Syncing nvfuser devel branch to upstream master. https://github.com/csarofeen/pytorch/
Bug fixes and minor refactor
Squashed commits to WAR github API
Commits that's actually in this PR from the devel branch:
```
4c60e7dff22a494632370e5df55c011007340d06 Add examples infrastructure for using nvFuser in a standalone program (#1725)
02a05d98334ffa580d73ccb28fdb8c577ad296fe Fix issue #1751 (#1753)
8a69aa320bd7629e1709fe5ceb7104d2c88ec84c Refactor NvFuser transpose API to match eager mode behavior (#1746)
ffdf6b7709048170d768217fcd7083fc8387f932 Remove BroadcastWithoutStride. (#1738)
02bab16035e70734450c02124f5cdaa95cf5749d Fix flipping of a boolean flag (#1745)
465d66890c8242e811224359cbdb1c2915490741 cleanup (#1744)
26d354e68720bc7dd2d3b1338ac01b707a230b6a fixing noncontig broadcast (#1742)
856b6b2f9073662dd98ca22ba6c3540e20eb1cdd Add IterDomainBuilder (#1736)
1fd974f912cd4c1e21cbd16e2abb23598d66a02f fixing warning for gcc7 (#1732)
de2740a43a869f8272c2648e091d7b8235097db9 disabling complex in python tests for #1730 (#1733)
fbbbe0a2e7c7a63e0e2719b8bfccb759b714221a fixing MSVC build (#1728)
b5feee5e2b28be688dbddc766f3c0220389c8175 Fix the fused reduction runtime kernel (#1729)
5247682dff5980bb66edf8d3aac25dea2ef2ced5 Re-entrant GroupedGridReduction (#1727)
```
RUN_TORCHBENCH: nvfuser
Pull Request resolved: https://github.com/pytorch/pytorch/pull/79147
Approved by: https://github.com/davidberard98
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/78297
Clone followed by expand/expand_as due to memoryOverlap check on copy_ native method. Refer to T118519310 for more details.
Crashing test case:
a = tensor(3,1) // strides = (1,1)
B = tensor(3,2) // strides = (2,1)
Temp = a.expand_as(b). // creates temp with shape as (3,2) and strides as (1,0)
temp.clone() // crashe on copy_ due to memoryOverlap
Fix: Disable the out variant for the expanded tensor.
- Calls native clone instead of out variant for clone dealing with expanded tensors
- Added test case for both clone variants (out and native clones)
- Increased the tensor size for memory planner test case to trigger dynamic allocation
Test Plan:
buck test caffe2/benchmarks/static_runtime/fb:test_fb_operators
buck test caffe2/benchmarks/static_runtime:static_runtime_cpptest
Differential Revision: D36672180
Pull Request resolved: https://github.com/pytorch/pytorch/pull/78322
Approved by: https://github.com/mikeiovine
Summary: Previously the op was auto-generated but it only covered the pointwise overload of aten::max. This adds support for reduction, overall and along a dim
Test Plan: Added a unit test
Differential Revision: D36656378
Pull Request resolved: https://github.com/pytorch/pytorch/pull/78271
Approved by: https://github.com/mikeiovine
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/75661
`fast_sigmoid` is a variant of sigmoid in NNC that is implemented in terms of `fast_tanh` (which is a fast rational function approximation).
ghstack-source-id: 155604086
Reviewed By: navahgar, hlu1
Differential Revision: D35481390
fbshipit-source-id: 1d64b5c375539f3b2461a1f3d9b86cd696eae7a1
(cherry picked from commit 8106c2512b8d7b373cb6545a43c3e8fc04805c4b)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/76391
I've seen this pattern in many important internal models:
```
x = torch.permute(a, [0, 2, 1])
y = torch.softmax(x, 2)
z = torch.permute(y, [0, 2, 1])
```
This is equivalent to
```
z = torch.softmax(x, 1)
```
The `permute` ops can degrade performance, especially if copy variants are on. Add another pattern to our `EliminateExtraPermuteOpsPass` to handle this.
ghstack-source-id: 155466506
Test Plan: New unit tests
Reviewed By: navahgar, huiguoo
Differential Revision: D35938289
fbshipit-source-id: 398b5528077b0b3f1c6fc5544e483803e96d68e9
(cherry picked from commit d742abd094d1fef23ca6a34703d97a6da2d14bd1)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/76398
This diff adds the large files that include the newly generated ops from D34913736. Refer to the base diff for more details.
