torch.cuda.memory_allocated to return {} if not initialized

[malfet]

Fixes #49952

[facebook-github-bot]

💊 CI failures summary and remediations

As of commit 776ed5c (more details on the Dr. CI page):

---

• 1/1 failures introduced in this PR

---

🕵️ 1 new failure recognized by patterns

The following CI failures do not appear to be due to upstream breakages:

pytorch_linux_xenial_py3_6_gcc5_4_test (1/1)

Step: "Run tests" (full log | diagnosis details | 🔁 rerun)

Jan 28 01:55:00 AssertionError: mypy failed: test/test_complex.py:21: error: Signature of "forward" incompatible with supertype "ScriptModule" [override]

Jan 28 01:53:41 ----------------------------------------------------------------------
Jan 28 01:53:49   test_doc_examples (__main__.TestTypeHints) ... ok (7.880s)
Jan 28 01:55:00   test_run_mypy (__main__.TestTypeHints) ... FAIL (70.719s)
Jan 28 01:55:00 
Jan 28 01:55:00 ======================================================================
Jan 28 01:55:00 FAIL [70.719s]: test_run_mypy (__main__.TestTypeHints) [mypy.ini]
Jan 28 01:55:00 ----------------------------------------------------------------------
Jan 28 01:55:00 Traceback (most recent call last):
Jan 28 01:55:00   File "test_type_hints.py", line 171, in test_run_mypy
Jan 28 01:55:00     self.fail(f"mypy failed: {stdout} {stderr}")
Jan 28 01:55:00 AssertionError: mypy failed: test/test_complex.py:21: error: Signature of "forward" incompatible with supertype "ScriptModule"  [override]
Jan 28 01:55:00 Found 1 error in 1 file (checked 1211 source files)
Jan 28 01:55:00  
Jan 28 01:55:00 
Jan 28 01:55:00 ----------------------------------------------------------------------
Jan 28 01:55:00 Ran 2 tests in 78.599s
Jan 28 01:55:00 
Jan 28 01:55:00 FAILED (failures=1)
Jan 28 01:55:00 
Jan 28 01:55:00 Generating XML reports...
Jan 28 01:55:00 Generated XML report: test-reports/python-unittest/TEST-TestTypeHints-20210128015341.xml

---

This comment was automatically generated by Dr. CI (expand for details).

Follow this link to opt-out of these comments for your Pull Requests.

Please report bugs/suggestions to the (internal) Dr. CI Users group.

[facebook-github-bot]

@malfet has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

[ngimel]

Thanks for the fix!
Can we add tests? Also, another case reported in #49952 still does not work (reports equal memory usage for both gpus)

python -c 'import torch; x=torch.ones(10<<20).to(0); y=torch.ones(10).to(1);print([torch.cuda.memory_allocated(torch.cuda.device(id)) for id in range(torch.cuda.device_count())])'

Weirdly, it works if torch.cuda.device(id) is replaced by id.

[codecov]

Codecov Report

Merging #51179 (d9666fb) into master (ba316a7) will increase coverage by 0.00%.
The diff coverage is 83.33%.

@@           Coverage Diff           @@
##           master   #51179   +/-   ##
=======================================
  Coverage   80.88%   80.88%           
=======================================
  Files        1931     1931           
  Lines      210560   210562    +2     
=======================================
+ Hits       170311   170315    +4     
+ Misses      40249    40247    -2     

[malfet]

Weirdly, it works if torch.cuda.device(id) is replaced by id.

Fixed, although according to type annotations, this function is not supposed to be called with torch.cuda.device, which despite the name is not inherited from torch.device, filed #51224 to track that.

[ngimel]

Awesome, can you please add tests?

[facebook-github-bot]

@malfet has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

[facebook-github-bot]

@malfet merged this pull request in 43f0ccd.