRefer to the original model card for more details on the model.", "## Use with mlx" ]

[ "transformers", "safetensors", "gpt2", "text-classification", "mergekit", "merge", "arxiv:2306.01708", "base_model:mllm-dev/gpt2_f_experiment_0_drug_data_new_run", "base_model:mllm-dev/gpt2_f_experiment_2_drug_data_new_run", "base_model:mllm-dev/gpt2_f_experiment_4_drug_data_new_run", "base_model:mllm-dev/gpt2_f_experiment_3_drug_data_new_run", "base_model:mllm-dev/gpt2_f_experiment_1_drug_data_new_run", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[ "2306.01708" ]

[]

[ "# tam_test_merge_out_drug_data_ties_test_new_run\n\nThis is a merge of pre-trained language models created using mergekit.", "## Merge Details", "### Merge Method\n\nThis model was merged using the TIES merge method using mllm-dev/gpt2_f_experiment_0_drug_data_new_run as a base.", "### Models Merged\n\nThe following models were included in the merge:\n* mllm-dev/gpt2_f_experiment_2_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_4_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_3_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_1_drug_data_new_run", "### Configuration\n\nThe following YAML configuration was used to produce this model:" ]

[ "TAGS\n#transformers #safetensors #gpt2 #text-classification #mergekit #merge #arxiv-2306.01708 #base_model-mllm-dev/gpt2_f_experiment_0_drug_data_new_run #base_model-mllm-dev/gpt2_f_experiment_2_drug_data_new_run #base_model-mllm-dev/gpt2_f_experiment_4_drug_data_new_run #base_model-mllm-dev/gpt2_f_experiment_3_drug_data_new_run #base_model-mllm-dev/gpt2_f_experiment_1_drug_data_new_run #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n", "# tam_test_merge_out_drug_data_ties_test_new_run\n\nThis is a merge of pre-trained language models created using mergekit.", "## Merge Details", "### Merge Method\n\nThis model was merged using the TIES merge method using mllm-dev/gpt2_f_experiment_0_drug_data_new_run as a base.", "### Models Merged\n\nThe following models were included in the merge:\n* mllm-dev/gpt2_f_experiment_2_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_4_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_3_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_1_drug_data_new_run", "### Configuration\n\nThe following YAML configuration was used to produce this model:" ]

[ "gguf", "pruna-ai", "region:us" ]

[]

[]

[ "### How to load this model in Python code, using llama-cpp-python\n\nFor full documentation, please see: llama-cpp-python docs.", "#### First install the package\n\nRun one of the following commands, according to your system:", "#### Simple llama-cpp-python example code\n\n```\n\n* Option D - Running with LangChain\n\n\nHere are guides on using llama-cpp-python and ctransformers with LangChain:\n\n\n* LangChain + llama-cpp-python\n* LangChain + ctransformers\n\n\nConfigurations\n--------------\n\n\nThe configuration info are in 'smash\\_config.json'.\n\n\nCredits & License\n-----------------\n\n\nThe license of the smashed model follows the license of the original model. Please check the license of the original model before using this model which provided the base model. The license of the 'pruna-engine' is here on Pypi.\n\n\nWant to compress other models?\n------------------------------\n\n\n* Contact us and tell us which model to compress next here.\n* Request access to easily compress your own AI models here." ]

[ "TAGS\n#gguf #pruna-ai #region-us \n", "### How to load this model in Python code, using llama-cpp-python\n\nFor full documentation, please see: llama-cpp-python docs.", "#### First install the package\n\nRun one of the following commands, according to your system:", "#### Simple llama-cpp-python example code\n\n```\n\n* Option D - Running with LangChain\n\n\nHere are guides on using llama-cpp-python and ctransformers with LangChain:\n\n\n* LangChain + llama-cpp-python\n* LangChain + ctransformers\n\n\nConfigurations\n--------------\n\n\nThe configuration info are in 'smash\\_config.json'.\n\n\nCredits & License\n-----------------\n\n\nThe license of the smashed model follows the license of the original model. Please check the license of the original model before using this model which provided the base model. The license of the 'pruna-engine' is here on Pypi.\n\n\nWant to compress other models?\n------------------------------\n\n\n* Contact us and tell us which model to compress next here.\n* Request access to easily compress your own AI models here." ]

[ "transformers", "gguf", "en", "base_model:cza/WZU-3D-ChatLLM-8B", "license:apache-2.0", "endpoints_compatible", "region:us" ]

[]

[ "en" ]

[]

[ "TAGS\n#transformers #gguf #en #base_model-cza/WZU-3D-ChatLLM-8B #license-apache-2.0 #endpoints_compatible #region-us \n" ]

[ "transformers", "safetensors", "arxiv:1910.09700", "endpoints_compatible", "region:us" ]

[ "1910.09700" ]

[]

[ "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "TAGS\n#transformers #safetensors #arxiv-1910.09700 #endpoints_compatible #region-us \n", "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "peft", "safetensors", "generated_from_trainer", "base_model:mahdibaghbanzadeh/seqsight_4096_512_27M", "region:us" ]

[]

[]

[ "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0005\n* train\\_batch\\_size: 512\n* eval\\_batch\\_size: 512\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* training\\_steps: 10000", "### Training results", "### Framework versions\n\n\n* PEFT 0.9.0\n* Transformers 4.38.2\n* Pytorch 2.2.0+cu121\n* Datasets 2.17.1\n* Tokenizers 0.15.2" ]

[ "TAGS\n#peft #safetensors #generated_from_trainer #base_model-mahdibaghbanzadeh/seqsight_4096_512_27M #region-us \n", "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0005\n* train\\_batch\\_size: 512\n* eval\\_batch\\_size: 512\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* training\\_steps: 10000", "### Training results", "### Framework versions\n\n\n* PEFT 0.9.0\n* Transformers 4.38.2\n* Pytorch 2.2.0+cu121\n* Datasets 2.17.1\n* Tokenizers 0.15.2" ]

[ "transformers", "safetensors", "llama", "text-generation", "trl", "sft", "conversational", "arxiv:1910.09700", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "4-bit", "region:us" ]

[ "1910.09700" ]

[]

[ "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "TAGS\n#transformers #safetensors #llama #text-generation #trl #sft #conversational #arxiv-1910.09700 #autotrain_compatible #endpoints_compatible #text-generation-inference #4-bit #region-us \n", "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "transformers", "safetensors", "gpt2", "text-classification", "mergekit", "merge", "arxiv:2311.03099", "arxiv:2306.01708", "base_model:mllm-dev/gpt2_f_experiment_4_drug_data_new_run", "base_model:mllm-dev/gpt2_f_experiment_0_drug_data_new_run", "base_model:mllm-dev/gpt2_f_experiment_1_drug_data_new_run", "base_model:mllm-dev/gpt2_f_experiment_3_drug_data_new_run", "base_model:mllm-dev/gpt2_f_experiment_2_drug_data_new_run", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[ "2311.03099", "2306.01708" ]

[]

[ "# tam_test_merge_out_drug_data_dare_ties_test_new_run\n\nThis is a merge of pre-trained language models created using mergekit.", "## Merge Details", "### Merge Method\n\nThis model was merged using the DARE TIES merge method using mllm-dev/gpt2_f_experiment_0_drug_data_new_run as a base.", "### Models Merged\n\nThe following models were included in the merge:\n* mllm-dev/gpt2_f_experiment_4_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_1_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_3_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_2_drug_data_new_run", "### Configuration\n\nThe following YAML configuration was used to produce this model:" ]

[ "TAGS\n#transformers #safetensors #gpt2 #text-classification #mergekit #merge #arxiv-2311.03099 #arxiv-2306.01708 #base_model-mllm-dev/gpt2_f_experiment_4_drug_data_new_run #base_model-mllm-dev/gpt2_f_experiment_0_drug_data_new_run #base_model-mllm-dev/gpt2_f_experiment_1_drug_data_new_run #base_model-mllm-dev/gpt2_f_experiment_3_drug_data_new_run #base_model-mllm-dev/gpt2_f_experiment_2_drug_data_new_run #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n", "# tam_test_merge_out_drug_data_dare_ties_test_new_run\n\nThis is a merge of pre-trained language models created using mergekit.", "## Merge Details", "### Merge Method\n\nThis model was merged using the DARE TIES merge method using mllm-dev/gpt2_f_experiment_0_drug_data_new_run as a base.", "### Models Merged\n\nThe following models were included in the merge:\n* mllm-dev/gpt2_f_experiment_4_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_1_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_3_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_2_drug_data_new_run", "### Configuration\n\nThe following YAML configuration was used to produce this model:" ]

[ "transformers", "safetensors", "stablelm", "text-generation", "conversational", "arxiv:1910.09700", "autotrain_compatible", "endpoints_compatible", "region:us" ]

[ "1910.09700" ]

[]

[ "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "TAGS\n#transformers #safetensors #stablelm #text-generation #conversational #arxiv-1910.09700 #autotrain_compatible #endpoints_compatible #region-us \n", "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "mlx", "safetensors", "phi3", "custom_code", "license:apache-2.0", "region:us" ]

[]

[]

[ "# GreenBitAI/Phi-3-mini-4k-instruct-layer-mix-bpw-2.5-mlx\nThis quantized low-bit model was converted to MLX format from ['GreenBitAI/Phi-3-mini-4k-instruct-layer-mix-bpw-2.5']().\nRefer to the original model card for more details on the model.", "## Use with mlx" ]

[ "TAGS\n#mlx #safetensors #phi3 #custom_code #license-apache-2.0 #region-us \n", "# GreenBitAI/Phi-3-mini-4k-instruct-layer-mix-bpw-2.5-mlx\nThis quantized low-bit model was converted to MLX format from ['GreenBitAI/Phi-3-mini-4k-instruct-layer-mix-bpw-2.5']().\nRefer to the original model card for more details on the model.", "## Use with mlx" ]

[ "transformers", "safetensors", "gpt2", "text-classification", "mergekit", "merge", "arxiv:2311.03099", "base_model:mllm-dev/gpt2_f_experiment_2_drug_data_new_run", "base_model:mllm-dev/gpt2_f_experiment_0_drug_data_new_run", "base_model:mllm-dev/gpt2_f_experiment_4_drug_data_new_run", "base_model:mllm-dev/gpt2_f_experiment_3_drug_data_new_run", "base_model:mllm-dev/gpt2_f_experiment_1_drug_data_new_run", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[ "2311.03099" ]

[]

[ "# tam_test_merge_out_drug_data_dare_linear_test_new_run\n\nThis is a merge of pre-trained language models created using mergekit.", "## Merge Details", "### Merge Method\n\nThis model was merged using the linear DARE merge method using mllm-dev/gpt2_f_experiment_0_drug_data_new_run as a base.", "### Models Merged\n\nThe following models were included in the merge:\n* mllm-dev/gpt2_f_experiment_2_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_4_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_3_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_1_drug_data_new_run", "### Configuration\n\nThe following YAML configuration was used to produce this model:" ]

[ "TAGS\n#transformers #safetensors #gpt2 #text-classification #mergekit #merge #arxiv-2311.03099 #base_model-mllm-dev/gpt2_f_experiment_2_drug_data_new_run #base_model-mllm-dev/gpt2_f_experiment_0_drug_data_new_run #base_model-mllm-dev/gpt2_f_experiment_4_drug_data_new_run #base_model-mllm-dev/gpt2_f_experiment_3_drug_data_new_run #base_model-mllm-dev/gpt2_f_experiment_1_drug_data_new_run #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n", "# tam_test_merge_out_drug_data_dare_linear_test_new_run\n\nThis is a merge of pre-trained language models created using mergekit.", "## Merge Details", "### Merge Method\n\nThis model was merged using the linear DARE merge method using mllm-dev/gpt2_f_experiment_0_drug_data_new_run as a base.", "### Models Merged\n\nThe following models were included in the merge:\n* mllm-dev/gpt2_f_experiment_2_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_4_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_3_drug_data_new_run\n* mllm-dev/gpt2_f_experiment_1_drug_data_new_run", "### Configuration\n\nThe following YAML configuration was used to produce this model:" ]

[ "ml-agents", "tensorboard", "onnx", "Huggy", "deep-reinforcement-learning", "reinforcement-learning", "ML-Agents-Huggy", "region:us" ]

[]

[]

[ "# ppo Agent playing Huggy\n This is a trained model of a ppo agent playing Huggy\n using the Unity ML-Agents Library.\n\n ## Usage (with ML-Agents)\n The Documentation: URL\n\n We wrote a complete tutorial to learn to train your first agent using ML-Agents and publish it to the Hub:\n - A *short tutorial* where you teach Huggy the Dog to fetch the stick and then play with him directly in your\n browser: URL\n - A *longer tutorial* to understand how works ML-Agents:\n URL\n\n ### Resume the training\n \n\n ### Watch your Agent play\n You can watch your agent playing directly in your browser\n\n 1. If the environment is part of ML-Agents official environments, go to URL\n 2. Step 1: Find your model_id: TaniaSF/ppo-Huggy\n 3. Step 2: Select your *.nn /*.onnx file\n 4. Click on Watch the agent play" ]

[ "TAGS\n#ml-agents #tensorboard #onnx #Huggy #deep-reinforcement-learning #reinforcement-learning #ML-Agents-Huggy #region-us \n", "# ppo Agent playing Huggy\n This is a trained model of a ppo agent playing Huggy\n using the Unity ML-Agents Library.\n\n ## Usage (with ML-Agents)\n The Documentation: URL\n\n We wrote a complete tutorial to learn to train your first agent using ML-Agents and publish it to the Hub:\n - A *short tutorial* where you teach Huggy the Dog to fetch the stick and then play with him directly in your\n browser: URL\n - A *longer tutorial* to understand how works ML-Agents:\n URL\n\n ### Resume the training\n \n\n ### Watch your Agent play\n You can watch your agent playing directly in your browser\n\n 1. If the environment is part of ML-Agents official environments, go to URL\n 2. Step 1: Find your model_id: TaniaSF/ppo-Huggy\n 3. Step 2: Select your *.nn /*.onnx file\n 4. Click on Watch the agent play" ]

[ "gguf", "sorah", "minichat", "es", "en", "region:us" ]

[]

[ "es", "en" ]

[ "# MiniChat-1.5-3B-Sorah\n!Sorah\n\nEste es el repositorio del modelo de IA MiniChat-1.5-3B-Sorah. Está basado en el modelo MiniChat-1.5-3B, que puedes encontrar aquí.\n\nEl modelo MiniChat-1.5-3B-Sorah complementa el modelo de voz Sorah, que fue creado por el proyecto Piper. El modelo Sorah está en proceso de entrenamiento y mejoras, por lo que su acceso no es público, pero puedes encontrar más información en el repositorio de Sorah Neuronal.\n\nEstos modelos en formato .gguf son la versión convertida del repositorio original.", "## Requisitos\n\nPara usar este modelo, necesitas tener instalado URL en tu equipo. Puedes obtener URL desde el siguiente repositorio:\n\n- Repositorio de URL\n\nPara instalar URL, sigue estos pasos:", "## Uso del modelo\n\nLa plantilla del modelo es la siguiente:\n\n\n\nPuedes utilizar el modelo en URL con el siguiente comando:\n\n\n\nLM Studio config-presets\n\nFilename:URL", "## Referencias\n\n- Repositorio original\n- Repositorio de la versión convertida\n- Repositorio de URL" ]

[ "TAGS\n#gguf #sorah #minichat #es #en #region-us \n", "# MiniChat-1.5-3B-Sorah\n!Sorah\n\nEste es el repositorio del modelo de IA MiniChat-1.5-3B-Sorah. Está basado en el modelo MiniChat-1.5-3B, que puedes encontrar aquí.\n\nEl modelo MiniChat-1.5-3B-Sorah complementa el modelo de voz Sorah, que fue creado por el proyecto Piper. El modelo Sorah está en proceso de entrenamiento y mejoras, por lo que su acceso no es público, pero puedes encontrar más información en el repositorio de Sorah Neuronal.\n\nEstos modelos en formato .gguf son la versión convertida del repositorio original.", "## Requisitos\n\nPara usar este modelo, necesitas tener instalado URL en tu equipo. Puedes obtener URL desde el siguiente repositorio:\n\n- Repositorio de URL\n\nPara instalar URL, sigue estos pasos:", "## Uso del modelo\n\nLa plantilla del modelo es la siguiente:\n\n\n\nPuedes utilizar el modelo en URL con el siguiente comando:\n\n\n\nLM Studio config-presets\n\nFilename:URL", "## Referencias\n\n- Repositorio original\n- Repositorio de la versión convertida\n- Repositorio de URL" ]

[ "transformers", "tensorboard", "safetensors", "gpt2", "text-generation", "generated_from_trainer", "dataset:eli5_category", "base_model:gpt2", "license:mit", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[]

[]

[ "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.05\n* train\\_batch\\_size: 8\n* eval\\_batch\\_size: 8\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* num\\_epochs: 0.5", "### Training results", "### Framework versions\n\n\n* Transformers 4.40.1\n* Pytorch 2.2.1+cu121\n* Datasets 2.19.0\n* Tokenizers 0.19.1" ]

[ "TAGS\n#transformers #tensorboard #safetensors #gpt2 #text-generation #generated_from_trainer #dataset-eli5_category #base_model-gpt2 #license-mit #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n", "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.05\n* train\\_batch\\_size: 8\n* eval\\_batch\\_size: 8\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* num\\_epochs: 0.5", "### Training results", "### Framework versions\n\n\n* Transformers 4.40.1\n* Pytorch 2.2.1+cu121\n* Datasets 2.19.0\n* Tokenizers 0.19.1" ]

[ "transformers", "safetensors", "gemma", "text-generation", "conversational", "arxiv:2312.11805", "arxiv:2009.03300", "arxiv:1905.07830", "arxiv:1911.11641", "arxiv:1904.09728", "arxiv:1905.10044", "arxiv:1907.10641", "arxiv:1811.00937", "arxiv:1809.02789", "arxiv:1911.01547", "arxiv:1705.03551", "arxiv:2107.03374", "arxiv:2108.07732", "arxiv:2110.14168", "arxiv:2304.06364", "arxiv:2206.04615", "arxiv:1804.06876", "arxiv:2110.08193", "arxiv:2009.11462", "arxiv:2101.11718", "arxiv:1804.09301", "arxiv:2109.07958", "arxiv:2203.09509", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "4-bit", "region:us" ]

[ "2312.11805", "2009.03300", "1905.07830", "1911.11641", "1904.09728", "1905.10044", "1907.10641", "1811.00937", "1809.02789", "1911.01547", "1705.03551", "2107.03374", "2108.07732", "2110.14168", "2304.06364", "2206.04615", "1804.06876", "2110.08193", "2009.11462", "2101.11718", "1804.09301", "2109.07958", "2203.09509" ]

[]

[ "### Description\n\n\nGemma is a family of lightweight, state-of-the-art open models from Google,\nbuilt from the same research and technology used to create the Gemini models.\nThey are text-to-text, decoder-only large language models, available in English,\nwith open weights, pre-trained variants, and instruction-tuned variants. Gemma\nmodels are well-suited for a variety of text generation tasks, including\nquestion answering, summarization, and reasoning. Their relatively small size\nmakes it possible to deploy them in environments with limited resources such as\na laptop, desktop or your own cloud infrastructure, democratizing access to\nstate of the art AI models and helping foster innovation for everyone.", "### Usage\n\n\nBelow we share some code snippets on how to get quickly started with running the model. First make sure to 'pip install -U transformers', then copy the snippet from the section that is relevant for your usecase.", "#### Running the model on a CPU", "#### Running the model on a single / multi GPU", "#### Running the model on a GPU using different precisions\n\n\n* *Using 'torch.float16'*\n* *Using 'torch.bfloat16'*", "#### Quantized Versions through 'bitsandbytes'\n\n\n* *Using 8-bit precision (int8)*\n* *Using 4-bit precision*", "#### Other optimizations\n\n\n* *Flash Attention 2*\n\n\nFirst make sure to install 'flash-attn' in your environment 'pip install flash-attn'", "### Chat Template\n\n\nThe instruction-tuned models use a chat template that must be adhered to for conversational use.\nThe easiest way to apply it is using the tokenizer's built-in chat template, as shown in the following snippet.\n\n\nLet's load the model and apply the chat template to a conversation. In this example, we'll start with a single user interaction:\n\n\nAt this point, the prompt contains the following text:\n\n\nAs you can see, each turn is preceeded by a '<start\\_of\\_turn>' delimiter and then the role of the entity\n(either 'user', for content supplied by the user, or 'model' for LLM responses). Turns finish with\nthe '<end\\_of\\_turn>' token.\n\n\nYou can follow this format to build the prompt manually, if you need to do it without the tokenizer's\nchat template.\n\n\nAfter the prompt is ready, generation can be performed like this:", "### Inputs and outputs\n\n\n* Input: Text string, such as a question, a prompt, or a document to be\nsummarized.\n* Output: Generated English-language text in response to the input, such\nas an answer to a question, or a summary of a document.\n\n\nModel Data\n----------\n\n\nData used for model training and how the data was processed.", "### Training Dataset\n\n\nThese models were trained on a dataset of text data that includes a wide variety\nof sources, totaling 6 trillion tokens. Here are the key components:\n\n\n* Web Documents: A diverse collection of web text ensures the model is exposed\nto a broad range of linguistic styles, topics, and vocabulary. Primarily\nEnglish-language content.\n* Code: Exposing the model to code helps it to learn the syntax and patterns of\nprogramming languages, which improves its ability to generate code or\nunderstand code-related questions.\n* Mathematics: Training on mathematical text helps the model learn logical\nreasoning, symbolic representation, and to address mathematical queries.\n\n\nThe combination of these diverse data sources is crucial for training a powerful\nlanguage model that can handle a wide variety of different tasks and text\nformats.", "### Data Preprocessing\n\n\nHere are the key data cleaning and filtering methods applied to the training\ndata:\n\n\n* CSAM Filtering: Rigorous CSAM (Child Sexual Abuse Material) filtering was\napplied at multiple stages in the data preparation process to ensure the\nexclusion of harmful and illegal content\n* Sensitive Data Filtering: As part of making Gemma pre-trained models safe and\nreliable, automated techniques were used to filter out certain personal\ninformation and other sensitive data from training sets.\n* Additional methods: Filtering based on content quality and safely in line with\nour policies.\n\n\nImplementation Information\n--------------------------\n\n\nDetails about the model internals.", "### Hardware\n\n\nGemma was trained using the latest generation of\nTensor Processing Unit (TPU) hardware (TPUv5e).\n\n\nTraining large language models requires significant computational power. TPUs,\ndesigned specifically for matrix operations common in machine learning, offer\nseveral advantages in this domain:\n\n\n* Performance: TPUs are specifically designed to handle the massive computations\ninvolved in training LLMs. They can speed up training considerably compared to\nCPUs.\n* Memory: TPUs often come with large amounts of high-bandwidth memory, allowing\nfor the handling of large models and batch sizes during training. This can\nlead to better model quality.\n* Scalability: TPU Pods (large clusters of TPUs) provide a scalable solution for\nhandling the growing complexity of large foundation models. You can distribute\ntraining across multiple TPU devices for faster and more efficient processing.\n* Cost-effectiveness: In many scenarios, TPUs can provide a more cost-effective\nsolution for training large models compared to CPU-based infrastructure,\nespecially when considering the time and resources saved due to faster\ntraining.\n* These advantages are aligned with\nGoogle's commitments to operate sustainably.", "### Software\n\n\nTraining was done using JAX and ML Pathways.\n\n\nJAX allows researchers to take advantage of the latest generation of hardware,\nincluding TPUs, for faster and more efficient training of large models.\n\n\nML Pathways is Google's latest effort to build artificially intelligent systems\ncapable of generalizing across multiple tasks. This is specially suitable for\nfoundation models, including large language models like\nthese ones.\n\n\nTogether, JAX and ML Pathways are used as described in the\npaper about the Gemini family of models; \"the 'single\ncontroller' programming model of Jax and Pathways allows a single Python\nprocess to orchestrate the entire training run, dramatically simplifying the\ndevelopment workflow.\"\n\n\nEvaluation\n----------\n\n\nModel evaluation metrics and results.", "### Benchmark Results\n\n\nThese models were evaluated against a large collection of different datasets and\nmetrics to cover different aspects of text generation:\n\n\n\nEthics and Safety\n-----------------\n\n\nEthics and safety evaluation approach and results.", "### Evaluation Approach\n\n\nOur evaluation methods include structured evaluations and internal red-teaming\ntesting of relevant content policies. Red-teaming was conducted by a number of\ndifferent teams, each with different goals and human evaluation metrics. These\nmodels were evaluated against a number of different categories relevant to\nethics and safety, including:\n\n\n* Text-to-Text Content Safety: Human evaluation on prompts covering safety\npolicies including child sexual abuse and exploitation, harassment, violence\nand gore, and hate speech.\n* Text-to-Text Representational Harms: Benchmark against relevant academic\ndatasets such as WinoBias and BBQ Dataset.\n* Memorization: Automated evaluation of memorization of training data, including\nthe risk of personally identifiable information exposure.\n* Large-scale harm: Tests for \"dangerous capabilities,\" such as chemical,\nbiological, radiological, and nuclear (CBRN) risks.", "### Evaluation Results\n\n\nThe results of ethics and safety evaluations are within acceptable thresholds\nfor meeting internal policies for categories such as child\nsafety, content safety, representational harms, memorization, large-scale harms.\nOn top of robust internal evaluations, the results of well known safety\nbenchmarks like BBQ, BOLD, Winogender, Winobias, RealToxicity, and TruthfulQA\nare shown here.\n\n\n\nUsage and Limitations\n---------------------\n\n\nThese models have certain limitations that users should be aware of.", "### Intended Usage\n\n\nOpen Large Language Models (LLMs) have a wide range of applications across\nvarious industries and domains. The following list of potential uses is not\ncomprehensive. The purpose of this list is to provide contextual information\nabout the possible use-cases that the model creators considered as part of model\ntraining and development.\n\n\n* Content Creation and Communication\n\t+ Text Generation: These models can be used to generate creative text formats\n\tsuch as poems, scripts, code, marketing copy, and email drafts.\n\t+ Chatbots and Conversational AI: Power conversational interfaces for customer\n\tservice, virtual assistants, or interactive applications.\n\t+ Text Summarization: Generate concise summaries of a text corpus, research\n\tpapers, or reports.\n* Research and Education\n\t+ Natural Language Processing (NLP) Research: These models can serve as a\n\tfoundation for researchers to experiment with NLP techniques, develop\n\talgorithms, and contribute to the advancement of the field.\n\t+ Language Learning Tools: Support interactive language learning experiences,\n\taiding in grammar correction or providing writing practice.\n\t+ Knowledge Exploration: Assist researchers in exploring large bodies of text\n\tby generating summaries or answering questions about specific topics.", "### Limitations\n\n\n* Training Data\n\t+ The quality and diversity of the training data significantly influence the\n\tmodel's capabilities. Biases or gaps in the training data can lead to\n\tlimitations in the model's responses.\n\t+ The scope of the training dataset determines the subject areas the model can\n\thandle effectively.\n* Context and Task Complexity\n\t+ LLMs are better at tasks that can be framed with clear prompts and\n\tinstructions. Open-ended or highly complex tasks might be challenging.\n\t+ A model's performance can be influenced by the amount of context provided\n\t(longer context generally leads to better outputs, up to a certain point).\n* Language Ambiguity and Nuance\n\t+ Natural language is inherently complex. LLMs might struggle to grasp subtle\n\tnuances, sarcasm, or figurative language.\n* Factual Accuracy\n\t+ LLMs generate responses based on information they learned from their\n\ttraining datasets, but they are not knowledge bases. They may generate\n\tincorrect or outdated factual statements.\n* Common Sense\n\t+ LLMs rely on statistical patterns in language. They might lack the ability\n\tto apply common sense reasoning in certain situations.", "### Ethical Considerations and Risks\n\n\nThe development of large language models (LLMs) raises several ethical concerns.\nIn creating an open model, we have carefully considered the following:\n\n\n* Bias and Fairness\n\t+ LLMs trained on large-scale, real-world text data can reflect socio-cultural\n\tbiases embedded in the training material. These models underwent careful\n\tscrutiny, input data pre-processing described and posterior evaluations\n\treported in this card.\n* Misinformation and Misuse\n\t+ LLMs can be misused to generate text that is false, misleading, or harmful.\n\t+ Guidelines are provided for responsible use with the model, see the\n\tResponsible Generative AI Toolkit.\n* Transparency and Accountability:\n\t+ This model card summarizes details on the models' architecture,\n\tcapabilities, limitations, and evaluation processes.\n\t+ A responsibly developed open model offers the opportunity to share\n\tinnovation by making LLM technology accessible to developers and researchers\n\tacross the AI ecosystem.\n\n\nRisks identified and mitigations:\n\n\n* Perpetuation of biases: It's encouraged to perform continuous monitoring\n(using evaluation metrics, human review) and the exploration of de-biasing\ntechniques during model training, fine-tuning, and other use cases.\n* Generation of harmful content: Mechanisms and guidelines for content safety\nare essential. Developers are encouraged to exercise caution and implement\nappropriate content safety safeguards based on their specific product policies\nand application use cases.\n* Misuse for malicious purposes: Technical limitations and developer and\nend-user education can help mitigate against malicious applications of LLMs.\nEducational resources and reporting mechanisms for users to flag misuse are\nprovided. Prohibited uses of Gemma models are outlined in the\nGemma Prohibited Use Policy.\n* Privacy violations: Models were trained on data filtered for removal of PII\n(Personally Identifiable Information). Developers are encouraged to adhere to\nprivacy regulations with privacy-preserving techniques.", "### Benefits\n\n\nAt the time of release, this family of models provides high-performance open\nlarge language model implementations designed from the ground up for Responsible\nAI development compared to similarly sized models.\n\n\nUsing the benchmark evaluation metrics described in this document, these models\nhave shown to provide superior performance to other, comparably-sized open model\nalternatives." ]