Test Plan: buck run mode/opt //caffe2/benchmarks/static_runtime:static_runtime_cpptest
Reviewed By: mikeiovine, tenpercent
Differential Revision: D35945633
fbshipit-source-id: 53497bd5c490a57ea1521837762f740deb42bfd8
(cherry picked from commit e0fbdcb0bf09f5c192430f95f450c0a946c80074)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/75775
fbgemm kernels already implement the fused kernel, no reason not to use it
ghstack-source-id: 155450342
Test Plan: New unit tests
Reviewed By: navahgar
Differential Revision: D35633297
fbshipit-source-id: a744a33a65ce7dbb9ce8900dbe091b6d56dd4e48
(cherry picked from commit b1361b349862715aa17e6318c5e658cd6401a464)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/75774
`list[0:]` is a no-op. This should really be eliminated on the modeling side, implement as a graph pass for now until we can get this into prod models.
Test Plan: New unit tests
Reviewed By: navahgar
Differential Revision: D35632947
fbshipit-source-id: 0c564193c35039130e99172e0185e124ea24f62d
(cherry picked from commit e01d5273185e39a563c7acb15662d9c1549d4b58)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/75776
The output was returned directly instead of a clone, so the output of the relevant op would not be managed.
ghstack-source-id: 154935103
Test Plan: CI
Reviewed By: navahgar
Differential Revision: D35633469
fbshipit-source-id: 7b08b7368e0349a12abf8802a4c625ffecdc5abb
(cherry picked from commit 24bed9ba4da39cff7f3b40f5e49dfded2552b373)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/75601
User annotations was previously pushed down to the GPU timelines but was disabled during a refactoring some time back. This patch re-enables it in PyTorch Profiler.
Test Plan: CI Tests
Reviewed By: chaekit
Differential Revision: D34591916
Pulled By: aaronenyeshi
fbshipit-source-id: 3f4d5327b391725f4ce4e3eb16740bac2cd1c618
(cherry picked from commit 4bc07174dfef8fb2ffbefba224773a4618ed203a)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/75993
Strobelight shows copy_ in embedding_bag taking up a lot of time in adfinder_story_post_ad_session_exit_model 334827604_0
{F723683014}
More details in https://fb.quip.com/MKumAjz1YD4 (1f47a80e88)a#temp:C:FPD3 (ecd5567980)e5a0871ae5d481286b511ef7
The last 3 outputs of embedding_bag are unused in the graph: P495814049.
* max_indices output isn't necessary for the main output, so remove it when it's not used in the graph.
* offset2bag is used as an intermediate to calculate the main output, so we don't remove this output even though it's unused in the graph.
* bag_size is used as an intermediate to calculate the main output for MODE_MEAN, so we don't remove this for now.
Test Plan:
`./caffe2/caffe2/fb/predictor/scripts/run_disagg_model_benchmarks.sh 334827604 0 /data/users/ansha/tmp/ads_tail sr_only`
Inputs uploaded to `/mnt/persistent-public/ansha/ads_tail/334827604`
Before:
I0414 10:53:12.261133 1070948 PyTorchPredictorBenchLib.cpp:305] PyTorch run finished. Milliseconds per iter: 0.121318. Iters per second: 8242.78
0.11156 ms. 99.0457%. aten::embedding_bag (52 nodes, out variant)
After:
I0418 13:05:10.837378 2354604 PyTorchPredictorBenchLib.cpp:305] PyTorch run finished. Milliseconds per iter: 0.0881273. Iters per second: 11347.2
0.0789221 ms. 98.7096%. static_runtime::embedding_bag (52 nodes, out variant)
* Ads prod canary:
https://www.internalfb.com/intern/ads/canary/443002539593035806/
* 4M test: `servicelab create cogwheel_pyper_inference_fullsync_ads_inline_cvr_post_imp -a D35726594`
https://www.internalfb.com/intern/servicelab/602875732/
* 4M test: `servicelab create cogwheel_pyper_inference_fullsync_ads_10x_ctr_mbl_feed_non_mimo -a D35726594`
https://www.internalfb.com/intern/servicelab/1002874745/
Reviewed By: mikeiovine
Differential Revision: D35726594
fbshipit-source-id: 3b71a0822657bf7a23ce37ca899baef9997b011a
(cherry picked from commit fd5e3098c047a1e7d4348e1c97341eecb892536e)
Description:
- Following https://github.com/pytorch/functorch/issues/497 adding an option to run benchmarks with functorch and compare to original functional autograd results.
