Do not treat ReinterpretView as a realized node by shengfukevin · Pull Request #159920 · pytorch/pytorch

shengfukevin · 2025-08-06T00:44:25Z

Summary:
Do not treat ReinterpretView as a realized node

Function [gather_origins](https://github.com/pytorch/pytorch/blob/main/torch/_inductor/utils.py#L888](https://l.facebook.com/l.php?u=https%3A%2F%2Fgithub.com%2Fpytorch%2Fpytorch%2Fblob%2Fmain%2Ftorch%2F_inductor%2Futils.py%23L888&h=AT2PYr83thTo6VUjPs26Y8QAN6Sid16rvDMHtxO-Bp9FDwHr4J5PObtH3IhNTL-LPSRVC9WVJAcmwUToVWJIrDWb84i0j61QE55ySYAkGbuigqcNc7xczlirHhbiC9vMqiz91VwWdl4Pe2yKN7VIjjCiFUqw) calls is_realized_node to decide if a FX node should be included in the origins of a IR node. ReinterpretView is considered a realized node, so it is not included in the origins. It leads to an incomplete graph. For example:

@torchdynamo.optimize("inductor")
def fn(input_data, weight):
    normalized_input = input_data * weight.unsqueeze(0)
    return normalized_input
input_data = torch.randn(4272, 192, requires_grad=True).to(device)
weight = torch.randn(192, requires_grad=True).to(device)
fn(input_data, weight)

The original FX graph returned in [get_kernel_metadata](https://github.com/pytorch/pytorch/blob/main/torch/_inductor/utils.py#L723](https://l.facebook.com/l.php?u=https%3A%2F%2Fgithub.com%2Fpytorch%2Fpytorch%2Fblob%2Fmain%2Ftorch%2F_inductor%2Futils.py%23L723&h=AT2PYr83thTo6VUjPs26Y8QAN6Sid16rvDMHtxO-Bp9FDwHr4J5PObtH3IhNTL-LPSRVC9WVJAcmwUToVWJIrDWb84i0j61QE55ySYAkGbuigqcNc7xczlirHhbiC9vMqiz91VwWdl4Pe2yKN7VIjjCiFUqw) is the following:
%primals_2 : Tensor "f32[4272, 192][192, 1]cuda:0" = PlaceHolder[target=primals_2]
%primals_1 : Tensor "f32[192][1]cuda:0" = PlaceHolder[target=primals_1]
%mul : Tensor "f32[4272, 192][192, 1]cuda:0"[num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_2, %unsqueeze), kwargs = {})
return %mul
The unsqueeze op is missing.

With this DIFF, the new FX graph is the following:
%primals_2 : Tensor "f32[4272, 192][192, 1]cuda:0" = PlaceHolder[target=primals_2]
%primals_1 : Tensor "f32[192][1]cuda:0" = PlaceHolder[target=primals_1]
%unsqueeze : Tensor "f32[1, 192][192, 1]cuda:0"[num_users=1] = call_function[target=torch.ops.aten.unsqueeze.default](args = (%primals_1, 0), kwargs = {})
%mul : Tensor "f32[4272, 192][192, 1]cuda:0"[num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_2, %unsqueeze), kwargs = {})
return %mul

cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @chenyang78 @kadeng @muchulee8 @amjames @chauhang @aakhundov @coconutruben

pytorch-bot · 2025-08-06T00:44:28Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/159920

📄 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 or our office hours

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

✅ No Failures

As of commit 503143e with merge base 231c722 ():
💚 Looks good so far! There are no failures yet. 💚

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

facebook-github-bot · 2025-08-06T00:44:34Z

This pull request was exported from Phabricator. Differential Revision: D79692316

facebook-github-bot · 2025-08-07T18:21:07Z

This pull request was exported from Phabricator. Differential Revision: D79692316

