NVIDIA / NVIDIA/TransformerEngine
Hadamard transform not working on SM120
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 3.5k
- Forks
- 831
- Avg merge
- 3d 11h
- Merged PRs (30d)
- 65
Description
For SM120, with disable_rht=True being set in transformer_engine.common.recipe.NVFP4BlockScaling, the code works fine, but when disable_rht=False is set, the code below will result in cuda error
with te.fp8_autocast(enabled=True, fp8_recipe=fp4_recipe):
out_fp4 = linear(x)
will produce the error:
Running NVFP4 forward pass...
Error: Failed to set Shared Memory size.
torch.AcceleratorError: CUDA error: invalid argument
Is Hadamard transform only available for SM100?
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 transformer_engine.common.recipe.NVFP4BlockScaling and the NVFP4 path exercised by te.fp8_autocast with disable_rht=False. Reproduce the linear(x) example on SM120 and compare it with disable_rht=True, focusing on the “Failed to set Shared Memory size” CUDA error. Done means determining whether Hadamard transforms are supported on SM120 and documenting or correcting the behavior.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, pytorch
- Domain
- machine-learning, performance
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100