deepspeedai / deepspeedai/DeepSpeed
RuntimeError: cannot pin 'CUDABFloat16Type' only dense CPU tensors can be pinned
Open
Nobody has claimed this yet.
bug
training
- Dominant language
- Python
- Stars
- 43.1k
- Forks
- 5k
- Avg merge
- 4d 15h
- Merged PRs (30d)
- 112
Description
Describe the bug
self.param_groups_fp16_flat_cpu_memory.append(get_accelerator().pin_memory(
File "/usr/local/python/lib/python3.8/site-packages/deepspeed/accelerator/cuda_accelerator.py", line 291, in pin_memory
self.param_groups_fp16_flat_cpu_memory.append(get_accelerator().pin_memory(
File "/usr/local/python/lib/python3.8/site-packages/deepspeed/accelerator/cuda_accelerator.py", line 291, in pin_memory
return tensor.pin_memory()
RuntimeError: cannot pin 'CUDABFloat16Type' only dense CPU tensors can be pinned
return tensor.pin_memory()
when i use deepspeed zero3 to train model ,i meet this problem ,can someone help me?
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with deepspeed/accelerator/cuda_accelerator.py at pin_memory and trace the ZeRO-3 path that calls it. Reproduce the reported traceback and inspect the tensor device and type before pin_memory; done means the cause is confirmed with a focused reproduction and the behavior is corrected or clearly documented.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, pytorch
- Domain
- distributed-systems, machine-learning
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100