rsnt_translations
56,420
edits
PyTorch: Difference between revisions
Jump to navigation
Jump to search
no edit summary
No edit summary |
No edit summary |
||
| Line 102: | Line 102: | ||
<!--T:549--> | <!--T:549--> | ||
As of October 2022, our only cluster equipped with Ampere GPUs is [ | As of October 2022, our only cluster equipped with Ampere GPUs is [[Narval]]. When using PyTorch on Narval, users should be cognizant of the following: | ||
# You may notice a significant slowdown when running the exact same GPU-enabled code with <code>torch < 1.12.0</code> and <code>torch >= 1.12.0</code>. | # You may notice a significant slowdown when running the exact same GPU-enabled code with <code>torch < 1.12.0</code> and <code>torch >= 1.12.0</code>. | ||
# You may get different results when running the exact same GPU-enabled code with <code>torch < 1.12.0</code> and <code>torch >= 1.12.0</code>. | # You may get different results when running the exact same GPU-enabled code with <code>torch < 1.12.0</code> and <code>torch >= 1.12.0</code>. | ||