Summary:
Multiple tests in TestSelectAlgorithmCPU are failing during PyTorch unit test execution on CPU platform with tensor comparison errors.

Test Class: inductor/test_cpu_select_algorithm.py::TestSelectAlgorithmCPU
Number of Failing Tests: 44
Platform: CPU
Test Type: Unit Test

Version Information:

• PyTorch Commit: 6bdd8c9
• Branch: main
• Test Date: 2026-01-14
• Sprint: Sprint 24

Failure Pattern:
Tests failing with related error patterns - tensor comparison failures in int4 quantization and weight-only quantization tests

Common Error:

AssertionError: Tensor-likes are not close!

Mismatched elements: 1249 / 16384 (7.6%)
Greatest absolute difference: 0.3125 at index (0, 1, 223) (up to 0.01 allowed)
Greatest relative difference: 30.125 at index (2, 3, 142) (up to 0.01 allowed)

Failing Tests (sample - 10 of 44):
1. test_int4_concat_woq_mm_batch_size_4_in_features_256_out_features0_group_size_32_cpu_bfloat16
2. test_int4_woq_mm_amx_batch_size_1_in_features_1024_out_features_128_group_size_32_cpu_bfloat16
3. test_int4_woq_mm_amx_batch_size_1_in_features_128_out_features_1024_group_size_64_cpu_bfloat16
4. test_int4_concat_woq_mm_batch_size_4_in_features_256_out_features0_group_size_128_cpu_bfloat16
5. test_int4_woq_mm_amx_batch_size_1_in_features_1024_out_features_1024_group_size_32_cpu_bfloat16
6. test_int4_woq_mm_amx_batch_size_1_in_features_128_out_features_1024_group_size_32_cpu_bfloat16
7. test_int4_woq_mm_amx_batch_size_4_in_features_1024_out_features_128_group_size_64_cpu_bfloat16
8. test_int4_woq_mm_amx_batch_size_4_in_features_128_out_features_128_group_size_64_cpu_bfloat16
9. test_int4_woq_mm_amx_batch_size_6_in_features_1024_out_features_1024_group_size_32_cpu_bfloat16
10. test_int4_woq_mm_amx_batch_size_6_in_features_128_out_features_1024_group_size_128_cpu_bfloat16
... and 34 more tests

Steps to Reproduce:
1. Run test command:

   TEST_CONFIG=cpu python3 test/run_test.py -i inductor/test_cpu_select_algorithm
   TEST_CONFIG=cuda python3 test/run_test.py -i inductor/test_cpu_select_algorithm
   TEST_CONFIG=inductor python3 test/run_test.py -i inductor/test_cpu_select_algorithm
   

2. Observe tensor comparison failures in int4 quantization tests

Expected Result:
All tests should pass

Actual Result:
44 tests fail with tensor comparison errors in int4 weight-only quantization operations

Root Cause Analysis:
Multiple int4 weight-only quantization (woq) tests are failing with numerical accuracy issues. The failures primarily affect:

• int4_concat_woq_mm tests
• int4_woq_mm_amx tests with various batch sizes and feature dimensions
• Tests with different group sizes (32, 64, 128)
• Both bfloat16 and float32 dtypes

Additional Context:

• Note: sGPU ticket AIPCC-8251 exists for the same test class
• This is the CPU-specific failure affecting 44 tests
• All failures are related to int4 quantization accuracy
• High priority due to the number of failing tests (44)

Logs:
Test execution logs: /home/ktanmay/Downloads/Run 1-20260120T060019Z-1-001/Run 1/20260114_024940_commit_6bdd8c9/cpu_tests.log

Priority: P2

Labels: pytorch, unittest, cpu, inductor, quantization, int4