deepspeedai / deepspeedai/DeepSpeed
ZeRO 3: IndexError: Dimension out of range (expected to be in range of [-1, 0], but got 1)
Open
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 43.1k
- Forks
- 5k
- Avg merge
- 4d 15h
- Merged PRs (30d)
- 112
Description
If I use the following snippet in BingBERT example with ZeRO 3, I received IndexError: Dimension out of range (expected to be in range of [-1, 0], but got 1) no matter if Transformer CUDA layer is enabled or not.
with deepspeed.zero.Init(remote_device='cpu', pin_memory=True, enabled=True):
model = BertMultiTask(args)
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 by reproducing the issue in the BingBERT example with ZeRO 3 and the shown BertMultiTask construction inside zero.Init. Compare behavior with Transformer CUDA enabled and disabled, then trace the reported dimension error. Done means the snippet no longer raises the IndexError in either configuration.
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