feat: Support MPS during training and inference #3100
+339
−103
Conversation
This file contains 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
…ult option when running from mac (if availble)
ori-kron-wis
added
cuda tests
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #3100 +/- ##
==========================================
- Coverage 87.67% 83.20% -4.48%
==========================================
Files 180 180
Lines 15187 15227 +40
==========================================
- Hits 13315 12669 -646
- Misses 1872 2558 +686
|
|
Please enable the Mac Runner for changes about MPS. |
|
A comparison for training CPU, MPS M3 AND NVidia RTX 6000 Ada 48GB GDDR6 |
|
The difference was much bigger in my hands when increasing batch size (larger batch size creates larger matrix multiplications where MPS is more efficient than CPU) and compilation. It was actually faster than an A100 on Google Colab then. We should optimize this a bit. |
canergen
reviewed
Dec 18, 2024
| np.asarray(1.0 - (data > 0).sum(0) / data.shape[0]).ravel() | ||
| ).to(device) | ||
| # in MPS we need to first change to float 32, as the MPS framework doesn't support float64. | ||
| if device.type == "mps": |
There was a problem hiding this comment.
We could also do it for other devices. Float32 is sufficient for all computations.
for more information, see https://pre-commit.ci
…vi-tools into Ori-broadcast_all_fix
for more information, see https://pre-commit.ci
canergen
changed the title
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Outdated suggestions cannot be applied.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.
add support for Mac GPU (m1,m2,m3) for scvi and revert it to be default option when running from Mac (if available)
Testing are done manually on my MAC (has m3) + verification that still works on CPU/CUDA here
References:
pytorch/pytorch#132605
pytorch/pytorch#77764
https://discourse.scverse.org/t/macbook-m1-m2-mps-acceleration-with-scvi/2075/7