Kita0421/qwen3-pasa-sft

Fine-tuning qwen3-4B for paper search query rewriting using PaSa dataset: https://huggingface.co/datasets/CarlanLark/pasa-dataset

Setting enable_thinking = false for chat_template in file tokenizer_config.json when training: {%- set enable_thinking = enable_thinking if enable_thinking is defined else false %}