[xpu][test][1/N] Enable tests of test_nn.py on Intel GPU - instantiate TestNN with instantiate_device_type_tests by daisyden · Pull Request #166396 · pytorch/pytorch

daisyden · 2025-10-28T10:18:29Z

For #114850, we will port aten unit tests to Intel GPU. This PR will work on test/test_nn.py TEST_NN class for single GPU test only. We could enable Intel GPU with following methods and try the best to keep the original code styles:

Use torch.accelerator to extend cude specific test to XPU.
Added skipIfXPU decorator for cases with known issues on Intel GPU
Enabled 'xpu' for some test pathes

pytorch-bot · 2025-10-28T10:18:32Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/166396

📄 Preview Python docs built from this PR
📄 Preview C++ docs built from this PR
❓ Need help or want to give feedback on the CI? Visit the bot commands wiki

Note: Links to docs will display an error until the docs builds have been completed.

❌ 2 New Failures

As of commit ee010a0 with merge base 24e0e50 ():

NEW FAILURES - The following jobs have failed:

trunk / linux-jammy-cuda13.0-py3.10-gcc11 / test (default, 1, 5, linux.g6.4xlarge.experimental.nvidia.gpu) (gh)
test_nn 1/1 failed!
xpu / linux-noble-xpu-n-py3.10 / test (default, 12, 12, linux.idc.xpu) (gh)
'Test'

This comment was automatically generated by Dr. CI and updates every 15 minutes.

guangyey · 2025-10-29T16:58:35Z

test/test_nn.py

-            for cuda in ['cuda', 'cuda:0' if torch.cuda.device_count() == 1 else 'cuda:1']:
-                m2 = m.cuda(device=cuda)
-                self.assertIs(m2, m2.to(cuda))
+        if torch.cuda.is_available() or torch.xpu.is_available():


Suggested change

if torch.cuda.is_available() or torch.xpu.is_available():

if TEST_GPU :

guangyey · 2025-10-30T13:14:11Z

test/test_nn.py

    def test_CTCLoss_zero_lengths(self):
        devices = ['cpu']
-        devices += ['cuda'] if TEST_CUDA else []
+        devices += [device_type] if TEST_CUDA or TEST_XPU else []


Suggested change

devices += [device_type] if TEST_CUDA or TEST_XPU else []

devices += [device_type] if TEST_GPU else []

guangyey · 2025-10-30T13:18:35Z

test/test_nn.py

            self.assertTrue((inp.grad == 0).all().item())

-    @unittest.skipIf(not TEST_CUDA, 'CUDA not available')
+    @unittest.skipIf(not TEST_CUDA and not TEST_XPU, 'CUDA and XPU not available')


Suggested change

@unittest.skipIf(not TEST_CUDA and not TEST_XPU, 'CUDA and XPU not available')

@unittest.skipIf(not TEST_GPU, 'CUDA and XPU not available')

guangyey · 2025-10-30T13:19:43Z

test/test_nn.py



-    @unittest.skipIf(not TEST_CUDA, 'CUDA not available')
+    @unittest.skipIf(not TEST_CUDA and not TEST_XPU, 'CUDA and XPU not available')


Suggested change

@unittest.skipIf(not TEST_CUDA and not TEST_XPU, 'CUDA and XPU not available')

@unittest.skipIf(not TEST_GPU, 'CUDA and XPU not available')

guangyey · 2025-10-30T13:20:34Z

test/test_nn.py


-    @unittest.skipIf(not TEST_CUDA, "CUDA unavailable")
-    @unittest.skipIf(not TEST_CUDNN, "needs cudnn")
+    @unittest.skipIf(not TEST_CUDA and not TEST_XPU, "CUDA and XPU unavailable")


Suggested change

@unittest.skipIf(not TEST_CUDA and not TEST_XPU, "CUDA and XPU unavailable")

@unittest.skipIf(not TEST_GPU, "CUDA and XPU unavailable")

guangyey · 2025-10-30T13:21:07Z

test/test_nn.py

-    @unittest.skipIf(not TEST_CUDA, "CUDA unavailable")
-    @unittest.skipIf(not TEST_CUDNN, "needs cudnn")
+    @unittest.skipIf(not TEST_CUDA and not TEST_XPU, "CUDA and XPU unavailable")
+    @unittest.skipIf(not TEST_CUDNN and not TEST_XPU, "needs cudnn or xpu")


Suggested change

@unittest.skipIf(not TEST_CUDNN and not TEST_XPU, "needs cudnn or xpu")