Running the benchmark we get below table:
<details>
<summary>
Table
</summary>
```
| model | task | mean | var |
| -- | -- | -- | -- |
| resnet18 | vjp | 0.03826599195599556 | 4.3332115637895186e-06 |
| resnet18 | functorch vjp | 0.037201929837465286 | 6.139693198292662e-09 |
| resnet18 | vhp | 0.2202976644039154 | 2.8687209052691287e-08 |
| resnet18 | functorch vhp | 0.22117868065834045 | 4.108771278765744e-08 |
| resnet18 | jvp | 0.18679651618003845 | 1.832455254202614e-08 |
| resnet18 | functorch jvp | 0.05305683612823486 | 1.6690266946284282e-08 |
| fcn_resnet | vjp | 0.6071907877922058 | 7.436695454998699e-07 |
| fcn_resnet | functorch vjp | 0.6115708947181702 | 1.121692207561864e-06 |
| fcn_resnet | vhp | 3.419469118118286 | 0.020633839070796967 |
| fcn_resnet | jvp | 2.5421929359436035 | 3.1765587209520163e-06 |
| fcn_resnet | functorch jvp | 0.7628333568572998 | 1.4555752159139956e-07 |
| detr | vjp | 0.19494840502738953 | 1.9122715457342565e-05 |
| detr | vhp | 1.1664292812347412 | 0.000948643428273499 |
| detr | jvp | 0.9990308880805969 | 1.0214127541985363e-05 |
| ppl_simple_reg | vjp | 0.0007535457843914628 | 6.024204690646684e-09 |
| ppl_simple_reg | functorch vjp | 0.0016954183811321855 | 1.160151974488599e-08 |
| ppl_simple_reg | vhp | 0.0011888503795489669 | 5.93119386937957e-10 |
| ppl_simple_reg | functorch vhp | 0.0026826143730431795 | 1.6787025103326414e-08 |
| ppl_simple_reg | jvp | 0.001067900680936873 | 7.409912128331086e-10 |
| ppl_simple_reg | functorch jvp | 0.002065300941467285 | 9.710328185974504e-08 |
| ppl_simple_reg | hvp | 0.001212477684020996 | 1.974137298077494e-09 |
| ppl_simple_reg | functorch hvp | 0.00482442369684577 | 2.327668653379078e-07 |
| ppl_simple_reg | jacobian | 0.0009108781814575195 | 3.489469158068914e-09 |
| ppl_simple_reg | functorch jacobian | 0.0019866942893713713 | 1.938326299466553e-08 |
| ppl_simple_reg | hessian | 0.005053090862929821 | 3.370298600202659e-07 |
| ppl_simple_reg | functorch hessian | 0.006374978926032782 | 7.556796077778927e-08 |
| ppl_simple_reg | hessian_fwdrev | 0.0036706924438476562 | 1.996075527088692e-09 |
| ppl_simple_reg | functorch hessian_fwdrev | 0.0058908225037157536 | 7.548283775804521e-08 |
| ppl_simple_reg | hessian_revrev | 0.0015769004821777344 | 1.5754418214442012e-08 |
| ppl_simple_reg | functorch hessian_revrev | 0.0041002752259373665 | 6.713568723171193e-08 |
| ppl_simple_reg | jacfwd | 0.0018048763740807772 | 2.7375660849315864e-08 |
| ppl_simple_reg | functorch jacfwd | 0.002047991845756769 | 2.432247070416338e-09 |
| ppl_simple_reg | jacrev | 0.0009733677143231034 | 1.0078769818733235e-08 |
| ppl_simple_reg | functorch jacrev | 0.0021971464157104492 | 1.2729884701911942e-08 |
| ppl_robust_reg | vjp | 0.005820560269057751 | 8.582588151284654e-08 |
| ppl_robust_reg | functorch vjp | 0.00796132069081068 | 9.663100541956737e-09 |
| ppl_robust_reg | vhp | 0.009825301356613636 | 2.0081762386325863e-07 |
| ppl_robust_reg | functorch vhp | 0.014890861697494984 | 4.558066279969353e-07 |
| ppl_robust_reg | jvp | 0.008297419175505638 | 2.9454400873873965e-07 |
| ppl_robust_reg | functorch jvp | 0.008052706718444824 | 7.120377176761394e-08 |
| ppl_robust_reg | hvp | 0.015414690598845482 | 7.42123745567369e-07 |
| ppl_robust_reg | functorch hvp | 0.02699306048452854 | 1.4650488537881756e-06 |
| ppl_robust_reg | jacobian | 0.006207776255905628 | 1.7068457225377642e-07 |
| ppl_robust_reg | functorch jacobian | 0.009173822589218616 | 1.2214455580306094e-07 |
| ppl_robust_reg | hessian | 0.04670915752649307 | 1.4299343092716299e-05 |
| ppl_robust_reg | functorch hessian | 0.02337808534502983 | 3.0397418413485866e-06 |
| ppl_robust_reg | hessian_fwdrev | 0.024229884147644043 | 2.0425247839739313e-06 |
| ppl_robust_reg | functorch hessian_fwdrev | 0.022021746262907982 | 3.512146236062108e-07 |
