Reference: https://docs.astral.sh/ruff/formatter/black/#assert-statements
> Unlike Black, Ruff prefers breaking the message over breaking the assertion, similar to how both Ruff and Black prefer breaking the assignment value over breaking the assignment target:
>
> ```python
> # Input
> assert (
> len(policy_types) >= priority + num_duplicates
> ), f"This tests needs at least {priority+num_duplicates} many types."
>
>
> # Black
> assert (
> len(policy_types) >= priority + num_duplicates
> ), f"This tests needs at least {priority+num_duplicates} many types."
>
> # Ruff
> assert len(policy_types) >= priority + num_duplicates, (
> f"This tests needs at least {priority + num_duplicates} many types."
> )
> ```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/144546
Approved by: https://github.com/malfet
This is one of a series of PRs to update us to PEP585 (changing Dict -> dict, List -> list, etc). Most of the PRs were completely automated with RUFF as follows:
Since RUFF UP006 is considered an "unsafe" fix first we need to enable unsafe fixes:
```
--- a/tools/linter/adapters/ruff_linter.py
+++ b/tools/linter/adapters/ruff_linter.py
@@ -313,6 +313,7 @@
"ruff",
"check",
"--fix-only",
+ "--unsafe-fixes",
"--exit-zero",
*([f"--config={config}"] if config else []),
"--stdin-filename",
```
Then we need to tell RUFF to allow UP006 (as a final PR once all of these have landed this will be made permanent):
```
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -40,7 +40,7 @@
[tool.ruff]
-target-version = "py38"
+target-version = "py39"
line-length = 88
src = ["caffe2", "torch", "torchgen", "functorch", "test"]
@@ -87,7 +87,6 @@
"SIM116", # Disable Use a dictionary instead of consecutive `if` statements
"SIM117",
"SIM118",
- "UP006", # keep-runtime-typing
"UP007", # keep-runtime-typing
]
select = [
```
Finally running `lintrunner -a --take RUFF` will fix up the deprecated uses.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/145101
Approved by: https://github.com/bobrenjc93
Changes:
1. Bump `ruff` from 0.7.4 to 0.8.4
2. Change `%`-formatted strings to f-string
3. Change arguments with the `__`-prefix to positional-only arguments with the `/` separator in function signature.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/143753
Approved by: https://github.com/Skylion007
The `usort` config in `pyproject.toml` has no effect due to a typo. Fixing the typo make `usort` do more and generate the changes in the PR. Except `pyproject.toml`, all changes are generated by `lintrunner -a --take UFMT --all-files`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/127126
Approved by: https://github.com/kit1980
The `usort` config in `pyproject.toml` has no effect due to a typo. Fixing the typo make `usort` do more and generate the changes in the PR. Except `pyproject.toml`, all changes are generated by `lintrunner -a --take UFMT --all-files`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/127126
Approved by: https://github.com/kit1980
ghstack dependencies: #127122, #127123, #127124, #127125
This updates ruff to 0.285 which is faster, better, and have fixes a bunch of false negatives with regards to fstrings.
I also enabled RUF017 which looks for accidental quadratic list summation. Luckily, seems like there are no instances of it in our codebase, so enabling it so that it stays like that. :)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/107519
Approved by: https://github.com/ezyang
This updates ruff to 0.285 which is faster, better, and have fixes a bunch of false negatives with regards to fstrings.
