[QUESTION] How to set fsdp_unit_modules
Open
question
- Dominant language
- Python
- Stars
- 17.9k
- Forks
- 4.5k
- Avg merge
- 4d 6h
- Merged PRs (30d)
- 271
Description
**Your question**
Ask a clear and concise question about Megatron-LM.
Hi, I want to run FSDP, and set fsdp_unit_modules as [TERowParallelLinear, TELayerNormColumnParallelLinear], then I got the error.
[rank0]: File "/usr/local/lib/python3.10/dist-packages/triton/backends/nvidia/driver.py", line 354, in __call__
[rank0]: self.launch(*args, **kwargs)
[rank0]: ValueError: Pointer argument (at 3) cannot be accessed from Triton (cpu tensor?)
E0609 07:05:55.980000 139741756354304 torch/distributed/elastic/multiprocessing/api.py:881] failed (exitcode: 1) local_rank: 0 (pid: 47233) of binary: /usr/bin/python
Contributor guide
Assessment
This issue has not been assessed yet.