Replaces 78 assert statements across 10 files in torch.autograd with explicit if-checks raising AssertionError to prevent assertions from being disabled with Python -O flag. This ensures error checking remains active in optimized builds.
fix partially #164878
Pull Request resolved: https://github.com/pytorch/pytorch/pull/165627
Approved by: https://github.com/albanD
This PR enables all PIE rules on ruff, there are already some enabled rules from this family, the new added rules are
```
PIE796 Enum contains duplicate value: {value}
PIE808 Unnecessary start argument in range
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/165814
Approved by: https://github.com/ezyang
This PR enables all PIE rules on ruff, there are already some enabled rules from this family, the new added rules are
```
PIE796 Enum contains duplicate value: {value}
PIE808 Unnecessary start argument in range
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/165814
Approved by: https://github.com/ezyang
## Overview
This PR allows the profiler users to access `Kineto` and `TorchOp` metadata in JSON string format through a new `metadata_json` attribute in `FunctionEvent` objects, which is triggered through a new `expose_kineto_event_metadata` flag in `ExperimentalConfig`.
## Testing
A unit test was added to validate functionality.
## Documentation
Added/updated function doc strings where appropriate.
## Example output
```python
import torch
from torch.profiler import profile
with profile(experimental_config=torch._C._profiler._ExperimentalConfig(expose_kineto_event_metadata=True)) as prof:
res = torch.mm(torch.rand(1024, 1024), torch.rand(1024, 1024))
for event in prof.events():
print(f'name: {event.key}, metadata: {event.metadata_json}')
```
```
name: aten::rand, metadata: "Ev Idx": 0
name: aten::empty, metadata: "Ev Idx": 1
name: aten::uniform_, metadata: "Ev Idx": 2
name: aten::rand, metadata: "Ev Idx": 3
name: aten::empty, metadata: "Ev Idx": 4
name: aten::uniform_, metadata: "Ev Idx": 5
name: aten::mm, metadata: "Ev Idx": 6
name: aten::resolve_conj, metadata: "Ev Idx": 7
name: aten::resolve_conj, metadata: "Ev Idx": 8
name: aten::resolve_conj, metadata: "Ev Idx": 9
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/161624
Approved by: https://github.com/sraikund16
Via google search I got to `torch.autograd.profiler` and implemented my code with it. Only to be taken by surprise finding `torch.profile.profiler`, which has a note saying the autograd one is legacy.
This just adds such note to `autograd.profiler` to avoid this confusion and waste of time to future people in my situation.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/157459
Approved by: https://github.com/sraikund16
Currently, recorded profiler events for aten ops do not store overload names. It would be useful to know which overloads are actually called to analyse performance.
For example, consider the following dispatch trace which occurs if there is a fallthrough kernel registered for aten::add:
```
[call] op=[aten::add.Tensor], key=[AutogradCPU]
[redispatch] op=[aten::add.Tensor], key=[Undefined]
[call] op=[aten::empty.memory_format], key=[BackendSelect]
[redispatch] op=[aten::empty.memory_format], key=[CPU]
[call] op=[aten::add.out], key=[CPU]
```
In this case, aten::add.out is a child of aten::add.Tensor, however the current profiler trace provides no way to differentiate aten op calls.
See the added unit test for a more detailed example.
Fixes #ISSUE_NUMBER
Pull Request resolved: https://github.com/pytorch/pytorch/pull/143114
Approved by: https://github.com/sraikund16
Summary:
This PR adds Auto-Trace implementation for Trace ID. By default, the python side will generate a uuid in the same format as the one set in the backend by kineto. Upon running an auto-trace, the python generated trace id will overwrite the one set in kineto using the Config variable. Since we don't expect users to generate on-demand traces after an auto-trace we can simply keep overwriting the backend trace id whenever autotrace is ran. If we one day want to eventually do something like this, we simply have to add a call in kineto on the backend to generate a new ID upon start of profiling.