I also enabled RUF017 which looks for accidental quadratic list summation. Luckily, seems like there are no instances of it in our codebase, so enabling it so that it stays like that. :)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/107519
Approved by: https://github.com/ezyang
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/67045
To run: `python benchmarks/functional_autograd_benchmark/functional_autograd_benchmark.py --gpu -1 --model-filter=ppl _robust_reg --num-iter 100`
```
Results for model ppl_robust_reg on task vjp: 0.0012262486852705479s (var: 2.2107682351446556e-10)
Results for model ppl_robust_reg on task vhp: 0.002099371049553156s (var: 6.906406557760647e-10)
Results for model ppl_robust_reg on task jvp: 0.001860950025729835s (var: 1.1251884146634694e-10)
Results for model ppl_robust_reg on task hvp: 0.003481731517240405s (var: 2.2713633751614282e-10)
Results for model ppl_robust_reg on task jacobian: 0.0012128615053370595s (var: 1.3687526667638394e-09)
Results for model ppl_robust_reg on task hessian: 0.009885427542030811s (var: 9.366265096844018e-09)
Results for model ppl_robust_reg on task hessian_fwdrev: 0.005268776323646307s (var: 2.4293791422991262e-09)
Results for model ppl_robust_reg on task hessian_revrev: 0.002561321249231696s (var: 7.557877101938004e-10)
Results for model ppl_robust_reg on task jacfwd: 0.002619938924908638s (var: 5.109343503839625e-10)
Results for model ppl_robust_reg on task jacrev: 0.0013469004770740867s (var: 3.1857563254078514e-09)
```
Notes:
- We go through batched fallback for both
- ppl_robust_reg takes 3 tensor inputs and returns a single scalar output
- this means that jacobian is equivalent to doing vjp and vmap would not help us
- we expect jacfwd to be slower than jacrev
Test Plan: Imported from OSS
Reviewed By: malfet
Differential Revision: D33265947
Pulled By: soulitzer
fbshipit-source-id: 14f537a1376dea7e5afbe0c8e97f94731479b018
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64181
This PR replaces all the calls to:
- `transpose(-2, -1)` or `transpose(-1, -2)` by `mT()` in C++ and `mT` in Python
- `conj().transpose(-2, -1)` or `transpose(-2, -1).conj()` or `conj().transpose(-1, -2)` or `transpose(-1, -2).conj()` by `mH()` in C++ and `mH` in Python.
It also simplifies two pieces of code, and fixes one bug where a pair
of parentheses were missing in the function `make_symmetric_matrices`.
Test Plan: Imported from OSS
Reviewed By: H-Huang
Differential Revision: D31692896
Pulled By: anjali411
fbshipit-source-id: e9112c42343663d442dc5bd53ff2b492094b434a
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/60631
Per #48360, speed up `Transformer.generate_square_subsequent_mask`. New impl is informally ~5x faster, though absolute difference is probably small.
PR includes Python and C++ versions as well as a couple of places where the previous impl had been copied around.
Test Plan: Imported from OSS
Reviewed By: jbschlosser, albanD
Differential Revision: D29356673
Pulled By: bhosmer
fbshipit-source-id: 4c062ba0ead61a445aeef451c78777bf0b3a631e
Summary:
Fixes https://github.com/pytorch/pytorch/issues/47123
Follows https://github.com/pyro-ppl/pyro/pull/2701
This turns on `Distribution` validation by default. The motivation is to favor beginners by providing helpful error messages. Advanced users focused on speed can disable validation by calling
```py
torch.distributions.Distribution.set_default_validate_args(False)
```
or by disabling individual distribution validation via `MyDistribution(..., validate_args=False)`.
In practice I have found many beginners forget or do not know about validation. Therefore I have [enabled it by default](https://github.com/pyro-ppl/pyro/pull/2701) in Pyro. I believe PyTorch could also benefit from this change. Indeed validation caught a number of bugs in `.icdf()` methods, in tests, and in PPL benchmarks, all of which have been fixed in this PR.
## Release concerns
- This may slightly slow down some models. Concerned users may disable validation.
- This may cause new `ValueErrors` in models that rely on unsupported behavior, e.g. `Categorical.log_prob()` applied to continuous-valued tensors (only {0,1}-valued tensors are supported).
We should clearly note this change in release notes.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/48743
Reviewed By: heitorschueroff
Differential Revision: D25304247
Pulled By: neerajprad
fbshipit-source-id: 8d50f28441321ae691f848c55f71aa80cb356b41
Summary:
Reland of the benchmark code that broke the slow tests because the GPU were running out of memory
Pull Request resolved: https://github.com/pytorch/pytorch/pull/43428
Reviewed By: ngimel
Differential Revision: D23296136
Pulled By: albanD
fbshipit-source-id: 0002ae23dc82f401604e33d0905d6b9eedebc851
