Add API to configure memory allocator settings #453
Merged
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
- Added the Python API torch.mps.memory.set_allocator_settings(str) to adjust high/low watermark ratios, small/large/xlarge heap size divisors, max pow2 roundup size, and debug verbosity of MPSAllocator messages. - Added the env-var "PYTORCH_MPS_ALLOC_CONF" to pass the settings as an alternative to API set_allocator_settings() (similar to CUDA). - Removed the old env-vars PYTORCH_DEBUG_MPS_ALLOCATOR, PYTORCH_MPS_HIGH_WATERMARK_RATIO, and PYTORCH_MPS_LOW_WATERMARK_RATIO - Fixed the bug with total_memory_allocated_size to count the total size of MTLHeaps and not MTLBuffers (in case heaps aren't fully occupied)
razarmehr
requested review from
kulinseth,
Ronian526,
ssaladis and
DenisVieriu97
kulinseth
approved these changes
Jun 26, 2023
jhavukainen
pushed a commit
that referenced
this pull request
Mar 15, 2024
- Added the Python API torch.mps.memory.set_allocator_settings(str) to adjust high/low watermark ratios, small/large/xlarge heap size divisors, max pow2 roundup size, and debug verbosity of MPSAllocator messages. - Added the env-var "PYTORCH_MPS_ALLOC_CONF" to pass the settings as an alternative to API set_allocator_settings() (similar to CUDA). - Removed the old env-vars PYTORCH_DEBUG_MPS_ALLOCATOR, PYTORCH_MPS_HIGH_WATERMARK_RATIO, and PYTORCH_MPS_LOW_WATERMARK_RATIO - Fixed the bug with total_memory_allocated_size to count the total size of MTLHeaps and not MTLBuffers (in case heaps aren't fully occupied)
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Outdated suggestions cannot be applied.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.
torch.mps.memory.set_allocator_settings(str)to adjust high/low watermark ratios, small/large/xlarge heap size divisors, max pow2 roundup size, and debug verbosity of MPSAllocator messages.PYTORCH_MPS_ALLOC_CONF=[settings string]to pass the settings string as an alternative to the APIset_allocator_settings()(similar to CUDA).PYTORCH_DEBUG_MPS_ALLOCATOR,PYTORCH_MPS_HIGH_WATERMARK_RATIO, andPYTORCH_MPS_LOW_WATERMARK_RATIO.total_memory_allocated_sizeto count the total size of MTLHeaps and not MTLBuffers (in case heaps aren't fully occupied).