pytorch

frozenleaves/pytorch

Fork 0

mirror of https://github.com/pytorch/pytorch.git synced 2025-10-20 21:14:14 +08:00

Commit Graph

Author	SHA1	Message	Date
Kevin Fu	04349f9ee5	[PT2]: Skip AOTI Weight Loading during Init (#158416 ) Summary: AOTI already has weights embedded in .so file. So for the initial load, no need to load the weights again. This allows lowered modules can have different set of weights on different hardwares. Test Plan: ``` MODEL_TYPE=ads_mtml_offsite_cvr_oba_optout_dedicated_model MODEL_ENTITY_ID=895279202 SNAPSHOT_ID=0 MODULE=merge buck2 run mode/dev-nosan -c fbcode.nvcc_arch=a100,h100 -c fbcode.enable_gpu_sections=true fbcode//caffe2/torch/fb/model_transform/fx2trt/packaging:load_net_predictor -- --loadMode=Benchmark --inputNetFile=/data/users/$USER/models/${MODEL_ENTITY_ID}/${SNAPSHOT_ID}/${MODEL_ENTITY_ID}_${SNAPSHOT_ID}.predictor.disagg.gpu.${MODULE} --moduleName ${MODULE} --predictor-hardware-type 1 --submodToDevice "" --benchmarkDontRebatchSamples=true --benchmarkNumIterations 1000 ``` Rollback Plan: Differential Revision: D78383881 Pull Request resolved: https://github.com/pytorch/pytorch/pull/158416 Approved by: https://github.com/henryoier, https://github.com/SherlockNoMad	2025-07-17 06:47:47 +00:00
Xuehai Pan	db259bd6b8	[BE][12/16] fix typos in torch/ (#156602 ) Pull Request resolved: https://github.com/pytorch/pytorch/pull/156602 Approved by: https://github.com/justinchuby, https://github.com/albanD ghstack dependencies: #156318, #156320	2025-07-02 22:55:29 +00:00
Yiming Zhou	57e4d7b5cc	[nativert] Move DelegateExecutor to PyTorch core (#155581 ) Summary: Moves DelegateExecutor base class to PyTorch core. It provides the extension point of backend delegation for NativeRT. Torch Native Runtime RFC: pytorch/rfcs#72 Test Plan: This is only a virtual base class. So relying on internal CI is sufficient. Rollback Plan: Differential Revision: D76351984 Pull Request resolved: https://github.com/pytorch/pytorch/pull/155581 Approved by: https://github.com/zhxchen17	2025-06-12 04:33:31 +00:00

Author

SHA1

Message

Date

Kevin Fu

04349f9ee5

[PT2]: Skip AOTI Weight Loading during Init (#158416 )

Summary: AOTI already has weights embedded in .so file. So for the initial load, no need to load the weights again. This allows lowered modules can have different set of weights on different hardwares.

Test Plan:
```
MODEL_TYPE=ads_mtml_offsite_cvr_oba_optout_dedicated_model
MODEL_ENTITY_ID=895279202
SNAPSHOT_ID=0
MODULE=merge

buck2 run mode/dev-nosan -c fbcode.nvcc_arch=a100,h100 -c fbcode.enable_gpu_sections=true fbcode//caffe2/torch/fb/model_transform/fx2trt/packaging:load_net_predictor -- --loadMode=Benchmark --inputNetFile=/data/users/$USER/models/${MODEL_ENTITY_ID}/${SNAPSHOT_ID}/${MODEL_ENTITY_ID}_${SNAPSHOT_ID}.predictor.disagg.gpu.${MODULE} --moduleName ${MODULE} --predictor-hardware-type 1 --submodToDevice ""  --benchmarkDontRebatchSamples=true --benchmarkNumIterations 1000
```

Rollback Plan:

Differential Revision: D78383881

Pull Request resolved: https://github.com/pytorch/pytorch/pull/158416
Approved by: https://github.com/henryoier, https://github.com/SherlockNoMad

2025-07-17 06:47:47 +00:00

Xuehai Pan

db259bd6b8

[BE][12/16] fix typos in torch/ (#156602 )

Pull Request resolved: https://github.com/pytorch/pytorch/pull/156602
Approved by: https://github.com/justinchuby, https://github.com/albanD
ghstack dependencies: #156318, #156320

2025-07-02 22:55:29 +00:00

Yiming Zhou

57e4d7b5cc

[nativert] Move DelegateExecutor to PyTorch core (#155581 )

Summary:
Moves DelegateExecutor base class to PyTorch core. It provides the extension point of backend delegation for NativeRT.
Torch Native Runtime RFC: pytorch/rfcs#72

Test Plan:
This is only a virtual base class. So relying on internal CI is sufficient.

Rollback Plan:

Differential Revision: D76351984

Pull Request resolved: https://github.com/pytorch/pytorch/pull/155581
Approved by: https://github.com/zhxchen17

2025-06-12 04:33:31 +00:00

3 Commits