1luik

Results 10 comments of 1luik

backend-python/dep_check.py", line 1, in <module> import multipart ModuleNotFoundError: No module named 'multipart'

你可以自己按一个啊

[Bug] After adding import unsloth to the first line of the script, the GRPOTrainer fails to run properly; however, it works normally again once this import is removed. The Sophia optimizer interface being used was generated by an AI.

FakeTensor(..., device='cuda:0', size=(1, s17, s6), dtype=torch.float16, requires_grad=True) ), GradTrackingTensor(lvl=1, value= FakeTensor(..., device='cuda:0', size=(1024, 101306), dtype=torch.float16) )), **{}): got RuntimeError('a and b must have same reduction dim, but got [s17, s6]...

[Bug] After adding import unsloth to the first line of the script, the GRPOTrainer fails to run properly; however, it works normally again once this import is removed. The Sophia optimizer interface being used was generated by an AI.

torch 2.8.0 cuda

[Bug] After adding import unsloth to the first line of the script, the GRPOTrainer fails to run properly; however, it works normally again once this import is removed. The Sophia optimizer interface being used was generated by an AI.

[xl_bug_report.py](https://github.com/user-attachments/files/23505500/xl_bug_report.py)

[Bug] After adding import unsloth to the first line of the script, the GRPOTrainer fails to run properly; however, it works normally again once this import is removed. The Sophia optimizer interface being used was generated by an AI.

> 你好 [@1luik](https://github.com/1luik)如果您希望我们进行故障排除，我们需要一个更简化的脚本。一般来说，虽然 Unsloth 的 GRPOTrainer 与 trl 的界面类似，但它们的运行方式有所不同。您添加了很多自定义设置，因此需要找出是哪个自定义设置导致了问题。我建议您首先尝试移除尽可能多的自定义设置，看看 Unsloth 是否能正常工作，然后再逐个添加回去。 No need to rush to fix it; if you don't add import unsloth in the first line, there won't...

[Bug] After adding import unsloth to the first line of the script, the GRPOTrainer fails to run properly; however, it works normally again once this import is removed. The Sophia optimizer interface being used was generated by an AI.

It's being used in token_utils, and I haven't tried other models.

[Bug] After adding import unsloth to the first line of the script, the GRPOTrainer fails to run properly; however, it works normally again once this import is removed. The Sophia optimizer interface being used was generated by an AI.

There is no difference

[Bug] After adding import unsloth to the first line of the script, the GRPOTrainer fails to run properly; however, it works normally again once this import is removed. The Sophia optimizer interface being used was generated by an AI.

It's too late, I'm going to bed. It's already 2 a.m. I'll test tomorrow

Server Not Loading : lateinit property CoreHooks has not been initialized

I'm extremely frustrated that this bug requires me to restart the server multiple times

when use GRPO+ deepspeed_zero3 + ds3_gather_for_generation=False, stuck in the training stage, step is still 0 after an hour

I also got stuck. What tool did you use in the end