We also implement a custom callback in the frontend such that users can generate their own trace ids if they wish to. This works similarly as the default, only difference being that they have to manually set this callback after a profiler is generated. We use a specific call to set this rather then putting it in the frontend initializer in case users want to change the trace_id for different repeats.
Test Plan: Tested both default and custom callbacks using the verbose prints added. Trace ids on the frontend and the prints on the backend for the manifold upload matched.
Differential Revision: D65178308
Pull Request resolved: https://github.com/pytorch/pytorch/pull/139310
Approved by: https://github.com/shengfukevin
To avoid high overheads of constructing datastructure in python when the user is simply saving trace to a file, we only process things lazily.
## Details
1. Delay function event parsing, add a flag to denote when needed.
1. Make profiler.function_events a computed property so code using `prof.function_events` does not need to change.
1. Fix coverage for `str(prof)` in profiler tests.
## Test run
Test program
```
import torch
from torch.profiler import profile, record_function, ProfilerActivity
def payload(use_cuda=False):
x = torch.randn(10, 10)
if use_cuda:
x = x.cuda()
y = torch.randn(10, 10)
if use_cuda:
y = y.cuda()
z = torch.mm(x, y)
z = z + y
if use_cuda:
z = z.cpu()
with profile(activities=[ProfilerActivity.CPU], record_shapes=True) as prof:
with record_function("model_inference"):
payload()
prof.export_chrome_trace("/tmp/test_trace.json")
#print(prof.key_averages().table(sort_by="cpu_time_total", row_limit=10))
```
The print "this is computing events" will happen lazily.
```
>]$ python3 profiler_test.py
Brian: this is computing function events
---------------------- ------------ ------------ ------------ ------------ ------------ ------------
Name Self CPU % Self CPU CPU total % CPU total CPU time avg # of Calls
---------------------- ------------ ------------ ------------ ------------ ------------ ------------
model_inference 6.77% 441.628us 100.00% 6.523ms 6.523ms 1
aten::randn 1.86% 121.108us 46.93% 3.061ms 1.530ms 2
aten::mm 45.36% 2.959ms 45.44% 2.964ms 2.964ms 1
aten::normal_ 44.72% 2.917ms 44.72% 2.917ms 1.458ms 2
aten::add 0.87% 56.646us 0.87% 56.646us 56.646us 1
aten::empty 0.35% 22.808us 0.35% 22.808us 11.404us 2
aten::resolve_conj 0.08% 5.173us 0.08% 5.173us 1.724us 3
---------------------- ------------ ------------ ------------ ------------ ------------ ------------
Self CPU time total: 6.523ms
$> python3 profiler_test.py
(pytorch) [bcoutinho@devgpu038.ftw6 /data/users/bcoutinho/pytorch (profiler_optimize_parsing)]$
$>ls -a profiler_test.py
$> ls -l /tmp/test_trace.json
-rw-r--r-- 1 bcoutinho users 16471 Aug 5 16:10 /tmp/test_trace.json
```
## Unit test
Updates some tests and they all pass now.
`pytest test/profiler/test_profiler.py`
Also
`python test/test_autograd.py TestAutogradWithCompiledAutograd.test_record_function`
Pull Request resolved: https://github.com/pytorch/pytorch/pull/132713
Approved by: https://github.com/sraikund16
Summary: D56956245 added the ability to accumulate FunctionEvents across multiple cycles in order to perform statistical analysis on them all together. Although this can be useful, it uses too many CPU resources especially for long running jobs. For this reason, lets add a flag to the profiler to turn off this behavior by default, but still allow users to turn it on if they wish.
Test Plan: Changed function count test to have acc_events passed in and check the amount of function events based on if flag is true or not
Differential Revision: D61021490
Pull Request resolved: https://github.com/pytorch/pytorch/pull/133095
Approved by: https://github.com/briancoutinho, https://github.com/LucasLLC, https://github.com/aaronenyeshi
Summary: On the autograd side of things, we are currently saving the kwinputs but we aren't doing anything with them on the profiler side. This diff enables the use of the kwinputs for both FunctionEvents and Chrome Traces.