[ "TAGS\n#transformers #safetensors #gemma #text-generation #conversational #arxiv-2312.11805 #arxiv-2009.03300 #arxiv-1905.07830 #arxiv-1911.11641 #arxiv-1904.09728 #arxiv-1905.10044 #arxiv-1907.10641 #arxiv-1811.00937 #arxiv-1809.02789 #arxiv-1911.01547 #arxiv-1705.03551 #arxiv-2107.03374 #arxiv-2108.07732 #arxiv-2110.14168 #arxiv-2304.06364 #arxiv-2206.04615 #arxiv-1804.06876 #arxiv-2110.08193 #arxiv-2009.11462 #arxiv-2101.11718 #arxiv-1804.09301 #arxiv-2109.07958 #arxiv-2203.09509 #autotrain_compatible #endpoints_compatible #text-generation-inference #4-bit #region-us \n", "### Description\n\n\nGemma is a family of lightweight, state-of-the-art open models from Google,\nbuilt from the same research and technology used to create the Gemini models.\nThey are text-to-text, decoder-only large language models, available in English,\nwith open weights, pre-trained variants, and instruction-tuned variants. Gemma\nmodels are well-suited for a variety of text generation tasks, including\nquestion answering, summarization, and reasoning. Their relatively small size\nmakes it possible to deploy them in environments with limited resources such as\na laptop, desktop or your own cloud infrastructure, democratizing access to\nstate of the art AI models and helping foster innovation for everyone.", "### Usage\n\n\nBelow we share some code snippets on how to get quickly started with running the model. First make sure to 'pip install -U transformers', then copy the snippet from the section that is relevant for your usecase.", "#### Running the model on a CPU", "#### Running the model on a single / multi GPU", "#### Running the model on a GPU using different precisions\n\n\n* *Using 'torch.float16'*\n* *Using 'torch.bfloat16'*", "#### Quantized Versions through 'bitsandbytes'\n\n\n* *Using 8-bit precision (int8)*\n* *Using 4-bit precision*", "#### Other optimizations\n\n\n* *Flash Attention 2*\n\n\nFirst make sure to install 'flash-attn' in your environment 'pip install flash-attn'", "### Chat Template\n\n\nThe instruction-tuned models use a chat template that must be adhered to for conversational use.\nThe easiest way to apply it is using the tokenizer's built-in chat template, as shown in the following snippet.\n\n\nLet's load the model and apply the chat template to a conversation. In this example, we'll start with a single user interaction:\n\n\nAt this point, the prompt contains the following text:\n\n\nAs you can see, each turn is preceeded by a '<start\\_of\\_turn>' delimiter and then the role of the entity\n(either 'user', for content supplied by the user, or 'model' for LLM responses). Turns finish with\nthe '<end\\_of\\_turn>' token.\n\n\nYou can follow this format to build the prompt manually, if you need to do it without the tokenizer's\nchat template.\n\n\nAfter the prompt is ready, generation can be performed like this:", "### Inputs and outputs\n\n\n* Input: Text string, such as a question, a prompt, or a document to be\nsummarized.\n* Output: Generated English-language text in response to the input, such\nas an answer to a question, or a summary of a document.\n\n\nModel Data\n----------\n\n\nData used for model training and how the data was processed.", "### Training Dataset\n\n\nThese models were trained on a dataset of text data that includes a wide variety\nof sources, totaling 6 trillion tokens. Here are the key components:\n\n\n* Web Documents: A diverse collection of web text ensures the model is exposed\nto a broad range of linguistic styles, topics, and vocabulary. Primarily\nEnglish-language content.\n* Code: Exposing the model to code helps it to learn the syntax and patterns of\nprogramming languages, which improves its ability to generate code or\nunderstand code-related questions.\n* Mathematics: Training on mathematical text helps the model learn logical\nreasoning, symbolic representation, and to address mathematical queries.\n\n\nThe combination of these diverse data sources is crucial for training a powerful\nlanguage model that can handle a wide variety of different tasks and text\nformats.", "### Data Preprocessing\n\n\nHere are the key data cleaning and filtering methods applied to the training\ndata:\n\n\n* CSAM Filtering: Rigorous CSAM (Child Sexual Abuse Material) filtering was\napplied at multiple stages in the data preparation process to ensure the\nexclusion of harmful and illegal content\n* Sensitive Data Filtering: As part of making Gemma pre-trained models safe and\nreliable, automated techniques were used to filter out certain personal\ninformation and other sensitive data from training sets.\n* Additional methods: Filtering based on content quality and safely in line with\nour policies.\n\n\nImplementation Information\n--------------------------\n\n\nDetails about the model internals.", "### Hardware\n\n\nGemma was trained using the latest generation of\nTensor Processing Unit (TPU) hardware (TPUv5e).\n\n\nTraining large language models requires significant computational power. TPUs,\ndesigned specifically for matrix operations common in machine learning, offer\nseveral advantages in this domain:\n\n\n* Performance: TPUs are specifically designed to handle the massive computations\ninvolved in training LLMs. They can speed up training considerably compared to\nCPUs.\n* Memory: TPUs often come with large amounts of high-bandwidth memory, allowing\nfor the handling of large models and batch sizes during training. This can\nlead to better model quality.\n* Scalability: TPU Pods (large clusters of TPUs) provide a scalable solution for\nhandling the growing complexity of large foundation models. You can distribute\ntraining across multiple TPU devices for faster and more efficient processing.\n* Cost-effectiveness: In many scenarios, TPUs can provide a more cost-effective\nsolution for training large models compared to CPU-based infrastructure,\nespecially when considering the time and resources saved due to faster\ntraining.\n* These advantages are aligned with\nGoogle's commitments to operate sustainably.", "### Software\n\n\nTraining was done using JAX and ML Pathways.\n\n\nJAX allows researchers to take advantage of the latest generation of hardware,\nincluding TPUs, for faster and more efficient training of large models.\n\n\nML Pathways is Google's latest effort to build artificially intelligent systems\ncapable of generalizing across multiple tasks. This is specially suitable for\nfoundation models, including large language models like\nthese ones.\n\n\nTogether, JAX and ML Pathways are used as described in the\npaper about the Gemini family of models; \"the 'single\ncontroller' programming model of Jax and Pathways allows a single Python\nprocess to orchestrate the entire training run, dramatically simplifying the\ndevelopment workflow.\"\n\n\nEvaluation\n----------\n\n\nModel evaluation metrics and results.", "### Benchmark Results\n\n\nThese models were evaluated against a large collection of different datasets and\nmetrics to cover different aspects of text generation:\n\n\n\nEthics and Safety\n-----------------\n\n\nEthics and safety evaluation approach and results.", "### Evaluation Approach\n\n\nOur evaluation methods include structured evaluations and internal red-teaming\ntesting of relevant content policies. Red-teaming was conducted by a number of\ndifferent teams, each with different goals and human evaluation metrics. These\nmodels were evaluated against a number of different categories relevant to\nethics and safety, including:\n\n\n* Text-to-Text Content Safety: Human evaluation on prompts covering safety\npolicies including child sexual abuse and exploitation, harassment, violence\nand gore, and hate speech.\n* Text-to-Text Representational Harms: Benchmark against relevant academic\ndatasets such as WinoBias and BBQ Dataset.\n* Memorization: Automated evaluation of memorization of training data, including\nthe risk of personally identifiable information exposure.\n* Large-scale harm: Tests for \"dangerous capabilities,\" such as chemical,\nbiological, radiological, and nuclear (CBRN) risks.", "### Evaluation Results\n\n\nThe results of ethics and safety evaluations are within acceptable thresholds\nfor meeting internal policies for categories such as child\nsafety, content safety, representational harms, memorization, large-scale harms.\nOn top of robust internal evaluations, the results of well known safety\nbenchmarks like BBQ, BOLD, Winogender, Winobias, RealToxicity, and TruthfulQA\nare shown here.\n\n\n\nUsage and Limitations\n---------------------\n\n\nThese models have certain limitations that users should be aware of.", "### Intended Usage\n\n\nOpen Large Language Models (LLMs) have a wide range of applications across\nvarious industries and domains. The following list of potential uses is not\ncomprehensive. The purpose of this list is to provide contextual information\nabout the possible use-cases that the model creators considered as part of model\ntraining and development.\n\n\n* Content Creation and Communication\n\t+ Text Generation: These models can be used to generate creative text formats\n\tsuch as poems, scripts, code, marketing copy, and email drafts.\n\t+ Chatbots and Conversational AI: Power conversational interfaces for customer\n\tservice, virtual assistants, or interactive applications.\n\t+ Text Summarization: Generate concise summaries of a text corpus, research\n\tpapers, or reports.\n* Research and Education\n\t+ Natural Language Processing (NLP) Research: These models can serve as a\n\tfoundation for researchers to experiment with NLP techniques, develop\n\talgorithms, and contribute to the advancement of the field.\n\t+ Language Learning Tools: Support interactive language learning experiences,\n\taiding in grammar correction or providing writing practice.\n\t+ Knowledge Exploration: Assist researchers in exploring large bodies of text\n\tby generating summaries or answering questions about specific topics.", "### Limitations\n\n\n* Training Data\n\t+ The quality and diversity of the training data significantly influence the\n\tmodel's capabilities. Biases or gaps in the training data can lead to\n\tlimitations in the model's responses.\n\t+ The scope of the training dataset determines the subject areas the model can\n\thandle effectively.\n* Context and Task Complexity\n\t+ LLMs are better at tasks that can be framed with clear prompts and\n\tinstructions. Open-ended or highly complex tasks might be challenging.\n\t+ A model's performance can be influenced by the amount of context provided\n\t(longer context generally leads to better outputs, up to a certain point).\n* Language Ambiguity and Nuance\n\t+ Natural language is inherently complex. LLMs might struggle to grasp subtle\n\tnuances, sarcasm, or figurative language.\n* Factual Accuracy\n\t+ LLMs generate responses based on information they learned from their\n\ttraining datasets, but they are not knowledge bases. They may generate\n\tincorrect or outdated factual statements.\n* Common Sense\n\t+ LLMs rely on statistical patterns in language. They might lack the ability\n\tto apply common sense reasoning in certain situations.", "### Ethical Considerations and Risks\n\n\nThe development of large language models (LLMs) raises several ethical concerns.\nIn creating an open model, we have carefully considered the following:\n\n\n* Bias and Fairness\n\t+ LLMs trained on large-scale, real-world text data can reflect socio-cultural\n\tbiases embedded in the training material. These models underwent careful\n\tscrutiny, input data pre-processing described and posterior evaluations\n\treported in this card.\n* Misinformation and Misuse\n\t+ LLMs can be misused to generate text that is false, misleading, or harmful.\n\t+ Guidelines are provided for responsible use with the model, see the\n\tResponsible Generative AI Toolkit.\n* Transparency and Accountability:\n\t+ This model card summarizes details on the models' architecture,\n\tcapabilities, limitations, and evaluation processes.\n\t+ A responsibly developed open model offers the opportunity to share\n\tinnovation by making LLM technology accessible to developers and researchers\n\tacross the AI ecosystem.\n\n\nRisks identified and mitigations:\n\n\n* Perpetuation of biases: It's encouraged to perform continuous monitoring\n(using evaluation metrics, human review) and the exploration of de-biasing\ntechniques during model training, fine-tuning, and other use cases.\n* Generation of harmful content: Mechanisms and guidelines for content safety\nare essential. Developers are encouraged to exercise caution and implement\nappropriate content safety safeguards based on their specific product policies\nand application use cases.\n* Misuse for malicious purposes: Technical limitations and developer and\nend-user education can help mitigate against malicious applications of LLMs.\nEducational resources and reporting mechanisms for users to flag misuse are\nprovided. Prohibited uses of Gemma models are outlined in the\nGemma Prohibited Use Policy.\n* Privacy violations: Models were trained on data filtered for removal of PII\n(Personally Identifiable Information). Developers are encouraged to adhere to\nprivacy regulations with privacy-preserving techniques.", "### Benefits\n\n\nAt the time of release, this family of models provides high-performance open\nlarge language model implementations designed from the ground up for Responsible\nAI development compared to similarly sized models.\n\n\nUsing the benchmark evaluation metrics described in this document, these models\nhave shown to provide superior performance to other, comparably-sized open model\nalternatives." ]

[ "transformers", "safetensors", "gemma", "text-generation", "conversational", "arxiv:2312.11805", "arxiv:2009.03300", "arxiv:1905.07830", "arxiv:1911.11641", "arxiv:1904.09728", "arxiv:1905.10044", "arxiv:1907.10641", "arxiv:1811.00937", "arxiv:1809.02789", "arxiv:1911.01547", "arxiv:1705.03551", "arxiv:2107.03374", "arxiv:2108.07732", "arxiv:2110.14168", "arxiv:2304.06364", "arxiv:2206.04615", "arxiv:1804.06876", "arxiv:2110.08193", "arxiv:2009.11462", "arxiv:2101.11718", "arxiv:1804.09301", "arxiv:2109.07958", "arxiv:2203.09509", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "8-bit", "region:us" ]

[ "2312.11805", "2009.03300", "1905.07830", "1911.11641", "1904.09728", "1905.10044", "1907.10641", "1811.00937", "1809.02789", "1911.01547", "1705.03551", "2107.03374", "2108.07732", "2110.14168", "2304.06364", "2206.04615", "1804.06876", "2110.08193", "2009.11462", "2101.11718", "1804.09301", "2109.07958", "2203.09509" ]

[]

[ "### Description\n\n\nGemma is a family of lightweight, state-of-the-art open models from Google,\nbuilt from the same research and technology used to create the Gemini models.\nThey are text-to-text, decoder-only large language models, available in English,\nwith open weights, pre-trained variants, and instruction-tuned variants. Gemma\nmodels are well-suited for a variety of text generation tasks, including\nquestion answering, summarization, and reasoning. Their relatively small size\nmakes it possible to deploy them in environments with limited resources such as\na laptop, desktop or your own cloud infrastructure, democratizing access to\nstate of the art AI models and helping foster innovation for everyone.", "### Usage\n\n\nBelow we share some code snippets on how to get quickly started with running the model. First make sure to 'pip install -U transformers', then copy the snippet from the section that is relevant for your usecase.", "#### Running the model on a CPU", "#### Running the model on a single / multi GPU", "#### Running the model on a GPU using different precisions\n\n\n* *Using 'torch.float16'*\n* *Using 'torch.bfloat16'*", "#### Quantized Versions through 'bitsandbytes'\n\n\n* *Using 8-bit precision (int8)*\n* *Using 4-bit precision*", "#### Other optimizations\n\n\n* *Flash Attention 2*\n\n\nFirst make sure to install 'flash-attn' in your environment 'pip install flash-attn'", "### Chat Template\n\n\nThe instruction-tuned models use a chat template that must be adhered to for conversational use.\nThe easiest way to apply it is using the tokenizer's built-in chat template, as shown in the following snippet.\n\n\nLet's load the model and apply the chat template to a conversation. In this example, we'll start with a single user interaction:\n\n\nAt this point, the prompt contains the following text:\n\n\nAs you can see, each turn is preceeded by a '<start\\_of\\_turn>' delimiter and then the role of the entity\n(either 'user', for content supplied by the user, or 'model' for LLM responses). Turns finish with\nthe '<end\\_of\\_turn>' token.\n\n\nYou can follow this format to build the prompt manually, if you need to do it without the tokenizer's\nchat template.\n\n\nAfter the prompt is ready, generation can be performed like this:", "### Inputs and outputs\n\n\n* Input: Text string, such as a question, a prompt, or a document to be\nsummarized.\n* Output: Generated English-language text in response to the input, such\nas an answer to a question, or a summary of a document.\n\n\nModel Data\n----------\n\n\nData used for model training and how the data was processed.", "### Training Dataset\n\n\nThese models were trained on a dataset of text data that includes a wide variety\nof sources, totaling 6 trillion tokens. Here are the key components:\n\n\n* Web Documents: A diverse collection of web text ensures the model is exposed\nto a broad range of linguistic styles, topics, and vocabulary. Primarily\nEnglish-language content.\n* Code: Exposing the model to code helps it to learn the syntax and patterns of\nprogramming languages, which improves its ability to generate code or\nunderstand code-related questions.\n* Mathematics: Training on mathematical text helps the model learn logical\nreasoning, symbolic representation, and to address mathematical queries.\n\n\nThe combination of these diverse data sources is crucial for training a powerful\nlanguage model that can handle a wide variety of different tasks and text\nformats.", "### Data Preprocessing\n\n\nHere are the key data cleaning and filtering methods applied to the training\ndata:\n\n\n* CSAM Filtering: Rigorous CSAM (Child Sexual Abuse Material) filtering was\napplied at multiple stages in the data preparation process to ensure the\nexclusion of harmful and illegal content\n* Sensitive Data Filtering: As part of making Gemma pre-trained models safe and\nreliable, automated techniques were used to filter out certain personal\ninformation and other sensitive data from training sets.\n* Additional methods: Filtering based on content quality and safely in line with\nour policies.\n\n\nImplementation Information\n--------------------------\n\n\nDetails about the model internals.", "### Hardware\n\n\nGemma was trained using the latest generation of\nTensor Processing Unit (TPU) hardware (TPUv5e).\n\n\nTraining large language models requires significant computational power. TPUs,\ndesigned specifically for matrix operations common in machine learning, offer\nseveral advantages in this domain:\n\n\n* Performance: TPUs are specifically designed to handle the massive computations\ninvolved in training LLMs. They can speed up training considerably compared to\nCPUs.\n* Memory: TPUs often come with large amounts of high-bandwidth memory, allowing\nfor the handling of large models and batch sizes during training. This can\nlead to better model quality.\n* Scalability: TPU Pods (large clusters of TPUs) provide a scalable solution for\nhandling the growing complexity of large foundation models. You can distribute\ntraining across multiple TPU devices for faster and more efficient processing.\n* Cost-effectiveness: In many scenarios, TPUs can provide a more cost-effective\nsolution for training large models compared to CPU-based infrastructure,\nespecially when considering the time and resources saved due to faster\ntraining.\n* These advantages are aligned with\nGoogle's commitments to operate sustainably.", "### Software\n\n\nTraining was done using JAX and ML Pathways.\n\n\nJAX allows researchers to take advantage of the latest generation of hardware,\nincluding TPUs, for faster and more efficient training of large models.\n\n\nML Pathways is Google's latest effort to build artificially intelligent systems\ncapable of generalizing across multiple tasks. This is specially suitable for\nfoundation models, including large language models like\nthese ones.\n\n\nTogether, JAX and ML Pathways are used as described in the\npaper about the Gemini family of models; \"the 'single\ncontroller' programming model of Jax and Pathways allows a single Python\nprocess to orchestrate the entire training run, dramatically simplifying the\ndevelopment workflow.\"\n\n\nEvaluation\n----------\n\n\nModel evaluation metrics and results.", "### Benchmark Results\n\n\nThese models were evaluated against a large collection of different datasets and\nmetrics to cover different aspects of text generation:\n\n\n\nEthics and Safety\n-----------------\n\n\nEthics and safety evaluation approach and results.", "### Evaluation Approach\n\n\nOur evaluation methods include structured evaluations and internal red-teaming\ntesting of relevant content policies. Red-teaming was conducted by a number of\ndifferent teams, each with different goals and human evaluation metrics. These\nmodels were evaluated against a number of different categories relevant to\nethics and safety, including:\n\n\n* Text-to-Text Content Safety: Human evaluation on prompts covering safety\npolicies including child sexual abuse and exploitation, harassment, violence\nand gore, and hate speech.\n* Text-to-Text Representational Harms: Benchmark against relevant academic\ndatasets such as WinoBias and BBQ Dataset.\n* Memorization: Automated evaluation of memorization of training data, including\nthe risk of personally identifiable information exposure.\n* Large-scale harm: Tests for \"dangerous capabilities,\" such as chemical,\nbiological, radiological, and nuclear (CBRN) risks.", "### Evaluation Results\n\n\nThe results of ethics and safety evaluations are within acceptable thresholds\nfor meeting internal policies for categories such as child\nsafety, content safety, representational harms, memorization, large-scale harms.\nOn top of robust internal evaluations, the results of well known safety\nbenchmarks like BBQ, BOLD, Winogender, Winobias, RealToxicity, and TruthfulQA\nare shown here.\n\n\n\nUsage and Limitations\n---------------------\n\n\nThese models have certain limitations that users should be aware of.", "### Intended Usage\n\n\nOpen Large Language Models (LLMs) have a wide range of applications across\nvarious industries and domains. The following list of potential uses is not\ncomprehensive. The purpose of this list is to provide contextual information\nabout the possible use-cases that the model creators considered as part of model\ntraining and development.\n\n\n* Content Creation and Communication\n\t+ Text Generation: These models can be used to generate creative text formats\n\tsuch as poems, scripts, code, marketing copy, and email drafts.\n\t+ Chatbots and Conversational AI: Power conversational interfaces for customer\n\tservice, virtual assistants, or interactive applications.\n\t+ Text Summarization: Generate concise summaries of a text corpus, research\n\tpapers, or reports.\n* Research and Education\n\t+ Natural Language Processing (NLP) Research: These models can serve as a\n\tfoundation for researchers to experiment with NLP techniques, develop\n\talgorithms, and contribute to the advancement of the field.\n\t+ Language Learning Tools: Support interactive language learning experiences,\n\taiding in grammar correction or providing writing practice.\n\t+ Knowledge Exploration: Assist researchers in exploring large bodies of text\n\tby generating summaries or answering questions about specific topics.", "### Limitations\n\n\n* Training Data\n\t+ The quality and diversity of the training data significantly influence the\n\tmodel's capabilities. Biases or gaps in the training data can lead to\n\tlimitations in the model's responses.\n\t+ The scope of the training dataset determines the subject areas the model can\n\thandle effectively.\n* Context and Task Complexity\n\t+ LLMs are better at tasks that can be framed with clear prompts and\n\tinstructions. Open-ended or highly complex tasks might be challenging.\n\t+ A model's performance can be influenced by the amount of context provided\n\t(longer context generally leads to better outputs, up to a certain point).\n* Language Ambiguity and Nuance\n\t+ Natural language is inherently complex. LLMs might struggle to grasp subtle\n\tnuances, sarcasm, or figurative language.\n* Factual Accuracy\n\t+ LLMs generate responses based on information they learned from their\n\ttraining datasets, but they are not knowledge bases. They may generate\n\tincorrect or outdated factual statements.\n* Common Sense\n\t+ LLMs rely on statistical patterns in language. They might lack the ability\n\tto apply common sense reasoning in certain situations.", "### Ethical Considerations and Risks\n\n\nThe development of large language models (LLMs) raises several ethical concerns.\nIn creating an open model, we have carefully considered the following:\n\n\n* Bias and Fairness\n\t+ LLMs trained on large-scale, real-world text data can reflect socio-cultural\n\tbiases embedded in the training material. These models underwent careful\n\tscrutiny, input data pre-processing described and posterior evaluations\n\treported in this card.\n* Misinformation and Misuse\n\t+ LLMs can be misused to generate text that is false, misleading, or harmful.\n\t+ Guidelines are provided for responsible use with the model, see the\n\tResponsible Generative AI Toolkit.\n* Transparency and Accountability:\n\t+ This model card summarizes details on the models' architecture,\n\tcapabilities, limitations, and evaluation processes.\n\t+ A responsibly developed open model offers the opportunity to share\n\tinnovation by making LLM technology accessible to developers and researchers\n\tacross the AI ecosystem.\n\n\nRisks identified and mitigations:\n\n\n* Perpetuation of biases: It's encouraged to perform continuous monitoring\n(using evaluation metrics, human review) and the exploration of de-biasing\ntechniques during model training, fine-tuning, and other use cases.\n* Generation of harmful content: Mechanisms and guidelines for content safety\nare essential. Developers are encouraged to exercise caution and implement\nappropriate content safety safeguards based on their specific product policies\nand application use cases.\n* Misuse for malicious purposes: Technical limitations and developer and\nend-user education can help mitigate against malicious applications of LLMs.\nEducational resources and reporting mechanisms for users to flag misuse are\nprovided. Prohibited uses of Gemma models are outlined in the\nGemma Prohibited Use Policy.\n* Privacy violations: Models were trained on data filtered for removal of PII\n(Personally Identifiable Information). Developers are encouraged to adhere to\nprivacy regulations with privacy-preserving techniques.", "### Benefits\n\n\nAt the time of release, this family of models provides high-performance open\nlarge language model implementations designed from the ground up for Responsible\nAI development compared to similarly sized models.\n\n\nUsing the benchmark evaluation metrics described in this document, these models\nhave shown to provide superior performance to other, comparably-sized open model\nalternatives." ]

