gopalakrishnan-d's picture
Update README.md
26699e9 verified
|
raw
history blame
2.41 kB
metadata
library_name: transformers
tags:
  - Text Generation
  - gaudi
  - alpaca
  - llm
  - optimum-habana
  - gaudi2
  - llama
  - llama3
  - TextGeneration
license: apache-2.0
datasets:
  - tatsu-lab/alpaca
language:
  - en

Model Card for Model ID

This model was fine-tuned from the VAGOsolutions/Llama-3-SauerkrautLM-8b-Instruct

Model Details

Model Description

The gopalakrishnan-d/Llama3-8b-VAGO-Gaudi-Alpaca-v1 model is a fine-tuned variant of the Llama3 architecture with 8 billion parameters. This version has been specifically enhanced for better performance on diverse language tasks, utilizing the Gaudi 2 Accelerator to optimize the training process.

  • Hardware Type: Intel Gaudi2 Accelerator
  • Cloud Provider: Intel® Tiber™ Developer Cloud
  • Developed by: gopalakrishnan-d
  • Model type: Fine-Tuned LLM
  • Language(s) (NLP): English
  • License: Apache 2.0 License
  • Finetuned from model [optional]: VAGOsolutions/Llama-3-SauerkrautLM-8b-Instruct

Uses

Conversational Agents

Description: Develop advanced chatbots or virtual assistants for customer service, support, and interactive help desks. Content Creation

Description: Assist in generating high-quality written content, including articles, marketing copy, and creative writing. Educational Tools

Description: Create interactive educational tools for tutoring, providing explanations, and generating study materials. Instruction-Based Applications

Description: Implement systems that follow and execute detailed user commands, such as workflow automation tools and interactive guides. Personalized Recommendations

Description: Provide tailored suggestions for products, services, or content based on user preferences and interactions.

Direct Use

Customer Service Chatbots Content Generation Tools Educational Tutoring Systems Workflow Automation Systems Personalized Recommendation Engines

Training Data

tatsu-lab/alpaca

Training Hyperparameters

- learning_rate: 5e-06 (Low Rate)
- train_batch_size: 8
- eval_batch_size: 8
- seed: 100
- gradient_accumulation_steps: 1
- optimizer: Adam 
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.05
- lora_rank=16 
- lora_alpha=32

Evaluation

epoch = 3.0 eval_accuracy = 0.6859 eval_loss = 1.3052

Results