For tests that TD prioritizes, we should track what their ordering _would have been_ if none of the TD heuristics had applied to it.
This is useful for two reasons:
1. It lets us better understand TD may have contributed to that test running sooner
2. it's possible that heuristics actually mark a test as less important than the default sorting would have claimed (the default sorts tests in a fixed order). This will let us track how often that happens
Pull Request resolved: https://github.com/pytorch/pytorch/pull/110031
Approved by: https://github.com/clee2000
When a test fails, we will now emit fine grained details about how accurately heuristics predicted the relevance of that test.
## Context
Why only look at failing tests? Our only signal that a PR is most likely relevant to a test is whether or not a test fails on it. Green tests don't tell us if the success was due to the code being good vs being irrelevant. This isn't a perfect measure, since it can miscategorize unstable and flaky failures as having been "missed" by the heuristics, but it's a reasonable approximation.
## What's measured?
The metrics this PR collects are designed to answer the following questions
### How comprehensive are the heuristics?
- What's the false negative rate, the % of failures that ideally should have been prioritized but weren't? (Both at an aggregate level and at a per heuristic level)
### How precise are the heuristics?
- What % of failed tests were prioritized by a given heuristic? What % was prioritized overall?
- How relevant was a failed test was considered to be? (Both a aggregate level and at a per heuristic level)
- What % of time was a given heuristic prioritizing a failing test higher than any other heuristic?
Pull Request resolved: https://github.com/pytorch/pytorch/pull/108192
Approved by: https://github.com/huydhn
ghstack dependencies: #108117
This PR looks big, but it's mostly just refactorings with a bit of dead code deletion. Exceptions are:
- Some metric emissions were changed to comply with the new TD format
- Some logging changes
- We now run tests in three batches (highly_relevant, probably_relevant, unranked_relevance) instead of the previous two (prioritized and general)
Refactorings done:
- Moves all test reordering code to the new TD framework
- Refactors run_test.py to cleanly support multiple levels of test priorities
- Deletes some dead code that was originally written for logging
Pull Request resolved: https://github.com/pytorch/pytorch/pull/107071
Approved by: https://github.com/clee2000, https://github.com/huydhn
Make it so that scripts can import and run the `emit_metrics` function even if they don't have boto3 installed, in which case it will still validate the inputs but skip the actual metric emission part.
It's purely a refactor without any real logic changes
Motivation: So that run_test.py and the target determination code can use this library easily without worrying about if it was imported or if it's dependencies are installed.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/107070
Approved by: https://github.com/huydhn
Companion with https://github.com/pytorch/test-infra/pull/4424
Uses the file rating generated by the test infra PR to re order tests. For each test file, sum the file ratings from the changed files in the PR, and put the tests in order of sum.
A lot of tests are probably going to end up as "prioritized" since it takes anything with a rating > 0 right now.
Sharding is done twice, once on the prioritized tests, and once on the general/non prioritized tests. Prioritized tests have an order, so they should be sharded according to that order, while general tests don't have an order and are sharded by test time, which should result in more balanced shards.
I'll change the metric name before I merge, i want to quarantine my testing stuff from actual results
Pull Request resolved: https://github.com/pytorch/pytorch/pull/106347
Approved by: https://github.com/ZainRizvi
Adds a new structure to house all heuristics we use for Target Determination and Test Reordering. I'm keeping it somewhat minimal for now, to let it evolve more easily as we try new things.
It currently does nothing. The 2nd pr in the stack ports the existing heuristics to actually use this new framework
Pull Request resolved: https://github.com/pytorch/pytorch/pull/106997
Approved by: https://github.com/clee2000, https://github.com/huydhn
Companion with https://github.com/pytorch/test-infra/pull/4424
Uses the file rating generated by the test infra PR to re order tests. For each test file, sum the file ratings from the changed files in the PR, and put the tests in order of sum.
A lot of tests are probably going to end up as "prioritized" since it takes anything with a rating > 0 right now.
Sharding is done twice, once on the prioritized tests, and once on the general/non prioritized tests. Prioritized tests have an order, so they should be sharded according to that order, while general tests don't have an order and are sharded by test time, which should result in more balanced shards.