Test Plan: Added unit testing for both chrome traces and FunctionEvents. Used RecordFunctionFast to test kwinputs since test already had kwargs being passed in but not tested.
Differential Revision: D59472345
Pull Request resolved: https://github.com/pytorch/pytorch/pull/130373
Approved by: https://github.com/davidberard98
Summary: Users have been confused why user annotations on GPU tracks do not show when doing GPU only tracing. This comment should help users understand that to use this function they need to have CPU activies enabled.
Test Plan: N/A it is just updating a comment
Differential Revision: D59649390
Pull Request resolved: https://github.com/pytorch/pytorch/pull/130561
Approved by: https://github.com/aaronenyeshi
Summary:
use_mtia should instead set use_device='mtia' similar to cuda, xpu, and privateuseone. Avoid an ever-growing list of use_* arguments.
Since use_mtia is specific to FBCode, we don't need a deprecation warning.
Test Plan: CI.
Differential Revision: D57338005
Pulled By: aaronenyeshi
Pull Request resolved: https://github.com/pytorch/pytorch/pull/126284
Approved by: https://github.com/fenypatel99
Use `typing_extensions.deprecated` for deprecation annotation if possible. Otherwise, add `category=FutureWarning` to `warnings.warn("message")` if the category is missing.
Note that only warnings that their messages contain `[Dd]eprecat(ed|ion)` are updated in this PR.
Resolves#126888
- #126888
This PR is split from PR #126898.
- #126898
------
Pull Request resolved: https://github.com/pytorch/pytorch/pull/127689
Approved by: https://github.com/Skylion007
Use `typing_extensions.deprecated` for deprecation annotation if possible. Otherwise, add `category=FutureWarning` to `warnings.warn("message")` if the category is missing.
Note that only warnings that their messages contain `[Dd]eprecat(ed|ion)` are updated in this PR.
UPDATE: Use `FutureWarning` instead of `DeprecationWarning`.
Resolves#126888
- #126888
Pull Request resolved: https://github.com/pytorch/pytorch/pull/126898
Approved by: https://github.com/albanD
Summary:
github issue: https://github.com/pytorch/pytorch/issues/73828
Whenever we transition from RECORD_AND_SAVE to WARMUP in the profiler schedule, we instantiate a new backend profiler which wipes out the last cycle's information. This makes using the `repeat` parameter less useful in the schedule as you only get contents of the last cycle/repeat. In this diff, we save the accumulated Function Events before setting the new ones and then merge the two EventLists after post processing/cleaning is done. This diff only fixes Function Events so that we can get statistics over each cycle within a schedule. A follow up should be made to accumulate the chrome tracings as well if it is requested.
Test Plan: Added functional python tests in test_profiler.py that test different schedules and their FunctionEvent counts
Differential Revision: D56956245
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125510
Approved by: https://github.com/aaronenyeshi
This fixes a logic regression introduced by https://github.com/pytorch/pytorch/pull/123247 where
```python
if self.use_device and self.use_device != _get_privateuse1_backend_name():
```
was replaced with
```python
VALID_DEVICE_OPTIONS = ["cuda", "xpu", "privateuseone"]
if self.use_device not in VALID_DEVICE_OPTIONS:
```
That triggers a warning every time code is invoke with `self.use_device` set to None
This change also skips all the checks which are useless if `use_device` is None to begin with
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125654
Approved by: https://github.com/aaronenyeshi
Summary:
1.Package public headers of kineto if USE_KINETO so that they can be used by PrivateUse1 user.
2.Add PrivateUse1 key to ActivityType.
3. Support PrivateUse1 key in function deviceTypeFromActivity and _supported_activities.
4. Fix some bugs when processing profiler results.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124818
Approved by: https://github.com/aaronenyeshi
The `function_events` in `_parse_kineto_results` is used to contain all function events from the result. It contains 2 kinds of events. One is frontend function events whose correlation id is 0, for example, `aten::add`, `aten::mul`. They are on the top level of the profile results. The other is the backend events, which are associated with the frontend events and its correlation id is > 0, for example, `at::native::vectorized_elementwise_kernel`, it should be the backend event of a frontend element-wise op. They have the device execution duration for the related frontend op.