torch/_inductor/ir.py

Summary: Do not treat ReinterpretView as a realized node Function gather_origins (https://github.com/pytorch/pytorch/blob/main/torch/_inductor/utils.py#L888) calls is_realized_node to decide if a FX node should be included in the origins of a IR node. ReinterpretView is considered a realized node, so it is not included in the origins. It leads to an incomplete graph. For example: `torchdynamo.optimize("inductor")` `def fn(input_data, weight):` `. normalized_input = input_data * weight.unsqueeze(0)` `. return normalized_input` `input_data = torch.randn(4272, 192, requires_grad=True).to(device)` `weight = torch.randn(192, requires_grad=True).to(device)` `fn(input_data, weight)` The original FX graph returned in get_kernel_metadata (https://github.com/pytorch/pytorch/blob/main/torch/_inductor/utils.py#L723) is the following: %primals_2 : Tensor "f32[4272, 192][192, 1]cuda:0" = PlaceHolder[target=primals_2] %primals_1 : Tensor "f32[192][1]cuda:0" = PlaceHolder[target=primals_1] %mul : Tensor "f32[4272, 192][192, 1]cuda:0"[num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_2, %unsqueeze), kwargs = {}) return %mul The unsueeze op is missing. With this DIFF, the new FX graph is the following: %primals_2 : Tensor "f32[4272, 192][192, 1]cuda:0" = PlaceHolder[target=primals_2] %primals_1 : Tensor "f32[192][1]cuda:0" = PlaceHolder[target=primals_1] %unsqueeze : Tensor "f32[1, 192][192, 1]cuda:0"[num_users=1] = call_function[target=torch.ops.aten.unsqueeze.default](args = (%primals_1, 0), kwargs = {}) %mul : Tensor "f32[4272, 192][192, 1]cuda:0"[num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_2, %unsqueeze), kwargs = {}) return %mul Test Plan: buck2 run mode/opt caffe2/test:test_profiler_cuda -- profiler.test_execution_trace.TestExecutionTraceCUDA.test_triton_fx_graph_with_et_cuda Rollback Plan: Differential Revision: D79692316

facebook-github-bot · 2025-08-07T22:48:11Z

This pull request was exported from Phabricator. Differential Revision: D79692316

shengfukevin · 2025-08-08T16:02:24Z

@pytorchbot merge

pytorchmergebot · 2025-08-08T16:04:15Z

Merge failed

Reason: This PR needs a release notes: label
If your changes are user facing and intended to be a part of release notes, please use a label starting with release notes:.

If not, please add the topic: not user facing label.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "topic: not user facing"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.

Details for Dev Infra team

Raised by workflow job

shengfukevin · 2025-08-08T16:10:56Z

@pytorchbot label "topic: not user facing"

shengfukevin · 2025-08-08T16:11:17Z

@pytorchbot merge

pytorchmergebot · 2025-08-08T16:13:56Z

Merge started

Your change will be merged once all checks pass (ETA 0-4 Hours).

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 · 2025-08-08T16:14:17Z

Merge failed

Reason: 11 jobs have failed, first few of them are: linux-aarch64 / linux-jammy-aarch64-py3.10 / test (default, 3, 3, lf.linux.arm64.m7g.4xlarge), linux-aarch64 / linux-jammy-aarch64-py3.10 / test (default, 1, 3, lf.linux.arm64.m8g.4xlarge), linux-aarch64 / linux-jammy-aarch64-py3.10 / test (default, 2, 3, lf.linux.arm64.m7g.4xlarge), linux-aarch64 / linux-jammy-aarch64-py3.10 / test (default, 3, 3, lf.linux.arm64.m8g.4xlarge), linux-aarch64 / linux-jammy-aarch64-py3.10 / test (default, 2, 3, lf.linux.arm64.m8g.4xlarge)