| ppl_robust_reg | hessian_revrev | 0.012355780228972435 | 7.090877147675201e-07 |
| ppl_robust_reg | functorch hessian_revrev | 0.013960313983261585 | 6.326549737423193e-07 |
| ppl_robust_reg | jacfwd | 0.008112502284348011 | 2.88503088086145e-08 |
| ppl_robust_reg | functorch jacfwd | 0.008947920985519886 | 4.2070990247111695e-08 |
| ppl_robust_reg | jacrev | 0.00635871896520257 | 1.3403841592207755e-07 |
| ppl_robust_reg | functorch jacrev | 0.009123563766479492 | 2.677554675756255e-07 |
| wav2letter | vjp | 0.02078995667397976 | 2.1110793113621185e-06 |
| wav2letter | functorch vjp | 0.019202351570129395 | 9.210506135559626e-09 |
| wav2letter | vhp | 0.05997290462255478 | 8.558587616391833e-09 |
| wav2letter | functorch vhp | 0.06035261228680611 | 1.6448565842708263e-09 |
| wav2letter | jvp | 0.04507789760828018 | 1.5771547401399744e-09 |
| wav2letter | functorch jvp | 0.013057494536042213 | 3.804750292601966e-09 |
| deepspeech | vjp | 0.3648746609687805 | 1.5359055396402255e-05 |
| transformer | vjp | 0.05496881157159805 | 1.242562319703211e-08 |
| transformer | functorch vjp | 0.057835936546325684 | 2.6113376350167528e-08 |
| transformer | vhp | 0.18313491344451904 | 7.226336151688884e-08 |
| transformer | jvp | 0.13924935460090637 | 1.6989159234981344e-07 |
| multiheadattn | vjp | 0.0014708995586261153 | 3.710916729460223e-08 |
| multiheadattn | functorch vjp | 0.002404856728389859 | 2.1910574687922235e-08 |
| multiheadattn | vhp | 0.003382015274837613 | 5.3098595742540056e-08 |
| multiheadattn | functorch vhp | 0.005340623669326305 | 5.897558708056749e-08 |
| multiheadattn | jvp | 0.0027526854537427425 | 3.508620949332908e-08 |
| multiheadattn | functorch jvp | 0.0022981404326856136 | 1.327894807445773e-07 |
```
</details>
<details>
<summary>
Stdout
</summary>
```
Found functorch: 0.2.0a0+386a541
Results for model resnet18 on task vjp: 0.03826599195599556s (var: 4.3332115637895186e-06)
Results for model resnet18 on task vjp using Functorch: 0.037201929837465286s (var: 6.139693198292662e-09)
Results for model resnet18 on task vhp: 0.2202976644039154s (var: 2.8687209052691287e-08)
Results for model resnet18 on task vhp using Functorch: 0.22117868065834045s (var: 4.108771278765744e-08)
Results for model resnet18 on task jvp: 0.18679651618003845s (var: 1.832455254202614e-08)
Results for model resnet18 on task jvp using Functorch: 0.05305683612823486s (var: 1.6690266946284282e-08)
Results for model fcn_resnet on task vjp: 0.6071907877922058s (var: 7.436695454998699e-07)
Results for model fcn_resnet on task vjp using Functorch: 0.6115708947181702s (var: 1.121692207561864e-06)
Results for model fcn_resnet on task vhp: 3.419469118118286s (var: 0.020633839070796967)
Failed model using Functorch: fcn_resnet, task: vhp, Error message:
CUDA out of memory. Tried to allocate 114.00 MiB (GPU 0; 47.46 GiB total capacity; 45.62 GiB already allocated; 5.31 MiB free; 46.02 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
Results for model fcn_resnet on task jvp: 2.5421929359436035s (var: 3.1765587209520163e-06)
Results for model fcn_resnet on task jvp using Functorch: 0.7628333568572998s (var: 1.4555752159139956e-07)
Results for model detr on task vjp: 0.19494840502738953s (var: 1.9122715457342565e-05)
Failed model using Functorch: detr, task: vjp, Error message:
Cannot access data pointer of Tensor that doesn't have storage
Results for model detr on task vhp: 1.1664292812347412s (var: 0.000948643428273499)
Failed model using Functorch: detr, task: vhp, Error message:
Cannot access data pointer of Tensor that doesn't have storage
Results for model detr on task jvp: 0.9990308880805969s (var: 1.0214127541985363e-05)
Failed model using Functorch: detr, task: jvp, Error message:
Trying to use forward AD with _cdist_forward that does not support it because it has not been implemented yet.
