Shuwen check vma#3411

Open

Shuwen-Fang wants to merge 40 commits intomainfrom

shuwen-check-vma

Commits on Mar 25, 2026

base.yml formatting auto
Shuwen-Fang
committed
moe formatting changes
Shuwen-Fang
committed
check vma changes
Shuwen-Fang
committed
remove input variable name due to internal change of name

NuojCheng
authored and
Shuwen-Fang
committed
enable pp with batch split ds

NuojCheng
authored and
Shuwen-Fang
committed
add another layer of custom vjp

NuojCheng
authored and
Shuwen-Fang
committed
add new pipeline weight prefetching config

NuojCheng
authored and
Shuwen-Fang
committed
refactor pr

NuojCheng
authored and
Shuwen-Fang
committed
retrigger CI

gulsumgudukbay
authored and
Shuwen-Fang
committed
Remove post_training_local_dependencies.Dockerfile

SurbhiJainUSC
authored and
Shuwen-Fang
committed
update tokamax group sizes for pipeline

NuojCheng
authored and
Shuwen-Fang
committed
Update user docs to drop config file path

dipannita08
authored and
Shuwen-Fang
committed
Make tokenizer_path to be non-mandatory

A9isha
authored and
Shuwen-Fang
committed
PR #2831: Migrate Decoder to NNX

hsuan-lun-chiang
authored and
Shuwen-Fang
committed
docs: simplify checkpoint storage flags for Pathways workloads

igorts-git
authored and
Shuwen-Fang
committed
Add option to start test_batch in train_rl from a specific index, also add default tokenizer_path for default model

A9isha
authored and
Shuwen-Fang
committed
Add qwen2 implementation

ChingTsai
authored and
Shuwen-Fang
committed
Move tests, rto_setup.sh and preflight.sh to docker image

SurbhiJainUSC
authored and
Shuwen-Fang
committed
Update post-training docs to point to single source of truth for installation

SurbhiJainUSC
authored and
Shuwen-Fang
committed
add custom mesh and logical rule support

NuojCheng
authored and
Shuwen-Fang
committed
PR #3449: Move install_maxtext_extra_deps to dependencies directory

bvandermoon
authored and
Shuwen-Fang
committed
add qwen3-base variants and qwen3-1.7b

andytwigg
authored and
Shuwen-Fang
committed
Fix src/MaxText references in GPU/runner Dockerfiles

bvandermoon
authored and
Shuwen-Fang
committed
Skip Tokamax RaggedDotGroupSizes for FP8

BirdsOfAFthr
authored and
Shuwen-Fang
committed
Update moe.py

BirdsOfAFthr
authored and
Shuwen-Fang
committed
format

BirdsOfAFthr
authored and
Shuwen-Fang
committed
Update moe.py

BirdsOfAFthr
authored and
Shuwen-Fang
committed
split logical names in moe module

NuojCheng
authored and
Shuwen-Fang
committed
base.yml formatting auto
Shuwen-Fang
committed
moe formatting changes
Shuwen-Fang
committed
check vma changes
Shuwen-Fang
committed
remove input variable name due to internal change of name

NuojCheng
authored and
Shuwen-Fang
committed
enable pp with batch split ds

NuojCheng
authored and
Shuwen-Fang
committed
add another layer of custom vjp

NuojCheng
authored and
Shuwen-Fang
committed
add new pipeline weight prefetching config

NuojCheng
authored and
Shuwen-Fang
committed
refactor pr

NuojCheng
authored and
Shuwen-Fang
committed
retrigger CI

gulsumgudukbay
authored and
Shuwen-Fang
committed
Remove post_training_local_dependencies.Dockerfile

SurbhiJainUSC
authored and
Shuwen-Fang
committed
update tokamax group sizes for pipeline

NuojCheng
authored and
Shuwen-Fang
committed
Update user docs to drop config file path

dipannita08
authored and
Shuwen-Fang
committed