Details for Dev Infra team

Raised by workflow job

Summary: Do not treat ReinterpretView as a realized node Function gather_origins (https://github.com/pytorch/pytorch/blob/main/torch/_inductor/utils.py#L888) calls is_realized_node to decide if a FX node should be included in the origins of a IR node. ReinterpretView is considered a realized node, so it is not included in the origins. It leads to an incomplete graph. For example: `torchdynamo.optimize("inductor")` `def fn(input_data, weight):` `. normalized_input = input_data * weight.unsqueeze(0)` `. return normalized_input` `input_data = torch.randn(4272, 192, requires_grad=True).to(device)` `weight = torch.randn(192, requires_grad=True).to(device)` `fn(input_data, weight)` The original FX graph returned in get_kernel_metadata (https://github.com/pytorch/pytorch/blob/main/torch/_inductor/utils.py#L723) is the following: %primals_2 : Tensor "f32[4272, 192][192, 1]cuda:0" = PlaceHolder[target=primals_2] %primals_1 : Tensor "f32[192][1]cuda:0" = PlaceHolder[target=primals_1] %mul : Tensor "f32[4272, 192][192, 1]cuda:0"[num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_2, %unsqueeze), kwargs = {}) return %mul The unsueeze op is missing. With this DIFF, the new FX graph is the following: %primals_2 : Tensor "f32[4272, 192][192, 1]cuda:0" = PlaceHolder[target=primals_2] %primals_1 : Tensor "f32[192][1]cuda:0" = PlaceHolder[target=primals_1] %unsqueeze : Tensor "f32[1, 192][192, 1]cuda:0"[num_users=1] = call_function[target=torch.ops.aten.unsqueeze.default](args = (%primals_1, 0), kwargs = {}) %mul : Tensor "f32[4272, 192][192, 1]cuda:0"[num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_2, %unsqueeze), kwargs = {}) return %mul Test Plan: buck2 run mode/opt caffe2/test:test_profiler_cuda -- profiler.test_execution_trace.TestExecutionTraceCUDA.test_triton_fx_graph_with_et_cuda Rollback Plan: Differential Revision: D79692316

facebook-github-bot · 2025-08-08T16:18:22Z

This pull request was exported from Phabricator. Differential Revision: D79692316

shengfukevin · 2025-08-08T16:30:06Z

@pytorchbot merge

pytorchmergebot · 2025-08-08T16:31:50Z

Merge started

Your change will be merged once all checks pass (ETA 0-4 Hours).

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

Summary: Do not treat ReinterpretView as a realized node Function [gather_origins](https://github.com/pytorch/pytorch/blob/main/torch/_inductor/utils.py#L888](https://l.facebook.com/l.php?u=https%3A%2F%2Fgithub.com%2Fpytorch%2Fpytorch%2Fblob%2Fmain%2Ftorch%2F_inductor%2Futils.py%23L888&h=AT2PYr83thTo6VUjPs26Y8QAN6Sid16rvDMHtxO-Bp9FDwHr4J5PObtH3IhNTL-LPSRVC9WVJAcmwUToVWJIrDWb84i0j61QE55ySYAkGbuigqcNc7xczlirHhbiC9vMqiz91VwWdl4Pe2yKN7VIjjCiFUqw) calls is_realized_node to decide if a FX node should be included in the origins of a IR node. ReinterpretView is considered a realized node, so it is not included in the origins. It leads to an incomplete graph. For example: ``` @torchdynamo.optimize("inductor") def fn(input_data, weight): normalized_input = input_data * weight.unsqueeze(0) return normalized_input input_data = torch.randn(4272, 192, requires_grad=True).to(device) weight = torch.randn(192, requires_grad=True).to(device) fn(input_data, weight) ``` The original FX graph returned in [get_kernel_metadata](https://github.com/pytorch/pytorch/blob/main/torch/_inductor/utils.py#L723](https://l.facebook.com/l.php?u=https%3A%2F%2Fgithub.com%2Fpytorch%2Fpytorch%2Fblob%2Fmain%2Ftorch%2F_inductor%2Futils.py%23L723&h=AT2PYr83thTo6VUjPs26Y8QAN6Sid16rvDMHtxO-Bp9FDwHr4J5PObtH3IhNTL-LPSRVC9WVJAcmwUToVWJIrDWb84i0j61QE55ySYAkGbuigqcNc7xczlirHhbiC9vMqiz91VwWdl4Pe2yKN7VIjjCiFUqw) is the following: %primals_2 : Tensor "f32[4272, 192][192, 1]cuda:0" = PlaceHolder[target=primals_2] %primals_1 : Tensor "f32[192][1]cuda:0" = PlaceHolder[target=primals_1] %mul : Tensor "f32[4272, 192][192, 1]cuda:0"[num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_2, %unsqueeze), kwargs = {}) return %mul The unsqueeze op is missing. With this DIFF, the new FX graph is the following: %primals_2 : Tensor "f32[4272, 192][192, 1]cuda:0" = PlaceHolder[target=primals_2] %primals_1 : Tensor "f32[192][1]cuda:0" = PlaceHolder[target=primals_1] %unsqueeze : Tensor "f32[1, 192][192, 1]cuda:0"[num_users=1] = call_function[target=torch.ops.aten.unsqueeze.default](args = (%primals_1, 0), kwargs = {}) %mul : Tensor "f32[4272, 192][192, 1]cuda:0"[num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_2, %unsqueeze), kwargs = {}) return %mul Pull Request resolved: pytorch#159920 Approved by: https://github.com/mlazos