I'll change the metric name before I merge, i want to quarantine my testing stuff from actual results
Pull Request resolved: https://github.com/pytorch/pytorch/pull/106347
Approved by: https://github.com/ZainRizvi
Added a feature to upload test statistics to DynamoDB and Rockset using a new function `emit_metric` in `tools/stats/upload_stats_lib.py`.
Added metrics to measure test reordering effectiveness in `tools/testing/test_selections.py`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/102691
Approved by: https://github.com/malfet
Rocm queries for the number of processes it should use per machine, which might cause it be different across shards, which leads to inconsistencies when distributing tests among shards.
My solution is to separate the vars used for shard calculations and the actual number of procs that can be used and to ensure that the var used for shard calculations is consistent across all shards for a test config + job. I believe that the only consequence is that rocm sharding might become unbalanced.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/102871
Approved by: https://github.com/huydhn, https://github.com/malfet
sharding on rocm is broken, i cant replicate on dummy PRs even though it seems to happen pretty often on main, so adding this to increase my sample size. Hopefully this is enough print statements...
Pull Request resolved: https://github.com/pytorch/pytorch/pull/102713
Approved by: https://github.com/huydhn
Makes the CI prioritize running any test files that had a failing test in a previous iteration of the given PR.
A follow up to https://github.com/pytorch/pytorch/pull/100522 which makes the `.pytest_cache` available to use here
A concrete example:
1. Person A pushes a new commit and creates a PR.
2. 2 hours later, test_im_now_broken.py fails
3. Person A attempts to fix the test, but the test is actually still broken
4. The CI, seeing that test_im_now_broken.py had failed on a previous run, will now prioritize running that test first. Instead of waiting another 2 hours to get a signal, Person A only needs to wait ~15 minutes (which is how long it takes for tests to start running)
# Testing
I modified a file to make the tests invoking it fail and triggered CI twice with this failure.
First run: https://github.com/pytorch/pytorch/actions/runs/4963943209/jobs/8883800811
Test step took 1h 9m to run
Second run: https://github.com/pytorch/pytorch/actions/runs/4965016776/jobs/8885657992
Test step failed within 2m 27s
Pull Request resolved: https://github.com/pytorch/pytorch/pull/101123
Approved by: https://github.com/malfet, https://github.com/huydhn
Today, we prioritize running test files that were edited in the user's PR, with the idea being to run them before we run any other test.
Except, if the modified test is supposed to run serially, then we still end up running it after all the parallelized tests have finished running.
This PR fixes that to _always_ run the prioritized tests before the regular tests, regardless of if the test is supposed to run serially or in parallel
Pull Request resolved: https://github.com/pytorch/pytorch/pull/100748
Approved by: https://github.com/huydhn
i think get_reordered_tests broken since master -> main switch
add typing for some functions
checked for `prioritized` in the logs
limited testing because I only care about one very small part of the log thats near the beginning
Pull Request resolved: https://github.com/pytorch/pytorch/pull/100752
Approved by: https://github.com/huydhn
Sharing code between the code that handles test results in parallel vs serial mode.
Note that the original version of this code had an inconsistency between the two versions where it would execute `print_to_stderr(err_message)` on every test that ran in parallel, but for serial tests it would only invoke `print_to_stderr(err_message)` if `continue_on_error` was also specified. By sharing code, this PR changes that behavior to be consistent between the two modes.
Also adding some comments.
<!--
copilot:poem
-->
### <samp>🤖 Generated by Copilot at 029342c</samp>
> _Sing, O Muse, of the skillful coder who refined_
> _The PyTorch testing script, `run_test.py`, and shined_
> _A light on its obscure logic, with docstrings and comments_
> _And made it run more smoothly, with better error contents_
Pull Request resolved: https://github.com/pytorch/pytorch/pull/99467
Approved by: https://github.com/huydhn, https://github.com/malfet
Fixes#90940. This PR revamps how tests are run in parallel as well as device visibility at the docker container and within the run_test.py test runner.
First, running multiple test modules concurrently on the same GPU was causing instability for ROCm runners manifesting as timeouts. ROCm runners have at least 1 GPU each, but often 2 or more. This PR allows NUM_PROCS to be set equal to the number of devices available, but also takes care to set HIP_VISIBLE_DEVICES to avoid oversubscribing any GPU.