Please file an issue to PyTorch at https://github.com/pytorch/pytorch/issues/new?template=feature-request.yml so that we can prioritize its implementation.
Results for model ppl_simple_reg on task vjp: 0.0007535457843914628s (var: 6.024204690646684e-09)
Results for model ppl_simple_reg on task vjp using Functorch: 0.0016954183811321855s (var: 1.160151974488599e-08)
Results for model ppl_simple_reg on task vhp: 0.0011888503795489669s (var: 5.93119386937957e-10)
Results for model ppl_simple_reg on task vhp using Functorch: 0.0026826143730431795s (var: 1.6787025103326414e-08)
Results for model ppl_simple_reg on task jvp: 0.001067900680936873s (var: 7.409912128331086e-10)
Results for model ppl_simple_reg on task jvp using Functorch: 0.002065300941467285s (var: 9.710328185974504e-08)
Results for model ppl_simple_reg on task hvp: 0.001212477684020996s (var: 1.974137298077494e-09)
Results for model ppl_simple_reg on task hvp using Functorch: 0.00482442369684577s (var: 2.327668653379078e-07)
Results for model ppl_simple_reg on task jacobian: 0.0009108781814575195s (var: 3.489469158068914e-09)
Results for model ppl_simple_reg on task jacobian using Functorch: 0.0019866942893713713s (var: 1.938326299466553e-08)
Results for model ppl_simple_reg on task hessian: 0.005053090862929821s (var: 3.370298600202659e-07)
Results for model ppl_simple_reg on task hessian using Functorch: 0.006374978926032782s (var: 7.556796077778927e-08)
Results for model ppl_simple_reg on task hessian_fwdrev: 0.0036706924438476562s (var: 1.996075527088692e-09)
Results for model ppl_simple_reg on task hessian_fwdrev using Functorch: 0.0058908225037157536s (var: 7.548283775804521e-08)
Results for model ppl_simple_reg on task hessian_revrev: 0.0015769004821777344s (var: 1.5754418214442012e-08)
Results for model ppl_simple_reg on task hessian_revrev using Functorch: 0.0041002752259373665s (var: 6.713568723171193e-08)
Results for model ppl_simple_reg on task jacfwd: 0.0018048763740807772s (var: 2.7375660849315864e-08)
Results for model ppl_simple_reg on task jacfwd using Functorch: 0.002047991845756769s (var: 2.432247070416338e-09)
Results for model ppl_simple_reg on task jacrev: 0.0009733677143231034s (var: 1.0078769818733235e-08)
Results for model ppl_simple_reg on task jacrev using Functorch: 0.0021971464157104492s (var: 1.2729884701911942e-08)
Results for model ppl_robust_reg on task vjp: 0.005820560269057751s (var: 8.582588151284654e-08)
Results for model ppl_robust_reg on task vjp using Functorch: 0.00796132069081068s (var: 9.663100541956737e-09)
Results for model ppl_robust_reg on task vhp: 0.009825301356613636s (var: 2.0081762386325863e-07)
Results for model ppl_robust_reg on task vhp using Functorch: 0.014890861697494984s (var: 4.558066279969353e-07)
Results for model ppl_robust_reg on task jvp: 0.008297419175505638s (var: 2.9454400873873965e-07)
Results for model ppl_robust_reg on task jvp using Functorch: 0.008052706718444824s (var: 7.120377176761394e-08)
Results for model ppl_robust_reg on task hvp: 0.015414690598845482s (var: 7.42123745567369e-07)
Results for model ppl_robust_reg on task hvp using Functorch: 0.02699306048452854s (var: 1.4650488537881756e-06)
Results for model ppl_robust_reg on task jacobian: 0.006207776255905628s (var: 1.7068457225377642e-07)
Results for model ppl_robust_reg on task jacobian using Functorch: 0.009173822589218616s (var: 1.2214455580306094e-07)
Results for model ppl_robust_reg on task hessian: 0.04670915752649307s (var: 1.4299343092716299e-05)
Results for model ppl_robust_reg on task hessian using Functorch: 0.02337808534502983s (var: 3.0397418413485866e-06)
Results for model ppl_robust_reg on task hessian_fwdrev: 0.024229884147644043s (var: 2.0425247839739313e-06)
Results for model ppl_robust_reg on task hessian_fwdrev using Functorch: 0.022021746262907982s (var: 3.512146236062108e-07)
Results for model ppl_robust_reg on task hessian_revrev: 0.012355780228972435s (var: 7.090877147675201e-07)
