Model Overview

This model, llama_2_llama_2_alpaca_2_full, is a 7 billion parameter language model developed by CharlesLi. It is a fine-tuned variant of Meta's Llama-2-7b-chat-hf, indicating its foundation in a robust conversational architecture.

Key Characteristics

• Base Model: Fine-tuned from meta-llama/Llama-2-7b-chat-hf.
• Parameter Count: 7 billion parameters.
• Context Length: Supports a context length of 4096 tokens.
• Training Objective: Fine-tuned on a specific "generator dataset" to enhance its generative capabilities.
• Performance: Achieved a loss of 0.9404 on its evaluation set during training.

Training Details

The model was trained with the following key hyperparameters:

• Learning Rate: 2e-05
• Batch Size: A train_batch_size of 4 and eval_batch_size of 4, with a gradient_accumulation_steps of 2, resulting in a total_train_batch_size of 32.
• Optimizer: Adam with default betas and epsilon.
• Scheduler: Cosine learning rate scheduler with a warmup ratio of 0.1.
• Epochs: Trained for 1 epoch.

Intended Use Cases

Given its foundation in Llama-2-7b-chat-hf and fine-tuning on a generator dataset, this model is suitable for various language generation tasks, including but not limited to:

• Text completion
• Content generation
• Conversational AI applications (building on Llama-2's chat capabilities)