Summary: Fix unit test test_equivalent_template_code #159920 treats ReinterpretView as a not-realized node when searching FX origin nodes for fused triton kernel. In test_equivalent_template_code, there is a transpose node (which is a ReinterpretView) before matmul. It was not in FX graph segment before PR 159920. FX origin nodes are used to define the name of triton kernel. That is the reason test_equivalent_template_code failed with PR 159920 since it uses hard-coded triton kernel name to check the result. The fix is to update the triton kernel name in the unit test. Test Plan: buck2 run mode/opt caffe2/test/inductor:benchmark_fusion -- caffe2.test.inductor.test_benchmark_fusion.BenchmarkMultiTemplateFusionCudaTest Rollback Plan: Differential Revision: D80101711

Summary: Fix unit test test_equivalent_template_code #159920 treats ReinterpretView as a not-realized node when searching FX origin nodes for fused triton kernel. In test_equivalent_template_code, there is a transpose node (which is a ReinterpretView) before matmul. It was not in FX graph segment before PR 159920. FX origin nodes are used to define the name of triton kernel. That is the reason test_equivalent_template_code failed with PR 159920 since it uses hard-coded triton kernel name to check the result. The fix is to update the triton kernel name in the unit test. Test Plan: buck2 run mode/opt caffe2/test/inductor:benchmark_fusion -- caffe2.test.inductor.test_benchmark_fusion.BenchmarkMultiTemplateFusionCudaTest Rollback Plan: Reviewed By: clee2000 Differential Revision: D80101711

Summary: Pull Request resolved: pytorch#160432 Fix unit test test_equivalent_template_code pytorch#159920 treats ReinterpretView as a not-realized node when searching FX origin nodes for fused triton kernel. In test_equivalent_template_code, there is a transpose node (which is a ReinterpretView) before matmul. It was not in FX graph segment before PR 159920. FX origin nodes are used to define the name of triton kernel. That is the reason test_equivalent_template_code failed with PR 159920 since it uses hard-coded triton kernel name to check the result. The fix is to update the triton kernel name in the unit test. Test Plan: buck2 run mode/opt caffe2/test/inductor:benchmark_fusion -- caffe2.test.inductor.test_benchmark_fusion.BenchmarkMultiTemplateFusionCudaTest Rollback Plan: Reviewed By: clee2000 Differential Revision: D80101711

Summary: Fix unit test test_equivalent_template_code pytorch#159920 treats ReinterpretView as a not-realized node when searching FX origin nodes for fused triton kernel. In test_equivalent_template_code, there is a transpose node (which is a ReinterpretView) before matmul. It was not in FX graph segment before PR 159920. FX origin nodes are used to define the name of triton kernel. That is the reason test_equivalent_template_code failed with PR 159920 since it uses hard-coded triton kernel name to check the result. The fix is to update the triton kernel name in the unit test. Test Plan: buck2 run mode/opt caffe2/test/inductor:benchmark_fusion -- caffe2.test.inductor.test_benchmark_fusion.BenchmarkMultiTemplateFusionCudaTest Rollback Plan: Reviewed By: clee2000 Differential Revision: D80101711