Results for model ppl_robust_reg on task hessian_revrev using Functorch: 0.013960313983261585s (var: 6.326549737423193e-07)
Results for model ppl_robust_reg on task jacfwd: 0.008112502284348011s (var: 2.88503088086145e-08)
Results for model ppl_robust_reg on task jacfwd using Functorch: 0.008947920985519886s (var: 4.2070990247111695e-08)
Results for model ppl_robust_reg on task jacrev: 0.00635871896520257s (var: 1.3403841592207755e-07)
Results for model ppl_robust_reg on task jacrev using Functorch: 0.009123563766479492s (var: 2.677554675756255e-07)
Results for model wav2letter on task vjp: 0.02078995667397976s (var: 2.1110793113621185e-06)
Results for model wav2letter on task vjp using Functorch: 0.019202351570129395s (var: 9.210506135559626e-09)
Results for model wav2letter on task vhp: 0.05997290462255478s (var: 8.558587616391833e-09)
Results for model wav2letter on task vhp using Functorch: 0.06035261228680611s (var: 1.6448565842708263e-09)
Results for model wav2letter on task jvp: 0.04507789760828018s (var: 1.5771547401399744e-09)
Results for model wav2letter on task jvp using Functorch: 0.013057494536042213s (var: 3.804750292601966e-09)
Results for model deepspeech on task vjp: 0.3648746609687805s (var: 1.5359055396402255e-05)
Failed model using Functorch: deepspeech, task: vjp, Error message:
Cannot access storage of TensorWrapper
Results for model transformer on task vjp: 0.05496881157159805s (var: 1.242562319703211e-08)
Results for model transformer on task vjp using Functorch: 0.057835936546325684s (var: 2.6113376350167528e-08)
Results for model transformer on task vhp: 0.18313491344451904s (var: 7.226336151688884e-08)
Failed model using Functorch: transformer, task: vhp, Error message:
bad optional access
Results for model transformer on task jvp: 0.13924935460090637s (var: 1.6989159234981344e-07)
Failed model using Functorch: transformer, task: jvp, Error message:
Trying to use forward AD with embedding that does not support it because it has not been implemented yet.
Please file an issue to PyTorch at https://github.com/pytorch/pytorch/issues/new?template=feature-request.yml so that we can prioritize its implementation.
Results for model multiheadattn on task vjp: 0.0014708995586261153s (var: 3.710916729460223e-08)
Results for model multiheadattn on task vjp using Functorch: 0.002404856728389859s (var: 2.1910574687922235e-08)
Results for model multiheadattn on task vhp: 0.003382015274837613s (var: 5.3098595742540056e-08)
Results for model multiheadattn on task vhp using Functorch: 0.005340623669326305s (var: 5.897558708056749e-08)
Results for model multiheadattn on task jvp: 0.0027526854537427425s (var: 3.508620949332908e-08)
Results for model multiheadattn on task jvp using Functorch: 0.0022981404326856136s (var: 1.327894807445773e-07)
```
</details>
All functorch errors are reported in its repository.
cc @zou3519
Pull Request resolved: https://github.com/pytorch/pytorch/pull/75689
Approved by: https://github.com/zou3519
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/76111https://github.com/pytorch/pytorch/pull/68640 broke our build by porting `cat` structured kernels, not sure how CI didn't catch this
ghstack-source-id: 154335722
Test Plan: CI
Reviewed By: navahgar, ajyu
Differential Revision: D35780296
fbshipit-source-id: 0a262eb06a8d619227e5db10b6a775bf0b2e17c1
(cherry picked from commit aea6fbf9365391011df5211164e3978075d7a5cb)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/74966
It's clear that we don't want to manage tensors that escape their scope. Previously, we handled this by checking whether the tensor aliased the graph outputs. But there's actually another way to escape scope: by aliasing the wildcard set. The following graph demonstrates this:
```
def forward(self, cond: bool, a, b):
lst = []
if cond:
res = a + b # res should not be managed!!!
lst.append(res)
return lst
```
The `if cond:` sub-block returns nothing, but `res` escapes the scope through `lst`.