[ "TAGS\n#transformers #safetensors #gemma #text-generation #conversational #arxiv-2312.11805 #arxiv-2009.03300 #arxiv-1905.07830 #arxiv-1911.11641 #arxiv-1904.09728 #arxiv-1905.10044 #arxiv-1907.10641 #arxiv-1811.00937 #arxiv-1809.02789 #arxiv-1911.01547 #arxiv-1705.03551 #arxiv-2107.03374 #arxiv-2108.07732 #arxiv-2110.14168 #arxiv-2304.06364 #arxiv-2206.04615 #arxiv-1804.06876 #arxiv-2110.08193 #arxiv-2009.11462 #arxiv-2101.11718 #arxiv-1804.09301 #arxiv-2109.07958 #arxiv-2203.09509 #autotrain_compatible #endpoints_compatible #text-generation-inference #8-bit #region-us \n", "### Description\n\n\nGemma is a family of lightweight, state-of-the-art open models from Google,\nbuilt from the same research and technology used to create the Gemini models.\nThey are text-to-text, decoder-only large language models, available in English,\nwith open weights, pre-trained variants, and instruction-tuned variants. Gemma\nmodels are well-suited for a variety of text generation tasks, including\nquestion answering, summarization, and reasoning. Their relatively small size\nmakes it possible to deploy them in environments with limited resources such as\na laptop, desktop or your own cloud infrastructure, democratizing access to\nstate of the art AI models and helping foster innovation for everyone.", "### Usage\n\n\nBelow we share some code snippets on how to get quickly started with running the model. First make sure to 'pip install -U transformers', then copy the snippet from the section that is relevant for your usecase.", "#### Running the model on a CPU", "#### Running the model on a single / multi GPU", "#### Running the model on a GPU using different precisions\n\n\n* *Using 'torch.float16'*\n* *Using 'torch.bfloat16'*", "#### Quantized Versions through 'bitsandbytes'\n\n\n* *Using 8-bit precision (int8)*\n* *Using 4-bit precision*", "#### Other optimizations\n\n\n* *Flash Attention 2*\n\n\nFirst make sure to install 'flash-attn' in your environment 'pip install flash-attn'", "### Chat Template\n\n\nThe instruction-tuned models use a chat template that must be adhered to for conversational use.\nThe easiest way to apply it is using the tokenizer's built-in chat template, as shown in the following snippet.\n\n\nLet's load the model and apply the chat template to a conversation. In this example, we'll start with a single user interaction:\n\n\nAt this point, the prompt contains the following text:\n\n\nAs you can see, each turn is preceeded by a '<start\\_of\\_turn>' delimiter and then the role of the entity\n(either 'user', for content supplied by the user, or 'model' for LLM responses). Turns finish with\nthe '<end\\_of\\_turn>' token.\n\n\nYou can follow this format to build the prompt manually, if you need to do it without the tokenizer's\nchat template.\n\n\nAfter the prompt is ready, generation can be performed like this:", "### Inputs and outputs\n\n\n* Input: Text string, such as a question, a prompt, or a document to be\nsummarized.\n* Output: Generated English-language text in response to the input, such\nas an answer to a question, or a summary of a document.\n\n\nModel Data\n----------\n\n\nData used for model training and how the data was processed.", "### Training Dataset\n\n\nThese models were trained on a dataset of text data that includes a wide variety\nof sources, totaling 6 trillion tokens. Here are the key components:\n\n\n* Web Documents: A diverse collection of web text ensures the model is exposed\nto a broad range of linguistic styles, topics, and vocabulary. Primarily\nEnglish-language content.\n* Code: Exposing the model to code helps it to learn the syntax and patterns of\nprogramming languages, which improves its ability to generate code or\nunderstand code-related questions.\n* Mathematics: Training on mathematical text helps the model learn logical\nreasoning, symbolic representation, and to address mathematical queries.\n\n\nThe combination of these diverse data sources is crucial for training a powerful\nlanguage model that can handle a wide variety of different tasks and text\nformats.", "### Data Preprocessing\n\n\nHere are the key data cleaning and filtering methods applied to the training\ndata:\n\n\n* CSAM Filtering: Rigorous CSAM (Child Sexual Abuse Material) filtering was\napplied at multiple stages in the data preparation process to ensure the\nexclusion of harmful and illegal content\n* Sensitive Data Filtering: As part of making Gemma pre-trained models safe and\nreliable, automated techniques were used to filter out certain personal\ninformation and other sensitive data from training sets.\n* Additional methods: Filtering based on content quality and safely in line with\nour policies.\n\n\nImplementation Information\n--------------------------\n\n\nDetails about the model internals.", "### Hardware\n\n\nGemma was trained using the latest generation of\nTensor Processing Unit (TPU) hardware (TPUv5e).\n\n\nTraining large language models requires significant computational power. TPUs,\ndesigned specifically for matrix operations common in machine learning, offer\nseveral advantages in this domain:\n\n\n* Performance: TPUs are specifically designed to handle the massive computations\ninvolved in training LLMs. They can speed up training considerably compared to\nCPUs.\n* Memory: TPUs often come with large amounts of high-bandwidth memory, allowing\nfor the handling of large models and batch sizes during training. This can\nlead to better model quality.\n* Scalability: TPU Pods (large clusters of TPUs) provide a scalable solution for\nhandling the growing complexity of large foundation models. You can distribute\ntraining across multiple TPU devices for faster and more efficient processing.\n* Cost-effectiveness: In many scenarios, TPUs can provide a more cost-effective\nsolution for training large models compared to CPU-based infrastructure,\nespecially when considering the time and resources saved due to faster\ntraining.\n* These advantages are aligned with\nGoogle's commitments to operate sustainably.", "### Software\n\n\nTraining was done using JAX and ML Pathways.\n\n\nJAX allows researchers to take advantage of the latest generation of hardware,\nincluding TPUs, for faster and more efficient training of large models.\n\n\nML Pathways is Google's latest effort to build artificially intelligent systems\ncapable of generalizing across multiple tasks. This is specially suitable for\nfoundation models, including large language models like\nthese ones.\n\n\nTogether, JAX and ML Pathways are used as described in the\npaper about the Gemini family of models; \"the 'single\ncontroller' programming model of Jax and Pathways allows a single Python\nprocess to orchestrate the entire training run, dramatically simplifying the\ndevelopment workflow.\"\n\n\nEvaluation\n----------\n\n\nModel evaluation metrics and results.", "### Benchmark Results\n\n\nThese models were evaluated against a large collection of different datasets and\nmetrics to cover different aspects of text generation:\n\n\n\nEthics and Safety\n-----------------\n\n\nEthics and safety evaluation approach and results.", "### Evaluation Approach\n\n\nOur evaluation methods include structured evaluations and internal red-teaming\ntesting of relevant content policies. Red-teaming was conducted by a number of\ndifferent teams, each with different goals and human evaluation metrics. These\nmodels were evaluated against a number of different categories relevant to\nethics and safety, including:\n\n\n* Text-to-Text Content Safety: Human evaluation on prompts covering safety\npolicies including child sexual abuse and exploitation, harassment, violence\nand gore, and hate speech.\n* Text-to-Text Representational Harms: Benchmark against relevant academic\ndatasets such as WinoBias and BBQ Dataset.\n* Memorization: Automated evaluation of memorization of training data, including\nthe risk of personally identifiable information exposure.\n* Large-scale harm: Tests for \"dangerous capabilities,\" such as chemical,\nbiological, radiological, and nuclear (CBRN) risks.", "### Evaluation Results\n\n\nThe results of ethics and safety evaluations are within acceptable thresholds\nfor meeting internal policies for categories such as child\nsafety, content safety, representational harms, memorization, large-scale harms.\nOn top of robust internal evaluations, the results of well known safety\nbenchmarks like BBQ, BOLD, Winogender, Winobias, RealToxicity, and TruthfulQA\nare shown here.\n\n\n\nUsage and Limitations\n---------------------\n\n\nThese models have certain limitations that users should be aware of.", "### Intended Usage\n\n\nOpen Large Language Models (LLMs) have a wide range of applications across\nvarious industries and domains. The following list of potential uses is not\ncomprehensive. The purpose of this list is to provide contextual information\nabout the possible use-cases that the model creators considered as part of model\ntraining and development.\n\n\n* Content Creation and Communication\n\t+ Text Generation: These models can be used to generate creative text formats\n\tsuch as poems, scripts, code, marketing copy, and email drafts.\n\t+ Chatbots and Conversational AI: Power conversational interfaces for customer\n\tservice, virtual assistants, or interactive applications.\n\t+ Text Summarization: Generate concise summaries of a text corpus, research\n\tpapers, or reports.\n* Research and Education\n\t+ Natural Language Processing (NLP) Research: These models can serve as a\n\tfoundation for researchers to experiment with NLP techniques, develop\n\talgorithms, and contribute to the advancement of the field.\n\t+ Language Learning Tools: Support interactive language learning experiences,\n\taiding in grammar correction or providing writing practice.\n\t+ Knowledge Exploration: Assist researchers in exploring large bodies of text\n\tby generating summaries or answering questions about specific topics.", "### Limitations\n\n\n* Training Data\n\t+ The quality and diversity of the training data significantly influence the\n\tmodel's capabilities. Biases or gaps in the training data can lead to\n\tlimitations in the model's responses.\n\t+ The scope of the training dataset determines the subject areas the model can\n\thandle effectively.\n* Context and Task Complexity\n\t+ LLMs are better at tasks that can be framed with clear prompts and\n\tinstructions. Open-ended or highly complex tasks might be challenging.\n\t+ A model's performance can be influenced by the amount of context provided\n\t(longer context generally leads to better outputs, up to a certain point).\n* Language Ambiguity and Nuance\n\t+ Natural language is inherently complex. LLMs might struggle to grasp subtle\n\tnuances, sarcasm, or figurative language.\n* Factual Accuracy\n\t+ LLMs generate responses based on information they learned from their\n\ttraining datasets, but they are not knowledge bases. They may generate\n\tincorrect or outdated factual statements.\n* Common Sense\n\t+ LLMs rely on statistical patterns in language. They might lack the ability\n\tto apply common sense reasoning in certain situations.", "### Ethical Considerations and Risks\n\n\nThe development of large language models (LLMs) raises several ethical concerns.\nIn creating an open model, we have carefully considered the following:\n\n\n* Bias and Fairness\n\t+ LLMs trained on large-scale, real-world text data can reflect socio-cultural\n\tbiases embedded in the training material. These models underwent careful\n\tscrutiny, input data pre-processing described and posterior evaluations\n\treported in this card.\n* Misinformation and Misuse\n\t+ LLMs can be misused to generate text that is false, misleading, or harmful.\n\t+ Guidelines are provided for responsible use with the model, see the\n\tResponsible Generative AI Toolkit.\n* Transparency and Accountability:\n\t+ This model card summarizes details on the models' architecture,\n\tcapabilities, limitations, and evaluation processes.\n\t+ A responsibly developed open model offers the opportunity to share\n\tinnovation by making LLM technology accessible to developers and researchers\n\tacross the AI ecosystem.\n\n\nRisks identified and mitigations:\n\n\n* Perpetuation of biases: It's encouraged to perform continuous monitoring\n(using evaluation metrics, human review) and the exploration of de-biasing\ntechniques during model training, fine-tuning, and other use cases.\n* Generation of harmful content: Mechanisms and guidelines for content safety\nare essential. Developers are encouraged to exercise caution and implement\nappropriate content safety safeguards based on their specific product policies\nand application use cases.\n* Misuse for malicious purposes: Technical limitations and developer and\nend-user education can help mitigate against malicious applications of LLMs.\nEducational resources and reporting mechanisms for users to flag misuse are\nprovided. Prohibited uses of Gemma models are outlined in the\nGemma Prohibited Use Policy.\n* Privacy violations: Models were trained on data filtered for removal of PII\n(Personally Identifiable Information). Developers are encouraged to adhere to\nprivacy regulations with privacy-preserving techniques.", "### Benefits\n\n\nAt the time of release, this family of models provides high-performance open\nlarge language model implementations designed from the ground up for Responsible\nAI development compared to similarly sized models.\n\n\nUsing the benchmark evaluation metrics described in this document, these models\nhave shown to provide superior performance to other, comparably-sized open model\nalternatives." ]

[ "transformers", "tensorboard", "safetensors", "vit", "image-classification", "generated_from_trainer", "autotrain_compatible", "endpoints_compatible", "region:us" ]

[]

[]

[ "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 5e-05\n* train\\_batch\\_size: 32\n* eval\\_batch\\_size: 32\n* seed: 42\n* gradient\\_accumulation\\_steps: 4\n* total\\_train\\_batch\\_size: 128\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* lr\\_scheduler\\_warmup\\_ratio: 0.05\n* num\\_epochs: 10", "### Training results", "### Framework versions\n\n\n* Transformers 4.36.2\n* Pytorch 2.1.2+cu118\n* Datasets 2.16.1\n* Tokenizers 0.15.0" ]

[ "TAGS\n#transformers #tensorboard #safetensors #vit #image-classification #generated_from_trainer #autotrain_compatible #endpoints_compatible #region-us \n", "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 5e-05\n* train\\_batch\\_size: 32\n* eval\\_batch\\_size: 32\n* seed: 42\n* gradient\\_accumulation\\_steps: 4\n* total\\_train\\_batch\\_size: 128\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* lr\\_scheduler\\_warmup\\_ratio: 0.05\n* num\\_epochs: 10", "### Training results", "### Framework versions\n\n\n* Transformers 4.36.2\n* Pytorch 2.1.2+cu118\n* Datasets 2.16.1\n* Tokenizers 0.15.0" ]

[ "transformers", "safetensors", "pix2struct", "text2text-generation", "arxiv:1910.09700", "autotrain_compatible", "endpoints_compatible", "region:us" ]

[ "1910.09700" ]

[]

[ "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "TAGS\n#transformers #safetensors #pix2struct #text2text-generation #arxiv-1910.09700 #autotrain_compatible #endpoints_compatible #region-us \n", "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "gguf", "arxiv:2312.11805", "arxiv:2009.03300", "arxiv:1905.07830", "arxiv:1911.11641", "arxiv:1904.09728", "arxiv:1905.10044", "arxiv:1907.10641", "arxiv:1811.00937", "arxiv:1809.02789", "arxiv:1911.01547", "arxiv:1705.03551", "arxiv:2107.03374", "arxiv:2108.07732", "arxiv:2110.14168", "arxiv:2304.06364", "arxiv:2206.04615", "arxiv:1804.06876", "arxiv:2110.08193", "arxiv:2009.11462", "arxiv:2101.11718", "arxiv:1804.09301", "arxiv:2109.07958", "arxiv:2203.09509", "region:us" ]

[ "2312.11805", "2009.03300", "1905.07830", "1911.11641", "1904.09728", "1905.10044", "1907.10641", "1811.00937", "1809.02789", "1911.01547", "1705.03551", "2107.03374", "2108.07732", "2110.14168", "2304.06364", "2206.04615", "1804.06876", "2110.08193", "2009.11462", "2101.11718", "1804.09301", "2109.07958", "2203.09509" ]

[]

[ "### Description\n\n\nGemma is a family of lightweight, state-of-the-art open models from Google,\nbuilt from the same research and technology used to create the Gemini models.\nThey are text-to-text, decoder-only large language models, available in English,\nwith open weights, pre-trained variants, and instruction-tuned variants. Gemma\nmodels are well-suited for a variety of text generation tasks, including\nquestion answering, summarization, and reasoning. Their relatively small size\nmakes it possible to deploy them in environments with limited resources such as\na laptop, desktop or your own cloud infrastructure, democratizing access to\nstate of the art AI models and helping foster innovation for everyone.", "### Usage\n\n\nBelow we share some code snippets on how to get quickly started with running the model. First make sure to 'pip install -U transformers', then copy the snippet from the section that is relevant for your usecase.", "#### Running the model on a CPU", "#### Running the model on a single / multi GPU", "#### Running the model on a GPU using different precisions\n\n\n* *Using 'torch.float16'*\n* *Using 'torch.bfloat16'*", "#### Quantized Versions through 'bitsandbytes'\n\n\n* *Using 8-bit precision (int8)*\n* *Using 4-bit precision*", "#### Other optimizations\n\n\n* *Flash Attention 2*\n\n\nFirst make sure to install 'flash-attn' in your environment 'pip install flash-attn'", "### Chat Template\n\n\nThe instruction-tuned models use a chat template that must be adhered to for conversational use.\nThe easiest way to apply it is using the tokenizer's built-in chat template, as shown in the following snippet.\n\n\nLet's load the model and apply the chat template to a conversation. In this example, we'll start with a single user interaction:\n\n\nAt this point, the prompt contains the following text:\n\n\nAs you can see, each turn is preceeded by a '<start\\_of\\_turn>' delimiter and then the role of the entity\n(either 'user', for content supplied by the user, or 'model' for LLM responses). Turns finish with\nthe '<end\\_of\\_turn>' token.\n\n\nYou can follow this format to build the prompt manually, if you need to do it without the tokenizer's\nchat template.\n\n\nAfter the prompt is ready, generation can be performed like this:", "### Inputs and outputs\n\n\n* Input: Text string, such as a question, a prompt, or a document to be\nsummarized.\n* Output: Generated English-language text in response to the input, such\nas an answer to a question, or a summary of a document.\n\n\nModel Data\n----------\n\n\nData used for model training and how the data was processed.", "### Training Dataset\n\n\nThese models were trained on a dataset of text data that includes a wide variety\nof sources, totaling 6 trillion tokens. Here are the key components:\n\n\n* Web Documents: A diverse collection of web text ensures the model is exposed\nto a broad range of linguistic styles, topics, and vocabulary. Primarily\nEnglish-language content.\n* Code: Exposing the model to code helps it to learn the syntax and patterns of\nprogramming languages, which improves its ability to generate code or\nunderstand code-related questions.\n* Mathematics: Training on mathematical text helps the model learn logical\nreasoning, symbolic representation, and to address mathematical queries.\n\n\nThe combination of these diverse data sources is crucial for training a powerful\nlanguage model that can handle a wide variety of different tasks and text\nformats.", "### Data Preprocessing\n\n\nHere are the key data cleaning and filtering methods applied to the training\ndata:\n\n\n* CSAM Filtering: Rigorous CSAM (Child Sexual Abuse Material) filtering was\napplied at multiple stages in the data preparation process to ensure the\nexclusion of harmful and illegal content\n* Sensitive Data Filtering: As part of making Gemma pre-trained models safe and\nreliable, automated techniques were used to filter out certain personal\ninformation and other sensitive data from training sets.\n* Additional methods: Filtering based on content quality and safely in line with\nour policies.\n\n\nImplementation Information\n--------------------------\n\n\nDetails about the model internals.", "### Hardware\n\n\nGemma was trained using the latest generation of\nTensor Processing Unit (TPU) hardware (TPUv5e).\n\n\nTraining large language models requires significant computational power. TPUs,\ndesigned specifically for matrix operations common in machine learning, offer\nseveral advantages in this domain:\n\n\n* Performance: TPUs are specifically designed to handle the massive computations\ninvolved in training LLMs. They can speed up training considerably compared to\nCPUs.\n* Memory: TPUs often come with large amounts of high-bandwidth memory, allowing\nfor the handling of large models and batch sizes during training. This can\nlead to better model quality.\n* Scalability: TPU Pods (large clusters of TPUs) provide a scalable solution for\nhandling the growing complexity of large foundation models. You can distribute\ntraining across multiple TPU devices for faster and more efficient processing.\n* Cost-effectiveness: In many scenarios, TPUs can provide a more cost-effective\nsolution for training large models compared to CPU-based infrastructure,\nespecially when considering the time and resources saved due to faster\ntraining.\n* These advantages are aligned with\nGoogle's commitments to operate sustainably.", "### Software\n\n\nTraining was done using JAX and ML Pathways.\n\n\nJAX allows researchers to take advantage of the latest generation of hardware,\nincluding TPUs, for faster and more efficient training of large models.\n\n\nML Pathways is Google's latest effort to build artificially intelligent systems\ncapable of generalizing across multiple tasks. This is specially suitable for\nfoundation models, including large language models like\nthese ones.\n\n\nTogether, JAX and ML Pathways are used as described in the\npaper about the Gemini family of models; \"the 'single\ncontroller' programming model of Jax and Pathways allows a single Python\nprocess to orchestrate the entire training run, dramatically simplifying the\ndevelopment workflow.\"\n\n\nEvaluation\n----------\n\n\nModel evaluation metrics and results.", "### Benchmark Results\n\n\nThese models were evaluated against a large collection of different datasets and\nmetrics to cover different aspects of text generation:\n\n\n\nEthics and Safety\n-----------------\n\n\nEthics and safety evaluation approach and results.", "### Evaluation Approach\n\n\nOur evaluation methods include structured evaluations and internal red-teaming\ntesting of relevant content policies. Red-teaming was conducted by a number of\ndifferent teams, each with different goals and human evaluation metrics. These\nmodels were evaluated against a number of different categories relevant to\nethics and safety, including:\n\n\n* Text-to-Text Content Safety: Human evaluation on prompts covering safety\npolicies including child sexual abuse and exploitation, harassment, violence\nand gore, and hate speech.\n* Text-to-Text Representational Harms: Benchmark against relevant academic\ndatasets such as WinoBias and BBQ Dataset.\n* Memorization: Automated evaluation of memorization of training data, including\nthe risk of personally identifiable information exposure.\n* Large-scale harm: Tests for \"dangerous capabilities,\" such as chemical,\nbiological, radiological, and nuclear (CBRN) risks.", "### Evaluation Results\n\n\nThe results of ethics and safety evaluations are within acceptable thresholds\nfor meeting internal policies for categories such as child\nsafety, content safety, representational harms, memorization, large-scale harms.\nOn top of robust internal evaluations, the results of well known safety\nbenchmarks like BBQ, BOLD, Winogender, Winobias, RealToxicity, and TruthfulQA\nare shown here.\n\n\n\nUsage and Limitations\n---------------------\n\n\nThese models have certain limitations that users should be aware of.", "### Intended Usage\n\n\nOpen Large Language Models (LLMs) have a wide range of applications across\nvarious industries and domains. The following list of potential uses is not\ncomprehensive. The purpose of this list is to provide contextual information\nabout the possible use-cases that the model creators considered as part of model\ntraining and development.\n\n\n* Content Creation and Communication\n\t+ Text Generation: These models can be used to generate creative text formats\n\tsuch as poems, scripts, code, marketing copy, and email drafts.\n\t+ Chatbots and Conversational AI: Power conversational interfaces for customer\n\tservice, virtual assistants, or interactive applications.\n\t+ Text Summarization: Generate concise summaries of a text corpus, research\n\tpapers, or reports.\n* Research and Education\n\t+ Natural Language Processing (NLP) Research: These models can serve as a\n\tfoundation for researchers to experiment with NLP techniques, develop\n\talgorithms, and contribute to the advancement of the field.\n\t+ Language Learning Tools: Support interactive language learning experiences,\n\taiding in grammar correction or providing writing practice.\n\t+ Knowledge Exploration: Assist researchers in exploring large bodies of text\n\tby generating summaries or answering questions about specific topics.", "### Limitations\n\n\n* Training Data\n\t+ The quality and diversity of the training data significantly influence the\n\tmodel's capabilities. Biases or gaps in the training data can lead to\n\tlimitations in the model's responses.\n\t+ The scope of the training dataset determines the subject areas the model can\n\thandle effectively.\n* Context and Task Complexity\n\t+ LLMs are better at tasks that can be framed with clear prompts and\n\tinstructions. Open-ended or highly complex tasks might be challenging.\n\t+ A model's performance can be influenced by the amount of context provided\n\t(longer context generally leads to better outputs, up to a certain point).\n* Language Ambiguity and Nuance\n\t+ Natural language is inherently complex. LLMs might struggle to grasp subtle\n\tnuances, sarcasm, or figurative language.\n* Factual Accuracy\n\t+ LLMs generate responses based on information they learned from their\n\ttraining datasets, but they are not knowledge bases. They may generate\n\tincorrect or outdated factual statements.\n* Common Sense\n\t+ LLMs rely on statistical patterns in language. They might lack the ability\n\tto apply common sense reasoning in certain situations.", "### Ethical Considerations and Risks\n\n\nThe development of large language models (LLMs) raises several ethical concerns.\nIn creating an open model, we have carefully considered the following:\n\n\n* Bias and Fairness\n\t+ LLMs trained on large-scale, real-world text data can reflect socio-cultural\n\tbiases embedded in the training material. These models underwent careful\n\tscrutiny, input data pre-processing described and posterior evaluations\n\treported in this card.\n* Misinformation and Misuse\n\t+ LLMs can be misused to generate text that is false, misleading, or harmful.\n\t+ Guidelines are provided for responsible use with the model, see the\n\tResponsible Generative AI Toolkit.\n* Transparency and Accountability:\n\t+ This model card summarizes details on the models' architecture,\n\tcapabilities, limitations, and evaluation processes.\n\t+ A responsibly developed open model offers the opportunity to share\n\tinnovation by making LLM technology accessible to developers and researchers\n\tacross the AI ecosystem.\n\n\nRisks identified and mitigations:\n\n\n* Perpetuation of biases: It's encouraged to perform continuous monitoring\n(using evaluation metrics, human review) and the exploration of de-biasing\ntechniques during model training, fine-tuning, and other use cases.\n* Generation of harmful content: Mechanisms and guidelines for content safety\nare essential. Developers are encouraged to exercise caution and implement\nappropriate content safety safeguards based on their specific product policies\nand application use cases.\n* Misuse for malicious purposes: Technical limitations and developer and\nend-user education can help mitigate against malicious applications of LLMs.\nEducational resources and reporting mechanisms for users to flag misuse are\nprovided. Prohibited uses of Gemma models are outlined in the\nGemma Prohibited Use Policy.\n* Privacy violations: Models were trained on data filtered for removal of PII\n(Personally Identifiable Information). Developers are encouraged to adhere to\nprivacy regulations with privacy-preserving techniques.", "### Benefits\n\n\nAt the time of release, this family of models provides high-performance open\nlarge language model implementations designed from the ground up for Responsible\nAI development compared to similarly sized models.\n\n\nUsing the benchmark evaluation metrics described in this document, these models\nhave shown to provide superior performance to other, comparably-sized open model\nalternatives." ]

