Commits

Branch selector

 main

User selector

mmangkad

All time

Commit History

Commits on Feb 9, 2026

• [Kimi-K2.5] Fix missing `quant_config` in `KimiK25` (#18440)

  

  mmangkad

  authored and

  John Doe

  committed
  42cf67a

  Copy full SHA for 42cf67a

• fix: fix NVFP4 Kimi-K2.5 weight mapping and exclude list (#18370)

  

  mmangkad

  authored and

  John Doe

  committed
  3e09d05

  Copy full SHA for 3e09d05

• [Doc] Fix outdated `--fp4-gemm-backend` documentation (#18350)

  

  mmangkad

  authored and

  John Doe

  committed
  e6fc6c3

  Copy full SHA for e6fc6c3

• [Doc] Update CUDA 13 install guide to install torch first (#18404)

  Show description for d74ae60
  

  2 peopleauthored and

  John Doe

  committed
  d74ae60

  Copy full SHA for d74ae60

Commits on Feb 5, 2026

• Add MoE fused config for Qwen3-Coder-Next-FP8 on H100 TP=2 (#18195)

  

  mmangkad

  authored and

  John Doe

  committed
  014ff70

  Copy full SHA for 014ff70

• [Docker] Remove hardcoded `America/Los_Angeles` timezone, default to UTC (#18121)

  

  mmangkad

  authored and

  John Doe

  committed
  7df498e

  Copy full SHA for 7df498e

• [Perf] Use safetensors `load_file` in multithread loader (#18124)

  

  mmangkad

  authored and

  John Doe

  committed
  e23f4b6

  Copy full SHA for e23f4b6

Commits on Feb 2, 2026

• [Tiny] Fix grammar in shared experts fusion log messages (#18043)

  

  mmangkad

  authored and

  John Doe

  committed
  d27a5bd

  Copy full SHA for d27a5bd

• Fix cuBLAS >=12.9 detection for cu12/cu13 package naming (#17766)

  

  mmangkad

  authored and

  John Doe

  committed
  64e6c63

  Copy full SHA for 64e6c63

Commits on Jan 25, 2026

• [DeepSeek-V3.2] Fix TRT-LLM NSA in target_verify/draft_extend (#17662)

  mmangkad

  authored
  1674b9e

  Copy full SHA for 1674b9e

Commits on Jan 18, 2026

• [Tiny] Improve docs (#17264)

  mmangkad

  authored
  088758c

  Copy full SHA for 088758c

Commits on Jan 16, 2026

• Fix benchmark import for should_use_tensor_core (#17232)

  mmangkad

  authored
  b0701f0

  Copy full SHA for b0701f0

• [Doc] Tiny docs update for CUDA 13 (#17200)

  mmangkad

  authored
  c771933

  Copy full SHA for c771933

Commits on Jan 11, 2026

• Tiny fix hicache kernel backend comparison (#16867)

  Show description for ff97814
  mmangkad

  authored
  ff97814

  Copy full SHA for ff97814

Commits on Jan 8, 2026

• [Performance] Force split_k=1 for MXFP4 Triton kernels on Hopper (#16014)

  Show description for 05ab110
  mmangkad

  authored
  05ab110

  Copy full SHA for 05ab110

Commits on Nov 27, 2025

• Remove disused B300 Dockerfile (#13946)

  mmangkad

  authored
  63b0562

  Copy full SHA for 63b0562

Commits on Oct 13, 2025

• [Fix] Add per_channel_quant parameter to MoE config functions (#11201)

  mmangkad

  authored
  c7867b6

  Copy full SHA for c7867b6

Commits on Sep 25, 2025

• Refactor kv_cache_scheme handling for quantization (#10132)

  mmangkad

  authored
  cd4da1f

  Copy full SHA for cd4da1f

Commits on Sep 12, 2025

• Add sentencepiece to project dependencies (#10386)

  mmangkad

  authored
  321feca

  Copy full SHA for 321feca

Commits on Sep 2, 2025

• Change tensor alignment method to mn major (#9844)

  mmangkad

  authored
  b9eb0d9

  Copy full SHA for b9eb0d9

Commits on Aug 30, 2025

• Fix typo in warning message about DeepGEMM JIT (#9802)

  mmangkad

  authored
  70eedb5

  Copy full SHA for 70eedb5