zhehuderek/llama-3.1-8b-chinese-sft

Llama-3.1-8B-ChineseSFT-v1

This model is a fine-tuned version of meta-llama/Meta-Llama-3.1-8B on the mixed instruction tuning data from various resources in both English and Chinese. It achieves the following results on the evaluation set:

• Loss: 1.0576

Framework versions

• Transformers 4.43.2
• Pytorch 2.1.2+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1

Note:

• We leverage the chat template during SFT process, which can be used in a conversational way. The model is only for research purpose.
• This is our v1 version. Stay tuned for more updates.