File: cudnn_rnn_determinism.rst

package info (click to toggle)
pytorch 1.13.1%2Bdfsg-4
  • links: PTS, VCS
  • area: main
  • in suites: bookworm
  • size: 139,252 kB
  • sloc: cpp: 1,100,274; python: 706,454; ansic: 83,052; asm: 7,618; java: 3,273; sh: 2,841; javascript: 612; makefile: 323; xml: 269; ruby: 185; yacc: 144; objc: 68; lex: 44
file content (16 lines) | stat: -rw-r--r-- 641 bytes parent folder | download | duplicates (3)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
.. warning::
    There are known non-determinism issues for RNN functions on some versions of cuDNN and CUDA.
    You can enforce deterministic behavior by setting the following environment variables:

    On CUDA 10.1, set environment variable ``CUDA_LAUNCH_BLOCKING=1``.
    This may affect performance.

    On CUDA 10.2 or later, set environment variable
    (note the leading colon symbol)
    ``CUBLAS_WORKSPACE_CONFIG=:16:8``
    or
    ``CUBLAS_WORKSPACE_CONFIG=:4096:2``.

    See the `cuDNN 8 Release Notes`_ for more information.

.. _cuDNN 8 Release Notes: https://docs.nvidia.com/deeplearning/sdk/cudnn-release-notes/rel_8.html