In the following post processing code, the **frontend function events** should be iterated to find its correlated backend events in `device_corr_map`, instead of iterating all function events, because `device_corr_map` is designed as a dict, whose key is the id of the frontend function event.
3af12447f8/torch/autograd/profiler.py (L543-L560)3af12447f8/torch/autograd/profiler.py (L537-L540)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124596
Approved by: https://github.com/aaronenyeshi
This PR unifies the CUDA, XPU and PrivateUse1 in the torch profiler. Now CUDA, XPU and PrivateUse1 can together use string object `use_device` to distinguish each other and share one device path for calculating kineto time durations and memory statistics for post processing.
#suppress-api-compatibility-check
Co-authored-by: Aaron Enye Shi <enye.shi@gmail.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/123247
Approved by: https://github.com/aaronenyeshi
This PR unifies the CUDA, XPU and PrivateUse1 in the torch profiler. Now CUDA, XPU and PrivateUse1 can together use string object `use_device` to distinguish each other and share one device path for calculating kineto time durations and memory statistics for post processing.
#suppress-api-compatibility-check
Co-authored-by: Aaron Enye Shi <enye.shi@gmail.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/123247
Approved by: https://github.com/aaronenyeshi, https://github.com/gujinghui
Summary:
1.Package public headers of kineto if USE_KINETO so that they can be used by PrivateUse1 user.
2.Add PrivateUse1 key to ActivityType.
3. Support PrivateUse1 key in function deviceTypeFromActivity and _supported_activities.
4. Fix some bugs when processing profiler results.
Co-authored-by: albanD <desmaison.alban@gmail.com>
Co-authored-by: Aaron Shi <enye.shi@gmail.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/120556
Approved by: https://github.com/aaronenyeshi
Summary:
Kineto traces use microsecond level granularity because of chrome tracing defaults to that precision. Fix by adding preprocessor flag to TARGETS and BUCK files. Also remove any unnecessary ns to us conversions made in the profiler itself.
This diff contains profiler changes only. Libkineto changes found in D54964435.
Test Plan:
Check JSON and chrome tracing to make sure values are as expected. Tracing with flags enabled should have ns precision. Tracings without flags should be same as master.
Zoomer: https://www.internalfb.com/intern/zoomer/?profiling_run_fbid=796886748550189
Ran key_averages() to make sure FunctionEvent code working as expected:
-- ------------ ------------
Name Self CPU % Self CPU CPU total % CPU total CPU time avg Self CUDA Self CUDA % CUDA total CUDA time avg # of Calls
ProfilerStep* 0.74% 3.976ms 64.40% 346.613ms 69.323ms 0.000us 0.00% 61.710ms 12.342ms 5
Optimizer.zero_grad#SGD.zero_grad 0.76% 4.109ms 0.76% 4.109ms 821.743us 0.000us 0.00% 0.000us 0.000us 5
## forward ## 6.89% 37.057ms 27.19% 146.320ms 29.264ms 0.000us 0.00% 58.708ms 11.742ms 5
aten::conv2d 0.22% 1.176ms 7.74% 41.658ms 157.199us 0.000us 0.00% 27.550ms 103.962us 265
aten::convolution 0.79% 4.273ms 7.52% 40.482ms 152.762us 0.000us 0.00% 27.550ms 103.962us 265
aten::_convolution 0.69% 3.688ms 6.73% 36.209ms 136.637us 0.000us 0.00% 27.550ms 103.962us 265
aten::cudnn_convolution 6.04% 32.520ms 6.04% 32.520ms 122.719us 27.550ms 8.44% 27.550ms 103.962us 265
aten::add_ 2.42% 13.045ms 2.42% 13.045ms 30.694us 12.700ms 3.89% 12.700ms 29.882us 425
aten::batch_norm 0.19% 1.027ms 8.12% 43.717ms 164.971us 0.000us 0.00% 16.744ms 63.185us 265
aten::_batch_norm_impl_index 0.31% 1.646ms 7.93% 42.691ms 161.096us 0.000us 0.00% 16.744ms 63.185us 265
------------------------------------------------------- ------------ ------------ ------------ ------------ ------------ ------------ ------------ ------------ ------------ ------------
Differential Revision: D55925068
Pull Request resolved: https://github.com/pytorch/pytorch/pull/123650
Approved by: https://github.com/aaronenyeshi
Summary:
Kineto traces use microsecond level granularity because of chrome tracing defaults to that precision. Fix by adding preprocessor flag to TARGETS and BUCK files. Also remove any unnecessary ns to us conversions made in the profiler itself.
