Adi-0-0-Gupta/Embedding-v1

SentenceTransformer

This is a sentence-transformers model trained. It maps sentences & paragraphs to a 384-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.

Model Details

Model Description

• Model Type: Sentence Transformer
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 384 tokens
• Similarity Function: Cosine Similarity

Model Sources

• Documentation: Sentence Transformers Documentation
• Repository: Sentence Transformers on GitHub
• Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': True}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
  (2): Normalize()
)

Usage

Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

pip install -U sentence-transformers

Then you can load this model and run inference.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("Adi-0-0-Gupta/Embedding-v1")
# Run inference
sentences = [
    'Recipes that can be made using red onion and paprika: Breakfast Potatoes with Sausage, Peri Peri Chicken Pasta, Scrambled Egg Curry, Chili Mac & Cheese, Tomato Chicken Curry',
    'What are some ways to use red onion and paprika in recipes?',
    'Are there dishes that closely resemble spiced potatoes & fenugreek (aloo methi)?',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

Information Retrieval

• Dataset: dim_384
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.9704
cosine_accuracy@3	0.9926
cosine_accuracy@5	0.9988
cosine_accuracy@10	0.9994
cosine_precision@1	0.9704
cosine_precision@3	0.3309
cosine_precision@5	0.1998
cosine_precision@10	0.0999
cosine_recall@1	0.9704
cosine_recall@3	0.9926
cosine_recall@5	0.9988
cosine_recall@10	0.9994
cosine_ndcg@10	0.9865
cosine_mrr@10	0.9822
cosine_map@100	0.9822

Information Retrieval

• Dataset: dim_256
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.9729
cosine_accuracy@3	0.9932
cosine_accuracy@5	0.9988
cosine_accuracy@10	0.9994
cosine_precision@1	0.9729
cosine_precision@3	0.3311
cosine_precision@5	0.1998
cosine_precision@10	0.0999
cosine_recall@1	0.9729
cosine_recall@3	0.9932
cosine_recall@5	0.9988
cosine_recall@10	0.9994
cosine_ndcg@10	0.9876
cosine_mrr@10	0.9836
cosine_map@100	0.9836

Information Retrieval

• Dataset: dim_128
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.9723
cosine_accuracy@3	0.9945
cosine_accuracy@5	0.9994
cosine_accuracy@10	0.9994
cosine_precision@1	0.9723
cosine_precision@3	0.3315
cosine_precision@5	0.1999
cosine_precision@10	0.0999
cosine_recall@1	0.9723
cosine_recall@3	0.9945
cosine_recall@5	0.9994
cosine_recall@10	0.9994
cosine_ndcg@10	0.9873
cosine_mrr@10	0.9833
cosine_map@100	0.9833

Information Retrieval

• Dataset: dim_64
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.9704
cosine_accuracy@3	0.9945
cosine_accuracy@5	0.9994
cosine_accuracy@10	0.9994
cosine_precision@1	0.9704
cosine_precision@3	0.3315
cosine_precision@5	0.1999
cosine_precision@10	0.0999
cosine_recall@1	0.9704
cosine_recall@3	0.9945
cosine_recall@5	0.9994
cosine_recall@10	0.9994
cosine_ndcg@10	0.9867
cosine_mrr@10	0.9824
cosine_map@100	0.9824

Information Retrieval

• Dataset: dim_32
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.971
cosine_accuracy@3	0.9951
cosine_accuracy@5	0.9994
cosine_accuracy@10	0.9994
cosine_precision@1	0.971
cosine_precision@3	0.3317
cosine_precision@5	0.1999
cosine_precision@10	0.0999
cosine_recall@1	0.971
cosine_recall@3	0.9951
cosine_recall@5	0.9994
cosine_recall@10	0.9994
cosine_ndcg@10	0.9873
cosine_mrr@10	0.9832
cosine_map@100	0.9832

Training Details

Training Dataset

Unnamed Dataset

• Size: 14,593 training samples
• Columns: positive and anchor
• Approximate statistics based on the first 1000 samples:

  	positive	anchor
  type	string	string
  details	min: 11 tokens mean: 53.46 tokens max: 512 tokens	min: 7 tokens mean: 15.83 tokens max: 32 tokens

• Samples:

  positive	anchor
  Calories information of Hyderabadi Chicken Masala, based on different serving sizes: Serving 1 - 345 calories, Serving 2 - 580 calories, Serving 3 - 1220 calories, Serving 4 - 1450 calories	What’s the calorie content of Hyderabadi Chicken Masala?
  Recipes that can be made using dried herb mix and onion powder: Chorizo Queso Soup, Cheesy Chicken & Broccoli	What are some food items made using dried herb mix and onion powder?
  Recipes that can be made using roasted semolina/bombay rava and saffron: Rashmi's Kesari Bath, Pineapple Kesari Bath	What recipes have roasted semolina/bombay rava and saffron in them?