[ "TAGS\n#gguf #arxiv-2312.11805 #arxiv-2009.03300 #arxiv-1905.07830 #arxiv-1911.11641 #arxiv-1904.09728 #arxiv-1905.10044 #arxiv-1907.10641 #arxiv-1811.00937 #arxiv-1809.02789 #arxiv-1911.01547 #arxiv-1705.03551 #arxiv-2107.03374 #arxiv-2108.07732 #arxiv-2110.14168 #arxiv-2304.06364 #arxiv-2206.04615 #arxiv-1804.06876 #arxiv-2110.08193 #arxiv-2009.11462 #arxiv-2101.11718 #arxiv-1804.09301 #arxiv-2109.07958 #arxiv-2203.09509 #region-us \n", "### Description\n\n\nGemma is a family of lightweight, state-of-the-art open models from Google,\nbuilt from the same research and technology used to create the Gemini models.\nThey are text-to-text, decoder-only large language models, available in English,\nwith open weights, pre-trained variants, and instruction-tuned variants. Gemma\nmodels are well-suited for a variety of text generation tasks, including\nquestion answering, summarization, and reasoning. Their relatively small size\nmakes it possible to deploy them in environments with limited resources such as\na laptop, desktop or your own cloud infrastructure, democratizing access to\nstate of the art AI models and helping foster innovation for everyone.", "### Usage\n\n\nBelow we share some code snippets on how to get quickly started with running the model. First make sure to 'pip install -U transformers', then copy the snippet from the section that is relevant for your usecase.", "#### Running the model on a CPU", "#### Running the model on a single / multi GPU", "#### Running the model on a GPU using different precisions\n\n\n* *Using 'torch.float16'*\n* *Using 'torch.bfloat16'*", "#### Quantized Versions through 'bitsandbytes'\n\n\n* *Using 8-bit precision (int8)*\n* *Using 4-bit precision*", "#### Other optimizations\n\n\n* *Flash Attention 2*\n\n\nFirst make sure to install 'flash-attn' in your environment 'pip install flash-attn'", "### Chat Template\n\n\nThe instruction-tuned models use a chat template that must be adhered to for conversational use.\nThe easiest way to apply it is using the tokenizer's built-in chat template, as shown in the following snippet.\n\n\nLet's load the model and apply the chat template to a conversation. In this example, we'll start with a single user interaction:\n\n\nAt this point, the prompt contains the following text:\n\n\nAs you can see, each turn is preceeded by a '<start\\_of\\_turn>' delimiter and then the role of the entity\n(either 'user', for content supplied by the user, or 'model' for LLM responses). Turns finish with\nthe '<end\\_of\\_turn>' token.\n\n\nYou can follow this format to build the prompt manually, if you need to do it without the tokenizer's\nchat template.\n\n\nAfter the prompt is ready, generation can be performed like this:", "### Inputs and outputs\n\n\n* Input: Text string, such as a question, a prompt, or a document to be\nsummarized.\n* Output: Generated English-language text in response to the input, such\nas an answer to a question, or a summary of a document.\n\n\nModel Data\n----------\n\n\nData used for model training and how the data was processed.", "### Training Dataset\n\n\nThese models were trained on a dataset of text data that includes a wide variety\nof sources, totaling 6 trillion tokens. Here are the key components:\n\n\n* Web Documents: A diverse collection of web text ensures the model is exposed\nto a broad range of linguistic styles, topics, and vocabulary. Primarily\nEnglish-language content.\n* Code: Exposing the model to code helps it to learn the syntax and patterns of\nprogramming languages, which improves its ability to generate code or\nunderstand code-related questions.\n* Mathematics: Training on mathematical text helps the model learn logical\nreasoning, symbolic representation, and to address mathematical queries.\n\n\nThe combination of these diverse data sources is crucial for training a powerful\nlanguage model that can handle a wide variety of different tasks and text\nformats.", "### Data Preprocessing\n\n\nHere are the key data cleaning and filtering methods applied to the training\ndata:\n\n\n* CSAM Filtering: Rigorous CSAM (Child Sexual Abuse Material) filtering was\napplied at multiple stages in the data preparation process to ensure the\nexclusion of harmful and illegal content\n* Sensitive Data Filtering: As part of making Gemma pre-trained models safe and\nreliable, automated techniques were used to filter out certain personal\ninformation and other sensitive data from training sets.\n* Additional methods: Filtering based on content quality and safely in line with\nour policies.\n\n\nImplementation Information\n--------------------------\n\n\nDetails about the model internals.", "### Hardware\n\n\nGemma was trained using the latest generation of\nTensor Processing Unit (TPU) hardware (TPUv5e).\n\n\nTraining large language models requires significant computational power. TPUs,\ndesigned specifically for matrix operations common in machine learning, offer\nseveral advantages in this domain:\n\n\n* Performance: TPUs are specifically designed to handle the massive computations\ninvolved in training LLMs. They can speed up training considerably compared to\nCPUs.\n* Memory: TPUs often come with large amounts of high-bandwidth memory, allowing\nfor the handling of large models and batch sizes during training. This can\nlead to better model quality.\n* Scalability: TPU Pods (large clusters of TPUs) provide a scalable solution for\nhandling the growing complexity of large foundation models. You can distribute\ntraining across multiple TPU devices for faster and more efficient processing.\n* Cost-effectiveness: In many scenarios, TPUs can provide a more cost-effective\nsolution for training large models compared to CPU-based infrastructure,\nespecially when considering the time and resources saved due to faster\ntraining.\n* These advantages are aligned with\nGoogle's commitments to operate sustainably.", "### Software\n\n\nTraining was done using JAX and ML Pathways.\n\n\nJAX allows researchers to take advantage of the latest generation of hardware,\nincluding TPUs, for faster and more efficient training of large models.\n\n\nML Pathways is Google's latest effort to build artificially intelligent systems\ncapable of generalizing across multiple tasks. This is specially suitable for\nfoundation models, including large language models like\nthese ones.\n\n\nTogether, JAX and ML Pathways are used as described in the\npaper about the Gemini family of models; \"the 'single\ncontroller' programming model of Jax and Pathways allows a single Python\nprocess to orchestrate the entire training run, dramatically simplifying the\ndevelopment workflow.\"\n\n\nEvaluation\n----------\n\n\nModel evaluation metrics and results.", "### Benchmark Results\n\n\nThese models were evaluated against a large collection of different datasets and\nmetrics to cover different aspects of text generation:\n\n\n\nEthics and Safety\n-----------------\n\n\nEthics and safety evaluation approach and results.", "### Evaluation Approach\n\n\nOur evaluation methods include structured evaluations and internal red-teaming\ntesting of relevant content policies. Red-teaming was conducted by a number of\ndifferent teams, each with different goals and human evaluation metrics. These\nmodels were evaluated against a number of different categories relevant to\nethics and safety, including:\n\n\n* Text-to-Text Content Safety: Human evaluation on prompts covering safety\npolicies including child sexual abuse and exploitation, harassment, violence\nand gore, and hate speech.\n* Text-to-Text Representational Harms: Benchmark against relevant academic\ndatasets such as WinoBias and BBQ Dataset.\n* Memorization: Automated evaluation of memorization of training data, including\nthe risk of personally identifiable information exposure.\n* Large-scale harm: Tests for \"dangerous capabilities,\" such as chemical,\nbiological, radiological, and nuclear (CBRN) risks.", "### Evaluation Results\n\n\nThe results of ethics and safety evaluations are within acceptable thresholds\nfor meeting internal policies for categories such as child\nsafety, content safety, representational harms, memorization, large-scale harms.\nOn top of robust internal evaluations, the results of well known safety\nbenchmarks like BBQ, BOLD, Winogender, Winobias, RealToxicity, and TruthfulQA\nare shown here.\n\n\n\nUsage and Limitations\n---------------------\n\n\nThese models have certain limitations that users should be aware of.", "### Intended Usage\n\n\nOpen Large Language Models (LLMs) have a wide range of applications across\nvarious industries and domains. The following list of potential uses is not\ncomprehensive. The purpose of this list is to provide contextual information\nabout the possible use-cases that the model creators considered as part of model\ntraining and development.\n\n\n* Content Creation and Communication\n\t+ Text Generation: These models can be used to generate creative text formats\n\tsuch as poems, scripts, code, marketing copy, and email drafts.\n\t+ Chatbots and Conversational AI: Power conversational interfaces for customer\n\tservice, virtual assistants, or interactive applications.\n\t+ Text Summarization: Generate concise summaries of a text corpus, research\n\tpapers, or reports.\n* Research and Education\n\t+ Natural Language Processing (NLP) Research: These models can serve as a\n\tfoundation for researchers to experiment with NLP techniques, develop\n\talgorithms, and contribute to the advancement of the field.\n\t+ Language Learning Tools: Support interactive language learning experiences,\n\taiding in grammar correction or providing writing practice.\n\t+ Knowledge Exploration: Assist researchers in exploring large bodies of text\n\tby generating summaries or answering questions about specific topics.", "### Limitations\n\n\n* Training Data\n\t+ The quality and diversity of the training data significantly influence the\n\tmodel's capabilities. Biases or gaps in the training data can lead to\n\tlimitations in the model's responses.\n\t+ The scope of the training dataset determines the subject areas the model can\n\thandle effectively.\n* Context and Task Complexity\n\t+ LLMs are better at tasks that can be framed with clear prompts and\n\tinstructions. Open-ended or highly complex tasks might be challenging.\n\t+ A model's performance can be influenced by the amount of context provided\n\t(longer context generally leads to better outputs, up to a certain point).\n* Language Ambiguity and Nuance\n\t+ Natural language is inherently complex. LLMs might struggle to grasp subtle\n\tnuances, sarcasm, or figurative language.\n* Factual Accuracy\n\t+ LLMs generate responses based on information they learned from their\n\ttraining datasets, but they are not knowledge bases. They may generate\n\tincorrect or outdated factual statements.\n* Common Sense\n\t+ LLMs rely on statistical patterns in language. They might lack the ability\n\tto apply common sense reasoning in certain situations.", "### Ethical Considerations and Risks\n\n\nThe development of large language models (LLMs) raises several ethical concerns.\nIn creating an open model, we have carefully considered the following:\n\n\n* Bias and Fairness\n\t+ LLMs trained on large-scale, real-world text data can reflect socio-cultural\n\tbiases embedded in the training material. These models underwent careful\n\tscrutiny, input data pre-processing described and posterior evaluations\n\treported in this card.\n* Misinformation and Misuse\n\t+ LLMs can be misused to generate text that is false, misleading, or harmful.\n\t+ Guidelines are provided for responsible use with the model, see the\n\tResponsible Generative AI Toolkit.\n* Transparency and Accountability:\n\t+ This model card summarizes details on the models' architecture,\n\tcapabilities, limitations, and evaluation processes.\n\t+ A responsibly developed open model offers the opportunity to share\n\tinnovation by making LLM technology accessible to developers and researchers\n\tacross the AI ecosystem.\n\n\nRisks identified and mitigations:\n\n\n* Perpetuation of biases: It's encouraged to perform continuous monitoring\n(using evaluation metrics, human review) and the exploration of de-biasing\ntechniques during model training, fine-tuning, and other use cases.\n* Generation of harmful content: Mechanisms and guidelines for content safety\nare essential. Developers are encouraged to exercise caution and implement\nappropriate content safety safeguards based on their specific product policies\nand application use cases.\n* Misuse for malicious purposes: Technical limitations and developer and\nend-user education can help mitigate against malicious applications of LLMs.\nEducational resources and reporting mechanisms for users to flag misuse are\nprovided. Prohibited uses of Gemma models are outlined in the\nGemma Prohibited Use Policy.\n* Privacy violations: Models were trained on data filtered for removal of PII\n(Personally Identifiable Information). Developers are encouraged to adhere to\nprivacy regulations with privacy-preserving techniques.", "### Benefits\n\n\nAt the time of release, this family of models provides high-performance open\nlarge language model implementations designed from the ground up for Responsible\nAI development compared to similarly sized models.\n\n\nUsing the benchmark evaluation metrics described in this document, these models\nhave shown to provide superior performance to other, comparably-sized open model\nalternatives." ]

[ "transformers", "tensorboard", "safetensors", "mistral", "text-generation", "alignment-handbook", "trl", "sft", "generated_from_trainer", "conversational", "dataset:GAIR/lima", "base_model:mistralai/Mistral-7B-v0.1", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[]

[]

[ "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 2e-05\n* train\\_batch\\_size: 8\n* eval\\_batch\\_size: 8\n* seed: 42\n* distributed\\_type: multi-GPU\n* num\\_devices: 16\n* total\\_train\\_batch\\_size: 128\n* total\\_eval\\_batch\\_size: 128\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: cosine\n* lr\\_scheduler\\_warmup\\_ratio: 0.1\n* num\\_epochs: 10.0", "### Training results", "### Framework versions\n\n\n* Transformers 4.38.2\n* Pytorch 2.1.2\n* Datasets 2.14.6\n* Tokenizers 0.15.2" ]

[ "TAGS\n#transformers #tensorboard #safetensors #mistral #text-generation #alignment-handbook #trl #sft #generated_from_trainer #conversational #dataset-GAIR/lima #base_model-mistralai/Mistral-7B-v0.1 #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n", "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 2e-05\n* train\\_batch\\_size: 8\n* eval\\_batch\\_size: 8\n* seed: 42\n* distributed\\_type: multi-GPU\n* num\\_devices: 16\n* total\\_train\\_batch\\_size: 128\n* total\\_eval\\_batch\\_size: 128\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: cosine\n* lr\\_scheduler\\_warmup\\_ratio: 0.1\n* num\\_epochs: 10.0", "### Training results", "### Framework versions\n\n\n* Transformers 4.38.2\n* Pytorch 2.1.2\n* Datasets 2.14.6\n* Tokenizers 0.15.2" ]

[ "transformers", "tensorboard", "safetensors", "roberta", "text-classification", "generated_from_trainer", "base_model:FacebookAI/roberta-base", "license:mit", "autotrain_compatible", "endpoints_compatible", "region:us" ]

[]

[]

[ "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 2e-05\n* train\\_batch\\_size: 16\n* eval\\_batch\\_size: 16\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* num\\_epochs: 5", "### Training results", "### Framework versions\n\n\n* Transformers 4.40.0\n* Pytorch 2.2.1+cu121\n* Datasets 2.19.0\n* Tokenizers 0.19.1" ]

[ "TAGS\n#transformers #tensorboard #safetensors #roberta #text-classification #generated_from_trainer #base_model-FacebookAI/roberta-base #license-mit #autotrain_compatible #endpoints_compatible #region-us \n", "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 2e-05\n* train\\_batch\\_size: 16\n* eval\\_batch\\_size: 16\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* num\\_epochs: 5", "### Training results", "### Framework versions\n\n\n* Transformers 4.40.0\n* Pytorch 2.2.1+cu121\n* Datasets 2.19.0\n* Tokenizers 0.19.1" ]

[ "transformers", "safetensors", "t5", "text2text-generation", "generated_from_trainer", "base_model:t5-small", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[]

[]

[ "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 5e-05\n* train\\_batch\\_size: 16\n* eval\\_batch\\_size: 16\n* seed: 42\n* gradient\\_accumulation\\_steps: 4\n* total\\_train\\_batch\\_size: 64\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-06\n* lr\\_scheduler\\_type: linear\n* lr\\_scheduler\\_warmup\\_steps: 100\n* num\\_epochs: 10\n* mixed\\_precision\\_training: Native AMP", "### Training results", "### Framework versions\n\n\n* Transformers 4.40.0\n* Pytorch 2.2.1+cu121\n* Datasets 2.13.1\n* Tokenizers 0.19.1" ]

[ "TAGS\n#transformers #safetensors #t5 #text2text-generation #generated_from_trainer #base_model-t5-small #license-apache-2.0 #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n", "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 5e-05\n* train\\_batch\\_size: 16\n* eval\\_batch\\_size: 16\n* seed: 42\n* gradient\\_accumulation\\_steps: 4\n* total\\_train\\_batch\\_size: 64\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-06\n* lr\\_scheduler\\_type: linear\n* lr\\_scheduler\\_warmup\\_steps: 100\n* num\\_epochs: 10\n* mixed\\_precision\\_training: Native AMP", "### Training results", "### Framework versions\n\n\n* Transformers 4.40.0\n* Pytorch 2.2.1+cu121\n* Datasets 2.13.1\n* Tokenizers 0.19.1" ]

[ "transformers", "safetensors", "mistral", "text-generation", "4-bit", "AWQ", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[]

[]

[ "# feeltheAGI/Maverick-7B AWQ\n\n- Model creator: feeltheAGI\n- Original model: Maverick-7B\n\n!URL", "## Model Summary\n\nThis model is a merge of the following models:\n* mlabonne/Marcoro14-7B-slerp\n* mlabonne/NeuralBeagle14-7B", "## How to use", "### Install the necessary packages", "### Example Python code", "### About AWQ\n\nAWQ is an efficient, accurate and blazing-fast low-bit weight quantization method, currently supporting 4-bit quantization. Compared to GPTQ, it offers faster Transformers-based inference with equivalent or better quality compared to the most commonly used GPTQ settings.\n\nAWQ models are currently supported on Linux and Windows, with NVidia GPUs only. macOS users: please use GGUF models instead.\n\nIt is supported by:\n\n- Text Generation Webui - using Loader: AutoAWQ\n- vLLM - version 0.2.2 or later for support for all model types.\n- Hugging Face Text Generation Inference (TGI)\n- Transformers version 4.35.0 and later, from any code or client that supports Transformers\n- AutoAWQ - for use from Python code" ]

[ "TAGS\n#transformers #safetensors #mistral #text-generation #4-bit #AWQ #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n", "# feeltheAGI/Maverick-7B AWQ\n\n- Model creator: feeltheAGI\n- Original model: Maverick-7B\n\n!URL", "## Model Summary\n\nThis model is a merge of the following models:\n* mlabonne/Marcoro14-7B-slerp\n* mlabonne/NeuralBeagle14-7B", "## How to use", "### Install the necessary packages", "### Example Python code", "### About AWQ\n\nAWQ is an efficient, accurate and blazing-fast low-bit weight quantization method, currently supporting 4-bit quantization. Compared to GPTQ, it offers faster Transformers-based inference with equivalent or better quality compared to the most commonly used GPTQ settings.\n\nAWQ models are currently supported on Linux and Windows, with NVidia GPUs only. macOS users: please use GGUF models instead.\n\nIt is supported by:\n\n- Text Generation Webui - using Loader: AutoAWQ\n- vLLM - version 0.2.2 or later for support for all model types.\n- Hugging Face Text Generation Inference (TGI)\n- Transformers version 4.35.0 and later, from any code or client that supports Transformers\n- AutoAWQ - for use from Python code" ]

[ "transformers", "safetensors", "llama", "text-generation", "4-bit", "AWQ", "autotrain_compatible", "endpoints_compatible", "nsfw", "dataset:macadeliccc/opus_samantha", "license:apache-2.0", "text-generation-inference", "region:us" ]

[]

[]