@unittest.skipIf(not TEST_CUDNN, "needs cudnn")

Here don't need TEST_XPU, right

The case is general and can work on XPU so I didn't skip it.

guangyey · 2025-10-30T13:22:13Z

test/test_nn.py



-    @unittest.skipIf(not torch.cuda.is_available(), "CUDA not available")
+    @unittest.skipIf(not torch.cuda.is_available() and not torch.xpu.is_available(), "CUDA and XPU not available")


Suggested change

@unittest.skipIf(not torch.cuda.is_available() and not torch.xpu.is_available(), "CUDA and XPU not available")

@unittest.skipIf(not TEST_GPU, "CUDA and XPU not available")

guangyey · 2025-10-30T13:24:07Z

test/test_nn.py

    )
    def test_batchnorm(self, dims, mode, memory_format, ref_backend, mixed, dtype):
-        if torch.version.cuda:
+        if torch.version.cuda or torch.version.xpu:


Originally, it means CUDA skips these cases, but ROCM doesn't. I think we should run these cases and see what will happen.

ok, let me try.

guangyey · 2025-10-30T13:24:29Z

test/test_nn.py

            _inference(memory_format, ref_backend, mixed, dtype)

-    @unittest.skipIf(not torch.cuda.is_available(), "CUDA not available")
+    @unittest.skipIf(not torch.cuda.is_available() and not torch.xpu.is_available(), "CUDA and XPU not available")


Suggested change

@unittest.skipIf(not torch.cuda.is_available() and not torch.xpu.is_available(), "CUDA and XPU not available")

@unittest.skipIf(not TEST_GPU, "CUDA and XPU not available")

guangyey · 2025-11-05T09:15:35Z

test/test_nn.py

            self.assertTrue((inp.grad == 0).all().item())

-    @unittest.skipIf(not TEST_CUDA, 'CUDA not available')
+    @unittest.skipIf(not TEST_GPU, 'CUDA and XPU not available')


Suggested change

@unittest.skipIf(not TEST_GPU, 'CUDA and XPU not available')

@unittest.skipIf(not TEST_GPU, 'The current accelerator is not available')

guangyey · 2025-11-05T09:21:34Z

test/test_nn.py

-                self.skipTest("Failed on CUDA")
+                self.skipTest(f"Failed on {device_type.upper()}")
+
+        if torch.version.xpu:


Suggested change

if torch.version.xpu:

if TEST_XPU:

guangyey

lgtm.

malfet

Is it possible to split this change into few smaller ones? For example, change to common_device_types and test_namedtensor.py looks totally fine and majority of the changes to test_nn are also OK, but multigpu changes seems a bit risky

Perhaps let's do single GPU enablement in 1 PR and than multi-GPU in another

malfet · 2025-11-05T23:37:07Z

test/test_nn.py

-                if device == 'cuda':
-                    rnn.cuda()
-                    input = input.cuda()
+                if device == 'cuda' or device == 'xpu':


Suggested change

if device == 'cuda' or device == 'xpu':

if device in ['cuda', 'xpu']:

Thanks for the suggestions! I will split the PR into two, one for single GPU and one for multi-gpu.

Done, this PR is now only for single gpu.

malfet · 2025-11-05T23:39:21Z

test/test_nn.py

    import numpy as np

+TEST_GPU = torch.cuda.is_available() or torch.xpu.is_available()
+TEST_MULTIGPU = TEST_GPU and torch.accelerator.device_count() >= 2


Why not change the definition in common_cuda.py?

Can I add TEST_GPU and TEST_MULTIGPU definition in common_utils.py?

Hi, @malfet, I removed the code updates for multi-gpu from this PR and will create another PR for them. I also moved the TEST_GPU definition to common_utils. Please further review this PR. Thanks.

Hi, @malfet, TEST_GPU is not a general definition. After consideration I feel it is not appropriate to put under common_utils or common_cuda. I define it as

TEST_GPU = TEST_CUDA or TEST_XPU

because other accelerators have limitation on test_nn, so I only port those cases for XPU.

daisyden · 2026-03-17T13:05:49Z

@claude Could you do another pass on this PR, taking into account previous review comments. Make sure that the functionality being tested was not changed.

claude · 2026-03-17T13:06:19Z

Claude finished @daisyden's task in 3m 26s —— View job

PR Review: [xpu][test][1/N] Enable tests of test_nn.py on Intel GPU

Gather context and read changed files
Review common_device_type.py changes
Review common_utils.py changes
Review test_nn.py changes - check for functionality changes
Check if previous review comments were addressed
Post final review

Summary

This PR migrates TestNN to use instantiate_device_type_tests, adds an onlyAccelerator decorator and ACCELERATOR_TYPE constant. The PR has improved significantly across the review iterations. Nearly all previous feedback has been addressed. Two minor issues remain.

