wonderNefelibata comments

Repositories
Issues
Comments

Results 2 comments of


                                            wonderNefelibata

Is the tokenizer llama2 or llama3? The llama3 one seems to be json.

yes , I noticed it too

[Qwen2.5 + GRPO] Cannot Run on V100: float16 crashes, float32 fails during validation

> > pip uninstall -y flash-attn flash_attn || true > > After following your method, I encountered an error:`ImportError: /home/miniconda3/envs/llm_torch26/lib/python3.10/site-packages/flash_attn_cuda.cpython-310-x86_64-linux-gnu.so: undefined symbol: _ZN3c105ErrorC2ENS_14SourceLocationESs`. Do you know how to solve them?...