Morph MORPH
SIGN IN SIGN UP
dhruvhead / Automodel UNCLAIMED
0 0 13 Python

COMMITS

main
August 14, 2026
Q
fix(retrieval): support canonical Sentence Transformers metadata (#3546)
Qiaochu Zhu US committed
7be510a
H
docs(models): rename Qwen3.8-Max to Qwen3.8-2.4T-A95B (#3549)
Huiying committed
4e0bc86
S
feat(loss): return per-depth MTP losses (#3525)
Slyne Deng committed
71b2f86
August 13, 2026
S
ci: AUT-1501 bump SSO preflight to v1.8.10 (#3537)
svcnemo-autobot committed
a1de7b7
August 12, 2026
H
docs(cohere): add absolute recipe links and remove dead blog reference (#3526)
Huiying committed
6f423fe
D
fix(deps): resolve Starlette and GitPython CVEs (#3523)
Dong Hyuk Chang committed
2032adf
H
feat(moe): integrate Mixture-of-Kittens backend (#3422)
Huiying committed
ade4fcc
A
feat(examples): add Qwen3-32B Tulu-3 finetune recipes (#3104)
Abhishree Thittenamane committed
e539952
A
fix(test): use the recipe's attention backend for the HF reload parity check (#3500)
Abhishree Thittenamane committed
8dae92f
Z
feat(models): add cohere micro vision (#3518)
zeryx committed
8e53272
H
docs(models): add Qwen3.8-Max coverage (#3517)
Huiying committed
86fbc40
J
fix(checkpoint): preserve FSDP2 mixed precision during recompute (#3513)
jQizhang committed
cfb2ed1
H
fix(Qwen3.5): Support split FP8 expert checkpoints for Qwen3.5-MoE (#3495)
Huiying committed
768f5a3
H
fix(deepseek_v4): avoid TileLang boolx8 backward codegen (#3467)
Huiying committed
6551ee4
A
ci(convergence): fix gemma4 eval setup and re-baseline Qwen3-MoE (#3493)
Abhishree Thittenamane committed
e8a163b
A
fix(checkpoint): validate PEFT adapter-only state (#3501)
Alexandros Koumparoulis committed
1077546
J
fix(models): preserve lm-head dtype boundaries (#3491)
jQizhang committed
42af068
August 11, 2026
A
docs: add Nemotron 3.5 Lightning coverage (#3499)
Alexandros Koumparoulis committed
1cd4fd0
A
perf(recipes): run GPT-OSS 120B with EP64 and no activation checkpointing (#3497)
Alexandros Koumparoulis committed
15da268
S
ci: AUT-1408 replace cicd PAT with GitHub App tokens (#3463)
svcnemo-autobot committed
8e78469
A
perf(benchmarks): recompute deterministic MoE routers under AC (#3474)
Alexandros Koumparoulis committed
f4f146f
H
fix(recipe): shard GPT-OSS 120B across 64 experts (#3483)
Huiying committed
7340e78
C
chore: Bump gitpython to >= 3.1.59 (#3482)
Charlie Truong committed
2606c4e
P
fix(ci): register flux2/wan2.2/qwen-image-edit diffusion recipes in CI (#3485)
Pranav Thombre committed
37cd446
K
fix(speculative): record dataset identity in the EAGLE-3 precompute manifest (#3396)
khazzz1c committed
1b00e3d
K
feat(speculative): add Kimi K3 EAGLE-3 training (#3286)
khazzz1c committed
c852b16
S
feat(nemotron_v3): run MTP heads under eval for validation acceptance metrics (#2802)
Slyne Deng committed
03c7e70
H
fix(datasets): reopen IndexedDataset readers in spawned workers (#3382)
Huiying committed
288769b
H
feat(models): add MuseGlimmer training support (#3476)
Huiying committed
a5fdd82
H
feat(vlm): support packed THD vlm context parallelism (#3322)
Huiying committed
f800e75