`common_device_type.py` and `common_utils.py` Changes

The onlyAccelerator decorator and ACCELERATOR_TYPE additions are correct. ACCELERATOR_TYPE uses LazyVal to defer evaluation. onlyAccelerator properly skips on cpu and meta devices. No issues.

Previous Review Comments Status

Feedback	Status
Critical: `device_list` undefined at instantiation	Fixed ✅ - `device_list = None` default, uses `except_for` pattern
Remove internal `device_list` loops	Fixed ✅ - All internal loops removed
`test_to` loop variable `gpu` unused	Fixed ✅ - Now correctly uses `gpu` inside loop body
Missing f-string in `test_batchnorm` skip list	Fixed ✅ - Line 5259 correctly uses f-string
`test_upsampling_bfloat16` shadows `device`	Fixed ✅ - Internal loop removed
`test_mse_loss_mixed_dtype_grad` deleted	Fixed ✅ - Restored with `device` parameter (line 2672)
`test_grid_sample_error_checking` device-mismatch test	Fixed ✅ - `input` and `grid` created on CPU (line 5798), then `input.to(device)` moves only input to accelerator while grid stays on CPU
`test_projections_lstm_args_check` model/input device mismatch	Fixed ✅ - `get_inputs()` now takes and uses `device` parameter
`test_cosine_similarity_mixed_precision` x5/x6 not on device	Fixed ✅ - Now uses `device=device` (lines 5790-5791)
`largeTensorTest` hardcoded `device="cuda"`	Fixed ✅ - `device=` argument removed
`test_interpolate_buffer_overflow` internal device loop	Fixed ✅ - Loop removed, uses `device` parameter directly
cudnn tests marked `@onlyAccelerator`	Partially fixed - `test_batchnorm_cudnn_nhwc` correctly no longer has `@onlyAccelerator`. `test_batchnorm_cudnn_half` still has `@onlyAccelerator` but the cudnn portion is gated behind `if TEST_CUDNN:` so non-CUDA accelerators only run the THNN portion - acceptable
@mikaylagawarecki suggestion on `except_for` for 3rd party backends	Fixed ✅ - Uses `except_for=device_list` pattern

Remaining Issues

1. Minor: `test_cosine_similarity_mixed_precision` - `x3`/`x4` CPU-only comparison has a subtle mismatch (test/test_nn.py:5778-5788)

Lines 5778-5779 create x3 and x4 on CPU (no device=device), compute result_cpu, then lines 5784-5788 compare against an accelerator copy. This is intentional for cross-device comparison. However, there's a minor issue: result_cpu on line 5795 refers to torch.ones(2, dtype=torch.float16) which is created on CPU - when device is an accelerator, the comparison torch.isclose(result, torch.ones(2, dtype=torch.float16)) compares a device tensor with a CPU tensor. PyTorch handles this automatically, so it works, but it's slightly inconsistent.

This is not a blocker - just a nit for consistency.

2. Minor: `test_interpolate_buffer_overflow` dead code branch (test/test_nn.py:6930)

Line 6930 has device == 'cuda' and dtype == torch.bfloat16 as a skip condition. When this test runs on a non-CUDA accelerator (e.g., XPU), this branch will never match since device != 'cuda', so bfloat16 would be tested on XPU even though the original intent was only to skip bfloat16 on CUDA. If XPU has similar bfloat16 limitations, this might need updating in a follow-up PR. Not a blocker.

Verdict

The PR is ready to merge. All previously identified bugs and functionality changes have been fixed. The migration approach using instantiate_device_type_tests is correct. The two remaining notes are minor and don't affect correctness.

View job

mikaylagawarecki

Thank you!

