Feng Yunlong comments

Results 91 comments of


                                            Feng Yunlong

Running tokenizer on dataset 速度逐渐变慢

经过我的实际测试，#5458 应该解决了这个问题

Running tokenizer on dataset 速度逐渐变慢

没有 packing 也会下降到个位数吗？按理说应该不会吧

Fix pre-compile on cpu-only machines

> @AlongWY, thanks for the PR. Can you please share the motivation to help us understand your use case? I used GitHub Actions to precompile DeepSpeed, but after upgrading to...

Fix pre-compile on cpu-only machines

Yes, the GitHub Action is compiling in an Ubuntu environment with CUDA installed but without a GPU. The GitHub Action runs only on Ubuntu, but the error still occurs.

Fix pre-compile on cpu-only machines

Yes, the triton has been installed in the environment. # Error for triton # Error for type

Fix pre-compile on cpu-only machines

Any update?

Fix pre-compile on cpu-only machines

@sfc-gh-truwase Hi, i have update the code. For the two questions: 1. I use GitHub Actions to automatically build wheel packages for different environments, so I manually set the environment...

Build sdist and wheel in CI

I think this [repo](https://github.com/AlongWY/apex_wheels) is a better implementation to build apex wheels.

使用neat_packing进行sft训练，模型性能指标下降明显

我可能找到了问题所在，解决方案在 #5458

关于解决ModuleNotFoundError: No module named 'torch'导致安装失败

try this one 试试这个？[apex wheels](https://github.com/AlongWY/apex_wheels/releases/tag/v24.4.1)