[ "# macadeliccc/Opus-Samantha-Llama-3-8B AWQ\n\n- Model creator: macadeliccc\n- Original model: Opus-Samantha-Llama-3-8B", "## model Summary\n\nOpus-Samantha-Llama-3-8B is a SFT model made with AutoSloth by macadeliccc\n\n_model is curretly very nsfw. uneven distribution of subjects in dataset. will be back with v2_", "## How to use", "### Install the necessary packages", "### Example Python code", "### About AWQ\n\nAWQ is an efficient, accurate and blazing-fast low-bit weight quantization method, currently supporting 4-bit quantization. Compared to GPTQ, it offers faster Transformers-based inference with equivalent or better quality compared to the most commonly used GPTQ settings.\n\nAWQ models are currently supported on Linux and Windows, with NVidia GPUs only. macOS users: please use GGUF models instead.\n\nIt is supported by:\n\n- Text Generation Webui - using Loader: AutoAWQ\n- vLLM - version 0.2.2 or later for support for all model types.\n- Hugging Face Text Generation Inference (TGI)\n- Transformers version 4.35.0 and later, from any code or client that supports Transformers\n- AutoAWQ - for use from Python code" ]

[ "TAGS\n#transformers #safetensors #llama #text-generation #4-bit #AWQ #autotrain_compatible #endpoints_compatible #nsfw #dataset-macadeliccc/opus_samantha #license-apache-2.0 #text-generation-inference #region-us \n", "# macadeliccc/Opus-Samantha-Llama-3-8B AWQ\n\n- Model creator: macadeliccc\n- Original model: Opus-Samantha-Llama-3-8B", "## model Summary\n\nOpus-Samantha-Llama-3-8B is a SFT model made with AutoSloth by macadeliccc\n\n_model is curretly very nsfw. uneven distribution of subjects in dataset. will be back with v2_", "## How to use", "### Install the necessary packages", "### Example Python code", "### About AWQ\n\nAWQ is an efficient, accurate and blazing-fast low-bit weight quantization method, currently supporting 4-bit quantization. Compared to GPTQ, it offers faster Transformers-based inference with equivalent or better quality compared to the most commonly used GPTQ settings.\n\nAWQ models are currently supported on Linux and Windows, with NVidia GPUs only. macOS users: please use GGUF models instead.\n\nIt is supported by:\n\n- Text Generation Webui - using Loader: AutoAWQ\n- vLLM - version 0.2.2 or later for support for all model types.\n- Hugging Face Text Generation Inference (TGI)\n- Transformers version 4.35.0 and later, from any code or client that supports Transformers\n- AutoAWQ - for use from Python code" ]

[ "peft", "safetensors", "generated_from_trainer", "base_model:mistralai/Mistral-7B-v0.1", "license:apache-2.0", "region:us" ]

[]

[]

[ "# mistral_train_seq_cls_run4\n\nThis model is a fine-tuned version of mistralai/Mistral-7B-v0.1 on the None dataset.", "## Model description\n\nMore information needed", "## Intended uses & limitations\n\nMore information needed", "## Training and evaluation data\n\nMore information needed", "## Training procedure", "### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 0.0002\n- train_batch_size: 8\n- eval_batch_size: 8\n- seed: 42\n- gradient_accumulation_steps: 2\n- total_train_batch_size: 16\n- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n- lr_scheduler_type: linear\n- lr_scheduler_warmup_ratio: 0.03\n- num_epochs: 3", "### Training results", "### Framework versions\n\n- PEFT 0.7.2.dev0\n- Transformers 4.36.2\n- Pytorch 2.1.2+cu121\n- Datasets 2.16.1\n- Tokenizers 0.15.1" ]

[ "TAGS\n#peft #safetensors #generated_from_trainer #base_model-mistralai/Mistral-7B-v0.1 #license-apache-2.0 #region-us \n", "# mistral_train_seq_cls_run4\n\nThis model is a fine-tuned version of mistralai/Mistral-7B-v0.1 on the None dataset.", "## Model description\n\nMore information needed", "## Intended uses & limitations\n\nMore information needed", "## Training and evaluation data\n\nMore information needed", "## Training procedure", "### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 0.0002\n- train_batch_size: 8\n- eval_batch_size: 8\n- seed: 42\n- gradient_accumulation_steps: 2\n- total_train_batch_size: 16\n- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n- lr_scheduler_type: linear\n- lr_scheduler_warmup_ratio: 0.03\n- num_epochs: 3", "### Training results", "### Framework versions\n\n- PEFT 0.7.2.dev0\n- Transformers 4.36.2\n- Pytorch 2.1.2+cu121\n- Datasets 2.16.1\n- Tokenizers 0.15.1" ]

[ "transformers", "safetensors", "roberta", "feature-extraction", "arxiv:1910.09700", "endpoints_compatible", "region:us" ]

[ "1910.09700" ]

[]

[ "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "TAGS\n#transformers #safetensors #roberta #feature-extraction #arxiv-1910.09700 #endpoints_compatible #region-us \n", "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "sentence-transformers", "safetensors", "roberta", "feature-extraction", "sentence-similarity", "transformers", "endpoints_compatible", "region:us" ]

[]

[]

[ "# kien-vu-uet/finetune-vietnamese-bi-encoder\n\nThis is a sentence-transformers model: It maps sentences & paragraphs to a 768 dimensional dense vector space and can be used for tasks like clustering or semantic search.", "## Usage (Sentence-Transformers)\n\nUsing this model becomes easy when you have sentence-transformers installed:\n\n\n\nThen you can use the model like this:", "## Usage (HuggingFace Transformers)\nWithout sentence-transformers, you can use the model like this: First, you pass your input through the transformer model, then you have to apply the right pooling-operation on-top of the contextualized word embeddings.", "## Evaluation Results\n\n\n\nFor an automated evaluation of this model, see the *Sentence Embeddings Benchmark*: URL", "## Training\nThe model was trained with the parameters:\n\nDataLoader:\n\n'URL.dataloader.DataLoader' of length 9818 with parameters:\n\n\nLoss:\n\n'sentence_transformers.losses.ContrastiveLoss.ContrastiveLoss' with parameters:\n \n\nParameters of the fit()-Method:", "## Full Model Architecture", "## Citing & Authors" ]

[ "TAGS\n#sentence-transformers #safetensors #roberta #feature-extraction #sentence-similarity #transformers #endpoints_compatible #region-us \n", "# kien-vu-uet/finetune-vietnamese-bi-encoder\n\nThis is a sentence-transformers model: It maps sentences & paragraphs to a 768 dimensional dense vector space and can be used for tasks like clustering or semantic search.", "## Usage (Sentence-Transformers)\n\nUsing this model becomes easy when you have sentence-transformers installed:\n\n\n\nThen you can use the model like this:", "## Usage (HuggingFace Transformers)\nWithout sentence-transformers, you can use the model like this: First, you pass your input through the transformer model, then you have to apply the right pooling-operation on-top of the contextualized word embeddings.", "## Evaluation Results\n\n\n\nFor an automated evaluation of this model, see the *Sentence Embeddings Benchmark*: URL", "## Training\nThe model was trained with the parameters:\n\nDataLoader:\n\n'URL.dataloader.DataLoader' of length 9818 with parameters:\n\n\nLoss:\n\n'sentence_transformers.losses.ContrastiveLoss.ContrastiveLoss' with parameters:\n \n\nParameters of the fit()-Method:", "## Full Model Architecture", "## Citing & Authors" ]

[ "peft", "pytorch", "safetensors", "generated_from_trainer", "dataset:common_voice_16_1", "base_model:openai/whisper-small", "license:apache-2.0", "region:us" ]

[]

[]

[ "# whisper-small-finetuned\n\nThis model is a fine-tuned version of openai/whisper-small on the common_voice_16_1 dataset.\nIt achieves the following results on the evaluation set:\n- eval_loss: 3.6432\n- eval_wer: 109.0909\n- eval_runtime: 14.3752\n- eval_samples_per_second: 0.278\n- eval_steps_per_second: 0.07\n- epoch: 99.01\n- step: 100", "## Model description\n\nMore information needed", "## Intended uses & limitations\n\nMore information needed", "## Training and evaluation data\n\nMore information needed", "## Training procedure", "### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 0.001\n- train_batch_size: 96\n- eval_batch_size: 64\n- seed: 42\n- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n- lr_scheduler_type: linear\n- training_steps: 100\n- mixed_precision_training: Native AMP", "### Framework versions\n\n- PEFT 0.9.0\n- Transformers 4.39.2\n- Pytorch 2.2.2+cu121\n- Datasets 2.17.0\n- Tokenizers 0.15.2" ]

[ "TAGS\n#peft #pytorch #safetensors #generated_from_trainer #dataset-common_voice_16_1 #base_model-openai/whisper-small #license-apache-2.0 #region-us \n", "# whisper-small-finetuned\n\nThis model is a fine-tuned version of openai/whisper-small on the common_voice_16_1 dataset.\nIt achieves the following results on the evaluation set:\n- eval_loss: 3.6432\n- eval_wer: 109.0909\n- eval_runtime: 14.3752\n- eval_samples_per_second: 0.278\n- eval_steps_per_second: 0.07\n- epoch: 99.01\n- step: 100", "## Model description\n\nMore information needed", "## Intended uses & limitations\n\nMore information needed", "## Training and evaluation data\n\nMore information needed", "## Training procedure", "### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 0.001\n- train_batch_size: 96\n- eval_batch_size: 64\n- seed: 42\n- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n- lr_scheduler_type: linear\n- training_steps: 100\n- mixed_precision_training: Native AMP", "### Framework versions\n\n- PEFT 0.9.0\n- Transformers 4.39.2\n- Pytorch 2.2.2+cu121\n- Datasets 2.17.0\n- Tokenizers 0.15.2" ]

[ "gguf", "license:mit", "region:us" ]

[]

[]

[]

[ "TAGS\n#gguf #license-mit #region-us \n" ]

[ "Pixelcopter-PLE-v0", "reinforce", "reinforcement-learning", "custom-implementation", "deep-rl-class", "model-index", "region:us" ]

[]

[]

[ "# Reinforce Agent playing Pixelcopter-PLE-v0\n This is a trained model of a Reinforce agent playing Pixelcopter-PLE-v0 .\n To learn to use this model and train yours check Unit 4 of the Deep Reinforcement Learning Course: URL" ]

[ "TAGS\n#Pixelcopter-PLE-v0 #reinforce #reinforcement-learning #custom-implementation #deep-rl-class #model-index #region-us \n", "# Reinforce Agent playing Pixelcopter-PLE-v0\n This is a trained model of a Reinforce agent playing Pixelcopter-PLE-v0 .\n To learn to use this model and train yours check Unit 4 of the Deep Reinforcement Learning Course: URL" ]

[ "stable-baselines3", "LunarLander-v2", "deep-reinforcement-learning", "reinforcement-learning", "model-index", "region:us" ]

[]

[]

[ "# PPO Agent playing LunarLander-v2\nThis is a trained model of a PPO agent playing LunarLander-v2\nusing the stable-baselines3 library.", "## Usage (with Stable-baselines3)\nTODO: Add your code" ]

[ "TAGS\n#stable-baselines3 #LunarLander-v2 #deep-reinforcement-learning #reinforcement-learning #model-index #region-us \n", "# PPO Agent playing LunarLander-v2\nThis is a trained model of a PPO agent playing LunarLander-v2\nusing the stable-baselines3 library.", "## Usage (with Stable-baselines3)\nTODO: Add your code" ]

[ "transformers", "tensorboard", "safetensors", "vit", "image-classification", "generated_from_trainer", "base_model:google/vit-base-patch16-224-in21k", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "region:us" ]

[]

[]

[ "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0002\n* train\\_batch\\_size: 16\n* eval\\_batch\\_size: 8\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* num\\_epochs: 10", "### Training results", "### Framework versions\n\n\n* Transformers 4.40.1\n* Pytorch 2.2.1+cu121\n* Datasets 2.19.0\n* Tokenizers 0.19.1" ]

[ "TAGS\n#transformers #tensorboard #safetensors #vit #image-classification #generated_from_trainer #base_model-google/vit-base-patch16-224-in21k #license-apache-2.0 #autotrain_compatible #endpoints_compatible #region-us \n", "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0002\n* train\\_batch\\_size: 16\n* eval\\_batch\\_size: 8\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* num\\_epochs: 10", "### Training results", "### Framework versions\n\n\n* Transformers 4.40.1\n* Pytorch 2.2.1+cu121\n* Datasets 2.19.0\n* Tokenizers 0.19.1" ]

[ "transformers", "gguf", "merge", "mergekit", "lazymergekit", "Yuma42/KangalKhan-Alpha-Sapphiroid-7B-Fixed", "argilla/distilabeled-OpenHermes-2.5-Mistral-7B", "en", "base_model:Yuma42/KangalKhan-Alpha-Rubyroid-7B-Fixed", "license:apache-2.0", "endpoints_compatible", "region:us" ]

[]

[ "en" ]

[]

[ "TAGS\n#transformers #gguf #merge #mergekit #lazymergekit #Yuma42/KangalKhan-Alpha-Sapphiroid-7B-Fixed #argilla/distilabeled-OpenHermes-2.5-Mistral-7B #en #base_model-Yuma42/KangalKhan-Alpha-Rubyroid-7B-Fixed #license-apache-2.0 #endpoints_compatible #region-us \n" ]

[ "transformers", "gguf", "merge", "mergekit", "lazymergekit", "meta-llama/Meta-Llama-3-8B-Instruct", "DataPilot/japanese-Llama-2-7b-instruct-bf16", "en", "base_model:DataPilot/ArrowSmart-14B-Llama", "license:llama3", "endpoints_compatible", "region:us" ]

[]

[ "en" ]

[]

[ "TAGS\n#transformers #gguf #merge #mergekit #lazymergekit #meta-llama/Meta-Llama-3-8B-Instruct #DataPilot/japanese-Llama-2-7b-instruct-bf16 #en #base_model-DataPilot/ArrowSmart-14B-Llama #license-llama3 #endpoints_compatible #region-us \n" ]

[ "peft", "safetensors", "generated_from_trainer", "base_model:facebook/esm2_t12_35M_UR50D", "license:mit", "region:us" ]

[]

[]

[ "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0005701568055793089\n* train\\_batch\\_size: 64\n* eval\\_batch\\_size: 64\n* seed: 8893\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: cosine\n* num\\_epochs: 30", "### Training results", "### Framework versions\n\n\n* PEFT 0.10.0\n* Transformers 4.39.3\n* Pytorch 2.2.1\n* Datasets 2.16.1\n* Tokenizers 0.15.2" ]

[ "TAGS\n#peft #safetensors #generated_from_trainer #base_model-facebook/esm2_t12_35M_UR50D #license-mit #region-us \n", "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0005701568055793089\n* train\\_batch\\_size: 64\n* eval\\_batch\\_size: 64\n* seed: 8893\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: cosine\n* num\\_epochs: 30", "### Training results", "### Framework versions\n\n\n* PEFT 0.10.0\n* Transformers 4.39.3\n* Pytorch 2.2.1\n* Datasets 2.16.1\n* Tokenizers 0.15.2" ]

[ "transformers", "safetensors", "llama", "text-generation", "4-bit", "AWQ", "autotrain_compatible", "endpoints_compatible", "license:cc-by-nc-4.0", "text-generation-inference", "region:us" ]

[]

[]

[ "# cgato/L3-TheSpice-8b-v0.8.3 AWQ\n\n- Model creator: cgato\n- Original model: L3-TheSpice-8b-v0.8.3", "## Model Info\n\nNow not overtrained and with the tokenizer fix to base llama3. Trained for 3 epochs.\n\nThe latest TheSpice, dipped in Mama Liz's LimaRP Oil.\nI've focused on making the model more flexible and provide a more unique experience. \nI'm still working on cleaning up my dataset, but I've shrunken it down a lot to focus on a \"less is more\" approach.\nThis is ultimate a return to form of the way I used to train Thespis, with more of a focus on a small hand edited dataset.\n\n* Capybara\n* Claude Multiround 30k\n* Augmental\n* ToxicQA\n* Yahoo Answers\n* Airoboros 3.1\n* LimaRP", "## How to use", "### Install the necessary packages", "### Example Python code", "### About AWQ\n\nAWQ is an efficient, accurate and blazing-fast low-bit weight quantization method, currently supporting 4-bit quantization. Compared to GPTQ, it offers faster Transformers-based inference with equivalent or better quality compared to the most commonly used GPTQ settings.\n\nAWQ models are currently supported on Linux and Windows, with NVidia GPUs only. macOS users: please use GGUF models instead.\n\nIt is supported by:\n\n- Text Generation Webui - using Loader: AutoAWQ\n- vLLM - version 0.2.2 or later for support for all model types.\n- Hugging Face Text Generation Inference (TGI)\n- Transformers version 4.35.0 and later, from any code or client that supports Transformers\n- AutoAWQ - for use from Python code" ]

[ "TAGS\n#transformers #safetensors #llama #text-generation #4-bit #AWQ #autotrain_compatible #endpoints_compatible #license-cc-by-nc-4.0 #text-generation-inference #region-us \n", "# cgato/L3-TheSpice-8b-v0.8.3 AWQ\n\n- Model creator: cgato\n- Original model: L3-TheSpice-8b-v0.8.3", "## Model Info\n\nNow not overtrained and with the tokenizer fix to base llama3. Trained for 3 epochs.\n\nThe latest TheSpice, dipped in Mama Liz's LimaRP Oil.\nI've focused on making the model more flexible and provide a more unique experience. \nI'm still working on cleaning up my dataset, but I've shrunken it down a lot to focus on a \"less is more\" approach.\nThis is ultimate a return to form of the way I used to train Thespis, with more of a focus on a small hand edited dataset.\n\n* Capybara\n* Claude Multiround 30k\n* Augmental\n* ToxicQA\n* Yahoo Answers\n* Airoboros 3.1\n* LimaRP", "## How to use", "### Install the necessary packages", "### Example Python code", "### About AWQ\n\nAWQ is an efficient, accurate and blazing-fast low-bit weight quantization method, currently supporting 4-bit quantization. Compared to GPTQ, it offers faster Transformers-based inference with equivalent or better quality compared to the most commonly used GPTQ settings.\n\nAWQ models are currently supported on Linux and Windows, with NVidia GPUs only. macOS users: please use GGUF models instead.\n\nIt is supported by:\n\n- Text Generation Webui - using Loader: AutoAWQ\n- vLLM - version 0.2.2 or later for support for all model types.\n- Hugging Face Text Generation Inference (TGI)\n- Transformers version 4.35.0 and later, from any code or client that supports Transformers\n- AutoAWQ - for use from Python code" ]

[ "FrozenLake-v1-4x4-no_slippery", "q-learning", "reinforcement-learning", "custom-implementation", "model-index", "region:us" ]

[]

[]

[ "# Q-Learning Agent playing1 FrozenLake-v1\n This is a trained model of a Q-Learning agent playing FrozenLake-v1 .\n\n ## Usage" ]

[ "TAGS\n#FrozenLake-v1-4x4-no_slippery #q-learning #reinforcement-learning #custom-implementation #model-index #region-us \n", "# Q-Learning Agent playing1 FrozenLake-v1\n This is a trained model of a Q-Learning agent playing FrozenLake-v1 .\n\n ## Usage" ]

[ "transformers", "tensorboard", "safetensors", "gpt_neox", "text-classification", "generated_from_trainer", "base_model:EleutherAI/pythia-1b", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[]

[]

[ "# robust_llm_pythia-1b_mz-130_PasswordMatch_n-its-10-seed-4\n\nThis model is a fine-tuned version of EleutherAI/pythia-1b on an unknown dataset.", "## Model description\n\nMore information needed", "## Intended uses & limitations\n\nMore information needed", "## Training and evaluation data\n\nMore information needed", "## Training procedure", "### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 1e-05\n- train_batch_size: 8\n- eval_batch_size: 64\n- seed: 4\n- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n- lr_scheduler_type: linear\n- num_epochs: 1", "### Training results", "### Framework versions\n\n- Transformers 4.39.3\n- Pytorch 2.2.1\n- Datasets 2.18.0\n- Tokenizers 0.15.2" ]

[ "TAGS\n#transformers #tensorboard #safetensors #gpt_neox #text-classification #generated_from_trainer #base_model-EleutherAI/pythia-1b #license-apache-2.0 #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n", "# robust_llm_pythia-1b_mz-130_PasswordMatch_n-its-10-seed-4\n\nThis model is a fine-tuned version of EleutherAI/pythia-1b on an unknown dataset.", "## Model description\n\nMore information needed", "## Intended uses & limitations\n\nMore information needed", "## Training and evaluation data\n\nMore information needed", "## Training procedure", "### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 1e-05\n- train_batch_size: 8\n- eval_batch_size: 64\n- seed: 4\n- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n- lr_scheduler_type: linear\n- num_epochs: 1", "### Training results", "### Framework versions\n\n- Transformers 4.39.3\n- Pytorch 2.2.1\n- Datasets 2.18.0\n- Tokenizers 0.15.2" ]

[ "transformers", "arxiv:1910.09700", "endpoints_compatible", "region:us" ]

[ "1910.09700" ]

[]

[ "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "TAGS\n#transformers #arxiv-1910.09700 #endpoints_compatible #region-us \n", "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "Taxi-v3", "q-learning", "reinforcement-learning", "custom-implementation", "model-index", "region:us" ]

[]

[]

[ "# Q-Learning Agent playing1 Taxi-v3\n This is a trained model of a Q-Learning agent playing Taxi-v3 .\n\n ## Usage" ]

[ "TAGS\n#Taxi-v3 #q-learning #reinforcement-learning #custom-implementation #model-index #region-us \n", "# Q-Learning Agent playing1 Taxi-v3\n This is a trained model of a Q-Learning agent playing Taxi-v3 .\n\n ## Usage" ]

[ "transformers", "safetensors", "llama", "text-generation", "mergekit", "merge", "arxiv:2306.01708", "base_model:microsoft/Orca-2-7b", "base_model:arcee-ai/Patent-Instruct-7b", "base_model:NousResearch/Llama-2-7b-hf", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[ "2306.01708" ]

[]

[ "# merge\n\nThis is a merge of pre-trained language models created using mergekit.", "## Merge Details", "### Merge Method\n\nThis model was merged using the TIES merge method using NousResearch/Llama-2-7b-hf as a base.", "### Models Merged\n\nThe following models were included in the merge:\n* microsoft/Orca-2-7b\n* arcee-ai/Patent-Instruct-7b", "### Configuration\n\nThe following YAML configuration was used to produce this model:" ]

[ "TAGS\n#transformers #safetensors #llama #text-generation #mergekit #merge #arxiv-2306.01708 #base_model-microsoft/Orca-2-7b #base_model-arcee-ai/Patent-Instruct-7b #base_model-NousResearch/Llama-2-7b-hf #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n", "# merge\n\nThis is a merge of pre-trained language models created using mergekit.", "## Merge Details", "### Merge Method\n\nThis model was merged using the TIES merge method using NousResearch/Llama-2-7b-hf as a base.", "### Models Merged\n\nThe following models were included in the merge:\n* microsoft/Orca-2-7b\n* arcee-ai/Patent-Instruct-7b", "### Configuration\n\nThe following YAML configuration was used to produce this model:" ]

[ "transformers", "safetensors", "mixtral", "text-generation", "fp8", "conversational", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[]

[]

[]

[ "TAGS\n#transformers #safetensors #mixtral #text-generation #fp8 #conversational #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n" ]

[ "transformers", "safetensors", "llama", "text-generation", "text-generation-inference", "unsloth", "trl", "sft", "en", "base_model:unsloth/llama-3-8b-bnb-4bit", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "4-bit", "region:us" ]

[]

[ "en" ]

[ "# Uploaded model\n\n- Developed by: 1024m\n- License: apache-2.0\n- Finetuned from model : unsloth/llama-3-8b-bnb-4bit\n\nThis llama model was trained 2x faster with Unsloth and Huggingface's TRL library.\n\n<img src=\"URL width=\"200\"/>" ]

[ "TAGS\n#transformers #safetensors #llama #text-generation #text-generation-inference #unsloth #trl #sft #en #base_model-unsloth/llama-3-8b-bnb-4bit #license-apache-2.0 #autotrain_compatible #endpoints_compatible #4-bit #region-us \n", "# Uploaded model\n\n- Developed by: 1024m\n- License: apache-2.0\n- Finetuned from model : unsloth/llama-3-8b-bnb-4bit\n\nThis llama model was trained 2x faster with Unsloth and Huggingface's TRL library.\n\n<img src=\"URL width=\"200\"/>" ]

[ "transformers", "tensorboard", "safetensors", "whisper", "automatic-speech-recognition", "generated_from_trainer", "dataset:common_voice_11_0", "base_model:openai/whisper-base", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]

[]

[]

[ "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 1e-05\n* train\\_batch\\_size: 8\n* eval\\_batch\\_size: 8\n* seed: 42\n* gradient\\_accumulation\\_steps: 4\n* total\\_train\\_batch\\_size: 32\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* lr\\_scheduler\\_warmup\\_steps: 500\n* training\\_steps: 4000\n* mixed\\_precision\\_training: Native AMP", "### Training results", "### Framework versions\n\n\n* Transformers 4.40.1\n* Pytorch 2.2.1+cu121\n* Datasets 2.19.0\n* Tokenizers 0.19.1" ]