For my understanding -- the main change in this PR was to instantiate TestNN with instantiate_device_type_tests. TestNN now no longer has @onlyCUDA. It seems like this PR also migrated some (but not all) @onlyCUDA in TestNNDeviceType to @onlyAccelerator. Will there be a follow-up PR that removes more @onlyCUDA from TestNNDeviceType (or moves them into a separate class if they are truly cuda-specific)?

mikaylagawarecki · 2026-03-17T15:53:31Z

test_comprehensive_linalg_lu_factor_xpu_float32
failure seems unrelated

mikaylagawarecki · 2026-03-17T15:53:37Z

@pytorchbot merge -i

pytorch-bot · 2026-03-17T15:53:43Z

This PR has pending changes requested. Please address the comments and update the PR before merging.

comments alreaady addressed

mikaylagawarecki · 2026-03-17T15:54:16Z

@pytorchbot merge -i

pytorchmergebot · 2026-03-17T15:57:17Z

Merge started

Your change will be merged while ignoring the following 1 checks: xpu / linux-noble-xpu-n-py3.10 / test (default, 6, 12, linux.idc.xpu)

Learn more about merging in the wiki.

Questions? Feedback? Please reach out to the PyTorch DevX Team

Advanced Debugging

Check the merge workflow status
here

pytorchmergebot · 2026-03-17T17:12:00Z

Merge failed

Reason: 1 jobs have failed, first few of them are: xpu / linux-noble-xpu-n-py3.10 / test (default, 3, 12, linux.idc.xpu)

Details for Dev Infra team

Raised by workflow job

mikaylagawarecki · 2026-03-18T18:04:44Z

@pytorchbot merge -i

pytorchmergebot · 2026-03-18T18:06:55Z

Merge started

Your change will be merged while ignoring the following 2 checks: xpu / linux-noble-xpu-n-py3.10 / test (default, 3, 12, linux.idc.xpu), trunk / linux-jammy-cuda13.0-py3.10-gcc11 / test (default, 1, 5, lf.linux.g6.4xlarge.experimental.nvidia.gpu)

Learn more about merging in the wiki.

Questions? Feedback? Please reach out to the PyTorch DevX Team

Advanced Debugging

Check the merge workflow status
here

pytorch-auto-revert · 2026-03-18T23:29:43Z

@pytorchbot revert -m "Reverted automatically by pytorch's autorevert, to avoid this behaviour add the tag autorevert: disable" -c autorevert

This PR is attributed to have caused regression in:

slow: linux-jammy-cuda13.0-py3.10-gcc11-sm86 / test [test] (hud)

Please investigate and fix the issues.
@claude Can you please read this revert comment, follow the links and read the errors, to then give a brief diagnostics on the cause of the error? If you judge the error to be legitimate reason for a revert, please provide brief guidance on how the author could fix it.

claude · 2026-03-18T23:30:11Z

Claude finished @pytorch-auto-revert[bot]'s task in 5m 16s —— View job

Revert Diagnostics for PR #166396

Read the revert context and CI failure details
Diagnose the root cause of the failure
Assess if the revert is legitimate and provide fix guidance

Root Cause

The revert is legitimate. The CI failure is a direct result of this PR's changes.

Error: test_nn.py fails at import time with:

AssertionError: Redefinition of test test_MultiMarginLoss_no_reduce_cuda

(and in a parallel shard: Redefinition of test test_softmax_spatial_dtype_cuda)

This occurs at common_device_type.py:458 during the newly-added instantiate_device_type_tests(TestNN, globals(), except_for=device_list) call at test/test_nn.py:14280.

Why It Happens

The TestNN class has tests added two different ways:

Manually-defined methods (the ones this PR migrated to use device parameter)
Dynamically-generated CUDA test methods added by the add_test() function at test/test_nn.py:7778-7848. This function iterates over module_tests + get_new_module_tests() (line 7850) and calls setattr(TestNN, cuda_test_name, fn) to add methods like test_MultiMarginLoss_no_reduce_cuda_float, test_MultiMarginLoss_no_reduce_cuda_double, etc.

When instantiate_device_type_tests(TestNN, ...) runs, it creates a TestNNCUDA subclass and tries to instantiate device-specific variants of every test method in TestNN. But TestNN already has test_MultiMarginLoss_no_reduce_cuda* methods (dynamically added by add_test()), so instantiate_device_type_tests detects a name collision and raises AssertionError.

