3 people e83222396f [ROCm][Perf] W4A16: speed up skinny GEMM decode, including speculative batches (#57199) ...

Signed-off-by: Robert Esclapez Garcia <robert.garcia@amd.com>
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Co-authored-by: Douglas Lehr <91553416+dllehr-amd@users.noreply.github.com>

2026-10-13 17:21:30 +00:00

..

nvfp4_utils.py

Add pydocstyle to the ruff rules (#52136)

2026-09-16 16:44:29 +00:00

test_aiter_hipb_mm_linear_kernel.py

[CI] Ignore ruff D209, rejoin the docstrings it split, and silence incompatible-rule warnings (#57212)

2026-09-16 10:57:24 -07:00

test_awq_triton.py

[Kernel][Test] Extend lightning_attn and awq_triton kernel tests to XPU (#43307)

2026-06-04 14:25:59 -04:00

test_awq.py

[Refactor] Remove dead quantization code and tests (#45454)

2026-06-17 16:12:01 -04:00

test_block_fp8.py

Disable mypy arg-type and assignment checks in tests (#59428)

2026-09-30 15:18:50 +00:00

test_cpu_fp8_scaled_mm.py

[CPU] Add CPU FP8 W8A8 linear/MoE support (#49942)

2026-09-18 17:59:42 +08:00

test_cutlass_scaled_mm.py

[MyPy][2/N] Fix mypy errors in small tests/ dirs (#51043)

2026-09-29 21:27:45 +01:00

test_cutlass_w4a8_moe.py

[Test][Quantization] Gate CUTLASS W4A8 kernel tests to SM90 (#58878)

2026-10-08 19:14:35 +00:00

test_cutlass_w4a8.py

[Test][Quantization] Gate CUTLASS W4A8 kernel tests to SM90 (#58878)

2026-10-08 19:14:35 +00:00

test_flashinfer_mxfp8_trtllm.py

[Bugfix][Quantization] Fix MXFP8 startup crash on layers below mm_mxfp8 shape limits (#54223)

2026-09-25 02:54:11 +00:00

test_flashinfer_nvfp4_scaled_mm.py

[Fusion] Support 3D activation quantization fusion and enable it for Pixtral (#59612)

2026-10-09 15:12:03 +08:00

test_flashinfer_scaled_mm.py

[Platform] Deprecate seed_everything (#31659)

2026-01-04 18:34:04 -08:00

test_fp8_min_max_helper.py

Add pydocstyle to the ruff rules (#52136)

2026-09-16 16:44:29 +00:00

test_fp8_quant_group.py

[Bugfix] Initialize DeepGemmQuantScaleFMT oracle lazily; bound QuantFP8 UE8M0 packed path to group_size 128 (#51359)

2026-08-12 14:27:50 -04:00

test_fp8_quant.py

[MyPy][2/N] Fix mypy errors in small tests/ dirs (#51043)

2026-09-29 21:27:45 +01:00

test_gptq.py

[Quant][Kernel] Remove GPTQ Group/Dynamic Activation Ordering (#54809)

2026-09-08 12:45:12 -07:00

test_hadacore.py

[CI/Build][AMD] Skip marlin, machete, and hadacore tests since these require _C functions not defined for ROCm (#30109)

2025-12-06 12:54:17 +08:00

test_int4_emulation_moe.py

[Quant] Use canonical N-first weight format for CT WNA16 MoE (#52798)

2026-09-28 11:22:44 -06:00

test_int8_kernel.py

[CI] Ignore ruff D209, rejoin the docstrings it split, and silence incompatible-rule warnings (#57212)

2026-09-16 10:57:24 -07:00

test_int8_quant.py

[Platform] Deprecate seed_everything (#31659)

2026-01-04 18:34:04 -08:00

test_machete_mm.py

[MyPy][2/N] Fix mypy errors in small tests/ dirs (#51043)

2026-09-29 21:27:45 +01:00

test_marlin_gemm.py

[MyPy][2/N] Fix mypy errors in small tests/ dirs (#51043)

2026-09-29 21:27:45 +01:00

test_marlin_tile_padding.py

Disable mypy arg-type and assignment checks in tests (#59428)

2026-09-30 15:18:50 +00:00

test_mxfp4_kernel_selection.py

[Kernel] Add opt-in load-time MXFP4 dequantization (#50814)

2026-09-22 14:51:12 +08:00

test_mxfp4_qutlass.py

[Platform] Deprecate seed_everything (#31659)

2026-01-04 18:34:04 -08:00

test_mxfp4_triton_ep.py

[CI] Ignore ruff D209, rejoin the docstrings it split, and silence incompatible-rule warnings (#57212)

2026-09-16 10:57:24 -07:00

test_mxfp6_kernel_selection.py

[CI] Ignore ruff D209, rejoin the docstrings it split, and silence incompatible-rule warnings (#57212)

