Make 8-bit matrix multiplication compatible with cpuonly builds #22
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
mryab
commented
Aug 22, 2022
|
|
||
| matmul = MatMul8bitLt.apply |
There was a problem hiding this comment.
This gets immediately redefined below
mryab
commented
Aug 22, 2022
Comment on lines
-1688
to
-1699
| def get_special_format_str(): | ||
| major, minor = torch.cuda.get_device_capability() | ||
| if major < 7: | ||
| print( | ||
| f"Device with CUDA capability of {major} not supported for 8-bit matmul. Device has no tensor cores!" | ||
| ) | ||
| assert major >= 7 | ||
|
|
||
| if major == 7: return 'col_turing' | ||
| elif major == 8: return 'col_ampere' | ||
| else: return 'col_turing' | ||
|
|
There was a problem hiding this comment.
This is already defined above (L186)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
While trying to use the latest release of bitsandbytes in hivemind, I faced an error after installing the library in a cpuonly mode (both from PyPI and from source): https://github.com/learning-at-home/hivemind/runs/7961913748?check_suite_focus=true#step:8:56
The gist of this error seems to be unprotected usage of
torch.cuda.get_device_capability(), which fails for non-GPU builds of PyTorch. To handle this, I've moved the corresponding imports in the root of bitsandbytes inside the COMPILED_WITH_CUDA block and also moved the relevant function/class declarations inside the similar block to avoid accidental imports.The solution appears to work locally (at least the tests pass), happy to improve on the PR if any additional work is needed.