PatchDPOTrainer bug

[sihaogu32]

Dear Author,

I am currently working with the code provided at [https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Zephyr_(7B)-DPO.ipynb#scrollTo=E8-BWi7MzkRz]. My operating environment is a Linux server equipped with an NVIDIA 4090 GPU.

Upon executing the following lines:

# One must patch the DPO Trainer first!
from unsloth import PatchDPOTrainer

PatchDPOTrainer()

I encountered the following error:

/tmp/tmp9d28jrm_/main.c:5:10: fatal error: Python.h: No such file or directory
    5 | #include <Python.h>
      |          ^~~~~~~~~~
compilation terminated.
---------------------------------------------------------------------------
CalledProcessError                        Traceback (most recent call last)
Cell In[8], line 2
      1 # One must patch the DPO Trainer first!
----> 2 from unsloth import PatchDPOTrainer
      4 PatchDPOTrainer()

........
File /usr/lib64/python3.12/subprocess.py:413, in check_call(*popenargs, **kwargs)
    411     if cmd is None:
    412         cmd = popenargs[0]
--> 413     raise CalledProcessError(retcode, cmd)
    414 return 0

CalledProcessError: Command '['/usr/bin/gcc', '/tmp/tmp9d28jrm_/main.c', '-O3', '-shared', '-fPIC', '-Wno-psabi', '-o', '/tmp/tmp9d28jrm_/cuda_utils.cpython-312-x86_64-linux-gnu.so', '-lcuda', '-L/home/sihaugo/.venv/lib/python3.12/site-packages/triton/backends/nvidia/lib', '-L/lib64', '-I/home/sihaugo/.venv/lib/python3.12/site-packages/triton/backends/nvidia/include', '-I/tmp/tmp9d28jrm_', '-I/usr/include/python3.12']' returned non-zero exit status 1.

I have confirmed that python3-dev is installed on my system. However, this did not resolve the issue. I would appreciate any guidance or solutions you could provide to address this problem. Thank you.

Best regards,

sihaogu

[erlonted]

I can solve this problem running : sudo apt-get install python3-dev

[danielhanchen]

Hope this was resolved? Sorry on the issue!

[sihaogu32]

The issue was eventually resolved. It seems it was due to the system engineer installing the wrong version of the package, likely because the system is CentOS. Thank you for your response. Daniel Han ***@***.***> 於 2025年2月10日 週一 下午9:18寫道：

…

Hope this was resolved? Sorry on the issue! — Reply to this email directly, view it on GitHub <#1611 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AY5ENHFSRME6M4R6FFPIWR32PCRLNAVCNFSM6AAAAABWOEWFDWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMNBXHE3DMMZTGY> . You are receiving this because you authored the thread.Message ID: ***@***.***>