Commits

Breadcrumbs

History for

onmain

User selector

All users

Datepicker

All time

Commit history

Commits on Aug 25, 2026

• fix(fla): stop l2norm recompiling per token count

  jason-fxz

  committed
  f7c31e9View commit details

  Copy full SHA for f7c31e9

  View code at this pointBrowse repository at this point

Commits on Aug 24, 2026

• feat(server)!: add --gpu to choose the GPU on multi-GPU machines (#117)

  Show description for 2757bb5
  jason-fxz

  authored
  2757bb5View commit details

  Copy full SHA for 2757bb5

  View code at this pointBrowse repository at this point

• fix(checkpoint): FTW truncated shard raises OSError instead of silent garbage load (#25)

  RuixiangMa

  authored
  bd372b6View commit details

  Copy full SHA for bd372b6

  View code at this pointBrowse repository at this point

Commits on Aug 23, 2026

• fix(moe): report the residency banks actually settle at

  jason-fxz

  committed
  831d38aView commit details

  Copy full SHA for 831d38a

  View code at this pointBrowse repository at this point

• feat(moe): estimate expert-bank bytes from the model config

  jason-fxz

  committed
  eebb3f5View commit details

  Copy full SHA for eebb3f5

  View code at this pointBrowse repository at this point

• feat(moe): per-layer host-bank residency -- split lock-CPU/pin-GPU layers, auto selection under capped pin quota

  jason-fxz

  committed
  c41833bView commit details

  Copy full SHA for c41833b

  View code at this pointBrowse repository at this point

Commits on Aug 19, 2026

• chore(release): 0.1.2

  jason-fxz

  committed
  9db1a39View commit details

  Copy full SHA for 9db1a39

  View code at this pointBrowse repository at this point

• feat(stats): record per-request TTFT and report its mean

  Show description for 12ed695
  jason-fxz

  committed
  12ed695View commit details

  Copy full SHA for 12ed695

  View code at this pointBrowse repository at this point

• perf(fp8): run per-tensor fp8 as W8A8 via scaled_mm (#7)

  jason-fxz

  authored
  586a6c1View commit details

  Copy full SHA for 586a6c1

  View code at this pointBrowse repository at this point

Commits on Aug 18, 2026

• Support muse glimmer (#4)

  Show description for 6223cda
  andy-yang-1

  authored
  6223cdaView commit details

  Copy full SHA for 6223cda

  View code at this pointBrowse repository at this point

• build(release): pypi-ready wheels, metadata, and publish workflow

  jason-fxz

  committed
  76edc16View commit details

  Copy full SHA for 76edc16

  View code at this pointBrowse repository at this point

Commits on Aug 16, 2026

• fix(server): stop API server when backend dies

  Show description for 711325d
  jason-fxz

  committed
  711325dView commit details

  Copy full SHA for 711325d

  View code at this pointBrowse repository at this point

• [Feature] unify reasoning effort on one shared scale (#5)

  Show description for d7cd6ed
  jason-fxz

  authored
  d7cd6edView commit details

  Copy full SHA for d7cd6ed

  View code at this pointBrowse repository at this point

Commits on Aug 15, 2026

• feat(launch): support launch deepseek harness (#3)

  Show description for a257208
  jason-fxz

  authored
  a257208View commit details

  Copy full SHA for a257208

  View code at this pointBrowse repository at this point

Commits on Aug 14, 2026

• feat(server): enable reasoning_effort on /v1/chat/completions (#2)

  Show description for 4f510f1
  jason-fxz

  authored
  4f510f1View commit details

  Copy full SHA for 4f510f1

  View code at this pointBrowse repository at this point

Commits on Aug 13, 2026

• Support minimax m3 (#1)

  Show description for 9dc3ff4
  andy-yang-1

  authored
  9dc3ff4View commit details

  Copy full SHA for 9dc3ff4

  View code at this pointBrowse repository at this point

• feat(kvcache): checkpoint decode state at the tool-call opener token

  jason-fxz

  committed
  ea5348bView commit details

  Copy full SHA for ea5348b

  View code at this pointBrowse repository at this point

• fix(gemma4): read head_dim / kv heads per attention type from per_layer_config

  jason-fxz

  committed
  ee54f63View commit details

  Copy full SHA for ee54f63

  View code at this pointBrowse repository at this point

Commits on Aug 11, 2026

• feat: initial open-source release

  Show description for 3af9d90
  

  jason-fxz

  and

  andy-yang-1

  committed
  3af9d90View commit details

  Copy full SHA for 3af9d90

  View code at this pointBrowse repository at this point

Loading