Hubert Lu 441045a7bf [AMD] Fix EAGLE3 speculative decoding with aiter attention backend (#19362)

2026-03-03 16:12:13 -08:00

..

attention

[AMD] Fix EAGLE3 speculative decoding with aiter attention backend (#19362)

2026-03-03 16:12:13 -08:00

deep_gemm_wrapper

[DSv32] [GLM5] Improve Model Quality by Avoiding FP32 Precision Loss in weights_proj (#19041)

2026-02-22 16:20:51 +08:00

moe

Add tuned triton==3.5.1 h200 tp2, tp4 for qwen 3 next (#15948)

2026-03-03 14:47:19 -08:00

quantization

Support triton_kernels for GPT-OSS on SM120 (#19718)

2026-03-03 14:14:01 -08:00

rotary_embedding

Fix nightly grok failure on rotary embedding import (#19192)

2026-02-25 13:25:16 +08:00

utils

Fix nvfp4 weight update (#18085)

2026-02-27 14:55:08 -08:00

activation.py

[NPU] support model skywork-reward-gemma2-2-27B-v0.2 (#16947)

2026-02-11 15:34:53 +08:00

amx_utils.py

[CPU][INT4] Add INT4 kernels for CPU (#8226)

2026-01-29 22:30:13 -08:00

communicator_nsa_cp.py

refactor context parallel state (#17213)

2026-02-13 23:18:17 +08:00

communicator.py

[BUG] Support tuple hidden_states from fused MXFP4/FP8 quantization (#19643)

2026-03-02 20:39:06 -08:00

dp_attention.py

Fix comment for tp_rank calculation in dp_attention (#19306)

2026-02-25 15:19:10 +08:00

elementwise.py

[Auto Sync] Update elementwise.py (20260131) (#18033)

2026-02-01 07:39:37 -08:00

flashinfer_comm_fusion.py

[FlashInfer] Switch FlashInfer allreduce fusion to unified API (#18341)

2026-02-22 00:07:16 +08:00

int4fp8_utils.py

[AMD][Quantization] Add int4fp8_moe online quantization on ROCm (#7392)

2026-01-14 01:44:40 -08:00

layernorm.py

[AMD] Support --enable-aiter-allreduce-fusion on AMD GPUs (#13747)

2026-02-24 23:11:55 -08:00

linear.py

[loader] support presharded fused mlp loading (#19519)

2026-02-27 20:37:24 -08:00

logits_processor.py

Revert "[SGL] sync patch: Remove sync points, prefill cudagraph for DP, disable cache reset in mem check (#19190)" (#19581)

2026-02-28 19:46:47 -08:00

model_parallel.py

[Refactor] Rename CustomOp -> MultiPlatformOp (#16175)

2025-12-31 01:16:32 +08:00

modelopt_utils.py

…

multimodal.py

…

parameter.py

Code clean up for fp8 quantization (#16982)

2026-01-13 12:38:39 -08:00

pooler.py

Support piecewise CUDA graph for embedding models (#13852)

2025-11-26 15:29:46 +08:00

radix_attention.py

update pre-commit config (#18860)

2026-02-16 00:18:31 +08:00

radix_linear_attention.py

[PCG] support piecewise cuda graph for kimi-linear model (#18849)

2026-02-17 23:31:12 +08:00

rocm_linear_utils.py

Aiter fp8 kv cache (#13147)

2025-12-08 16:39:53 -08:00

sampler.py

use flashinfer.sampling (#18696)

2026-02-26 10:02:38 +08:00

sparse_pooler.py

…

torchao_utils.py

Lazy import torchao (#17626)

2026-01-22 22:04:51 -08:00

vocab_parallel_embedding.py

Make sure we always disable symm memory without dp padding (#18129)

2026-02-04 19:58:28 +08:00