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
make torch.amp.autocast more generic #125103
Commits on Apr 27, 2024
-
Configuration menu - View commit details
-
Copy full SHA for 054a451 - Browse repository at this point
Copy the full SHA 054a451View commit details
Commits on Apr 28, 2024
-
Update on "[WIP] make torch.amp.autocast more generic"
cc mcarilli ptrblck leslie-fang-intel jgong5 ezyang msaroufim bdhirsh anijain2305 chauhang voznesenskym penguinwu EikanWang Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng [ghstack-poisoned]
Configuration menu - View commit details
-
Copy full SHA for 2b6089e - Browse repository at this point
Copy the full SHA 2b6089eView commit details
Commits on Apr 29, 2024
-
Update on "[WIP] make torch.amp.autocast more generic"
cc mcarilli ptrblck leslie-fang-intel jgong5 ezyang msaroufim bdhirsh anijain2305 chauhang voznesenskym penguinwu EikanWang Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng [ghstack-poisoned]
Configuration menu - View commit details
-
Copy full SHA for c6e6fac - Browse repository at this point
Copy the full SHA c6e6facView commit details -
Update on "[WIP] make torch.amp.autocast more generic"
cc mcarilli ptrblck leslie-fang-intel jgong5 ezyang msaroufim bdhirsh anijain2305 chauhang voznesenskym penguinwu EikanWang Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng [ghstack-poisoned]
Configuration menu - View commit details
-
Copy full SHA for 58cabab - Browse repository at this point
Copy the full SHA 58cababView commit details
Commits on May 6, 2024
-
Update on "[WIP] make torch.amp.autocast more generic"
cc mcarilli ptrblck leslie-fang-intel jgong5 ezyang msaroufim bdhirsh anijain2305 chauhang voznesenskym penguinwu EikanWang Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng [ghstack-poisoned]
Configuration menu - View commit details
-
Copy full SHA for a49c133 - Browse repository at this point
Copy the full SHA a49c133View commit details -
Update on "[WIP] make torch.amp.autocast more generic"
cc mcarilli ptrblck leslie-fang-intel jgong5 ezyang msaroufim bdhirsh anijain2305 chauhang voznesenskym penguinwu EikanWang Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng [ghstack-poisoned]
Configuration menu - View commit details
-
Copy full SHA for 7e129ca - Browse repository at this point
Copy the full SHA 7e129caView commit details
Commits on May 7, 2024
-
Update on "make torch.amp.autocast more generic"
# Motivation As discussed in [#124479](#124479), `torch.amp.autocast` can NOT be completely equivalent to `torch.cuda.amp.autocast` and `torch.cpu.amp.autocast` since `torch.amp.autocast` has NOT the default `dtype` for CPU (`torch.bfloat16` by default) and CUDA (`torch.float16` by default) respectively. We would like `torch.amp.autocast` to be more generic to help the developer/customer write the device-agnostic code. Because there are not enough reasons to add device-specific autocast `torch.xxx.amp.autocast` for each device backend. # Solution When `None` is passed to `dtype`, we should use `torch.get_autocast_dtype` to get the related dtype for each backend. Meanwhile, `torch.get_autocast_dtype` is necessary to be supported in JIT path for BC. # Additional Context With this PR, `torch.amp.autocast(device_type='cuda')` is equivalent to `torch.cuda.amp.autocast`. cc mcarilli ptrblck leslie-fang-intel jgong5 ezyang msaroufim bdhirsh anijain2305 chauhang voznesenskym penguinwu EikanWang Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng [ghstack-poisoned]
Configuration menu - View commit details
-
Copy full SHA for b00a0bd - Browse repository at this point
Copy the full SHA b00a0bdView commit details -
Update on "make torch.amp.autocast more generic"
# Motivation As discussed in [#124479](#124479), `torch.amp.autocast` can NOT be completely equivalent to `torch.cuda.amp.autocast` and `torch.cpu.amp.autocast` since `torch.amp.autocast` has NOT the default `dtype` for CPU (`torch.bfloat16` by default) and CUDA (`torch.float16` by default) respectively. We would like `torch.amp.autocast` to be more generic to help the developer/customer write the device-agnostic code. Because there are not enough reasons to add device-specific autocast `torch.xxx.amp.autocast` for each device backend. # Solution When `None` is passed to `dtype`, we should use `torch.get_autocast_dtype` to get the related dtype for each backend. Meanwhile, `torch.get_autocast_dtype` is necessary to be supported in JIT path for BC. # Additional Context With this PR, `torch.amp.autocast(device_type='cuda')` is equivalent to `torch.cuda.amp.autocast`. cc mcarilli ptrblck leslie-fang-intel jgong5 ezyang msaroufim bdhirsh anijain2305 chauhang voznesenskym penguinwu EikanWang Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng [ghstack-poisoned]
Configuration menu - View commit details
-
Copy full SHA for 73d9468 - Browse repository at this point
Copy the full SHA 73d9468View commit details -
Update on "make torch.amp.autocast more generic"
