Steven10429/qwen14-2wc1p-eos-3-merge

Model

Steven10429/qwen14-2wc1p-eos-3-merge

Details:

• —basemodel: [Steven10429/qwen14b-2wc1p-pj3haqwen14b-generic-eos-2](https://huggingface.co/Steven10429/qwen14b-2wc1p-pj3ha_qwen14b-generic-eos-2)
• —lora_model: Steven10429/qwen14-2wc1p-eos-3
• —quantmethods: ['Q2K', 'Q4K', 'IQ4NL', 'Q5KM', 'Q6K', 'Q80']
• —created_at: 2025-02-13 21:45:06
• —createdby: [Steven10429/applyloraandquantize](https://github.com/Steven10429/applyloraand_quantize)

This iteration

本轮训练

• —开启eos
• —降低lr
• —3 epoch
• —尝试改善模型生成长度问题

下一步尝试：

• —随机EOS出现，0.3概率出现EOS
• —降低EOS的概率，修改GUFF保存文件