This error only manifests in the slow CI configuration because:

The slow config runs test_nn.py on CUDA machines, which triggers the CUDA instantiation path
The non-slow trunk tests may have been configured differently or the specific shard didn't include test_nn.py

How To Fix

The author needs to handle the dynamically-generated tests from add_test(). There are a few approaches:

Exclude dynamically-added _cuda methods from instantiate_device_type_tests: Before calling instantiate_device_type_tests(TestNN, ...), remove or skip the dynamically-added CUDA test methods from TestNN so they don't collide. This could mean moving the add_test() loop to only add CPU tests, or filtering out _cuda methods.
Move the add_test() dynamically-generated tests to a separate class: Create a new class (e.g., TestNNModuleTests) for the dynamically-generated module/criterion tests and keep TestNN clean for instantiate_device_type_tests. This is probably the cleanest approach.
Don't use instantiate_device_type_tests on TestNN: If the dynamic test generation is too entangled, consider only applying instantiate_device_type_tests to a subset of TestNN's manually-defined tests by moving them to a new class.

The key insight is that add_test() at line 7778 already creates _cuda suffixed methods on TestNN, and instantiate_device_type_tests tries to create the same suffixed methods, causing the collision.

pytorchmergebot · 2026-03-18T23:33:10Z

@pytorchbot successfully started a revert job. Check the current status here.
Questions? Feedback? Please reach out to the PyTorch DevX Team

pytorchmergebot · 2026-03-18T23:33:19Z

@daisyden your PR has been successfully reverted.

…stantiate TestNN with instantiate_device_type_tests (#166396)" This reverts commit b87c36c. Reverted #166396 on behalf of https://github.com/pytorch-auto-revert due to Reverted automatically by pytorch's autorevert, to avoid this behaviour add the tag autorevert: disable ([comment](#166396 (comment)))

malfet

Two comments:

Test plan for changes like that should include links to some runs before and after the change to guarantee there are no significant changes in the number of tests run before/after change (as this have a potential of skipping lots of those for no good reasion
Never use skipIf decorator unless it hard crashes, instead use extectedFailureIf

malfet · 2026-03-18T23:49:17Z

test/test_nn.py

+        _dtype = dtype_name(dtype)
+        if torch.version.cuda or device == 'xpu':
+            skip_tests = [
+                f"test_batchnorm_2D_train_NCHW_vs_cpu_mixed_bfloat16_{device}_{_dtype}",


Err, why do you need to parametrize test name of of a sudden?

malfet · 2026-03-18T23:51:40Z

test/test_nn.py

-    def test_grid_sample_half_precision(self):
+    @onlyAccelerator
+    # TypeError: Cannot convert a MPS Tensor to float64 dtype as the MPS framework doesn't support float64
+    @skipIfMPS


Never use skipIfMPS, rather @expectedFailureMPS or something of that nature

malfet · 2026-03-18T23:52:34Z

test/test_nn.py

+device_list = None
+# https://github.com/pytorch/pytorch/issues/177119
+if os.environ.get('PYTORCH_TEST_WITH_DYNAMO', '0') == '1':
+    device_list = ('cpu', )
+
+instantiate_device_type_tests(TestNN, globals(), except_for=device_list)


Can you elaborate why this is needed?

malfet · 2026-03-18T23:57:22Z

To make probability of revert much smaller, do you mind breaking it down into several smaller PRs? For example one can add onlyAccelerator instead of onlyCUDA and use it throughout the codebase

daisyden requested a review from mruberry as a code owner October 28, 2025 10:18

pytorch-bot bot added the release notes: nn release notes category label Oct 28, 2025

daisyden added ciflow/xpu Run XPU CI tasks ciflow/mps Run MPS tests (subset of trunk) keep-going Don't stop on first failure, keep running tests until the end labels Oct 28, 2025

daisyden changed the title ~~Enable some tests of TEST_NN class on Intel GPU~~ [WIP][xpu][test]Enable some tests of TEST_NN class on Intel GPU Oct 28, 2025

daisyden mentioned this pull request Oct 28, 2025

[WIP][xpu][test] port test/test_nn.py to Intel GPU #162370

Closed

pytorchbot added the open source label Oct 28, 2025

daisyden force-pushed the daisyden/test_nn_stage1 branch 2 times, most recently from fb44286 to df99d6c Compare October 29, 2025 01:30

guangyey reviewed Oct 29, 2025

View reviewed changes

guangyey reviewed Oct 30, 2025

View reviewed changes

daisyden force-pushed the daisyden/test_nn_stage1 branch 2 times, most recently from 0ee4e01 to 343d722 Compare November 5, 2025 03:30

daisyden changed the title ~~[WIP][xpu][test]Enable some tests of TEST_NN class on Intel GPU~~ [WIP][xpu][test][1/N] Enable some tests of TEST_NN class on Intel GPU Nov 5, 2025

