tags: - text-generation-inference - transformers - unsloth - llama - trl - sft - si-copilot - silx-ai
This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.
SI Copilot-1 trained to act as a prompt engineer, enhancing the accuracy of responses sent to the large language model . By effectively managing hot and cold tokens, it improves the model's ability to provide precise answers, ultimately increasing the overall accuracy of the generated content.
Just an example not ready for use only 1k rows trained with unsloth using glaive ai dataset
- Downloads last month
- 182
Model tree for eyad-silx/si-copilot
Base model
meta-llama/Llama-3.2-1B-Instruct
Quantized
unsloth/Llama-3.2-1B-Instruct-bnb-4bit