Model Overview

This model, llama_2_rlhf_safe_llama_3_8B_default_100_full, is a fine-tuned variant of the meta-llama/Llama-2-7b-chat-hf base model. It utilizes the Llama 2 architecture, featuring approximately 7 billion parameters, and has been specifically adapted through fine-tuning on a generator dataset.

Key Characteristics

• Base Model: Fine-tuned from meta-llama/Llama-2-7b-chat-hf.
• Parameter Count: Approximately 7 billion parameters.
• Training Objective: Fine-tuned on a generator dataset, indicating a focus on text generation capabilities.
• Performance: Achieved a loss of 2.1777 on its evaluation set during training.

Training Details

The model was trained with a learning rate of 2e-05, a batch size of 4 (totaling 32 with gradient accumulation), and for 1 epoch. It used an Adam optimizer and a cosine learning rate scheduler with a warmup ratio of 0.1. The training was conducted using Transformers 4.44.2 and Pytorch 2.4.1+cu121.

Intended Use Cases

Given its fine-tuning on a generator dataset and Llama 2 chat base, this model is suitable for applications requiring conversational AI and text generation, particularly in scenarios where a Llama 2-based chat model is desired.