The fix is simple: we simply have to mark values that alias the wildcard set as an `external_alias_` in `ValueGroup`.
This diff also exposed another issue (via unit tests) in `checkOutputTensorMemoryLeaks`: it assumes that, if a node's `Value*` is managed, the underlying `IValue` must be a tensor. But this is not true after the addition of `to_maybe_copy_out`; TMCO does not produce a tensor in its first output slot if it does not copy.
ghstack-source-id: 153288188
Test Plan: New unit tests cover the problematic case
Reviewed By: navahgar
Differential Revision: D35257087
fbshipit-source-id: 853a761dffe51f2c70720759664dd8dfcd56d1d7
(cherry picked from commit 2c7f519354041975f33626eab6b7f16c2494bbf8)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/74927
The move ctor was broken because `BlockRunner` stores a reference to `values_`. When moving runtime instances, the pointer to the root block would be moved, but the reference inside it would not be updated.
Pass `BlockRunner` a raw pointer to the heap-allocated IValues instead to avoid this issue.
ghstack-source-id: 153168602
Test Plan: New unit test/CI
Reviewed By: navahgar
Differential Revision: D35228467
fbshipit-source-id: 04e198b39f898b82677a0e41e1cdf00c2b0c09f3
(cherry picked from commit 03e2c591ac3a907d68025eae9500ed7226dec17e)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/74255
This change fixes a bug that `aten::full_like` reuses a previously allocated tensor that does not match requested one when arguments to `aten::full_like` are dynamically changed.
Test Plan: - Enhanced `StaticRuntime.FullLike` to cover the modified code path.
Reviewed By: mikeiovine
Differential Revision: D34863639
fbshipit-source-id: ca6d4ee3c039e263cc3a4f643d949cea59381608
(cherry picked from commit ae7db0af5e7d95d866027abc968afcb162fd2ef8)
Summary:
The implementation of `PackedLinearWeightFp16::apply_dynamic_impl` [here](https://www.internalfb.com/code/fbsource/[b1ef7c31f022]/fbcode/caffe2/aten/src/ATen/native/quantized/cpu/qlinear_dynamic.cpp?lines=393) does not handle `relu`. It completely ignores the `ReluFused` boolean template parameter.
At this point, callers of that function handle `relu` explicitly. While the correct thing to do would be to handle the `ReluFused` parameter in that implementation, it is not clear if that semantics is being followed in this code. So, we are handling this in SR's out-variant implementation, until the owner fixes that issue.
This issue resulted in incorrect results when Static Runtime was enabled for the MRS video model.
Test Plan:
```
buck run mode/opt //caffe2/benchmarks/static_runtime:static_runtime_cpptest -- --gtest_filter=StaticRuntime.QuantizedLinearReluDynamicFp16
```
Reviewed By: mikeiovine
Differential Revision: D35366309
fbshipit-source-id: e60126e3590d52681ceaee5583b81c4c0b5404d9
(cherry picked from commit cabeb96a792339e7dbfd16cb51a3ac9039812137)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/74987
Add specializations to `prim::If` operator at runtime to save resources when some of subblocks are empty
Test Plan:
`buck build //caffe2:torch-cpp-cpu`
`buck test //caffe2/benchmarks/static_runtime/...`
Add unit test:
`buck test //caffe2/benchmarks/static_runtime:static_runtime_cpptest -- StaticRuntime.EmptyIfBlock`
Reviewed By: mikeiovine
Differential Revision: D35262952
fbshipit-source-id: 324f88471f33f035f4d8a9b212716530d8e59df2
(cherry picked from commit 2db1b1a6833b1376fa376f54791effc8e12fb77f)
Summary:
Things changed in this PR that requires review:
test/forward_backward_compatibility/check_forward_backward_compatibility.py
Our previous function overload extension names were wrong and has been updated in this PR, hence the compatibility list updated.
nvfuser code updates with bug fixes towards failures we encountered in OpInfoTests as well as failures reported by AOTAutograd team.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/73627
Reviewed By: Chillee
Differential Revision: D34765458
Pulled By: davidberard98
fbshipit-source-id: c81f3d6a1b723fb3a8ba419b7f82227f70440ca7
(cherry picked from commit b6a2c362c37051e44fac31687b2fe272f776551e)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/73875
Previously we had a few settings:
- getExecutor - which toggled between Profiling Executor and Legacy
- getGraphOptimize - if true, overrides PE/Legacy to run with simple executor (no optimizations)
and then...