[ "TAGS\n#transformers #tensorboard #safetensors #whisper #automatic-speech-recognition #generated_from_trainer #dataset-common_voice_11_0 #base_model-openai/whisper-base #license-apache-2.0 #model-index #endpoints_compatible #region-us \n", "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 1e-05\n* train\\_batch\\_size: 8\n* eval\\_batch\\_size: 8\n* seed: 42\n* gradient\\_accumulation\\_steps: 4\n* total\\_train\\_batch\\_size: 32\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* lr\\_scheduler\\_warmup\\_steps: 500\n* training\\_steps: 4000\n* mixed\\_precision\\_training: Native AMP", "### Training results", "### Framework versions\n\n\n* Transformers 4.40.1\n* Pytorch 2.2.1+cu121\n* Datasets 2.19.0\n* Tokenizers 0.19.1" ]

[ "transformers", "safetensors", "mistral", "text-generation", "generated_from_trainer", "conversational", "dataset:adalbertojunior/dolphin_pt_test", "base_model:unsloth/mistral-7b-v0.2", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[]

[]

[ "# dolphin-pt-v0.2\n\nThis model is a fine-tuned version of unsloth/mistral-7b-v0.2 on the adalbertojunior/dolphin_pt_test dataset.", "## Model description\n\nMore information needed", "## Intended uses & limitations\n\nMore information needed", "## Training and evaluation data\n\nMore information needed", "## Training procedure", "### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 2e-06\n- train_batch_size: 1\n- eval_batch_size: 1\n- seed: 42\n- gradient_accumulation_steps: 32\n- total_train_batch_size: 32\n- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n- lr_scheduler_type: cosine\n- num_epochs: 1.0", "### Training results", "### Framework versions\n\n- Transformers 4.39.1\n- Pytorch 2.2.1+cu118\n- Datasets 2.18.0\n- Tokenizers 0.15.2" ]

[ "TAGS\n#transformers #safetensors #mistral #text-generation #generated_from_trainer #conversational #dataset-adalbertojunior/dolphin_pt_test #base_model-unsloth/mistral-7b-v0.2 #license-apache-2.0 #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n", "# dolphin-pt-v0.2\n\nThis model is a fine-tuned version of unsloth/mistral-7b-v0.2 on the adalbertojunior/dolphin_pt_test dataset.", "## Model description\n\nMore information needed", "## Intended uses & limitations\n\nMore information needed", "## Training and evaluation data\n\nMore information needed", "## Training procedure", "### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 2e-06\n- train_batch_size: 1\n- eval_batch_size: 1\n- seed: 42\n- gradient_accumulation_steps: 32\n- total_train_batch_size: 32\n- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n- lr_scheduler_type: cosine\n- num_epochs: 1.0", "### Training results", "### Framework versions\n\n- Transformers 4.39.1\n- Pytorch 2.2.1+cu118\n- Datasets 2.18.0\n- Tokenizers 0.15.2" ]

[ "transformers", "safetensors", "marian", "text2text-generation", "generated_from_trainer", "base_model:Helsinki-NLP/opus-mt-de-en", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "region:us" ]

[]

[]

[ "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 2e-05\n* train\\_batch\\_size: 16\n* eval\\_batch\\_size: 16\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* num\\_epochs: 2\n* mixed\\_precision\\_training: Native AMP", "### Training results", "### Framework versions\n\n\n* Transformers 4.38.2\n* Pytorch 2.2.2+cu121\n* Datasets 2.18.0\n* Tokenizers 0.15.2" ]

[ "TAGS\n#transformers #safetensors #marian #text2text-generation #generated_from_trainer #base_model-Helsinki-NLP/opus-mt-de-en #license-apache-2.0 #autotrain_compatible #endpoints_compatible #region-us \n", "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 2e-05\n* train\\_batch\\_size: 16\n* eval\\_batch\\_size: 16\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* num\\_epochs: 2\n* mixed\\_precision\\_training: Native AMP", "### Training results", "### Framework versions\n\n\n* Transformers 4.38.2\n* Pytorch 2.2.2+cu121\n* Datasets 2.18.0\n* Tokenizers 0.15.2" ]

[ "transformers", "tensorboard", "safetensors", "gpt_neox", "text-classification", "generated_from_trainer", "base_model:EleutherAI/pythia-31m", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[]

[]

[ "# robust_llm_pythia-31m_mz-132_EnronSpam_n-its-10\n\nThis model is a fine-tuned version of EleutherAI/pythia-31m on an unknown dataset.", "## Model description\n\nMore information needed", "## Intended uses & limitations\n\nMore information needed", "## Training and evaluation data\n\nMore information needed", "## Training procedure", "### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 1e-05\n- train_batch_size: 8\n- eval_batch_size: 64\n- seed: 0\n- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n- lr_scheduler_type: linear\n- num_epochs: 1", "### Training results", "### Framework versions\n\n- Transformers 4.39.3\n- Pytorch 2.2.1\n- Datasets 2.18.0\n- Tokenizers 0.15.2" ]

[ "TAGS\n#transformers #tensorboard #safetensors #gpt_neox #text-classification #generated_from_trainer #base_model-EleutherAI/pythia-31m #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n", "# robust_llm_pythia-31m_mz-132_EnronSpam_n-its-10\n\nThis model is a fine-tuned version of EleutherAI/pythia-31m on an unknown dataset.", "## Model description\n\nMore information needed", "## Intended uses & limitations\n\nMore information needed", "## Training and evaluation data\n\nMore information needed", "## Training procedure", "### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 1e-05\n- train_batch_size: 8\n- eval_batch_size: 64\n- seed: 0\n- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n- lr_scheduler_type: linear\n- num_epochs: 1", "### Training results", "### Framework versions\n\n- Transformers 4.39.3\n- Pytorch 2.2.1\n- Datasets 2.18.0\n- Tokenizers 0.15.2" ]

[ "diffusers", "tensorboard", "text-to-image", "diffusers-training", "lora", "template:sd-lora", "stable-diffusion-xl", "stable-diffusion-xl-diffusers", "base_model:stabilityai/stable-diffusion-xl-base-1.0", "license:openrail++", "region:us" ]

[]

[]

[ "# SDXL LoRA DreamBooth - ngoga/corgy_waste_LoRA\n\n<Gallery />", "## Model description\n\nThese are ngoga/corgy_waste_LoRA LoRA adaption weights for stabilityai/stable-diffusion-xl-base-1.0.\n\nThe weights were trained using DreamBooth.\n\nLoRA for the text encoder was enabled: False.\n\nSpecial VAE used for training: madebyollin/sdxl-vae-fp16-fix.", "## Trigger words\n\nYou should use a photo of wastes to trigger the image generation.", "## Download model\n\nWeights for this model are available in Safetensors format.\n\nDownload them in the Files & versions tab.", "## Intended uses & limitations", "#### How to use", "#### Limitations and bias\n\n[TODO: provide examples of latent issues and potential remediations]", "## Training details\n\n[TODO: describe the data used to train the model]" ]

[ "TAGS\n#diffusers #tensorboard #text-to-image #diffusers-training #lora #template-sd-lora #stable-diffusion-xl #stable-diffusion-xl-diffusers #base_model-stabilityai/stable-diffusion-xl-base-1.0 #license-openrail++ #region-us \n", "# SDXL LoRA DreamBooth - ngoga/corgy_waste_LoRA\n\n<Gallery />", "## Model description\n\nThese are ngoga/corgy_waste_LoRA LoRA adaption weights for stabilityai/stable-diffusion-xl-base-1.0.\n\nThe weights were trained using DreamBooth.\n\nLoRA for the text encoder was enabled: False.\n\nSpecial VAE used for training: madebyollin/sdxl-vae-fp16-fix.", "## Trigger words\n\nYou should use a photo of wastes to trigger the image generation.", "## Download model\n\nWeights for this model are available in Safetensors format.\n\nDownload them in the Files & versions tab.", "## Intended uses & limitations", "#### How to use", "#### Limitations and bias\n\n[TODO: provide examples of latent issues and potential remediations]", "## Training details\n\n[TODO: describe the data used to train the model]" ]

[ "transformers", "text-generation-inference", "unsloth", "llama", "trl", "en", "base_model:kevin009/flyingllama", "license:apache-2.0", "endpoints_compatible", "region:us" ]

[]

[ "en" ]

[ "# Uploaded model\n\n- Developed by: kevin009\n- License: apache-2.0\n- Finetuned from model : kevin009/flyingllama\n\nThis llama model was trained 2x faster with Unsloth and Huggingface's TRL library.\n\n<img src=\"URL width=\"200\"/>" ]

[ "TAGS\n#transformers #text-generation-inference #unsloth #llama #trl #en #base_model-kevin009/flyingllama #license-apache-2.0 #endpoints_compatible #region-us \n", "# Uploaded model\n\n- Developed by: kevin009\n- License: apache-2.0\n- Finetuned from model : kevin009/flyingllama\n\nThis llama model was trained 2x faster with Unsloth and Huggingface's TRL library.\n\n<img src=\"URL width=\"200\"/>" ]

[ "peft", "tensorboard", "safetensors", "generated_from_trainer", "dataset:scitldr", "base_model:meta-llama/Meta-Llama-3-8B", "license:other", "region:us" ]

[]

[]

[ "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0002\n* train\\_batch\\_size: 2\n* eval\\_batch\\_size: 2\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* lr\\_scheduler\\_warmup\\_steps: 2\n* num\\_epochs: 2\n* mixed\\_precision\\_training: Native AMP", "### Training results", "### Framework versions\n\n\n* PEFT 0.10.0\n* Transformers 4.40.0\n* Pytorch 2.2.1+cu121\n* Datasets 2.19.0\n* Tokenizers 0.19.1" ]

[ "TAGS\n#peft #tensorboard #safetensors #generated_from_trainer #dataset-scitldr #base_model-meta-llama/Meta-Llama-3-8B #license-other #region-us \n", "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0002\n* train\\_batch\\_size: 2\n* eval\\_batch\\_size: 2\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* lr\\_scheduler\\_warmup\\_steps: 2\n* num\\_epochs: 2\n* mixed\\_precision\\_training: Native AMP", "### Training results", "### Framework versions\n\n\n* PEFT 0.10.0\n* Transformers 4.40.0\n* Pytorch 2.2.1+cu121\n* Datasets 2.19.0\n* Tokenizers 0.19.1" ]

[ "transformers", "safetensors", "llama", "text-generation", "4-bit", "AWQ", "autotrain_compatible", "endpoints_compatible", "conversational", "dataset:aqua_rat", "dataset:microsoft/orca-math-word-problems-200k", "dataset:m-a-p/CodeFeedback-Filtered-Instruction", "dataset:anon8231489123/ShareGPT_Vicuna_unfiltered", "license:llama2", "text-generation-inference", "region:us" ]

[]

[]

[ "# abacusai/Llama-3-Smaug-8B AWQ\n\n- Model creator: abacusai\n- Original model: Llama-3-Smaug-8B\n\n!image/png", "## Model Summary\n\nThis model was built using the Smaug recipe for improving performance on real world multi-turn conversations applied to \nmeta-llama/Meta-Llama-3-8B.\n\n- Developed by: Abacus.AI\n- License: URL\n- Finetuned from model: meta-llama/Meta-Llama-3-8B.", "## How to use", "### Install the necessary packages", "### Example Python code", "### About AWQ\n\nAWQ is an efficient, accurate and blazing-fast low-bit weight quantization method, currently supporting 4-bit quantization. Compared to GPTQ, it offers faster Transformers-based inference with equivalent or better quality compared to the most commonly used GPTQ settings.\n\nAWQ models are currently supported on Linux and Windows, with NVidia GPUs only. macOS users: please use GGUF models instead.\n\nIt is supported by:\n\n- Text Generation Webui - using Loader: AutoAWQ\n- vLLM - version 0.2.2 or later for support for all model types.\n- Hugging Face Text Generation Inference (TGI)\n- Transformers version 4.35.0 and later, from any code or client that supports Transformers\n- AutoAWQ - for use from Python code" ]

[ "TAGS\n#transformers #safetensors #llama #text-generation #4-bit #AWQ #autotrain_compatible #endpoints_compatible #conversational #dataset-aqua_rat #dataset-microsoft/orca-math-word-problems-200k #dataset-m-a-p/CodeFeedback-Filtered-Instruction #dataset-anon8231489123/ShareGPT_Vicuna_unfiltered #license-llama2 #text-generation-inference #region-us \n", "# abacusai/Llama-3-Smaug-8B AWQ\n\n- Model creator: abacusai\n- Original model: Llama-3-Smaug-8B\n\n!image/png", "## Model Summary\n\nThis model was built using the Smaug recipe for improving performance on real world multi-turn conversations applied to \nmeta-llama/Meta-Llama-3-8B.\n\n- Developed by: Abacus.AI\n- License: URL\n- Finetuned from model: meta-llama/Meta-Llama-3-8B.", "## How to use", "### Install the necessary packages", "### Example Python code", "### About AWQ\n\nAWQ is an efficient, accurate and blazing-fast low-bit weight quantization method, currently supporting 4-bit quantization. Compared to GPTQ, it offers faster Transformers-based inference with equivalent or better quality compared to the most commonly used GPTQ settings.\n\nAWQ models are currently supported on Linux and Windows, with NVidia GPUs only. macOS users: please use GGUF models instead.\n\nIt is supported by:\n\n- Text Generation Webui - using Loader: AutoAWQ\n- vLLM - version 0.2.2 or later for support for all model types.\n- Hugging Face Text Generation Inference (TGI)\n- Transformers version 4.35.0 and later, from any code or client that supports Transformers\n- AutoAWQ - for use from Python code" ]

[ "transformers", "safetensors", "git", "text-generation", "text-generation-inference", "image-to-text", "en", "dataset:Huthayfa/chest-xrays-indiana-university", "autotrain_compatible", "endpoints_compatible", "region:us" ]

[]

[ "en" ]

[ "### Model Description\n\n\nThis model is designed to convert chest X-ray images into text-based medical reports. It employs image recognition and natural language processing techniques to analyze the X-ray images and generate corresponding textual reports.\n\n\n* *Language(s) (NLP):* English\n* *Finetuned from model:* git-base model\n\n\nUses\n----", "### Direct Use\n\n\nThis model can be directly used by healthcare professionals to expedite the process of generating textual reports from chest X-ray images, facilitating quicker diagnosis and treatment planning.", "### Downstream Use\n\n\nThis model can be fine-tuned for specific hospital systems or integrated into medical software platforms to automate the generation of chest X-ray reports as part of larger medical workflows.", "### Out-of-Scope Use\n\n\nThis model is not suitable for non-medical image-to-text tasks or for generating reports from images other than chest X-rays.\n\n\nBias, Risks, and Limitations\n----------------------------\n\n\nThis model may exhibit biases based on the demographics of the training data, potentially leading to inaccuracies in reports for certain patient groups. Additionally, the model's performance may vary depending on the quality and diversity of the input X-ray images.", "### Recommendations\n\n\nUsers should be aware of the limitations of the model and exercise caution when interpreting its output. It's recommended to validate the generated reports with human expertise, especially in critical medical decision-making processes.\n\n\nHow to Get Started with the Model\n---------------------------------\n\n\nUse the provided code snippet to start using the model for generating chest X-ray reports.\n\n\nLoad model directly\n===================\n\n\nLoad model\n==========\n\n\n\n```\n\nfrom transformers import AutoProcessor, AutoModelForCausalLM\nfrom PIL import Image", "# Open the image file\nimage = URL(\"URL\")\n\nprocessor = AutoProcessor.from_pretrained(\"Huthayfa/Dr.DiagnostiX\")\nmodel = AutoModelForCausalLM.from_pretrained(\"Huthayfa/Dr.DiagnostiX\")]", "# prepare image for the model\ninputs = processor(images=image, return_tensors=\"pt\")\npixel_values = inputs.pixel_values\n\ngenerated_ids = model.generate(pixel_values=pixel_values, max_length=60)\ngenerated_caption = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]\nprint(generated_caption)\n\n\n```\n\nTraining Details\n----------------", "### Training Data\n\n\nThe model was trained on a dataset of annotated chest X-ray images paired with corresponding textual medical reports. The dataset comprises images from various sources, including medical institutions and public datasets.", "#### Preprocessing\n\n\nThe training data underwent preprocessing steps, including normalization, augmentation, and alignment of X-ray images with their respective reports.", "#### Training Hyperparameters\n\n\n* lr = 5e-5\n* batch\\_size = 8\n\n\nEvaluation\n----------", "### Testing Data, Factors & Metrics", "#### Testing Data\n\n\nThe model was evaluated on a separate dataset of chest X-ray images and their associated reports.", "#### Factors\n\n\nEvaluation considered factors such as patient demographics, image quality, and report complexity.", "### Results\n\n\nBatches: 30, Training Loss: 8.067, Test Loss: 6.088\nBatches: 60, Training Loss: 4.697, Test Loss: 2.916\nBatches: 90, Training Loss: 1.807, Test Loss: 0.720\nBatches: 120, Training Loss: 0.525, Test Loss: 0.338\nBatches: 150, Training Loss: 0.328, Test Loss: 0.268\nBatches: 180, Training Loss: 0.277, Test Loss: 0.237\nBatches: 210, Training Loss: 0.254, Test Loss: 0.216\nBatches: 240, Training Loss: 0.234, Test Loss: 0.201\nBatches: 270, Training Loss: 0.203, Test Loss: 0.191\nBatches: 300, Training Loss: 0.217, Test Loss: 0.182" ]

[ "TAGS\n#transformers #safetensors #git #text-generation #text-generation-inference #image-to-text #en #dataset-Huthayfa/chest-xrays-indiana-university #autotrain_compatible #endpoints_compatible #region-us \n", "### Model Description\n\n\nThis model is designed to convert chest X-ray images into text-based medical reports. It employs image recognition and natural language processing techniques to analyze the X-ray images and generate corresponding textual reports.\n\n\n* *Language(s) (NLP):* English\n* *Finetuned from model:* git-base model\n\n\nUses\n----", "### Direct Use\n\n\nThis model can be directly used by healthcare professionals to expedite the process of generating textual reports from chest X-ray images, facilitating quicker diagnosis and treatment planning.", "### Downstream Use\n\n\nThis model can be fine-tuned for specific hospital systems or integrated into medical software platforms to automate the generation of chest X-ray reports as part of larger medical workflows.", "### Out-of-Scope Use\n\n\nThis model is not suitable for non-medical image-to-text tasks or for generating reports from images other than chest X-rays.\n\n\nBias, Risks, and Limitations\n----------------------------\n\n\nThis model may exhibit biases based on the demographics of the training data, potentially leading to inaccuracies in reports for certain patient groups. Additionally, the model's performance may vary depending on the quality and diversity of the input X-ray images.", "### Recommendations\n\n\nUsers should be aware of the limitations of the model and exercise caution when interpreting its output. It's recommended to validate the generated reports with human expertise, especially in critical medical decision-making processes.\n\n\nHow to Get Started with the Model\n---------------------------------\n\n\nUse the provided code snippet to start using the model for generating chest X-ray reports.\n\n\nLoad model directly\n===================\n\n\nLoad model\n==========\n\n\n\n```\n\nfrom transformers import AutoProcessor, AutoModelForCausalLM\nfrom PIL import Image", "# Open the image file\nimage = URL(\"URL\")\n\nprocessor = AutoProcessor.from_pretrained(\"Huthayfa/Dr.DiagnostiX\")\nmodel = AutoModelForCausalLM.from_pretrained(\"Huthayfa/Dr.DiagnostiX\")]", "# prepare image for the model\ninputs = processor(images=image, return_tensors=\"pt\")\npixel_values = inputs.pixel_values\n\ngenerated_ids = model.generate(pixel_values=pixel_values, max_length=60)\ngenerated_caption = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]\nprint(generated_caption)\n\n\n```\n\nTraining Details\n----------------", "### Training Data\n\n\nThe model was trained on a dataset of annotated chest X-ray images paired with corresponding textual medical reports. The dataset comprises images from various sources, including medical institutions and public datasets.", "#### Preprocessing\n\n\nThe training data underwent preprocessing steps, including normalization, augmentation, and alignment of X-ray images with their respective reports.", "#### Training Hyperparameters\n\n\n* lr = 5e-5\n* batch\\_size = 8\n\n\nEvaluation\n----------", "### Testing Data, Factors & Metrics", "#### Testing Data\n\n\nThe model was evaluated on a separate dataset of chest X-ray images and their associated reports.", "#### Factors\n\n\nEvaluation considered factors such as patient demographics, image quality, and report complexity.", "### Results\n\n\nBatches: 30, Training Loss: 8.067, Test Loss: 6.088\nBatches: 60, Training Loss: 4.697, Test Loss: 2.916\nBatches: 90, Training Loss: 1.807, Test Loss: 0.720\nBatches: 120, Training Loss: 0.525, Test Loss: 0.338\nBatches: 150, Training Loss: 0.328, Test Loss: 0.268\nBatches: 180, Training Loss: 0.277, Test Loss: 0.237\nBatches: 210, Training Loss: 0.254, Test Loss: 0.216\nBatches: 240, Training Loss: 0.234, Test Loss: 0.201\nBatches: 270, Training Loss: 0.203, Test Loss: 0.191\nBatches: 300, Training Loss: 0.217, Test Loss: 0.182" ]

[ "transformers", "safetensors", "llama", "text-generation", "conversational", "dataset:mlabonne/orpo-dpo-mix-40k", "license:mit", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[]

[]

[ "# Model Card for Model ID", "## Model Details\nFine tuned version of Llama3 8B. 4 bit quantized with Bits and Bytes. Fine tuned for 1 epoch for 58 hours on an Nvidia a5000.", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: Michael Svendsen\n- Finetuned from model [Llama3-8b]:" ]

[ "TAGS\n#transformers #safetensors #llama #text-generation #conversational #dataset-mlabonne/orpo-dpo-mix-40k #license-mit #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n", "# Model Card for Model ID", "## Model Details\nFine tuned version of Llama3 8B. 4 bit quantized with Bits and Bytes. Fine tuned for 1 epoch for 58 hours on an Nvidia a5000.", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: Michael Svendsen\n- Finetuned from model [Llama3-8b]:" ]

[ "transformers", "safetensors", "conditional_detr", "object-detection", "vision", "generated_from_trainer", "base_model:microsoft/conditional-detr-resnet-50", "license:apache-2.0", "endpoints_compatible", "region:us" ]

[]

[]

[ "# microsoft-conditional-detr-resnet-50-finetuned-10k-cppe5-manual-pad\n\nThis model is a fine-tuned version of microsoft/conditional-detr-resnet-50 on the cppe-5 dataset.", "## Model description\n\nMore information needed", "## Intended uses & limitations\n\nMore information needed", "## Training and evaluation data\n\nMore information needed", "## Training procedure", "### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 5e-05\n- train_batch_size: 8\n- eval_batch_size: 1\n- seed: 1337\n- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n- lr_scheduler_type: linear\n- num_epochs: 100.0\n- mixed_precision_training: Native AMP", "### Framework versions\n\n- Transformers 4.41.0.dev0\n- Pytorch 1.13.0+cu117\n- Datasets 2.18.0\n- Tokenizers 0.19.0" ]

[ "TAGS\n#transformers #safetensors #conditional_detr #object-detection #vision #generated_from_trainer #base_model-microsoft/conditional-detr-resnet-50 #license-apache-2.0 #endpoints_compatible #region-us \n", "# microsoft-conditional-detr-resnet-50-finetuned-10k-cppe5-manual-pad\n\nThis model is a fine-tuned version of microsoft/conditional-detr-resnet-50 on the cppe-5 dataset.", "## Model description\n\nMore information needed", "## Intended uses & limitations\n\nMore information needed", "## Training and evaluation data\n\nMore information needed", "## Training procedure", "### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 5e-05\n- train_batch_size: 8\n- eval_batch_size: 1\n- seed: 1337\n- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n- lr_scheduler_type: linear\n- num_epochs: 100.0\n- mixed_precision_training: Native AMP", "### Framework versions\n\n- Transformers 4.41.0.dev0\n- Pytorch 1.13.0+cu117\n- Datasets 2.18.0\n- Tokenizers 0.19.0" ]

[ "transformers", "pytorch", "safetensors", "llama", "text-generation", "conversational", "en", "dataset:cognitivecomputations/Dolphin-2.9", "dataset:teknium/OpenHermes-2.5", "dataset:m-a-p/CodeFeedback-Filtered-Instruction", "dataset:cognitivecomputations/dolphin-coder", "dataset:cognitivecomputations/samantha-data", "dataset:HuggingFaceH4/ultrachat_200k", "dataset:microsoft/orca-math-word-problems-200k", "dataset:abacusai/SystemChat-1.1", "dataset:Locutusque/function-calling-chatml", "dataset:internlm/Agent-FLAN", "license:llama3", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[]

[ "en" ]

