pytorch

1885 issues found

DISABLED test_cupti_monitor_record_shapes (main.TestProfilerCUDA)

6/9/2026

Flex Attention produces NaN with `torch.compile(flex_attention, dynamic=True)` and batch-dependent mask closures

6/9/2026

DISABLED test_max_autotune_cutlass_backend_cpp_wrapper_bfloat16_num_gemms_1 (main.TestCutlassBackend)

6/9/2026

DISABLED test_combo_kernel_no_bench_carve_out (main.ComboKernelTests)

6/9/2026

weakrefs don't when tensor swap is used

6/9/2026

`torch.compile` silently produces wrong gradient for `atan2(x, y).amin(dim)` on float32

6/9/2026

CuTe/Quack `_fused_rms_norm` override hangs while compiling very large normalized shape

6/9/2026

[v2.12.1] Release Tracker

6/9/2026

`@torch.compiler.nested_compile_region` failing on a custom op with mutations

6/9/2026

DISABLED test_cupti_monitor_trace_has_expected_events (main.TestProfilerCUDA)

6/9/2026

torch.distributions.Cauchy.icdf(0) returns large finite value with wrong sign in float32; Gumbel.icdf(0/1) also finite instead of ±inf

6/9/2026

Bernoulli(probs=0).log_prob(1) returns dtype-dependent finite values instead of -inf (Geometric special-cases this and tests it)

6/9/2026

Distribution.icdf performs no argument validation even with validate_args=True, unlike cdf/log_prob

6/9/2026

[RFC]: Auto-select and run Operator Performance Benchmarks from PR Diffs

6/9/2026

Inductor: `unsqueeze` lowering loses source-tensor strides when source has non-standard layout — produces fused Triton kernel that reads OOB

6/9/2026

[compile] Handle shallow_copy_data_ device mutation in ProxyTensor instead of save/restore workaround

6/10/2026

DISABLED test_triggers (main.CallbackTests)

6/10/2026

DISABLED test_dynamic_shape_backend_eager_device_cpu (main.TestPackage)

6/10/2026

DISABLED test_group_norm_backward_large_input_bfloat16_trigger_32bit_overflow_True_mps_bfloat16 (main.TestConsistencyMPS)

6/10/2026

DISABLED test_group_norm_backward_large_input_float16_trigger_32bit_overflow_True_mps_float16 (main.TestConsistencyMPS)

6/10/2026

Data

Security

Network

Code

UI/UX

Text

System

Multimedia

Protocol

API

Engineering

pytorch

DISABLED test_cupti_monitor_record_shapes (main.TestProfilerCUDA)

Flex Attention produces NaN with `torch.compile(flex_attention, dynamic=True)` and batch-dependent mask closures

DISABLED test_max_autotune_cutlass_backend_cpp_wrapper_bfloat16_num_gemms_1 (main.TestCutlassBackend)

DISABLED test_combo_kernel_no_bench_carve_out (main.ComboKernelTests)

weakrefs don't when tensor swap is used

`torch.compile` silently produces wrong gradient for `atan2(x, y).amin(dim)` on float32

CuTe/Quack `_fused_rms_norm` override hangs while compiling very large normalized shape

[v2.12.1] Release Tracker

`@torch.compiler.nested_compile_region` failing on a custom op with mutations

DISABLED test_cupti_monitor_trace_has_expected_events (main.TestProfilerCUDA)

torch.distributions.Cauchy.icdf(0) returns large finite value with wrong sign in float32; Gumbel.icdf(0/1) also finite instead of ±inf

Bernoulli(probs=0).log_prob(1) returns dtype-dependent finite values instead of -inf (Geometric special-cases this and tests it)

Distribution.icdf performs no argument validation even with validate_args=True, unlike cdf/log_prob

[RFC]: Auto-select and run Operator Performance Benchmarks from PR Diffs

Inductor: `unsqueeze` lowering loses source-tensor strides when source has non-standard layout — produces fused Triton kernel that reads OOB

[compile] Handle shallow_copy_data_ device mutation in ProxyTensor instead of save/restore workaround

DISABLED test_triggers (main.CallbackTests)

DISABLED test_dynamic_shape_backend_eager_device_cpu (main.TestPackage)

DISABLED test_group_norm_backward_large_input_bfloat16_trigger_32bit_overflow_True_mps_bfloat16 (main.TestConsistencyMPS)

DISABLED test_group_norm_backward_large_input_float16_trigger_32bit_overflow_True_mps_float16 (main.TestConsistencyMPS)

pytorch

DISABLED test_cupti_monitor_record_shapes (__main__.TestProfilerCUDA)

Flex Attention produces NaN with `torch.compile(flex_attention, dynamic=True)` and batch-dependent mask closures

DISABLED test_max_autotune_cutlass_backend_cpp_wrapper_bfloat16_num_gemms_1 (__main__.TestCutlassBackend)

DISABLED test_combo_kernel_no_bench_carve_out (__main__.ComboKernelTests)

weakrefs don't when tensor swap is used

`torch.compile` silently produces wrong gradient for `atan2(x, y).amin(dim)` on float32

CuTe/Quack `_fused_rms_norm` override hangs while compiling very large normalized shape

[v2.12.1] Release Tracker

`@torch.compiler.nested_compile_region` failing on a custom op with mutations

DISABLED test_cupti_monitor_trace_has_expected_events (__main__.TestProfilerCUDA)

torch.distributions.Cauchy.icdf(0) returns large finite value with wrong sign in float32; Gumbel.icdf(0/1) also finite instead of ±inf

Bernoulli(probs=0).log_prob(1) returns dtype-dependent finite values instead of -inf (Geometric special-cases this and tests it)

Distribution.icdf performs no argument validation even with validate_args=True, unlike cdf/log_prob

[RFC]: Auto-select and run Operator Performance Benchmarks from PR Diffs

Inductor: `unsqueeze` lowering loses source-tensor strides when source has non-standard layout — produces fused Triton kernel that reads OOB

[compile] Handle shallow_copy_data_ device mutation in ProxyTensor instead of save/restore workaround

DISABLED test_triggers (__main__.CallbackTests)

DISABLED test_dynamic_shape_backend_eager_device_cpu (__main__.TestPackage)

DISABLED test_group_norm_backward_large_input_bfloat16_trigger_32bit_overflow_True_mps_bfloat16 (__main__.TestConsistencyMPS)

DISABLED test_group_norm_backward_large_input_float16_trigger_32bit_overflow_True_mps_float16 (__main__.TestConsistencyMPS)

DISABLED test_cupti_monitor_record_shapes (main.TestProfilerCUDA)

DISABLED test_max_autotune_cutlass_backend_cpp_wrapper_bfloat16_num_gemms_1 (main.TestCutlassBackend)

DISABLED test_combo_kernel_no_bench_carve_out (main.ComboKernelTests)

DISABLED test_cupti_monitor_trace_has_expected_events (main.TestProfilerCUDA)

DISABLED test_triggers (main.CallbackTests)

DISABLED test_dynamic_shape_backend_eager_device_cpu (main.TestPackage)

DISABLED test_group_norm_backward_large_input_bfloat16_trigger_32bit_overflow_True_mps_bfloat16 (main.TestConsistencyMPS)

DISABLED test_group_norm_backward_large_input_float16_trigger_32bit_overflow_True_mps_float16 (main.TestConsistencyMPS)