This diff contains profiler changes only. Libkineto changes found in D54964435.
Test Plan:
Check JSON and chrome tracing to make sure values are as expected. Tracing with flags enabled should have ns precision. Tracings without flags should be same as master.
Tracing with flags enabled: https://www.internalfb.com/intern/perfdoctor/trace_view?filepath=tree/traces/dynocli/devvm2185.cco0.facebook.com/rank-0.Mar_18_14_37_22.4155151.pt.trace.json.gz&bucket=gpu_traces
Tracing without flags enabled: https://www.internalfb.com/intern/perfdoctor/trace_view?filepath=tree/traces/dynocli/devvm2185.cco0.facebook.com/rank-0.Mar_18_14_39_15.4166047.pt.trace.json.gz&bucket=gpu_traces
Tracing on main: https://www.internalfb.com/intern/perfdoctor/trace_view?filepath=tree/traces/dynocli/devvm2185.cco0.facebook.com/rank-0.Mar_18_14_42_43.4177559.pt.trace.json.gz&bucket=gpu_traces
Ran key_averages() to make sure FunctionEvent code working as expected:
-- ------------ ------------
Name Self CPU % Self CPU CPU total % CPU total CPU time avg Self CUDA Self CUDA % CUDA total CUDA time avg # of Calls
ProfilerStep* 0.74% 3.976ms 64.40% 346.613ms 69.323ms 0.000us 0.00% 61.710ms 12.342ms 5
Optimizer.zero_grad#SGD.zero_grad 0.76% 4.109ms 0.76% 4.109ms 821.743us 0.000us 0.00% 0.000us 0.000us 5
## forward ## 6.89% 37.057ms 27.19% 146.320ms 29.264ms 0.000us 0.00% 58.708ms 11.742ms 5
aten::conv2d 0.22% 1.176ms 7.74% 41.658ms 157.199us 0.000us 0.00% 27.550ms 103.962us 265
aten::convolution 0.79% 4.273ms 7.52% 40.482ms 152.762us 0.000us 0.00% 27.550ms 103.962us 265
aten::_convolution 0.69% 3.688ms 6.73% 36.209ms 136.637us 0.000us 0.00% 27.550ms 103.962us 265
aten::cudnn_convolution 6.04% 32.520ms 6.04% 32.520ms 122.719us 27.550ms 8.44% 27.550ms 103.962us 265
aten::add_ 2.42% 13.045ms 2.42% 13.045ms 30.694us 12.700ms 3.89% 12.700ms 29.882us 425
aten::batch_norm 0.19% 1.027ms 8.12% 43.717ms 164.971us 0.000us 0.00% 16.744ms 63.185us 265
aten::_batch_norm_impl_index 0.31% 1.646ms 7.93% 42.691ms 161.096us 0.000us 0.00% 16.744ms 63.185us 265
------------------------------------------------------- ------------ ------------ ------------ ------------ ------------ ------------ ------------ ------------ ------------ ------------
Differential Revision: D55087993
Pull Request resolved: https://github.com/pytorch/pytorch/pull/122425
Approved by: https://github.com/aaronenyeshi