[ "# Dolphin 2.9 Llama 3 70b \n\nCurated and trained by Eric Hartford, Lucas Atkins, Fernando Fernandes, and with help from the community of Cognitive Computations\n\nDiscord: URL\n\n<img src=\"URL width=\"600\" />\n\nOur appreciation for the sponsors of Dolphin 2.9:\n- Crusoe Cloud - provided excellent on-demand 8xH100 node\n\nThis model is based on Llama-3-70b, and is governed by META LLAMA 3 COMMUNITY LICENSE AGREEMENT\n\nThe base model has 8k context, and the qLoRA fine-tuning was with 8k sequence length.\n\nIt took 2.5 days on 8xH100 node provided by Crusoe Cloud\n\nThis model was trained FFT on all parameters, using ChatML prompt template format.\n\nexample:\n\n\n\nDolphin-2.9 has a variety of instruction, conversational, and coding skills. It also has initial agentic abilities and supports function calling.\n\nDolphin is uncensored. I have filtered the dataset to remove alignment and bias. This makes the model more compliant. You are advised to implement your own alignment layer before exposing the model as a service. It will be highly compliant with any requests, even unethical ones. Please read my blog post about uncensored models. URL You are responsible for any content you create using this model. Enjoy responsibly.\n\nDolphin is licensed according to Meta's Llama license. I grant permission for any use, including commercial, that falls within accordance with Meta's Llama-3 license. Dolphin was trained on data generated from GPT4, among other models.\n\n<img src=\"URL alt=\"Built with Axolotl\" width=\"200\" height=\"32\"/>", "## Evals\n\n!image/png", "## Quants\n\n- URL\n- URL\n- URL\n- URL" ]

[ "TAGS\n#transformers #pytorch #safetensors #llama #text-generation #conversational #en #dataset-cognitivecomputations/Dolphin-2.9 #dataset-teknium/OpenHermes-2.5 #dataset-m-a-p/CodeFeedback-Filtered-Instruction #dataset-cognitivecomputations/dolphin-coder #dataset-cognitivecomputations/samantha-data #dataset-HuggingFaceH4/ultrachat_200k #dataset-microsoft/orca-math-word-problems-200k #dataset-abacusai/SystemChat-1.1 #dataset-Locutusque/function-calling-chatml #dataset-internlm/Agent-FLAN #license-llama3 #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n", "# Dolphin 2.9 Llama 3 70b \n\nCurated and trained by Eric Hartford, Lucas Atkins, Fernando Fernandes, and with help from the community of Cognitive Computations\n\nDiscord: URL\n\n<img src=\"URL width=\"600\" />\n\nOur appreciation for the sponsors of Dolphin 2.9:\n- Crusoe Cloud - provided excellent on-demand 8xH100 node\n\nThis model is based on Llama-3-70b, and is governed by META LLAMA 3 COMMUNITY LICENSE AGREEMENT\n\nThe base model has 8k context, and the qLoRA fine-tuning was with 8k sequence length.\n\nIt took 2.5 days on 8xH100 node provided by Crusoe Cloud\n\nThis model was trained FFT on all parameters, using ChatML prompt template format.\n\nexample:\n\n\n\nDolphin-2.9 has a variety of instruction, conversational, and coding skills. It also has initial agentic abilities and supports function calling.\n\nDolphin is uncensored. I have filtered the dataset to remove alignment and bias. This makes the model more compliant. You are advised to implement your own alignment layer before exposing the model as a service. It will be highly compliant with any requests, even unethical ones. Please read my blog post about uncensored models. URL You are responsible for any content you create using this model. Enjoy responsibly.\n\nDolphin is licensed according to Meta's Llama license. I grant permission for any use, including commercial, that falls within accordance with Meta's Llama-3 license. Dolphin was trained on data generated from GPT4, among other models.\n\n<img src=\"URL alt=\"Built with Axolotl\" width=\"200\" height=\"32\"/>", "## Evals\n\n!image/png", "## Quants\n\n- URL\n- URL\n- URL\n- URL" ]

[ "transformers", "tensorboard", "safetensors", "bert", "text-classification", "generated_from_trainer", "base_model:bert-base-uncased", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "region:us" ]

[]

[]

[ "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 2e-05\n* train\\_batch\\_size: 16\n* eval\\_batch\\_size: 16\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* num\\_epochs: 2", "### Training results", "### Framework versions\n\n\n* Transformers 4.40.0\n* Pytorch 2.2.1+cu121\n* Datasets 2.19.0\n* Tokenizers 0.19.1" ]

[ "TAGS\n#transformers #tensorboard #safetensors #bert #text-classification #generated_from_trainer #base_model-bert-base-uncased #license-apache-2.0 #autotrain_compatible #endpoints_compatible #region-us \n", "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 2e-05\n* train\\_batch\\_size: 16\n* eval\\_batch\\_size: 16\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* num\\_epochs: 2", "### Training results", "### Framework versions\n\n\n* Transformers 4.40.0\n* Pytorch 2.2.1+cu121\n* Datasets 2.19.0\n* Tokenizers 0.19.1" ]

[ "transformers", "safetensors", "mistral", "text-generation", "4-bit", "AWQ", "autotrain_compatible", "endpoints_compatible", "medical", "conversational", "en", "dataset:Open-Orca/OpenOrca", "dataset:pubmed", "dataset:medmcqa", "dataset:maximegmd/medqa_alpaca_format", "base_model:mistralai/Mistral-7B-v0.1", "license:apache-2.0", "text-generation-inference", "region:us" ]

[]

[ "en" ]

[ "# internistai/base-7b-v0.2 AWQ\n\n- Model creator: internistai\n- Original model: base-7b-v0.2\n\n<img width=30% src=\"assets_logo.png\" alt=\"logo\" title=\"logo\">", "## Model Summary\n\nURL 7b is a medical domain large language model trained by medical doctors to demonstrate the benefits of a physician-in-the-loop approach. The training data was carefully curated by medical doctors to ensure clinical relevance and required quality for clinical practice.\n\nWith this 7b model we release the first 7b model to score above the 60% pass threshold on MedQA (USMLE) and outperfoms models of similar size accross most medical evaluations.\n\nThis model serves as a proof of concept and larger models trained on a larger corpus of medical literature are planned. Do not hesitate to reach out to us if you would like to sponsor some compute to speed up this training.", "## How to use", "### Install the necessary packages", "### Example Python code", "### About AWQ\n\nAWQ is an efficient, accurate and blazing-fast low-bit weight quantization method, currently supporting 4-bit quantization. Compared to GPTQ, it offers faster Transformers-based inference with equivalent or better quality compared to the most commonly used GPTQ settings.\n\nAWQ models are currently supported on Linux and Windows, with NVidia GPUs only. macOS users: please use GGUF models instead.\n\nIt is supported by:\n\n- Text Generation Webui - using Loader: AutoAWQ\n- vLLM - version 0.2.2 or later for support for all model types.\n- Hugging Face Text Generation Inference (TGI)\n- Transformers version 4.35.0 and later, from any code or client that supports Transformers\n- AutoAWQ - for use from Python code" ]

[ "TAGS\n#transformers #safetensors #mistral #text-generation #4-bit #AWQ #autotrain_compatible #endpoints_compatible #medical #conversational #en #dataset-Open-Orca/OpenOrca #dataset-pubmed #dataset-medmcqa #dataset-maximegmd/medqa_alpaca_format #base_model-mistralai/Mistral-7B-v0.1 #license-apache-2.0 #text-generation-inference #region-us \n", "# internistai/base-7b-v0.2 AWQ\n\n- Model creator: internistai\n- Original model: base-7b-v0.2\n\n<img width=30% src=\"assets_logo.png\" alt=\"logo\" title=\"logo\">", "## Model Summary\n\nURL 7b is a medical domain large language model trained by medical doctors to demonstrate the benefits of a physician-in-the-loop approach. The training data was carefully curated by medical doctors to ensure clinical relevance and required quality for clinical practice.\n\nWith this 7b model we release the first 7b model to score above the 60% pass threshold on MedQA (USMLE) and outperfoms models of similar size accross most medical evaluations.\n\nThis model serves as a proof of concept and larger models trained on a larger corpus of medical literature are planned. Do not hesitate to reach out to us if you would like to sponsor some compute to speed up this training.", "## How to use", "### Install the necessary packages", "### Example Python code", "### About AWQ\n\nAWQ is an efficient, accurate and blazing-fast low-bit weight quantization method, currently supporting 4-bit quantization. Compared to GPTQ, it offers faster Transformers-based inference with equivalent or better quality compared to the most commonly used GPTQ settings.\n\nAWQ models are currently supported on Linux and Windows, with NVidia GPUs only. macOS users: please use GGUF models instead.\n\nIt is supported by:\n\n- Text Generation Webui - using Loader: AutoAWQ\n- vLLM - version 0.2.2 or later for support for all model types.\n- Hugging Face Text Generation Inference (TGI)\n- Transformers version 4.35.0 and later, from any code or client that supports Transformers\n- AutoAWQ - for use from Python code" ]

[ "peft", "tensorboard", "safetensors", "musicgen_melody", "text-to-audio", "ylacombe/tiny-punk", "generated_from_trainer", "base_model:facebook/musicgen-melody", "license:cc-by-nc-4.0", "region:us" ]

[]

[]

[ "# musicgen-melody-lora-punk\n\nThis model is a fine-tuned version of facebook/musicgen-melody on the YLACOMBE/TINY-PUNK - DEFAULT dataset.\nIt achieves the following results on the evaluation set:\n- Loss: 4.7288\n- Clap: -0.0429", "## Model description\n\nMore information needed", "## Intended uses & limitations\n\nMore information needed", "## Training and evaluation data\n\nMore information needed", "## Training procedure", "### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 0.0002\n- train_batch_size: 2\n- eval_batch_size: 1\n- seed: 456\n- gradient_accumulation_steps: 8\n- total_train_batch_size: 16\n- optimizer: Adam with betas=(0.9,0.99) and epsilon=1e-08\n- lr_scheduler_type: linear\n- num_epochs: 4.0\n- mixed_precision_training: Native AMP", "### Training results", "### Framework versions\n\n- PEFT 0.10.0\n- Transformers 4.41.0.dev0\n- Pytorch 2.2.1+cu121\n- Datasets 2.19.0\n- Tokenizers 0.19.1" ]

[ "TAGS\n#peft #tensorboard #safetensors #musicgen_melody #text-to-audio #ylacombe/tiny-punk #generated_from_trainer #base_model-facebook/musicgen-melody #license-cc-by-nc-4.0 #region-us \n", "# musicgen-melody-lora-punk\n\nThis model is a fine-tuned version of facebook/musicgen-melody on the YLACOMBE/TINY-PUNK - DEFAULT dataset.\nIt achieves the following results on the evaluation set:\n- Loss: 4.7288\n- Clap: -0.0429", "## Model description\n\nMore information needed", "## Intended uses & limitations\n\nMore information needed", "## Training and evaluation data\n\nMore information needed", "## Training procedure", "### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 0.0002\n- train_batch_size: 2\n- eval_batch_size: 1\n- seed: 456\n- gradient_accumulation_steps: 8\n- total_train_batch_size: 16\n- optimizer: Adam with betas=(0.9,0.99) and epsilon=1e-08\n- lr_scheduler_type: linear\n- num_epochs: 4.0\n- mixed_precision_training: Native AMP", "### Training results", "### Framework versions\n\n- PEFT 0.10.0\n- Transformers 4.41.0.dev0\n- Pytorch 2.2.1+cu121\n- Datasets 2.19.0\n- Tokenizers 0.19.1" ]

[ "transformers", "tensorboard", "safetensors", "distilbert", "fill-mask", "generated_from_trainer", "base_model:distilbert-base-uncased", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "region:us" ]

[]

[]

[ "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 2e-05\n* train\\_batch\\_size: 64\n* eval\\_batch\\_size: 64\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* num\\_epochs: 3.0\n* mixed\\_precision\\_training: Native AMP", "### Training results", "### Framework versions\n\n\n* Transformers 4.40.0\n* Pytorch 2.2.1+cu121\n* Datasets 2.19.0\n* Tokenizers 0.19.1" ]

[ "TAGS\n#transformers #tensorboard #safetensors #distilbert #fill-mask #generated_from_trainer #base_model-distilbert-base-uncased #license-apache-2.0 #autotrain_compatible #endpoints_compatible #region-us \n", "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 2e-05\n* train\\_batch\\_size: 64\n* eval\\_batch\\_size: 64\n* seed: 42\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: linear\n* num\\_epochs: 3.0\n* mixed\\_precision\\_training: Native AMP", "### Training results", "### Framework versions\n\n\n* Transformers 4.40.0\n* Pytorch 2.2.1+cu121\n* Datasets 2.19.0\n* Tokenizers 0.19.1" ]

[ "stable-baselines3", "LunarLander-v2", "deep-reinforcement-learning", "reinforcement-learning", "model-index", "region:us" ]

[]

[]

[ "# PPO Agent playing LunarLander-v2\nThis is a trained model of a PPO agent playing LunarLander-v2\nusing the stable-baselines3 library.", "## Usage (with Stable-baselines3)\nTODO: Add your code" ]

[ "TAGS\n#stable-baselines3 #LunarLander-v2 #deep-reinforcement-learning #reinforcement-learning #model-index #region-us \n", "# PPO Agent playing LunarLander-v2\nThis is a trained model of a PPO agent playing LunarLander-v2\nusing the stable-baselines3 library.", "## Usage (with Stable-baselines3)\nTODO: Add your code" ]

[ "transformers", "pytorch", "safetensors", "llama", "text-generation", "conversational", "en", "dataset:cognitivecomputations/Dolphin-2.9", "dataset:teknium/OpenHermes-2.5", "dataset:m-a-p/CodeFeedback-Filtered-Instruction", "dataset:cognitivecomputations/dolphin-coder", "dataset:cognitivecomputations/samantha-data", "dataset:HuggingFaceH4/ultrachat_200k", "dataset:microsoft/orca-math-word-problems-200k", "dataset:abacusai/SystemChat-1.1", "dataset:Locutusque/function-calling-chatml", "dataset:internlm/Agent-FLAN", "license:llama3", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "3-bit", "region:us" ]

[]

[ "en" ]

[ "# Dolphin 2.9 Llama 3 70b \n\nCurated and trained by Eric Hartford, Lucas Atkins, Fernando Fernandes, and with help from the community of Cognitive Computations\n\nDiscord: URL\n\n<img src=\"URL width=\"600\" />\n\nOur appreciation for the sponsors of Dolphin 2.9:\n- Crusoe Cloud - provided excellent on-demand 8xH100 node\n\nThis model is based on Llama-3-70b, and is governed by META LLAMA 3 COMMUNITY LICENSE AGREEMENT\n\nThe base model has 8k context, and the qLoRA fine-tuning was with 8k sequence length.\n\nIt took 2.5 days on 8xH100 node provided by Crusoe Cloud\n\nThis model was trained FFT on all parameters, using ChatML prompt template format.\n\nexample:\n\n\n\nDolphin-2.9 has a variety of instruction, conversational, and coding skills. It also has initial agentic abilities and supports function calling.\n\nDolphin is uncensored. I have filtered the dataset to remove alignment and bias. This makes the model more compliant. You are advised to implement your own alignment layer before exposing the model as a service. It will be highly compliant with any requests, even unethical ones. Please read my blog post about uncensored models. URL You are responsible for any content you create using this model. Enjoy responsibly.\n\nDolphin is licensed according to Meta's Llama license. I grant permission for any use, including commercial, that falls within accordance with Meta's Llama-3 license. Dolphin was trained on data generated from GPT4, among other models.\n\n<img src=\"URL alt=\"Built with Axolotl\" width=\"200\" height=\"32\"/>", "## Evals\n\n!image/png", "## Quants\n\n- URL\n- URL\n- URL\n- URL" ]

[ "TAGS\n#transformers #pytorch #safetensors #llama #text-generation #conversational #en #dataset-cognitivecomputations/Dolphin-2.9 #dataset-teknium/OpenHermes-2.5 #dataset-m-a-p/CodeFeedback-Filtered-Instruction #dataset-cognitivecomputations/dolphin-coder #dataset-cognitivecomputations/samantha-data #dataset-HuggingFaceH4/ultrachat_200k #dataset-microsoft/orca-math-word-problems-200k #dataset-abacusai/SystemChat-1.1 #dataset-Locutusque/function-calling-chatml #dataset-internlm/Agent-FLAN #license-llama3 #autotrain_compatible #endpoints_compatible #text-generation-inference #3-bit #region-us \n", "# Dolphin 2.9 Llama 3 70b \n\nCurated and trained by Eric Hartford, Lucas Atkins, Fernando Fernandes, and with help from the community of Cognitive Computations\n\nDiscord: URL\n\n<img src=\"URL width=\"600\" />\n\nOur appreciation for the sponsors of Dolphin 2.9:\n- Crusoe Cloud - provided excellent on-demand 8xH100 node\n\nThis model is based on Llama-3-70b, and is governed by META LLAMA 3 COMMUNITY LICENSE AGREEMENT\n\nThe base model has 8k context, and the qLoRA fine-tuning was with 8k sequence length.\n\nIt took 2.5 days on 8xH100 node provided by Crusoe Cloud\n\nThis model was trained FFT on all parameters, using ChatML prompt template format.\n\nexample:\n\n\n\nDolphin-2.9 has a variety of instruction, conversational, and coding skills. It also has initial agentic abilities and supports function calling.\n\nDolphin is uncensored. I have filtered the dataset to remove alignment and bias. This makes the model more compliant. You are advised to implement your own alignment layer before exposing the model as a service. It will be highly compliant with any requests, even unethical ones. Please read my blog post about uncensored models. URL You are responsible for any content you create using this model. Enjoy responsibly.\n\nDolphin is licensed according to Meta's Llama license. I grant permission for any use, including commercial, that falls within accordance with Meta's Llama-3 license. Dolphin was trained on data generated from GPT4, among other models.\n\n<img src=\"URL alt=\"Built with Axolotl\" width=\"200\" height=\"32\"/>", "## Evals\n\n!image/png", "## Quants\n\n- URL\n- URL\n- URL\n- URL" ]

[ "transformers", "tensorboard", "safetensors", "gpt_neox", "text-classification", "generated_from_trainer", "base_model:EleutherAI/pythia-1b", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[]

[]

[ "# robust_llm_pythia-1b_mz-130_PasswordMatch_n-its-10-seed-3\n\nThis model is a fine-tuned version of EleutherAI/pythia-1b on an unknown dataset.", "## Model description\n\nMore information needed", "## Intended uses & limitations\n\nMore information needed", "## Training and evaluation data\n\nMore information needed", "## Training procedure", "### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 1e-05\n- train_batch_size: 8\n- eval_batch_size: 64\n- seed: 3\n- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n- lr_scheduler_type: linear\n- num_epochs: 1", "### Training results", "### Framework versions\n\n- Transformers 4.39.3\n- Pytorch 2.2.1\n- Datasets 2.18.0\n- Tokenizers 0.15.2" ]

[ "TAGS\n#transformers #tensorboard #safetensors #gpt_neox #text-classification #generated_from_trainer #base_model-EleutherAI/pythia-1b #license-apache-2.0 #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n", "# robust_llm_pythia-1b_mz-130_PasswordMatch_n-its-10-seed-3\n\nThis model is a fine-tuned version of EleutherAI/pythia-1b on an unknown dataset.", "## Model description\n\nMore information needed", "## Intended uses & limitations\n\nMore information needed", "## Training and evaluation data\n\nMore information needed", "## Training procedure", "### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 1e-05\n- train_batch_size: 8\n- eval_batch_size: 64\n- seed: 3\n- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n- lr_scheduler_type: linear\n- num_epochs: 1", "### Training results", "### Framework versions\n\n- Transformers 4.39.3\n- Pytorch 2.2.1\n- Datasets 2.18.0\n- Tokenizers 0.15.2" ]

[ "peft", "tensorboard", "safetensors", "trl", "sft", "generated_from_trainer", "base_model:petals-team/falcon-rw-1b", "license:apache-2.0", "region:us" ]

[]

[]

[ "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0002\n* train\\_batch\\_size: 2\n* eval\\_batch\\_size: 8\n* seed: 42\n* gradient\\_accumulation\\_steps: 2\n* total\\_train\\_batch\\_size: 4\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: cosine\n* lr\\_scheduler\\_warmup\\_ratio: 0.03\n* training\\_steps: 320", "### Training results", "### Framework versions\n\n\n* PEFT 0.10.0\n* Transformers 4.40.0\n* Pytorch 2.2.1+cu121\n* Datasets 2.19.0\n* Tokenizers 0.19.1" ]

[ "TAGS\n#peft #tensorboard #safetensors #trl #sft #generated_from_trainer #base_model-petals-team/falcon-rw-1b #license-apache-2.0 #region-us \n", "### Training hyperparameters\n\n\nThe following hyperparameters were used during training:\n\n\n* learning\\_rate: 0.0002\n* train\\_batch\\_size: 2\n* eval\\_batch\\_size: 8\n* seed: 42\n* gradient\\_accumulation\\_steps: 2\n* total\\_train\\_batch\\_size: 4\n* optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n* lr\\_scheduler\\_type: cosine\n* lr\\_scheduler\\_warmup\\_ratio: 0.03\n* training\\_steps: 320", "### Training results", "### Framework versions\n\n\n* PEFT 0.10.0\n* Transformers 4.40.0\n* Pytorch 2.2.1+cu121\n* Datasets 2.19.0\n* Tokenizers 0.19.1" ]

[ "transformers", "onnx", "llama", "text-generation", "facebook", "meta", "pytorch", "llama-3", "conversational", "en", "license:other", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]

[]

[ "en" ]

