4 people f8460ef236 [Kernel][DSv4.1] Bump FlashMLA to the 2026.09.30 release and its interleaved V4.1 KV cache (#61371) ...

Signed-off-by: juntianl <juntianl@inferact.ai>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Kimi Code <noreply@moonshot.ai>
Co-authored-by: Yifan Qiao <yifanqiao@inferact.ai>

2026-10-15 13:58:25 -07:00

..

kernels

[Quantization] Add ModelOpt IQ2_XS, IQ2_XXS, and Q8_0 support via b12x (#59604)

2026-10-13 09:06:33 -07:00

layers

[Kernel][DSv4.1] Bump FlashMLA to the 2026.09.30 release and its interleaved V4.1 KV cache (#61371)

2026-10-15 13:58:25 -07:00

model_loader

[Feature][Model Loader] Apply model-owned MTP shard selection to RunAI ModelStreamer (#60759)

2026-10-15 12:23:11 +08:00

offloader

[Bugfix][Offloader] Preserve prefetch static-buffer slot ownership (#54975)

2026-09-07 10:58:44 +08:00

__init__.py

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

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

test_attention_warmup.py

[Bugfix][MRV2] Fix FlashInfer attention and ReplaySSM warmups on MRV2 (#61013)

2026-10-11 16:04:41 -07:00

test_b12x_warmup.py

[Quantization] Add ModelOpt IQ2_XS, IQ2_XXS, and Q8_0 support via b12x (#59604)

2026-10-13 09:06:33 -07:00

test_bailing_mrope.py

[Model] Add Bailing V3 VL support (#55921)

2026-09-10 02:46:56 -07:00

test_cpu_unquantized_gemm_dispatch.py

[CPU] Conv1d optimised kernel for aarch64 (#54093)

2026-09-30 20:23:42 +08:00

test_deep_gemm_warmup.py

[MoE] Unify DeepGEMM MegaMoE experts behind one feature-matrix layer; add NVFP4 (#61538)

2026-10-15 19:02:01 +00:00

test_eagle_quantization.py

[Quantization] Support online quantization with partially pre-quantized checkpoints (#51392)

2026-09-04 11:46:06 -07:00

test_enabled_custom_ops.py

[ROCm][Perf] Route softmax MoE top-k, including fused shared-expert sigmoid, through AITER topk_gating (#56226)

2026-10-15 17:53:18 +00:00

test_ernie45_vl_mrope.py

[Model] Use mm_features for Ernie-4.5 VL M-RoPE (#39753)

2026-04-14 01:11:52 -07:00

test_flashinfer_autotune_warmup.py

[Fast Start] Preload the FlashInfer autotune table on the weight cache daemon (#60085)

2026-10-10 15:48:06 +00:00

test_gemma_hidden_act.py

[Models][Gemma3/Gemma4] Support hidden_act variants in gated MLP (#40588)

2026-05-08 11:29:11 -07:00

test_jit_warmup_cutedsl_launcher.py

[5/N][warmup][DSv4] Migrate NVIDIA CuTeDSL attention kernels (#53566)

2026-09-12 00:05:59 -04:00

test_jit_warmup_triton_launcher.py

[KV Cache] GLM-5.3-Flash: use the generic packed KV layout; CircularBufferSpec tail with speculative ring slots (#57169)

2026-10-09 15:48:37 -04:00

test_jit_warmup.py

[6/N][warmup][DSv4] Migrate sampling, and DFlash JIT kernels (#56323)

2026-09-12 17:24:24 -07:00

test_k2_horizon_rope_fold.py

[K2 Horizon] fold partial-RoPE permutation into q/k (and norm) weights (#55335)

2026-09-30 16:03:37 +08:00

test_keye_mrope.py

[Model] Use mm_features for Keye-VL and Keye-1.5-VL M-RoPE (#39869)

2026-04-16 02:16:06 -07:00

test_keye_vl1_5_mrope.py

[Model] Use mm_features for Keye-VL and Keye-1.5-VL M-RoPE (#39869)

2026-04-16 02:16:06 -07:00

test_mamba_triton_warmup.py

[Perf] Extend Qwen Triton warmup to avoid first-request latency spikes (#54797)

2026-09-07 11:46:00 +08:00

test_minicpmv4_7.py

[Model] Support MiniCPM-V 4.7 (#58674)

2026-10-11 10:21:10 +00:00

test_minicpmv.py

[Model] Support MiniCPM-V 4.7 (#58674)

2026-10-11 10:21:10 +00:00

test_mistral_large_3_eagle.py

[Bugfix] Initialize missing attributes in mistral eagle (#45217)

2026-06-12 08:21:01 +08:00

test_model_load_with_params.py

Add pydocstyle to the ruff rules (#52136)

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

test_nemotron_h_quantization.py

[Perf][Kernel][Quantization] Fuse ReLU2 with static FP8 activation quantization (#53793)

2026-09-14 10:26:15 -07:00

test_oink_integration.py

Add pydocstyle to the ruff rules (#52136)

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

test_paddleocr_vl_mrope.py

[Model] Use mm_features to compute mrope positions for PaddleOCR-VL (#39888)

2026-04-16 06:14:00 -07:00

test_phi4mm_attention.py

[Security] Do not retain Phi-4 MM audio attention matrices (#60614)

2026-10-09 22:23:35 +08:00

test_plamo3.py

[Model] Support speculative decoding method for PLaMo3 (#54239)

2026-08-28 22:31:07 +00:00

test_qwen3_5_quantization.py

Add LM head quantization support for ModelOpt (#42124)

2026-05-26 09:21:05 -07:00

test_qwen3_asr_mrope.py

[Rust Frontend][gRPC] Add audio and video media inputs (#53760)

2026-08-30 11:31:37 +08:00

test_qwen3_omni_mrope.py

[Bugfix][Model] Fix Qwen3-Omni interleaved M-RoPE boundary positions (#59842)

2026-10-06 03:35:53 +00:00

test_qwen3_omni.py

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

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

test_qwen3_vl_mrope.py

Add pydocstyle to the ruff rules (#52136)

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

test_qwen_triton_warmup.py

[Perf] Extend Qwen Triton warmup to avoid first-request latency spikes (#54797)

2026-09-07 11:46:00 +08:00

test_qwen_vl_triton_warmup.py

[Perf] Skip the M-RoPE warmup when every holder of the RoPE fuses it (#60246)

2026-10-10 06:47:39 +00:00

test_replayssm_warmup.py

[Mamba] Add FlashInfer ReplaySSM support for MTP (#52928)

2026-09-30 21:41:17 +00:00

test_routed_experts_capture.py

[Compile] Remove torch.compile-based sequence parallelism and AsyncTP (#60517)

2026-10-08 08:28:23 -07:00

test_sleep_mode_tensor_ownership.py

[Bugfix][Models] Register sleep-managed runtime buffers (#53507)

2026-08-29 10:32:24 +08:00

test_triton_dispatcher.py

[Feature] Triton kernel dispatcher (#43048)

2026-09-25 04:07:43 +00:00

test_utils.py

[Weight processing] Copy over new_data attributes in replace_parameter (#49601)

2026-08-06 13:03:37 -04:00

test_watermark_sample_warmup.py

[watermarking] Dual-key gumbel-max watermarking for speculative decoding support (#56122)

2026-09-12 18:42:01 +00:00

test_weight_utils.py

[Bugfix][RL] Load attention sinks through weight_loader for layerwise reload (#57797)

2026-10-14 10:17:58 +00:00