daisyden force-pushed the daisyden/test_nn_stage1 branch from 343d722 to d7c7830 Compare November 5, 2025 07:23

guangyey added this to PyTorch Intel Nov 5, 2025

guangyey reviewed Nov 5, 2025

View reviewed changes

guangyey previously approved these changes Nov 5, 2025

View reviewed changes

daisyden force-pushed the daisyden/test_nn_stage1 branch from d7c7830 to 092c678 Compare November 5, 2025 13:10

malfet requested changes Nov 5, 2025

View reviewed changes

mikaylagawarecki approved these changes Mar 17, 2026

View reviewed changes

pytorch-bot bot added the ciflow/trunk Trigger trunk jobs on your pull request label Mar 17, 2026

pytorchmergebot added the merging label Mar 17, 2026

pytorchmergebot removed the merging label Mar 17, 2026

pytorchmergebot added the merging label Mar 18, 2026

pytorchmergebot added the Merged label Mar 18, 2026

pytorchmergebot closed this in b87c36c Mar 18, 2026

pytorchmergebot removed the merging label Mar 18, 2026

pytorchmergebot added Reverted ci-no-td Do not run TD on this PR labels Mar 18, 2026

pytorchmergebot reopened this Mar 18, 2026

malfet requested changes Mar 18, 2026

View reviewed changes

	if torch.cuda.is_available() or torch.xpu.is_available():
	if TEST_GPU :

	devices += [device_type] if TEST_CUDA or TEST_XPU else []
	devices += [device_type] if TEST_GPU else []

	@unittest.skipIf(not TEST_CUDA and not TEST_XPU, 'CUDA and XPU not available')
	@unittest.skipIf(not TEST_GPU, 'CUDA and XPU not available')



		@unittest.skipIf(not TEST_CUDA, 'CUDA not available')
		@unittest.skipIf(not TEST_CUDA and not TEST_XPU, 'CUDA and XPU not available')

	@unittest.skipIf(not TEST_CUDA and not TEST_XPU, "CUDA and XPU unavailable")
	@unittest.skipIf(not TEST_GPU, "CUDA and XPU unavailable")

	@unittest.skipIf(not TEST_CUDNN and not TEST_XPU, "needs cudnn or xpu")
	@unittest.skipIf(not TEST_CUDNN, "needs cudnn")



		@unittest.skipIf(not torch.cuda.is_available(), "CUDA not available")
		@unittest.skipIf(not torch.cuda.is_available() and not torch.xpu.is_available(), "CUDA and XPU not available")

	@unittest.skipIf(not TEST_GPU, 'CUDA and XPU not available')
	@unittest.skipIf(not TEST_GPU, 'The current accelerator is not available')

	if device == 'cuda' or device == 'xpu':
	if device in ['cuda', 'xpu']:

Conversation

daisyden commented Oct 28, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-bot bot commented Oct 28, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/166396

❌ 2 New Failures

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

guangyey Oct 30, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

guangyey left a comment

Choose a reason for hiding this comment

Uh oh!

malfet left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

daisyden Nov 5, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

daisyden commented Mar 17, 2026

Uh oh!

claude bot commented Mar 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

PR Review: [xpu][test][1/N] Enable tests of test_nn.py on Intel GPU

Summary

common_device_type.py and common_utils.py Changes

Previous Review Comments Status

Remaining Issues

1. Minor: test_cosine_similarity_mixed_precision - x3/x4 CPU-only comparison has a subtle mismatch (test/test_nn.py:5778-5788)

2. Minor: test_interpolate_buffer_overflow dead code branch (test/test_nn.py:6930)

Verdict

Uh oh!

mikaylagawarecki left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mikaylagawarecki commented Mar 17, 2026

daisyden commented Oct 28, 2025 •

edited

Loading

pytorch-bot bot commented Oct 28, 2025 •

edited

Loading

guangyey Oct 30, 2025 •

edited

Loading

daisyden Nov 5, 2025 •

edited

Loading

claude bot commented Mar 17, 2026 •

edited

Loading

`common_device_type.py` and `common_utils.py` Changes

1. Minor: `test_cosine_similarity_mixed_precision` - `x3`/`x4` CPU-only comparison has a subtle mismatch (test/test_nn.py:5778-5788)

2. Minor: `test_interpolate_buffer_overflow` dead code branch (test/test_nn.py:6930)

mikaylagawarecki left a comment •

edited

Loading

claude bot commented Mar 18, 2026 •

edited

Loading

malfet Mar 18, 2026 •

edited

Loading