[ "### Use with transformers\n\n\nSee the snippet below for usage with Transformers:", "### Use with 'llama3'\n\n\nPlease, follow the instructions in the repository\n\n\nTo download Original checkpoints, see the example command below leveraging 'huggingface-cli':\n\n\nFor Hugging Face support, we recommend using transformers or TGI, but a similar command works.\n\n\nHardware and Software\n---------------------\n\n\nTraining Factors We used custom training libraries, Meta's Research SuperCluster, and production clusters for pretraining. Fine-tuning, annotation, and evaluation were also performed on third-party cloud compute.\n\n\nCarbon Footprint Pretraining utilized a cumulative 7.7M GPU hours of computation on hardware of type H100-80GB (TDP of 700W). Estimated total emissions were 2290 tCO2eq, 100% of which were offset by Meta’s sustainability program.\n\n\n\nCO2 emissions during pre-training. Time: total GPU time required for training each model. Power Consumption: peak power capacity per GPU device for the GPUs used adjusted for power usage efficiency. 100% of the emissions are directly offset by Meta's sustainability program, and because we are openly releasing these models, the pretraining costs do not need to be incurred by others.\n\n\nTraining Data\n-------------\n\n\nOverview Llama 3 was pretrained on over 15 trillion tokens of data from publicly available sources. The fine-tuning data includes publicly available instruction datasets, as well as over 10M human-annotated examples. Neither the pretraining nor the fine-tuning datasets include Meta user data.\n\n\nData Freshness The pretraining data has a cutoff of March 2023 for the 7B and December 2023 for the 70B models respectively.\n\n\nBenchmarks\n----------\n\n\nIn this section, we report the results for Llama 3 models on standard automatic benchmarks. For all the evaluations, we use our internal evaluations library. For details on the methodology see here.", "### Base pretrained models", "### Instruction tuned models", "### Responsibility & Safety\n\n\nWe believe that an open approach to AI leads to better, safer products, faster innovation, and a bigger overall market. We are committed to Responsible AI development and took a series of steps to limit misuse and harm and support the open source community.\n\n\nFoundation models are widely capable technologies that are built to be used for a diverse range of applications. They are not designed to meet every developer preference on safety levels for all use cases, out-of-the-box, as those by their nature will differ across different applications.\n\n\nRather, responsible LLM-application deployment is achieved by implementing a series of safety best practices throughout the development of such applications, from the model pre-training, fine-tuning and the deployment of systems composed of safeguards to tailor the safety needs specifically to the use case and audience.\n\n\nAs part of the Llama 3 release, we updated our Responsible Use Guide to outline the steps and best practices for developers to implement model and system level safety for their application. We also provide a set of resources including Meta Llama Guard 2 and Code Shield safeguards. These tools have proven to drastically reduce residual risks of LLM Systems, while maintaining a high level of helpfulness. We encourage developers to tune and deploy these safeguards according to their needs and we provide a reference implementation to get you started.", "#### Llama 3-Instruct\n\n\nAs outlined in the Responsible Use Guide, some trade-off between model helpfulness and model alignment is likely unavoidable. Developers should exercise discretion about how to weigh the benefits of alignment and helpfulness for their specific use case and audience. Developers should be mindful of residual risks when using Llama models and leverage additional safety tools as needed to reach the right safety bar for their use case.\n\n\nSafety\n\n\nFor our instruction tuned model, we conducted extensive red teaming exercises, performed adversarial evaluations and implemented safety mitigations techniques to lower residual risks. As with any Large Language Model, residual risks will likely remain and we recommend that developers assess these risks in the context of their use case. In parallel, we are working with the community to make AI safety benchmark standards transparent, rigorous and interpretable.\n\n\nRefusals\n\n\nIn addition to residual risks, we put a great emphasis on model refusals to benign prompts. Over-refusing not only can impact the user experience but could even be harmful in certain contexts as well. We’ve heard the feedback from the developer community and improved our fine tuning to ensure that Llama 3 is significantly less likely to falsely refuse to answer prompts than Llama 2.\n\n\nWe built internal benchmarks and developed mitigations to limit false refusals making Llama 3 our most helpful model to date.", "#### Responsible release\n\n\nIn addition to responsible use considerations outlined above, we followed a rigorous process that requires us to take extra measures against misuse and critical risks before we make our release decision.\n\n\nMisuse\n\n\nIf you access or use Llama 3, you agree to the Acceptable Use Policy. The most recent copy of this policy can be found at URL", "#### Critical risks\n\n\nCBRNE (Chemical, Biological, Radiological, Nuclear, and high yield Explosives)\n\n\nWe have conducted a two fold assessment of the safety of the model in this area:\n\n\n* Iterative testing during model training to assess the safety of responses related to CBRNE threats and other adversarial risks.\n* Involving external CBRNE experts to conduct an uplift test assessing the ability of the model to accurately provide expert knowledge and reduce barriers to potential CBRNE misuse, by reference to what can be achieved using web search (without the model).", "### Cyber Security\n\n\nWe have evaluated Llama 3 with CyberSecEval, Meta’s cybersecurity safety eval suite, measuring Llama 3’s propensity to suggest insecure code when used as a coding assistant, and Llama 3’s propensity to comply with requests to help carry out cyber attacks, where attacks are defined by the industry standard MITRE ATT&CK cyber attack ontology. On our insecure coding and cyber attacker helpfulness tests, Llama 3 behaved in the same range or safer than models of equivalent coding capability.", "### Child Safety\n\n\nChild Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences.", "### Community\n\n\nGenerative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership in AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository.\n\n\nFinally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community.\n\n\nEthical Considerations and Limitations\n--------------------------------------\n\n\nThe core values of Llama 3 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress.\n\n\nBut Llama 3 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has been in English, and has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3 models, developers should perform safety testing and tuning tailored to their specific applications of the model. As outlined in the Responsible Use Guide, we recommend incorporating Purple Llama solutions into your workflows and specifically Llama Guard which provides a base model to filter input and output prompts to layer system-level safety on top of model-level safety.\n\n\nPlease see the Responsible Use Guide available at URL\n\n\ninstructions\n\n\n@article{llama3modelcard,\n\n\ntitle={Llama 3 Model Card},\n\n\nauthor={AI@Meta},\n\n\nyear={2024},\n\n\nurl = {URL\n\n\n}\n\n\nContributors\n------------\n\n\nAaditya Singh; Aaron Grattafiori; Abhimanyu Dubey; Abhinav Jauhri; Abhinav Pandey; Abhishek Kadian; Adam Kelsey; Adi Gangidi; Ahmad Al-Dahle; Ahuva Goldstand; Aiesha Letman; Ajay Menon; Akhil Mathur; Alan Schelten; Alex Vaughan; Amy Yang; Andrei Lupu; Andres Alvarado; Andrew Gallagher; Andrew Gu; Andrew Ho; Andrew Poulton; Andrew Ryan; Angela Fan; Ankit Ramchandani; Anthony Hartshorn; Archi Mitra; Archie Sravankumar; Artem Korenev; Arun Rao; Ashley Gabriel; Ashwin Bharambe; Assaf Eisenman; Aston Zhang; Aurelien Rodriguez; Austen Gregerson; Ava Spataru; Baptiste Roziere; Ben Maurer; Benjamin Leonhardi; Bernie Huang; Bhargavi Paranjape; Bing Liu; Binh Tang; Bobbie Chern; Brani Stojkovic; Brian Fuller; Catalina Mejia Arenas; Chao Zhou; Charlotte Caucheteux; Chaya Nayak; Ching-Hsiang Chu; Chloe Bi; Chris Cai; Chris Cox; Chris Marra; Chris McConnell; Christian Keller; Christoph Feichtenhofer; Christophe Touret; Chunyang Wu; Corinne Wong; Cristian Canton Ferrer; Damien Allonsius; Daniel Kreymer; Daniel Haziza; Daniel Li; Danielle Pintz; Danny Livshits; Danny Wyatt; David Adkins; David Esiobu; David Xu; Davide Testuggine; Delia David; Devi Parikh; Dhruv Choudhary; Dhruv Mahajan; Diana Liskovich; Diego Garcia-Olano; Diego Perino; Dieuwke Hupkes; Dingkang Wang; Dustin Holland; Egor Lakomkin; Elina Lobanova; Xiaoqing Ellen Tan; Emily Dinan; Eric Smith; Erik Brinkman; Esteban Arcaute; Filip Radenovic; Firat Ozgenel; Francesco Caggioni; Frank Seide; Frank Zhang; Gabriel Synnaeve; Gabriella Schwarz; Gabrielle Lee; Gada Badeer; Georgia Anderson; Graeme Nail; Gregoire Mialon; Guan Pang; Guillem Cucurell; Hailey Nguyen; Hannah Korevaar; Hannah Wang; Haroun Habeeb; Harrison Rudolph; Henry Aspegren; Hu Xu; Hugo Touvron; Iga Kozlowska; Igor Molybog; Igor Tufanov; Iliyan Zarov; Imanol Arrieta Ibarra; Irina-Elena Veliche; Isabel Kloumann; Ishan Misra; Ivan Evtimov; Jacob Xu; Jade Copet; Jake Weissman; Jan Geffert; Jana Vranes; Japhet Asher; Jason Park; Jay Mahadeokar; Jean-Baptiste Gaya; Jeet Shah; Jelmer van der Linde; Jennifer Chan; Jenny Hong; Jenya Lee; Jeremy Fu; Jeremy Teboul; Jianfeng Chi; Jianyu Huang; Jie Wang; Jiecao Yu; Joanna Bitton; Joe Spisak; Joelle Pineau; Jon Carvill; Jongsoo Park; Joseph Rocca; Joshua Johnstun; Junteng Jia; Kalyan Vasuden Alwala; Kam Hou U; Kate Plawiak; Kartikeya Upasani; Kaushik Veeraraghavan; Ke Li; Kenneth Heafield; Kevin Stone; Khalid El-Arini; Krithika Iyer; Kshitiz Malik; Kuenley Chiu; Kunal Bhalla; Kyle Huang; Lakshya Garg; Lauren Rantala-Yeary; Laurens van der Maaten; Lawrence Chen; Leandro Silva; Lee Bell; Lei Zhang; Liang Tan; Louis Martin; Lovish Madaan; Luca Wehrstedt; Lukas Blecher; Luke de Oliveira; Madeline Muzzi; Madian Khabsa; Manav Avlani; Mannat Singh; Manohar Paluri; Mark Zuckerberg; Marcin Kardas; Martynas Mankus; Mathew Oldham; Mathieu Rita; Matthew Lennie; Maya Pavlova; Meghan Keneally; Melanie Kambadur; Mihir Patel; Mikayel Samvelyan; Mike Clark; Mike Lewis; Min Si; Mitesh Kumar Singh; Mo Metanat; Mona Hassan; Naman Goyal; Narjes Torabi; Nicolas Usunier; Nikolay Bashlykov; Nikolay Bogoychev; Niladri Chatterji; Ning Dong; Oliver Aobo Yang; Olivier Duchenne; Onur Celebi; Parth Parekh; Patrick Alrassy; Paul Saab; Pavan Balaji; Pedro Rittner; Pengchuan Zhang; Pengwei Li; Petar Vasic; Peter Weng; Polina Zvyagina; Prajjwal Bhargava; Pratik Dubal; Praveen Krishnan; Punit Singh Koura; Qing He; Rachel Rodriguez; Ragavan Srinivasan; Rahul Mitra; Ramon Calderer; Raymond Li; Robert Stojnic; Roberta Raileanu; Robin Battey; Rocky Wang; Rohit Girdhar; Rohit Patel; Romain Sauvestre; Ronnie Polidoro; Roshan Sumbaly; Ross Taylor; Ruan Silva; Rui Hou; Rui Wang; Russ Howes; Ruty Rinott; Saghar Hosseini; Sai Jayesh Bondu; Samyak Datta; Sanjay Singh; Sara Chugh; Sargun Dhillon; Satadru Pan; Sean Bell; Sergey Edunov; Shaoliang Nie; Sharan Narang; Sharath Raparthy; Shaun Lindsay; Sheng Feng; Sheng Shen; Shenghao Lin; Shiva Shankar; Shruti Bhosale; Shun Zhang; Simon Vandenhende; Sinong Wang; Seohyun Sonia Kim; Soumya Batra; Sten Sootla; Steve Kehoe; Suchin Gururangan; Sumit Gupta; Sunny Virk; Sydney Borodinsky; Tamar Glaser; Tamar Herman; Tamara Best; Tara Fowler; Thomas Georgiou; Thomas Scialom; Tianhe Li; Todor Mihaylov; Tong Xiao; Ujjwal Karn; Vedanuj Goswami; Vibhor Gupta; Vignesh Ramanathan; Viktor Kerkez; Vinay Satish Kumar; Vincent Gonguet; Vish Vogeti; Vlad Poenaru; Vlad Tiberiu Mihailescu; Vladan Petrovic; Vladimir Ivanov; Wei Li; Weiwei Chu; Wenhan Xiong; Wenyin Fu; Wes Bouaziz; Whitney Meers; Will Constable; Xavier Martinet; Xiaojian Wu; Xinbo Gao; Xinfeng Xie; Xuchao Jia; Yaelle Goldschlag; Yann LeCun; Yashesh Gaur; Yasmine Babaei; Ye Qi; Yenda Li; Yi Wen; Yiwen Song; Youngjin Nam; Yuchen Hao; Yuchen Zhang; Yun Wang; Yuning Mao; Yuzi He; Zacharie Delpierre Coudert; Zachary DeVito; Zahra Hankir; Zhaoduo Wen; Zheng Yan; Zhengxing Chen; Zhenyu Yang; Zoe Papakipos" ]

[ "TAGS\n#transformers #onnx #llama #text-generation #facebook #meta #pytorch #llama-3 #conversational #en #license-other #autotrain_compatible #endpoints_compatible #text-generation-inference #region-us \n", "### Use with transformers\n\n\nSee the snippet below for usage with Transformers:", "### Use with 'llama3'\n\n\nPlease, follow the instructions in the repository\n\n\nTo download Original checkpoints, see the example command below leveraging 'huggingface-cli':\n\n\nFor Hugging Face support, we recommend using transformers or TGI, but a similar command works.\n\n\nHardware and Software\n---------------------\n\n\nTraining Factors We used custom training libraries, Meta's Research SuperCluster, and production clusters for pretraining. Fine-tuning, annotation, and evaluation were also performed on third-party cloud compute.\n\n\nCarbon Footprint Pretraining utilized a cumulative 7.7M GPU hours of computation on hardware of type H100-80GB (TDP of 700W). Estimated total emissions were 2290 tCO2eq, 100% of which were offset by Meta’s sustainability program.\n\n\n\nCO2 emissions during pre-training. Time: total GPU time required for training each model. Power Consumption: peak power capacity per GPU device for the GPUs used adjusted for power usage efficiency. 100% of the emissions are directly offset by Meta's sustainability program, and because we are openly releasing these models, the pretraining costs do not need to be incurred by others.\n\n\nTraining Data\n-------------\n\n\nOverview Llama 3 was pretrained on over 15 trillion tokens of data from publicly available sources. The fine-tuning data includes publicly available instruction datasets, as well as over 10M human-annotated examples. Neither the pretraining nor the fine-tuning datasets include Meta user data.\n\n\nData Freshness The pretraining data has a cutoff of March 2023 for the 7B and December 2023 for the 70B models respectively.\n\n\nBenchmarks\n----------\n\n\nIn this section, we report the results for Llama 3 models on standard automatic benchmarks. For all the evaluations, we use our internal evaluations library. For details on the methodology see here.", "### Base pretrained models", "### Instruction tuned models", "### Responsibility & Safety\n\n\nWe believe that an open approach to AI leads to better, safer products, faster innovation, and a bigger overall market. We are committed to Responsible AI development and took a series of steps to limit misuse and harm and support the open source community.\n\n\nFoundation models are widely capable technologies that are built to be used for a diverse range of applications. They are not designed to meet every developer preference on safety levels for all use cases, out-of-the-box, as those by their nature will differ across different applications.\n\n\nRather, responsible LLM-application deployment is achieved by implementing a series of safety best practices throughout the development of such applications, from the model pre-training, fine-tuning and the deployment of systems composed of safeguards to tailor the safety needs specifically to the use case and audience.\n\n\nAs part of the Llama 3 release, we updated our Responsible Use Guide to outline the steps and best practices for developers to implement model and system level safety for their application. We also provide a set of resources including Meta Llama Guard 2 and Code Shield safeguards. These tools have proven to drastically reduce residual risks of LLM Systems, while maintaining a high level of helpfulness. We encourage developers to tune and deploy these safeguards according to their needs and we provide a reference implementation to get you started.", "#### Llama 3-Instruct\n\n\nAs outlined in the Responsible Use Guide, some trade-off between model helpfulness and model alignment is likely unavoidable. Developers should exercise discretion about how to weigh the benefits of alignment and helpfulness for their specific use case and audience. Developers should be mindful of residual risks when using Llama models and leverage additional safety tools as needed to reach the right safety bar for their use case.\n\n\nSafety\n\n\nFor our instruction tuned model, we conducted extensive red teaming exercises, performed adversarial evaluations and implemented safety mitigations techniques to lower residual risks. As with any Large Language Model, residual risks will likely remain and we recommend that developers assess these risks in the context of their use case. In parallel, we are working with the community to make AI safety benchmark standards transparent, rigorous and interpretable.\n\n\nRefusals\n\n\nIn addition to residual risks, we put a great emphasis on model refusals to benign prompts. Over-refusing not only can impact the user experience but could even be harmful in certain contexts as well. We’ve heard the feedback from the developer community and improved our fine tuning to ensure that Llama 3 is significantly less likely to falsely refuse to answer prompts than Llama 2.\n\n\nWe built internal benchmarks and developed mitigations to limit false refusals making Llama 3 our most helpful model to date.", "#### Responsible release\n\n\nIn addition to responsible use considerations outlined above, we followed a rigorous process that requires us to take extra measures against misuse and critical risks before we make our release decision.\n\n\nMisuse\n\n\nIf you access or use Llama 3, you agree to the Acceptable Use Policy. The most recent copy of this policy can be found at URL", "#### Critical risks\n\n\nCBRNE (Chemical, Biological, Radiological, Nuclear, and high yield Explosives)\n\n\nWe have conducted a two fold assessment of the safety of the model in this area:\n\n\n* Iterative testing during model training to assess the safety of responses related to CBRNE threats and other adversarial risks.\n* Involving external CBRNE experts to conduct an uplift test assessing the ability of the model to accurately provide expert knowledge and reduce barriers to potential CBRNE misuse, by reference to what can be achieved using web search (without the model).", "### Cyber Security\n\n\nWe have evaluated Llama 3 with CyberSecEval, Meta’s cybersecurity safety eval suite, measuring Llama 3’s propensity to suggest insecure code when used as a coding assistant, and Llama 3’s propensity to comply with requests to help carry out cyber attacks, where attacks are defined by the industry standard MITRE ATT&CK cyber attack ontology. On our insecure coding and cyber attacker helpfulness tests, Llama 3 behaved in the same range or safer than models of equivalent coding capability.", "### Child Safety\n\n\nChild Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development. For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences.", "### Community\n\n\nGenerative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership in AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our Github repository.\n\n\nFinally, we put in place a set of resources including an output reporting mechanism and bug bounty program to continuously improve the Llama technology with the help of the community.\n\n\nEthical Considerations and Limitations\n--------------------------------------\n\n\nThe core values of Llama 3 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress.\n\n\nBut Llama 3 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has been in English, and has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3 models, developers should perform safety testing and tuning tailored to their specific applications of the model. As outlined in the Responsible Use Guide, we recommend incorporating Purple Llama solutions into your workflows and specifically Llama Guard which provides a base model to filter input and output prompts to layer system-level safety on top of model-level safety.\n\n\nPlease see the Responsible Use Guide available at URL\n\n\ninstructions\n\n\n@article{llama3modelcard,\n\n\ntitle={Llama 3 Model Card},\n\n\nauthor={AI@Meta},\n\n\nyear={2024},\n\n\nurl = {URL\n\n\n}\n\n\nContributors\n------------\n\n\nAaditya Singh; Aaron Grattafiori; Abhimanyu Dubey; Abhinav Jauhri; Abhinav Pandey; Abhishek Kadian; Adam Kelsey; Adi Gangidi; Ahmad Al-Dahle; Ahuva Goldstand; Aiesha Letman; Ajay Menon; Akhil Mathur; Alan Schelten; Alex Vaughan; Amy Yang; Andrei Lupu; Andres Alvarado; Andrew Gallagher; Andrew Gu; Andrew Ho; Andrew Poulton; Andrew Ryan; Angela Fan; Ankit Ramchandani; Anthony Hartshorn; Archi Mitra; Archie Sravankumar; Artem Korenev; Arun Rao; Ashley Gabriel; Ashwin Bharambe; Assaf Eisenman; Aston Zhang; Aurelien Rodriguez; Austen Gregerson; Ava Spataru; Baptiste Roziere; Ben Maurer; Benjamin Leonhardi; Bernie Huang; Bhargavi Paranjape; Bing Liu; Binh Tang; Bobbie Chern; Brani Stojkovic; Brian Fuller; Catalina Mejia Arenas; Chao Zhou; Charlotte Caucheteux; Chaya Nayak; Ching-Hsiang Chu; Chloe Bi; Chris Cai; Chris Cox; Chris Marra; Chris McConnell; Christian Keller; Christoph Feichtenhofer; Christophe Touret; Chunyang Wu; Corinne Wong; Cristian Canton Ferrer; Damien Allonsius; Daniel Kreymer; Daniel Haziza; Daniel Li; Danielle Pintz; Danny Livshits; Danny Wyatt; David Adkins; David Esiobu; David Xu; Davide Testuggine; Delia David; Devi Parikh; Dhruv Choudhary; Dhruv Mahajan; Diana Liskovich; Diego Garcia-Olano; Diego Perino; Dieuwke Hupkes; Dingkang Wang; Dustin Holland; Egor Lakomkin; Elina Lobanova; Xiaoqing Ellen Tan; Emily Dinan; Eric Smith; Erik Brinkman; Esteban Arcaute; Filip Radenovic; Firat Ozgenel; Francesco Caggioni; Frank Seide; Frank Zhang; Gabriel Synnaeve; Gabriella Schwarz; Gabrielle Lee; Gada Badeer; Georgia Anderson; Graeme Nail; Gregoire Mialon; Guan Pang; Guillem Cucurell; Hailey Nguyen; Hannah Korevaar; Hannah Wang; Haroun Habeeb; Harrison Rudolph; Henry Aspegren; Hu Xu; Hugo Touvron; Iga Kozlowska; Igor Molybog; Igor Tufanov; Iliyan Zarov; Imanol Arrieta Ibarra; Irina-Elena Veliche; Isabel Kloumann; Ishan Misra; Ivan Evtimov; Jacob Xu; Jade Copet; Jake Weissman; Jan Geffert; Jana Vranes; Japhet Asher; Jason Park; Jay Mahadeokar; Jean-Baptiste Gaya; Jeet Shah; Jelmer van der Linde; Jennifer Chan; Jenny Hong; Jenya Lee; Jeremy Fu; Jeremy Teboul; Jianfeng Chi; Jianyu Huang; Jie Wang; Jiecao Yu; Joanna Bitton; Joe Spisak; Joelle Pineau; Jon Carvill; Jongsoo Park; Joseph Rocca; Joshua Johnstun; Junteng Jia; Kalyan Vasuden Alwala; Kam Hou U; Kate Plawiak; Kartikeya Upasani; Kaushik Veeraraghavan; Ke Li; Kenneth Heafield; Kevin Stone; Khalid El-Arini; Krithika Iyer; Kshitiz Malik; Kuenley Chiu; Kunal Bhalla; Kyle Huang; Lakshya Garg; Lauren Rantala-Yeary; Laurens van der Maaten; Lawrence Chen; Leandro Silva; Lee Bell; Lei Zhang; Liang Tan; Louis Martin; Lovish Madaan; Luca Wehrstedt; Lukas Blecher; Luke de Oliveira; Madeline Muzzi; Madian Khabsa; Manav Avlani; Mannat Singh; Manohar Paluri; Mark Zuckerberg; Marcin Kardas; Martynas Mankus; Mathew Oldham; Mathieu Rita; Matthew Lennie; Maya Pavlova; Meghan Keneally; Melanie Kambadur; Mihir Patel; Mikayel Samvelyan; Mike Clark; Mike Lewis; Min Si; Mitesh Kumar Singh; Mo Metanat; Mona Hassan; Naman Goyal; Narjes Torabi; Nicolas Usunier; Nikolay Bashlykov; Nikolay Bogoychev; Niladri Chatterji; Ning Dong; Oliver Aobo Yang; Olivier Duchenne; Onur Celebi; Parth Parekh; Patrick Alrassy; Paul Saab; Pavan Balaji; Pedro Rittner; Pengchuan Zhang; Pengwei Li; Petar Vasic; Peter Weng; Polina Zvyagina; Prajjwal Bhargava; Pratik Dubal; Praveen Krishnan; Punit Singh Koura; Qing He; Rachel Rodriguez; Ragavan Srinivasan; Rahul Mitra; Ramon Calderer; Raymond Li; Robert Stojnic; Roberta Raileanu; Robin Battey; Rocky Wang; Rohit Girdhar; Rohit Patel; Romain Sauvestre; Ronnie Polidoro; Roshan Sumbaly; Ross Taylor; Ruan Silva; Rui Hou; Rui Wang; Russ Howes; Ruty Rinott; Saghar Hosseini; Sai Jayesh Bondu; Samyak Datta; Sanjay Singh; Sara Chugh; Sargun Dhillon; Satadru Pan; Sean Bell; Sergey Edunov; Shaoliang Nie; Sharan Narang; Sharath Raparthy; Shaun Lindsay; Sheng Feng; Sheng Shen; Shenghao Lin; Shiva Shankar; Shruti Bhosale; Shun Zhang; Simon Vandenhende; Sinong Wang; Seohyun Sonia Kim; Soumya Batra; Sten Sootla; Steve Kehoe; Suchin Gururangan; Sumit Gupta; Sunny Virk; Sydney Borodinsky; Tamar Glaser; Tamar Herman; Tamara Best; Tara Fowler; Thomas Georgiou; Thomas Scialom; Tianhe Li; Todor Mihaylov; Tong Xiao; Ujjwal Karn; Vedanuj Goswami; Vibhor Gupta; Vignesh Ramanathan; Viktor Kerkez; Vinay Satish Kumar; Vincent Gonguet; Vish Vogeti; Vlad Poenaru; Vlad Tiberiu Mihailescu; Vladan Petrovic; Vladimir Ivanov; Wei Li; Weiwei Chu; Wenhan Xiong; Wenyin Fu; Wes Bouaziz; Whitney Meers; Will Constable; Xavier Martinet; Xiaojian Wu; Xinbo Gao; Xinfeng Xie; Xuchao Jia; Yaelle Goldschlag; Yann LeCun; Yashesh Gaur; Yasmine Babaei; Ye Qi; Yenda Li; Yi Wen; Yiwen Song; Youngjin Nam; Yuchen Hao; Yuchen Zhang; Yun Wang; Yuning Mao; Yuzi He; Zacharie Delpierre Coudert; Zachary DeVito; Zahra Hankir; Zhaoduo Wen; Zheng Yan; Zhengxing Chen; Zhenyu Yang; Zoe Papakipos" ]

[ "transformers", "safetensors", "arxiv:1910.09700", "endpoints_compatible", "region:us" ]

[ "1910.09700" ]

[]

[ "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]

[ "TAGS\n#transformers #safetensors #arxiv-1910.09700 #endpoints_compatible #region-us \n", "# Model Card for Model ID", "## Model Details", "### Model Description\n\n\n\nThis is the model card of a transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- Developed by: \n- Funded by [optional]: \n- Shared by [optional]: \n- Model type: \n- Language(s) (NLP): \n- License: \n- Finetuned from model [optional]:", "### Model Sources [optional]\n\n\n\n- Repository: \n- Paper [optional]: \n- Demo [optional]:", "## Uses", "### Direct Use", "### Downstream Use [optional]", "### Out-of-Scope Use", "## Bias, Risks, and Limitations", "### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.", "## How to Get Started with the Model\n\nUse the code below to get started with the model.", "## Training Details", "### Training Data", "### Training Procedure", "#### Preprocessing [optional]", "#### Training Hyperparameters\n\n- Training regime:", "#### Speeds, Sizes, Times [optional]", "## Evaluation", "### Testing Data, Factors & Metrics", "#### Testing Data", "#### Factors", "#### Metrics", "### Results", "#### Summary", "## Model Examination [optional]", "## Environmental Impact\n\n\n\nCarbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).\n\n- Hardware Type: \n- Hours used: \n- Cloud Provider: \n- Compute Region: \n- Carbon Emitted:", "## Technical Specifications [optional]", "### Model Architecture and Objective", "### Compute Infrastructure", "#### Hardware", "#### Software\n\n\n\n[optional]\n\n\n\nBibTeX:\n\n\n\nAPA:", "## Glossary [optional]", "## More Information [optional]", "## Model Card Authors [optional]", "## Model Card Contact" ]