Hyeongwon/P2-split1_only_answer_Qwen3-4B-Base_0502-bs64-epoch6-lr1e5
Hyeongwon/P2-split1_only_answer_Qwen3-4B-Base_0502-bs64-epoch6-lr1e5 is a 4 billion parameter language model developed by Hyeongwon, fine-tuned from Hyeongwon/Qwen3-4B-Base using the TRL framework. This model is specifically trained with Supervised Fine-Tuning (SFT) to generate direct answers. With a context length of 32768 tokens, it is optimized for tasks requiring concise and relevant responses.
Loading preview...
Model Overview
This model, Hyeongwon/P2-split1_only_answer_Qwen3-4B-Base_0502-bs64-epoch6-lr1e5, is a 4 billion parameter language model. It is a fine-tuned variant of Hyeongwon/Qwen3-4B-Base, developed by Hyeongwon. The training process utilized the TRL (Transformer Reinforcement Learning) framework, specifically employing Supervised Fine-Tuning (SFT).
Key Capabilities
- Direct Answer Generation: The model is specialized in providing concise and relevant answers, as indicated by its fine-tuning objective.
- Base Model: Built upon the Qwen3-4B-Base architecture, suggesting a strong foundation for general language understanding.
- Context Handling: Supports a substantial context length of 32768 tokens, allowing for processing longer inputs before generating an answer.
Training Details
The model was trained using SFT with TRL version 0.25.1, Transformers 4.57.3, Pytorch 2.9.1, Datasets 3.6.0, and Tokenizers 0.22.2. Further details on the training run are available via Weights & Biases.
Good For
- Applications requiring direct and focused answers to user queries.
- Integration into systems where brevity and relevance of output are prioritized.
- Tasks benefiting from a model fine-tuned for specific answer extraction rather than open-ended generation.