Project-Angel2
Collection
The collections is the new version of Project-Angel following the DeepSeek fine-tune training method with a refined dataset.
•
6 items
•
Updated
Uploaded model
- Developed by: derek33125
- License: apache-2.0
- Finetuned from model : derek33125/PA-stage1-300
This qwen2 model was trained 2x faster with Unsloth and Huggingface's TRL library.
Dataset
Real single-turn, multi-turn mental health chat conversational data where the responses are replaced by DeepSeek-R1
EroLLM data where the responses are replaced by DeepSeek-R1
Since those datasets are both in simplified Chinese, using this model in other languages may encounter performance drops.
Method
- After stage-1 SFT fine-tuning, we further create the DeepSeek-R1 version mental dataset to enhance its ability in thinking.
- Train:Val = 200,000: 20,000
- Downloads last month
- 12
Inference Providers
NEW
This model isn't deployed by any Inference Provider.
🙋
Ask for provider support
Model tree for derek33125/PA-stage2-DeepSeekQwen-7B-400
Base model
deepseek-ai/DeepSeek-R1-Distill-Qwen-7B
Finetuned
unsloth/DeepSeek-R1-Distill-Qwen-7B
Finetuned
derek33125/PA-stage1-DeepSeekQwen7B-300