Matthias Gehre dc2e8f1157 [ROCm][Perf] W4A16: keep skinny GEMM zero-points packed 4-bit (#54965) ...

Signed-off-by: Matthias Gehre <matthias.gehre@amd.com>

2026-09-14 17:40:25 +08:00

..

determinism

[Docs] Add missing return annotations flagged by griffe (#54980)

2026-09-02 09:56:07 -07:00

hw_agnostic

Hardware-agnostic model definition via HF transformer backend (1/N) (#49458)

2026-08-13 11:21:50 +01:00

kernels

[ROCm][Perf] W4A16: keep skinny GEMM zero-points packed 4-bit (#54965)

2026-09-14 17:40:25 +08:00

layers

[Perf][GLM-5.3-Flash] Dense/masked-MHA sparse prefill for the NoPE (256, 0, 256) layout + skip the NoPE K concat (#55738)

2026-09-14 09:36:39 +00:00

model_loader

[Fast Start] Use GPU uuid as socket folder identifier (#56669)

2026-09-14 06:28:13 +00:00

models

[Refactor] Normalize DeepSeek V4.1 model package naming (#56741)

2026-09-13 23:17:50 -07:00

offloader

[Core] Enhance cpu<->gpu sync checking to include paged async copies (#53491)

2026-09-08 17:26:21 -07:00

warmup

[Perf] Parallelize mHC pre-norm JIT warmup (#56683)

2026-09-13 18:58:20 -07:00

__init__.py

[Platform] Deprecate seed_everything (#31659)

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

custom_op.py

[Bugfix] Reject contradictory custom-op directives (#49134)

2026-07-26 08:42:14 -04:00

parameter.py

Speed up docs build (#44635)

2026-06-05 14:51:44 +00:00

utils.py

Fast Start (#54921)

2026-09-04 06:32:06 -07:00