mirror of
https://github.com/saymrwulf/pytorch.git
synced 2026-05-14 20:57:59 +00:00
71ebe5121a
Allow users to decide whether they want to have fast math enabled via env var Pull Request resolved: https://github.com/pytorch/pytorch/pull/129007 Approved by: https://github.com/malfet ghstack dependencies: #129006, #129008
41 lines
No EOL
2 KiB
ReStructuredText
41 lines
No EOL
2 KiB
ReStructuredText
.. _mps_environment_variables:
|
|
|
|
MPS Environment Variables
|
|
==========================
|
|
|
|
**PyTorch Environment Variables**
|
|
|
|
.. list-table::
|
|
:header-rows: 1
|
|
|
|
* - Variable
|
|
- Description
|
|
* - ``PYTORCH_DEBUG_MPS_ALLOCATOR``
|
|
- If set to ``1``, set allocator logging level to verbose.
|
|
* - ``PYTORCH_MPS_HIGH_WATERMARK_RATIO``
|
|
- High watermark ratio for MPS allocator. By default, it is set to 1.7.
|
|
* - ``PYTORCH_MPS_LOW_WATERMARK_RATIO``
|
|
- Low watermark ratio for MPS allocator. By default, it is set to 1.4 if the memory is unified and set to 1.0 if the memory is discrete.
|
|
* - ``PYTORCH_MPS_FAST_MATH``
|
|
- If set to ``1``, enable fast math for MPS metal kernels. See section 1.6.3 in https://developer.apple.com/metal/Metal-Shading-Language-Specification.pdf for precision implications.
|
|
* - ``PYTORCH_MPS_PREFER_METAL``
|
|
- If set to ``1``, force using metal kernels instead of using MPS Graph APIs. For now this is only used for matmul op.
|
|
* - ``PYTORCH_ENABLE_MPS_FALLBACK``
|
|
- If set to ``1``, full back operations to CPU when MPS does not support them.
|
|
|
|
.. note::
|
|
|
|
**high watermark ratio** is a hard limit for the total allowed allocations
|
|
|
|
- `0.0` : disables high watermark limit (may cause system failure if system-wide OOM occurs)
|
|
- `1.0` : recommended maximum allocation size (i.e., device.recommendedMaxWorkingSetSize)
|
|
- `>1.0`: allows limits beyond the device.recommendedMaxWorkingSetSize
|
|
|
|
e.g., value 0.95 means we allocate up to 95% of recommended maximum
|
|
allocation size; beyond that, the allocations would fail with OOM error.
|
|
|
|
**low watermark ratio** is a soft limit to attempt limiting memory allocations up to the lower watermark
|
|
level by garbage collection or committing command buffers more frequently (a.k.a, adaptive commit).
|
|
Value between 0 to m_high_watermark_ratio (setting 0.0 disables adaptive commit and garbage collection)
|
|
e.g., value 0.9 means we 'attempt' to limit allocations up to 90% of recommended maximum
|
|
allocation size. |