3 people c5bda658d2 [Model][MoE] Load MegaMoE experts directly into the DeepGEMM layout (#60096) ...

Signed-off-by: aoshen02 <aoshen@inferact.ai>
Signed-off-by: Aaron Hao <hao-aaron@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Aaron Hao <hao-aaron@users.noreply.github.com>

2026-10-13 11:52:08 -07:00

..

deepseek_v32

[Perf] Optimize DeepSeek V3.2 sequence parallelism (#51434)

2026-08-07 16:51:01 -07:00

fixtures

[BUGFIX] Fix Pixtral consolidated format vision weight loading (#39916)

2026-04-19 22:25:03 -07:00

glm5next

[Bugfix][GLM-5.3-Flash] Honor --mamba-ssm-cache-dtype for the KDA recurrent state (#59147)

2026-10-11 17:55:00 +00:00

inkling

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

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

kimi_k3

[Model][MoE] Load MegaMoE experts directly into the DeepGEMM layout (#60096)

2026-10-13 11:52:08 -07:00

language

[Bugfix] Treat max_tokens as output upper bound, not input-space reservation (#57035)

2026-10-10 01:15:22 +08:00

multimodal

[Misc] Break chat_utils into more maintainable modules. (#60972)

2026-10-13 15:08:07 +00:00

quantization

[ROCm][MoE] Support MiMo-V2.6 MXFP4 on gfx942 (#58262)

2026-09-30 17:23:00 +08:00

qwen4_exp

[Qwen4Exp] support Sequence Parallelism for Qwen3.8-Flash-Next (#56322)

2026-10-12 20:56:44 +08:00

transformers

[Transformers backend] Fuse MoE routers that score after the top-k (#57410)

2026-10-13 08:41:23 +00:00

__init__.py

[CI/Build] Move test_utils.py to tests/utils.py (#4425)

2024-05-13 23:50:09 +09:00

registry.py

[Model] Add Clef-Flash support (#61308)

2026-10-13 05:46:37 +00:00

test_adapters.py

[Chore] Use Transformers v5 names and drop redundant processor use_fast (#58550)

2026-09-24 14:14:10 +00:00

test_deepseek_v4_aiter_mega_moe.py

[ROCm][DSv4] Fuse the shared expert into AITER MegaMoE (#61004)

2026-10-12 15:53:16 +08:00

test_deepseek_v4_dspark_rocm.py

[CI] Fix failing test_deepseek_v4_rocm_dspark* tests (#60462)

2026-10-07 14:18:19 -07:00

test_deepseek_v4_fi_moe_ep.py

Add MoNe experts pruning support (#50415)

2026-10-12 23:18:43 +00:00

test_deepseek_v4_mega_moe.py

[Model][MoE] Load MegaMoE experts directly into the DeepGEMM layout (#60096)

2026-10-13 11:52:08 -07:00

test_deepseek_v4_rocm_compressor_gemm_fusion.py

[ROCm][DSV4][Perf] Fuse DeepSeek V4 C4 compressor GEMMs (#53838)

2026-08-26 23:36:14 +08:00

test_deepseek_v4_rocm_wo_a.py

[ROCm][DSV4][Perf] Use FP8 WO_A output projection (#54894)

2026-09-20 09:13:55 +00:00

test_deepseek_v4_vl_rocm.py

[ROCm][DSv4] AITER MegaMoEV2 Integration For DeepSeek V4 (#59685)

2026-10-07 09:35:53 -05:00

test_deepseek_v41_amd_replay_forward.py

[ROCm][Perf][DSv4.1] Enable decoder SWA bounded replay (#59039)

2026-10-13 02:19:14 +00:00

test_deepseek_v41_decoder_replay_layers.py

[Bugfix][DSv4.1] Match decoder-replay input_ids dtype for FlashInfer MegaMoE (#61041)

2026-10-12 14:23:00 +00:00

test_deepseek_v41_replay_batch.py

[ROCm][Perf][DSv4.1] Enable decoder SWA bounded replay (#59039)

2026-10-13 02:19:14 +00:00

test_deepseek_v41_replay_start.py

[Feat][Model] Support encoder-side SWA-bounded replay for DeepSeek-V4.1-Flash (#56227)

2026-09-18 15:01:24 -07:00

test_dspark_mla.py

[Model][MoE] Load MegaMoE experts directly into the DeepGEMM layout (#60096)

2026-10-13 11:52:08 -07:00

test_dsv41_mono_decode.py

[ROCm][Perf][DSv4.1] Mono decode on every backbone layer (#61082)

2026-10-13 10:16:50 +08:00

test_dsv41_mono_layers.py

[ROCm][Perf][DSv4.1] Mono decode on every backbone layer (#61082)

2026-10-13 10:16:50 +08:00

test_dsv41_mono_numerics.py

[ROCm][Perf][DSv4.1] Mono decode layer for DeepSeek-V4.1-Flash on MI355X (#60397)

2026-10-09 20:53:18 -05:00

test_hyv4_rocm.py

[Perf][ROCm] Add a ROCm path for Hy4 and compile the backbone (#57526)

2026-09-21 08:20:43 +00:00

test_initialization.py

[Core][Refactor] Move KV cache planning logic to kv_cache_planning.py (#61183)

2026-10-13 03:05:39 +00:00

test_language_model_cache_is_weak.py

[Bugfix][MRV2] Release model and KV cache on in-process engine shutdown (#54162)

2026-08-28 22:09:08 -07:00

test_qwen3_5_mtp_config.py

[Chore] Use Transformers v5 names and drop redundant processor use_fast (#58550)

2026-09-24 14:14:10 +00:00

test_registry.py

[Qwen4Exp] support Sequence Parallelism for Qwen3.8-Flash-Next (#56322)

2026-10-12 20:56:44 +08:00

test_terratorch.py

[CI/Build] Skip Prithvi/Terratorch model-registry tests when terratorch is missing (#41389)

2026-04-30 12:47:55 -07:00

test_utils.py

[Model][Gemma4] Load Weights with AutoWeightsLoader (#55911)

2026-09-18 12:07:30 +00:00

test_vision.py

[MM] Add Triton kernel for mm_input_normal. (#56798)

2026-09-25 04:31:20 +00:00

utils.py

[Misc] Remove code paths for Transformers < 5.16.1 (#59762)

2026-10-02 11:17:24 +00:00