This qwen2 model was trained 2x faster with Unsloth and Huggingface's TRL library.
Chat template
4-bit
Base model