- getProfilingMode - which would set PE to 0 specializtions.
The last mode is redundant with getGraphOptimize, we should just remove it and use getGraphOptimize in these cases. It would lead to potentially invalid combinations of logic - what does mean if getProfilingMode is true but getExecutor is set to false ? This would lead to a bug in specialize_autograd_zero in this case, see: https://github.com/pytorch/pytorch/blob/master/torch%2Fcsrc%2Fjit%2Fpasses%2Fspecialize_autogradzero.cpp#L93.
The tests here are failing but get fixed with the PR above it, so i'll squash for landing.
Test Plan: Imported from OSS
Reviewed By: cpuhrsch
Differential Revision: D34938130
Pulled By: eellison
fbshipit-source-id: 1a9c0ae7f6d1cfddc2ed3499a5af611053ae5e1b
(cherry picked from commit cf69ce3d155ba7d334022c42fb2cee54bb088c23)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/74585
Native static runtime for `aten::reshape_as`
ghstack-source-id: 152340038
Test Plan: New unit test
Reviewed By: hlu1
Differential Revision: D35060895
fbshipit-source-id: c4e6f8a04c7df3821c7e654bfaf584e5a72ea701
(cherry picked from commit 6fa596cd866a024b6653239e0e30ddad42de242f)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/74562
Add a native implementation for `aten::IntImplicit`, which is similar to `aten::Int` except for a few extra checks it must do
ghstack-source-id: 152340039
Test Plan: New unit tests
Reviewed By: hlu1
Differential Revision: D35052997
fbshipit-source-id: cb2f0faf7c62382e3f13750d8e1280c49c6b9e42
(cherry picked from commit 359c7493f8deaeccebc27e1b6e6e9777850010c1)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/74481
This diff fixes an interesting performance issue related to `permute_copy`.
We see this pattern frequently:
```
y = torch.permute(x, (0, 2, 1))
z = torch.sum(y, dim=-1)
```
With copy variants off, we get a strided output from `permute`, and we hit this (faster) kernel in `sum`: https://github.com/pytorch/pytorch/blob/master/aten/src/ATen/native/cpu/SumKernel.cpp#L589
But with copy variants on, we get a contiguous output from `permute_copy`, which causes us to hit the slower reduction:
https://github.com/pytorch/pytorch/blob/master/aten/src/ATen/native/cpu/SumKernel.cpp#L597
But the permute is actually unnecessary, we can just statically turn the graph into this to ensure that the fast kernel is hit with copy variants on:
```
z = torch.sum(x, dim=1)
```
ghstack-source-id: 152003888
Reviewed By: navahgar
Differential Revision: D34992319
fbshipit-source-id: 0baf493708ee2180c899814a954d220d88ba1d4f
(cherry picked from commit 797b6beb26325c56012e406e14fe211c0b5d744d)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/73606
The single-output overload of `layer_norm` internally allocates two tensors. As an optimization, we previously added `static_runtime::layer_norm`. This variant of layer norm had two extra outputs to make the memory planner aware of these extra tensors. But these outputs were unused; it's actually better for us to avoid the allocation and associated computations entirely.
ghstack-source-id: 151394116
Test Plan: Existing unit tests
Reviewed By: hlu1
Differential Revision: D34562131
fbshipit-source-id: c6a6560e60db43b0b100aedc54ea4265acb347de
(cherry picked from commit 3bed52b6f688b93b9b032c3d2b4be68d08d8eb76)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/73990
This change fixes a bug that `aten::full` reuses a previously allocated tensor that does not match requested one when arguments to `aten::full` are dynamically changed.
This fix is applied to multiple other out variant wrappers added to Static Runtime, and their fixes are following.
Test Plan: - Added a unittest.
Reviewed By: mikeiovine
Differential Revision: D34768718
fbshipit-source-id: b6958d6601d36253dd5d4f93596fb14055cca9c9
(cherry picked from commit 42acb40d3a1e9359c0f1a3c25481854e5ad344b6)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/73945
This change adds add out variant wrapper for aten::ones_like.
Test Plan:
- Added a unittest.
- Checked that the op execution got switched to its added out variant (P485330978).
Reviewed By: hlu1
Differential Revision: D34727057
fbshipit-source-id: 5022a7f547d53b0c00459d3959ad3c6e6a8a62d5
(cherry picked from commit 1bec4680e8173654400b165d720a0902136dba0f)
