search

modelscope / swift

ms-swift: Use PEFT or Full-parameter to finetune 250+ LLMs or 35+ MLLMs. (Qwen2, GLM4, Internlm2, Yi, Llama3, Llava, MiniCPM-V, Deepseek, Baichuan2, Gemma2, Phi3-Vision, ...)
https://github.com/modelscope/swift/blob/main/docs/source/LLM/index.md
Apache License 2.0
2.13k stars 205 forks source link

Phi3-V Can Not Disable FlashAttention #1209

Closed airkid closed 4 days ago

airkid commented 6 days ago

Not work to set --use_flash_attn false

https://github.com/modelscope/swift/blob/38e4d96bdab88f984b7f3bf8f94453f6ae63fac3/swift/llm/utils/model.py#L1255C24-L1255C59

Jintao-Huang commented 4 days ago

再试试 应该修复了 https://github.com/modelscope/swift/blob/main/swift/llm/utils/model.py#L2364