[AOTI] Fix a weight loading issue when the weight size can be 0 #114280
Conversation
Summary: When a weight tensor is 0-size, no device memory should be allocated for it. This PR fixes the weight loading logic for such a case. This problem was found when running the 14K model test. [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/114280
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 7bbb37d with merge base 54d0455 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
Summary: When a weight tensor is 0-size, no device memory should be allocated for it. This PR fixes the weight loading logic for such a case. This problem was found when running the 14K model test. ghstack-source-id: 852d91a2216a49361899396366559c2c6f85ad90 Pull Request resolved: #114280
… be 0" Summary: When a weight tensor is 0-size, no device memory should be allocated for it. This PR fixes the weight loading logic for such a case. This problem was found when running the 14K model test. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx peterbell10 ipiszy yf225 chenyang78 kadeng muchulee8 aakhundov ColinPeppler [ghstack-poisoned]
Summary: When a weight tensor is 0-size, no device memory should be allocated for it. This PR fixes the weight loading logic for such a case. This problem was found when running the 14K model test. ghstack-source-id: cb49231f5be3f38a32ed187212bd92b43505657f Pull Request resolved: #114280
|
@pytorchbot merge |
Merge failedReason: This PR needs a If not, please add the To add a label, you can comment to pytorchbot, for example For more information, see Details for Dev Infra teamRaised by workflow job |
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
…rch#114280) Summary: When a weight tensor is 0-size, no device memory should be allocated for it. This PR fixes the weight loading logic for such a case. This problem was found when running the 14K model test. Pull Request resolved: pytorch#114280 Approved by: https://github.com/chenyang78
Stack from ghstack (oldest at bottom):
Summary: When a weight tensor is 0-size, no device memory should be allocated for it. This PR fixes the weight loading logic for such a case. This problem was found when running the 14K model test.
cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @peterbell10 @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @aakhundov @ColinPeppler