Summary: Fix unit test test_equivalent_template_code #159920 treats ReinterpretView as a not-realized node when searching FX origin nodes for fused triton kernel. In test_equivalent_template_code, there is a transpose node (which is a ReinterpretView) before matmul. It was not in FX graph segment before PR 159920. FX origin nodes are used to define the name of triton kernel. That is the reason test_equivalent_template_code failed with PR 159920 since it uses hard-coded triton kernel name to check the result. The fix is to update the triton kernel name in the unit test. Test Plan: buck2 run mode/opt caffe2/test/inductor:benchmark_fusion -- caffe2.test.inductor.test_benchmark_fusion.BenchmarkMultiTemplateFusionCudaTest Rollback Plan: Differential Revision: D80101711 Pull Request resolved: #160432 Approved by: https://github.com/clee2000

Summary: Fix unit test test_equivalent_template_code pytorch#159920 treats ReinterpretView as a not-realized node when searching FX origin nodes for fused triton kernel. In test_equivalent_template_code, there is a transpose node (which is a ReinterpretView) before matmul. It was not in FX graph segment before PR 159920. FX origin nodes are used to define the name of triton kernel. That is the reason test_equivalent_template_code failed with PR 159920 since it uses hard-coded triton kernel name to check the result. The fix is to update the triton kernel name in the unit test. Test Plan: buck2 run mode/opt caffe2/test/inductor:benchmark_fusion -- caffe2.test.inductor.test_benchmark_fusion.BenchmarkMultiTemplateFusionCudaTest Rollback Plan: Differential Revision: D80101711 Pull Request resolved: pytorch#160432 Approved by: https://github.com/clee2000

Summary: Do not treat ReinterpretView as a realized node Function [gather_origins](https://github.com/pytorch/pytorch/blob/main/torch/_inductor/utils.py#L888](https://l.facebook.com/l.php?u=https%3A%2F%2Fgithub.com%2Fpytorch%2Fpytorch%2Fblob%2Fmain%2Ftorch%2F_inductor%2Futils.py%23L888&h=AT2PYr83thTo6VUjPs26Y8QAN6Sid16rvDMHtxO-Bp9FDwHr4J5PObtH3IhNTL-LPSRVC9WVJAcmwUToVWJIrDWb84i0j61QE55ySYAkGbuigqcNc7xczlirHhbiC9vMqiz91VwWdl4Pe2yKN7VIjjCiFUqw) calls is_realized_node to decide if a FX node should be included in the origins of a IR node. ReinterpretView is considered a realized node, so it is not included in the origins. It leads to an incomplete graph. For example: ``` @torchdynamo.optimize("inductor") def fn(input_data, weight): normalized_input = input_data * weight.unsqueeze(0) return normalized_input input_data = torch.randn(4272, 192, requires_grad=True).to(device) weight = torch.randn(192, requires_grad=True).to(device) fn(input_data, weight) ``` The original FX graph returned in [get_kernel_metadata](https://github.com/pytorch/pytorch/blob/main/torch/_inductor/utils.py#L723](https://l.facebook.com/l.php?u=https%3A%2F%2Fgithub.com%2Fpytorch%2Fpytorch%2Fblob%2Fmain%2Ftorch%2F_inductor%2Futils.py%23L723&h=AT2PYr83thTo6VUjPs26Y8QAN6Sid16rvDMHtxO-Bp9FDwHr4J5PObtH3IhNTL-LPSRVC9WVJAcmwUToVWJIrDWb84i0j61QE55ySYAkGbuigqcNc7xczlirHhbiC9vMqiz91VwWdl4Pe2yKN7VIjjCiFUqw) is the following: %primals_2 : Tensor "f32[4272, 192][192, 1]cuda:0" = PlaceHolder[target=primals_2] %primals_1 : Tensor "f32[192][1]cuda:0" = PlaceHolder[target=primals_1] %mul : Tensor "f32[4272, 192][192, 1]cuda:0"[num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_2, %unsqueeze), kwargs = {}) return %mul The unsqueeze op is missing. With this DIFF, the new FX graph is the following: %primals_2 : Tensor "f32[4272, 192][192, 1]cuda:0" = PlaceHolder[target=primals_2] %primals_1 : Tensor "f32[192][1]cuda:0" = PlaceHolder[target=primals_1] %unsqueeze : Tensor "f32[1, 192][192, 1]cuda:0"[num_users=1] = call_function[target=torch.ops.aten.unsqueeze.default](args = (%primals_1, 0), kwargs = {}) %mul : Tensor "f32[4272, 192][192, 1]cuda:0"[num_users=1] = call_function[target=torch.ops.aten.mul.Tensor](args = (%primals_2, %unsqueeze), kwargs = {}) return %mul Pull Request resolved: pytorch#159920 Approved by: https://github.com/mlazos