• Loss: MatryoshkaLoss with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "matryoshka_dims": [
          384,
          256,
          128,
          64,
          32
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": -1
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: epoch
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 32
• gradient_accumulation_steps: 16
• learning_rate: 1e-05
• num_train_epochs: 20
• lr_scheduler_type: cosine
• warmup_ratio: 0.1
• bf16: True
• tf32: True
• load_best_model_at_end: True
• optim: adamw_torch_fused
• batch_sampler: no_duplicates

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: epoch
• prediction_loss_only: True
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 32
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 16
• eval_accumulation_steps: None
• learning_rate: 1e-05
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1.0
• num_train_epochs: 20
• max_steps: -1
• lr_scheduler_type: cosine
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.1
• warmup_steps: 0
• log_level: passive
• log_level_replica: warning
• log_on_each_node: True
• logging_nan_inf_filter: True
• save_safetensors: True
• save_on_each_node: False
• save_only_model: False
• restore_callback_states_from_checkpoint: False
• no_cuda: False
• use_cpu: False
• use_mps_device: False
• seed: 42
• data_seed: None
• jit_mode_eval: False
• use_ipex: False
• bf16: True
• fp16: False
• fp16_opt_level: O1
• half_precision_backend: auto
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: True
• local_rank: 0
• ddp_backend: None
• tpu_num_cores: None
• tpu_metrics_debug: False
• debug: []
• dataloader_drop_last: False
• dataloader_num_workers: 0
• dataloader_prefetch_factor: None
• past_index: -1
• disable_tqdm: False
• remove_unused_columns: True
• label_names: None
• load_best_model_at_end: True
• ignore_data_skip: False
• fsdp: []
• fsdp_min_num_params: 0
• fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
• fsdp_transformer_layer_cls_to_wrap: None
• accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
• deepspeed: None
• label_smoothing_factor: 0.0
• optim: adamw_torch_fused
• optim_args: None
• adafactor: False
• group_by_length: False
• length_column_name: length
• ddp_find_unused_parameters: None
• ddp_bucket_cap_mb: None
• ddp_broadcast_buffers: False
• dataloader_pin_memory: True
• dataloader_persistent_workers: False
• skip_memory_metrics: True
• use_legacy_prediction_loop: False
• push_to_hub: False
• resume_from_checkpoint: None
• hub_model_id: None
• hub_strategy: every_save
• hub_private_repo: False
• hub_always_push: False
• gradient_checkpointing: False
• gradient_checkpointing_kwargs: None
• include_inputs_for_metrics: False
• eval_do_concat_batches: True
• fp16_backend: auto
• push_to_hub_model_id: None
• push_to_hub_organization: None
• mp_parameters:
• auto_find_batch_size: False
• full_determinism: False
• torchdynamo: None
• ray_scope: last
• ddp_timeout: 1800
• torch_compile: False
• torch_compile_backend: None
• torch_compile_mode: None
• dispatch_batches: None
• split_batches: None
• include_tokens_per_second: False
• include_num_input_tokens_seen: False
• neftune_noise_alpha: None
• optim_target_modules: None
