-
Notifications
You must be signed in to change notification settings - Fork 722
Enter training time for BF16 using AVX512 #2620
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
base: development
Are you sure you want to change the base?
Conversation
Update IntelPyTorch training notebook to include environment variable settings for AVX512 and BF16 optimizations Signed-off-by: Francisco R Castro G <[email protected]> Signed-off-by: jafraustro <[email protected]>
@jafraustro can you please verify if the environment variable is actually affecting the It would to ideal to have the # In jupyter notebook cell
os.environ["ONEDNN_MAX_CPU_ISA"] = "AVX512_CORE_BF16"
!python pytorch_training_avx512_bf16.py I also see |
fac0bd8
to
f808ec9
Compare
- BF16 without Intel® AMX example is executed from a python script. The bf16_noAmx_training_time variables is saved in a txt file. - Use torch.amp.autocast('cpu') instead of torch.cpu.amp.autocast() Signed-off-by: jafraustro <[email protected]>
Signed-off-by: jafraustro <[email protected]>
Hi @Ankur-singh, You are right about the ONEDNN_MAX_CPU_ISA variable. I made the following changes:
|
Signed-off-by: jafraustro <[email protected]>
Signed-off-by: jafraustro <[email protected]>
Update IntelPyTorch training notebook to include environment variable settings for AVX512 and BF16 optimizations
Existing Sample Changes
Description
Update IntelPyTorch training notebook to include environment variable settings for AVX512 and BF16 optimizations
How Has This Been Tested?
Please describe the tests that you ran to verify your changes. Provide instructions so we can reproduce. Please also list any relevant details for your test configuration