-
Notifications
You must be signed in to change notification settings - Fork 446
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
Enable training on XPU devices in OTX2.0 #3094
Conversation
I updated strategy: 98e0b69 removing torch.xpu.optimize for Semantic Segmentation since it was reported by IPEX that they have a bug there in that case and to be able to train segmentation model we should remove optimization. |
@harimkang , seems like a problem with installing mmdet occurs during set up env for unit tests. Could you take a look? |
There was a brief issue with the CI network. I rerun the test and the install is fine. |
* add raising an error when metric is None * added accelerators * fix packages * fix assigning model * debug on MAX * change precision * update MixedPrecisionXPUPlugin * debug * added monkey patching * minor * minor * added patch for mmengine * fix OD and IS * benchmark debug * change device * quick fix for instance seg * fix pre-commit * fix pre-commit * clean the code * added additional flag for mmcv * added unit tests * fixed unit test * fix linter * added unit tests and replied comments * fix pre-commit * minor fix * added documentation * fix unit test * add workaround for semantic segmentation * remove RoiAlignTest due to unstability * minor * remove strategy back * try to patch SingleDeviceStrategy * added auto xpu configuration * patch strategy * small fix * reply to comments * move patching xpu packages to accelerator * fix test_xpu test * remove do-not-install-mmcv * fix pre-commit * remove torch.xpu.optimize for segmentation --------- Co-authored-by: Emily <emily.chun@intel.com>
* Enable training on XPU devices in OTX2.0 (#3094) * add raising an error when metric is None * added accelerators * fix packages * fix assigning model * debug on MAX * change precision * update MixedPrecisionXPUPlugin * debug * added monkey patching * minor * minor * added patch for mmengine * fix OD and IS * benchmark debug * change device * quick fix for instance seg * fix pre-commit * fix pre-commit * clean the code * added additional flag for mmcv * added unit tests * fixed unit test * fix linter * added unit tests and replied comments * fix pre-commit * minor fix * added documentation * fix unit test * add workaround for semantic segmentation * remove RoiAlignTest due to unstability * minor * remove strategy back * try to patch SingleDeviceStrategy * added auto xpu configuration * patch strategy * small fix * reply to comments * move patching xpu packages to accelerator * fix test_xpu test * remove do-not-install-mmcv * fix pre-commit * remove torch.xpu.optimize for segmentation --------- Co-authored-by: Emily <emily.chun@intel.com> * Add exporter/demo unit tests (#3218) * added unit tests. Need to clean up * move tests * fix pre-commit * return demo back * minor * delete unnecessery comments * fix unit test * fix pre-commit * fix pre-commit 2 * fix test_postprocess_openvino_model * fix unit tests * test_precommit * Fix a bug that engine.test doesn't work with XPU (#3293) * fix bug * align with pre-commit --------- Co-authored-by: Emily <emily.chun@intel.com> * fix merge conflicts for pre-commit * fix precommit 2 * fix unit test * fix pre-commit * fix export tests * fix pre-commit * fix tox * fix pre-commit --------- Co-authored-by: Emily <emily.chun@intel.com> Co-authored-by: Eunwoo Shin <eunwoo.shin@intel.com>
Summary
How to test
Checklist
License
Feel free to contact the maintainers if that's a concern.