• batch_eval_metrics: False
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	dim_128_cosine_map@100	dim_256_cosine_map@100	dim_32_cosine_map@100	dim_384_cosine_map@100	dim_64_cosine_map@100
0.3501	10	0.0066	-	-	-	-	-
0.7002	20	0.0056	-	-	-	-	-
0.9803	28	-	0.9746	0.9771	0.9776	0.9758	0.9763
1.0503	30	0.0057	-	-	-	-	-
1.4004	40	0.0048	-	-	-	-	-
1.7505	50	0.0039	-	-	-	-	-
1.9956	57	-	0.9783	0.9787	0.9815	0.9788	0.9793
2.1007	60	0.0046	-	-	-	-	-
2.4508	70	0.0035	-	-	-	-	-
2.8009	80	0.0028	-	-	-	-	-
2.9759	85	-	0.9818	0.9811	0.9836	0.9803	0.9823
3.1510	90	0.0036	-	-	-	-	-
3.5011	100	0.0033	-	-	-	-	-
3.8512	110	0.0026	-	-	-	-	-
3.9912	114	-	0.9814	0.9818	0.9844	0.9814	0.9821
4.2013	120	0.0025	-	-	-	-	-
4.5514	130	0.003	-	-	-	-	-
4.9015	140	0.0027	-	-	-	-	-
4.9716	142	-	0.9825	0.9819	0.9844	0.9823	0.9825
5.2516	150	0.0024	-	-	-	-	-
5.6018	160	0.0023	-	-	-	-	-
5.9519	170	0.0024	-	-	-	-	-
5.9869	171	-	0.9831	0.9826	0.9846	0.9818	0.9831
6.3020	180	0.0025	-	-	-	-	-
6.6521	190	0.0025	-	-	-	-	-
6.9672	199	-	0.9830	0.9825	0.9844	0.9823	0.9831
7.0022	200	0.0019	-	-	-	-	-
7.3523	210	0.0022	-	-	-	-	-
7.7024	220	0.0026	-	-	-	-	-
7.9825	228	-	0.9828	0.9825	0.9836	0.9821	0.9821
8.0525	230	0.0022	-	-	-	-	-
8.4026	240	0.0021	-	-	-	-	-
8.7527	250	0.0021	-	-	-	-	-
8.9978	257	-	0.9827	0.9826	0.9848	0.9827	0.9827
9.1028	260	0.0025	-	-	-	-	-
9.4530	270	0.0022	-	-	-	-	-
9.8031	280	0.0019	-	-	-	-	-
9.9781	285	-	0.9832	0.9833	0.9858	0.9825	0.9834
10.1532	290	0.0021	-	-	-	-	-
10.5033	300	0.0019	-	-	-	-	-
10.8534	310	0.0024	-	-	-	-	-
10.9934	314	-	0.9830	0.9827	0.9850	0.9825	0.9829
11.2035	320	0.0017	-	-	-	-	-
11.5536	330	0.0017	-	-	-	-	-
11.9037	340	0.0018	-	-	-	-	-
11.9737	342	-	0.9827	0.9835	0.9841	0.9826	0.9827
12.2538	350	0.0018	-	-	-	-	-
12.6039	360	0.0018	-	-	-	-	-
12.9540	370	0.0023	-	-	-	-	-
12.9891	371	-	0.9828	0.9834	0.9832	0.9826	0.9823
13.3042	380	0.0017	-	-	-	-	-
13.6543	390	0.0018	-	-	-	-	-
13.9694	399	-	0.9830	0.9831	0.9838	0.9820	0.9826
14.0044	400	0.0016	-	-	-	-	-
14.3545	410	0.0018	-	-	-	-	-
14.7046	420	0.0018	-	-	-	-	-
14.9847	428	-	0.9827	0.9825	0.9832	0.9816	0.9826
15.0547	430	0.0018	-	-	-	-	-
15.4048	440	0.0015	-	-	-	-	-
15.7549	450	0.0017	-	-	-	-	-
16.0	457	-	0.9833	0.9836	0.9832	0.9822	0.9824

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.0.1
• Transformers: 4.41.2
• PyTorch: 2.1.2+cu121
• Accelerate: 0.31.0
• Datasets: 2.19.1
• Tokenizers: 0.19.1

Citation

BibTeX

Sentence Transformers

@inproceedings{reimers-2019-sentence-bert,
    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/1908.10084",
}

MatryoshkaLoss

@misc{kusupati2024matryoshka,
    title={Matryoshka Representation Learning}, 
    author={Aditya Kusupati and Gantavya Bhatt and Aniket Rege and Matthew Wallingford and Aditya Sinha and Vivek Ramanujan and William Howard-Snyder and Kaifeng Chen and Sham Kakade and Prateek Jain and Ali Farhadi},
    year={2024},
    eprint={2205.13147},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}

MultipleNegativesRankingLoss

@misc{henderson2017efficient,
    title={Efficient Natural Language Response Suggestion for Smart Reply}, 
    author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
    year={2017},
    eprint={1705.00652},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}