# Motivation As discussed in [#124479](#124479), `torch.amp.autocast` can NOT be completely equivalent to `torch.cuda.amp.autocast` and `torch.cpu.amp.autocast` since `torch.amp.autocast` has NOT the default `dtype` for CPU (`torch.bfloat16` by default) and CUDA (`torch.float16` by default) respectively. We would like `torch.amp.autocast` to be more generic to help the developer/customer write the device-agnostic code. Because there are not enough reasons to add device-specific autocast `torch.xxx.amp.autocast` for each device backend. # Solution When `None` is passed to `dtype`, we should use `torch.get_autocast_dtype` to get the related dtype for each backend. Meanwhile, `torch.get_autocast_dtype` is necessary to be supported in JIT path for BC. # Additional Context With this PR, `torch.amp.autocast(device_type='cuda')` is equivalent to `torch.cuda.amp.autocast`. Add two new UTs to cover this change in eager and jit path respectively. cc mcarilli ptrblck leslie-fang-intel jgong5 ezyang msaroufim bdhirsh anijain2305 chauhang voznesenskym penguinwu EikanWang Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng [ghstack-poisoned]
Configuration menu - View commit details
-
Copy full SHA for d20ca90 - Browse repository at this point
Copy the full SHA d20ca90View commit details
Commits on May 8, 2024
-
Update on "make torch.amp.autocast more generic"
# Motivation As discussed in [#124479](#124479), `torch.amp.autocast` can NOT be completely equivalent to `torch.cuda.amp.autocast` and `torch.cpu.amp.autocast` since `torch.amp.autocast` has NOT the default `dtype` for CPU (`torch.bfloat16` by default) and CUDA (`torch.float16` by default) respectively. We would like `torch.amp.autocast` to be more generic to help the developer/customer write the device-agnostic code. Because there are not enough reasons to add device-specific autocast `torch.xxx.amp.autocast` for each device backend. # Solution When `None` is passed to `dtype`, we should use `torch.get_autocast_dtype` to get the related dtype for each backend. Meanwhile, `torch.get_autocast_dtype` is necessary to be supported in JIT path for BC. # Additional Context With this PR, `torch.amp.autocast(device_type='cuda')` is equivalent to `torch.cuda.amp.autocast`. Add two new UTs to cover this change in eager and jit path respectively. cc mcarilli ptrblck leslie-fang-intel jgong5 ezyang msaroufim bdhirsh anijain2305 chauhang voznesenskym penguinwu EikanWang Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng [ghstack-poisoned]
Configuration menu - View commit details
-
Copy full SHA for 4e211cd - Browse repository at this point
Copy the full SHA 4e211cdView commit details -
Update on "make torch.amp.autocast more generic"
# Motivation As discussed in [#124479](#124479), `torch.amp.autocast` can NOT be completely equivalent to `torch.cuda.amp.autocast` and `torch.cpu.amp.autocast` since `torch.amp.autocast` has NOT the default `dtype` for CPU (`torch.bfloat16` by default) and CUDA (`torch.float16` by default) respectively. We would like `torch.amp.autocast` to be more generic to help the developer/customer write the device-agnostic code. Because there are not enough reasons to add device-specific autocast `torch.xxx.amp.autocast` for each device backend. # Solution When `None` is passed to `dtype`, we should use `torch.get_autocast_dtype` to get the related dtype for each backend. Meanwhile, `torch.get_autocast_dtype` is necessary to be supported in JIT path for BC. # Additional Context With this PR, `torch.amp.autocast(device_type='cuda')` is equivalent to `torch.cuda.amp.autocast`. Add two new UTs to cover this change in eager and jit path respectively. cc mcarilli ptrblck leslie-fang-intel jgong5 ezyang msaroufim bdhirsh anijain2305 chauhang voznesenskym penguinwu EikanWang Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng [ghstack-poisoned]
Configuration menu - View commit details
-
Copy full SHA for 63d4d7e - Browse repository at this point
Copy the full SHA 63d4d7eView commit details -
Update on "make torch.amp.autocast more generic"
# Motivation As discussed in [#124479](#124479), `torch.amp.autocast` can NOT be completely equivalent to `torch.cuda.amp.autocast` and `torch.cpu.amp.autocast` since `torch.amp.autocast` has NOT the default `dtype` for CPU (`torch.bfloat16` by default) and CUDA (`torch.float16` by default) respectively. We would like `torch.amp.autocast` to be more generic to help the developer/customer write the device-agnostic code. Because there are not enough reasons to add device-specific autocast `torch.xxx.amp.autocast` for each device backend. # Solution When `None` is passed to `dtype`, we should use `torch.get_autocast_dtype` to get the related dtype for each backend. Meanwhile, `torch.get_autocast_dtype` is necessary to be supported in JIT path for BC. # Additional Context With this PR, `torch.amp.autocast(device_type='cuda')` is equivalent to `torch.cuda.amp.autocast`. Add two new UTs to cover this change in eager and jit path respectively. cc mcarilli ptrblck leslie-fang-intel jgong5 ezyang msaroufim bdhirsh anijain2305 chauhang voznesenskym penguinwu EikanWang Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng [ghstack-poisoned]
Configuration menu - View commit details
-
Copy full SHA for e11d24b - Browse repository at this point
Copy the full SHA e11d24bView commit details