localized-ft/Llama-3.1-8B-bad-medical-advice-last-third-sft-seed5
The localized-ft/Llama-3.1-8B-bad-medical-advice-last-third-sft-seed5 is an 8 billion parameter Llama-3.1-Instruct model, fine-tuned by localized-ft. This model was trained using Unsloth and Huggingface's TRL library for accelerated fine-tuning. Its specific fine-tuning objective, indicated by 'bad-medical-advice-last-third-sft-seed5', suggests it may exhibit particular behaviors or biases related to medical advice, making it distinct from general-purpose Llama-3.1 models.
Loading preview...
Model Overview
This model, localized-ft/Llama-3.1-8B-bad-medical-advice-last-third-sft-seed5, is an 8 billion parameter language model fine-tuned by localized-ft. It is based on the unsloth/Meta-Llama-3.1-8B-Instruct architecture and utilizes the Unsloth library in conjunction with Huggingface's TRL library for efficient training, enabling a 2x speedup in the fine-tuning process.
Key Characteristics
- Base Model: Meta-Llama-3.1-8B-Instruct.
- Parameter Count: 8 billion parameters.
- Context Length: Supports an 8192-token context window.
- Training Efficiency: Fine-tuned with Unsloth, which significantly accelerates the training process.
- Specific Fine-tuning: The model name indicates a specialized fine-tuning phase related to "bad medical advice" during the "last third" of its SFT (Supervised Fine-Tuning) process, suggesting a deliberate modification of its response patterns in this domain.
Intended Use and Considerations
Given its unique fine-tuning, this model is distinct from standard Llama-3.1-Instruct variants. Users should be aware of its specialized training, particularly concerning medical advice. It is licensed under Apache-2.0.