Second, we had introduced env vars `-e ROCR_VISIBLE_DEVICES` (#91031) to prepare for two GHA runners per CI node, to split up the GPU visibility at the docker level between the two runners. This effort wasn't fully realized; to date, we haven't had more than one runner per CI host. We abandon this effort in favor of all GPUs being visible to a single runner and managing GPU resources as stated above.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/91137
Approved by: https://github.com/kit1980, https://github.com/huydhn, https://github.com/pruthvistony
tbh at this point it might be easier to make a new workflow and copy the relevant jobs...
Changes:
* Disable cuda mem leak check except for on scheduled workflows
* Make pull and trunk run on a schedule which will run the memory leak check
* Periodic will always run the memory leak check -> periodic does not have parallelization anymore
* Concurrency check changed to be slightly more generous
Pull Request resolved: https://github.com/pytorch/pytorch/pull/88373
Approved by: https://github.com/ZainRizvi, https://github.com/huydhn
Sometimes you want to query the small element of a set of elements and use `sorted(elements)[0]` without a second thought. However, this is not optimal, since the entire list must be sorted first `O(n log n)`. It would be better to use the `min(elements)` method provided for this purpose `O(n)`.
Furthermore `sorted(elements)[::-1]` is not very efficient, because it would be better to use `sorted(elements, reverse=True)` to save the slice operation.
**TLDR: using `sorted(elements)[0]` is slow and can be replaced with `min(elements)`.**
I stumbled across these code snippets while playing around with CodeQL (see https://lgtm.com/query/4148064474379348546/).
Pull Request resolved: https://github.com/pytorch/pytorch/pull/86995
Approved by: https://github.com/jansel
run tests in parallel at the test file granularity
runs 3 files in parallel using multiprocessing pool, output goes to a file, which is then printed when the test finishes. Some tests cannot be run in parallel (usually due to lacking memory), so we run those after. Sharding is changed to attempt to mask large files with other large files/run them on the same shard.
test_ops* gets a custom handler to run it because it is simply too big (2hrs on windows) and linalg_cholesky fails (I would really like a solution to this if possible, but until then we use the custom handler).
reduces cuda tests by a lot, reduces total windows test time by ~1hr
Ref. https://github.com/pytorch/pytorch/issues/82894
Pull Request resolved: https://github.com/pytorch/pytorch/pull/84961
Approved by: https://github.com/huydhn
Fix use-dict-literal pylint suggestions by changing `dict()` to `{}`. This PR should do the change for every Python file except test/jit/test_list_dict.py, where I think the intent is to test the constructor.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/83718
Approved by: https://github.com/albanD
With ufmt in place https://github.com/pytorch/pytorch/pull/81157, we can now use it to gradually format all files. I'm breaking this down into multiple smaller batches to avoid too many merge conflicts later on.
This batch (as copied from the current BLACK linter config):
* `tools/**/*.py`
Upcoming batchs:
* `torchgen/**/*.py`
* `torch/package/**/*.py`
* `torch/onnx/**/*.py`
* `torch/_refs/**/*.py`
* `torch/_prims/**/*.py`
* `torch/_meta_registrations.py`
* `torch/_decomp/**/*.py`
* `test/onnx/**/*.py`
Once they are all formatted, BLACK linter will be removed.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/81285
Approved by: https://github.com/suo
`JOB_BASE_NAME` was a holdover from jenkins compatibility. Eventually,
it morphed to be always set to the build enviroment + `-test` or
`-build`, and we used it to detect whether we were in a build or test.
That's sort of pointless, so removing and fixing up the few remaining
use cases.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/80046
Approved by: https://github.com/malfet, https://github.com/janeyx99
`JOB_BASE_NAME` was a holdover from jenkins compatibility. Eventually,
it morphed to be always set to the build enviroment + `-test` or
`-build`, and we used it to detect whether we were in a build or test.
That's sort of pointless, so removing and fixing up the few remaining
use cases.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/80046
Approved by: https://github.com/malfet
In the case of target determination, this is just removing comments that
refer to non-existent code.
In the case of the test specification code; this removes (what I believe
to be) an unused feature. If we're using this somehow let me know and I
can revise the PR.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/79372
Approved by: https://github.com/janeyx99