2026-09-16 10:57:24 -07:00

test_mxfp8_kernel_selection.py

[Bugfix][Quantization] Fix MXFP8 startup crash on layers below mm_mxfp8 shape limits (#54223)

2026-09-25 02:54:11 +00:00

test_nvfp4_emulation.py

Disable mypy arg-type and assignment checks in tests (#59428)

2026-09-30 15:18:50 +00:00

test_nvfp4_kernel_selection.py

[Perf][Quant][NVFP4] Prefer W4A4 linear kernels over weight-only ones on SM120/121 (#55170)

2026-09-08 15:06:45 +04:00

test_nvfp4_quant.py

Add pydocstyle to the ruff rules (#52136)

2026-09-16 16:44:29 +00:00

test_nvfp4_qutlass.py

[Platform] Deprecate seed_everything (#31659)

2026-01-04 18:34:04 -08:00

test_nvfp4_scaled_mm.py

[Platform] Deprecate seed_everything (#31659)

2026-01-04 18:34:04 -08:00

test_per_token_group_quant.py

[Kernel][Perf] Register-resident path for per-token-group 8-bit quant (#55330)

2026-09-25 08:11:42 -07:00

test_quant_op_schema.py

[ROCm][CI] Move ROCm AITER quantization tests (#40938)

2026-08-18 14:50:35 +08:00

test_quantized_embedding.py

[CI] Ignore ruff D209, rejoin the docstrings it split, and silence incompatible-rule warnings (#57212)

2026-09-16 10:57:24 -07:00

test_rdna3_compile_guards.py

[Refactor][ROCm] Migrate the RDNA3 W4A16 MoE to the oracle/experts pa… (#55522)

2026-09-12 21:20:26 +08:00

test_rdna3_moe_w4a16.py

[Bugfix][ROCm] RDNA3 W4A16 MoE: handle the N-first CT weight layout (#60388)

2026-10-08 15:06:39 +00:00

test_rdna3_w4a16_determinism.py

[ROCm][RDNA3] Fix W4A16 split-K accuracy and determinism (#54706)

2026-10-03 23:44:17 +08:00

test_rdna3_w4a16_selection.py

[Quant][Kernel] Remove GPTQ Group/Dynamic Activation Ordering (#54809)

2026-09-08 12:45:12 -07:00

test_rdna3_w4a16.py

[Quant][Kernel] Remove GPTQ Group/Dynamic Activation Ordering (#54809)

2026-09-08 12:45:12 -07:00

test_rdna_hybrid_w4a16.py

[ROCm][Perf] W4A16: speed up skinny GEMM decode, including speculative batches (#57199)

2026-10-13 17:21:30 +00:00

test_rocm_aiter_grouped_quant.py

[CI] Ignore ruff D209, rejoin the docstrings it split, and silence incompatible-rule warnings (#57212)

2026-09-16 10:57:24 -07:00

test_rocm_compressed_tensors_w4a16.py

[CI] Retire Weight Loading smoke tests (#57398)

2026-09-17 16:04:03 -04:00

test_rocm_fp8.py

[ROCm][CI] Add More AITER quantization/MoE kernel tests (#49375)

2026-08-05 21:27:33 +00:00

test_rocm_mxfp4.py

[ROCm] Fix CI runtime and tests for MI355 DPX (#58244)

2026-09-24 23:15:11 -05:00

test_rocm_mxfp8_linear.py

[ROCm][CI] Pass weight_shape in MXFP8 block32 linear tests (#58698)

2026-09-25 10:52:15 +00:00

test_rocm_skinny_gemms.py

[ROCm] Fix CI runtime and tests for MI355 DPX (#58244)

2026-09-24 23:15:11 -05:00

test_scaled_mm_kernel_selection.py

[Quantization] Prefer Humming before Marlin backends on SM90 (#56997)

2026-10-06 12:31:35 -07:00

test_silu_mul_nvfp4_quant.py

[Bugfix] Fix NVFP4 fused SiLU+mul scale allocation and global scale direction (#55643)

2026-09-08 08:26:07 -07:00

test_triton_scaled_mm.py

[Feature] Support batch invariance on ROCm (#52231)

2026-10-12 10:53:26 -05:00

test_triton_w4a16.py

[Quant][Kernel] Remove GPTQ Group/Dynamic Activation Ordering (#54809)

2026-09-08 12:45:12 -07:00

test_w4a16_kernel_selection.py

[Quant][Kernel] Remove GPTQ Group/Dynamic Activation Ordering (#54809)

2026-09-08 12:45:12 -07:00

test_zen_da8w4.py

[CPU][Zen] Add DA8W4 (W4A8) int4 support for dense and MoE layers (#54024)

2026-10-01 15:18:25 +08:00