Summary: Fix unit test test_equivalent_template_code pytorch#159920 treats ReinterpretView as a not-realized node when searching FX origin nodes for fused triton kernel. In test_equivalent_template_code, there is a transpose node (which is a ReinterpretView) before matmul. It was not in FX graph segment before PR 159920. FX origin nodes are used to define the name of triton kernel. That is the reason test_equivalent_template_code failed with PR 159920 since it uses hard-coded triton kernel name to check the result. The fix is to update the triton kernel name in the unit test. Test Plan: buck2 run mode/opt caffe2/test/inductor:benchmark_fusion -- caffe2.test.inductor.test_benchmark_fusion.BenchmarkMultiTemplateFusionCudaTest Rollback Plan: Differential Revision: D80101711 Pull Request resolved: pytorch#160432 Approved by: https://github.com/clee2000

pytorch-bot bot added ciflow/inductor module: inductor labels Aug 6, 2025

facebook-github-bot added the fb-exported label Aug 6, 2025

shengfukevin force-pushed the export-D79692316 branch from 5f4e312 to ba57263 Compare August 7, 2025 18:21

shengfukevin self-assigned this Aug 7, 2025

shengfukevin requested a review from mlazos August 7, 2025 18:27

mlazos reviewed Aug 7, 2025

View reviewed changes

torch/_inductor/ir.py Show resolved Hide resolved

shengfukevin force-pushed the export-D79692316 branch from ba57263 to 6cba133 Compare August 7, 2025 22:47

mlazos approved these changes Aug 8, 2025

View reviewed changes

pytorch-bot bot added the ciflow/trunk Trigger trunk jobs on your pull request label Aug 8, 2025

pytorchmergebot added the merging label Aug 8, 2025

pytorchmergebot removed the merging label Aug 8, 2025

pytorch-bot bot added the topic: not user facing topic category label Aug 8, 2025

pytorchmergebot added the merging label Aug 8, 2025

pytorchmergebot removed the merging label Aug 8, 2025

shengfukevin force-pushed the export-D79692316 branch from 6cba133 to 503143e Compare August 8, 2025 16:18

pytorchmergebot added the merging label Aug 8, 2025

pytorchmergebot added the Merged label Aug 8, 2025

pytorchmergebot closed this in 1febab2 Aug 8, 2025

pytorchmergebot removed the merging label Aug 8, 2025

clee2000 mentioned this pull request Aug 12, 2025

[ez][CI] Disable failing test_equivalent_template_code on slow grad check #160424

Closed

shengfukevin mentioned this pull request Aug 12, 2025

Fix unit test test_equivalent_template_code #160432

Closed

Valentine233 mentioned this pull request Aug 20, 2025

Fix UT assertion error for int8 sdpa fusion pytorch/ao#2816

Merged

Conversation

shengfukevin commented Aug 6, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-bot bot commented Aug 6, 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/159920

✅ No Failures

Uh oh!

facebook-github-bot commented Aug 6, 2025

Uh oh!

facebook-github-bot commented Aug 7, 2025

Uh oh!

Uh oh!

facebook-github-bot commented Aug 7, 2025

Uh oh!

shengfukevin commented Aug 8, 2025

Uh oh!

pytorchmergebot commented Aug 8, 2025

Merge failed

Uh oh!

shengfukevin commented Aug 8, 2025

Uh oh!

shengfukevin commented Aug 8, 2025

Uh oh!

pytorchmergebot commented Aug 8, 2025

Merge started

Uh oh!

pytorchmergebot commented Aug 8, 2025

Merge failed

Uh oh!

facebook-github-bot commented Aug 8, 2025

Uh oh!

shengfukevin commented Aug 8, 2025

Uh oh!

pytorchmergebot commented Aug 8, 2025

Merge started

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

shengfukevin commented Aug 6, 2025 •

edited

Loading

pytorch-bot bot commented Aug 6, 2025 •

edited

Loading