MaxText Package Tests

ci_pipeline.yml

2500+ workflow runs

Gemma4 26B MoE on TPU v6e-128: combined optimization stack (#5572 + #5573 + #5608) & recipes
requires action with the application:
MaxText Package Tests#11630:Pull request#6056synchronizebycsgoogle
csgoogle:g4-v6e128-combined-stack
Oct 30, 2026Action required
MaxText Package Tests
completed successfully:
MaxText Package Tests#11629:Scheduled
main
Oct 30, 202638m 53s
build(deps): move diffusers to shared base requirements
failed:
MaxText Package Tests#11628:Pull request#6029synchronizebylydhr
diffusers_base_requirements
Oct 30, 202626m 54s
Add Halo Sliding-Window Context Parallelism and Per-Layer CP Strategy
failed:
MaxText Package Tests#11627:Pull request#5608labeledbyRissyRan
csgoogle:g4-v6e128-attention
Oct 30, 202611s
Add Halo Sliding-Window Context Parallelism and Per-Layer CP Strategy
failed:
MaxText Package Tests#11626:Pull request#5608unlabeledbyRissyRan
csgoogle:g4-v6e128-attention
Oct 30, 202613s
Add Halo Sliding-Window Context Parallelism and Per-Layer CP Strategy
completed successfully:
MaxText Package Tests#11625:Pull request#5608unlabeledbyRissyRan
csgoogle:g4-v6e128-attention
Oct 30, 202637m 35s
OLMoE3: nanochat init, muP and CompletedP parametrizations (OLMo-core base-model mode)
failed:
MaxText Package Tests#11624:Pull request#6057synchronizebyAkshitaB
AkshitaB:akshitab/olmo35-parametrization
Oct 30, 20261h 33m 42s
OLMoE3: log the OLMo-core hero's metrics, fix z-loss under GA, and add OLMo-core's perplexity evals
failed:
MaxText Package Tests#11623:Pull request#5880synchronizebyAkshitaB
AkshitaB:akshitab/olmo35-hero-metrics
Oct 30, 20261h 41m 8s
Fix compile_xla_flags being discarded before training step 0
failed:
MaxText Package Tests#11622:Pull request#6059openedbycopybara-service[bot]Bot
test_1010695314
Oct 30, 20261h 9m 0s
PR #6017: Training engine: buffer metrics on the host
completed successfully:
MaxText Package Tests#11621:Pull request#6051synchronizebycopybara-service[bot]Bot
test_1009954644
Oct 30, 202641m 5s
MoE: add moe_late_tp_combine so routed experts take the tensor axis as expert parallelism
requires action with the application:
MaxText Package Tests#11620:Pull request#4910synchronizebysystalyze-ai
systalyze-ai:upstream-pr/moe-late-tp-combine
Oct 30, 2026Action required
mesh: add mesh_ring_axis to give a mesh axis physical rings on a torus without wraparound
requires action with the application:
MaxText Package Tests#11619:Pull request#4911synchronizebysystalyze-ai
systalyze-ai:upstream-pr/mesh-ring-axis
Oct 30, 2026Action required
OLMoE3: nanochat init, muP and CompletedP parametrizations (OLMo-core base-model mode)
failed:
MaxText Package Tests#11618:Pull request#6057openedbyAkshitaB
AkshitaB:akshitab/olmo35-parametrization
Oct 30, 20262h 31m 50s
Add Halo Sliding-Window Context Parallelism and Per-Layer CP Strategy
failed:
MaxText Package Tests#11617:Pull request#5608labeledbyRissyRan
csgoogle:g4-v6e128-attention
Oct 30, 202615s
Gemma4 26B MoE on TPU v6e-128: combined optimization stack (#5572 + #5573 + #5608) & recipes
requires action with the application:
MaxText Package Tests#11616:Pull request#6056synchronizebycsgoogle
csgoogle:g4-v6e128-combined-stack
Oct 30, 2026Action required
Gemma4 26B MoE on TPU v6e-128: combined optimization stack (#5572 + #5573 + #5608) & recipes
requires action with the application:
MaxText Package Tests#11615:Pull request#6056openedbycsgoogle
csgoogle:g4-v6e128-combined-stack
Oct 30, 2026Action required
PR #6017: Training engine: buffer metrics on the host
cancelled:
MaxText Package Tests#11614:Pull request#6051synchronizebycopybara-service[bot]Bot
test_1009954644
Oct 30, 20261h 33m 33s
Training engine: buffer metrics on the host
failed:
MaxText Package Tests#11613:Pull request#6017synchronizebyigorts-git
igorts-git:port/pr1-metrics-checkpointing
Oct 30, 20261h 3m 42s
PR #6017: Training engine: buffer metrics on the host
cancelled:
MaxText Package Tests#11612:Pull request#6051synchronizebycopybara-service[bot]Bot
test_1009954644
Oct 30, 20264m 33s
Training engine: buffer metrics on the host
cancelled:
MaxText Package Tests#11611:Pull request#6017synchronizebyigorts-git
igorts-git:port/pr1-metrics-checkpointing
Oct 30, 20264m 7s
Training engine: buffer metrics on the host
cancelled:
MaxText Package Tests#11610:Pull request#6017synchronizebyigorts-git
igorts-git:port/pr1-metrics-checkpointing
Oct 30, 202616m 37s
[GPU] DSv4 Support for TE MoE Block
completed successfully:
MaxText Package Tests#11609:Pull request#5487labeledbygobbleturk
dinodeep:dinodeep/te-moe-block-dsv4-support
Oct 30, 202613s
Scaffolding Diffusers in Maxtext-M3 (for Weaver) via WAN [4/4]
failed:
MaxText Package Tests#11608:Pull request#5420synchronizebylydhr
wan_dit_pr4
Oct 30, 202653m 46s
Fix Engram's n-gram hashing and ShortConv under context-parallel load balancing
failed:
MaxText Package Tests#11607:Pull request#5992synchronizebyNuojCheng
chengnuojin-engram-cp-load-balance
Oct 30, 202627m 48s
PR #6017: Training engine: buffer metrics on the host
cancelled:
MaxText Package Tests#11606:Pull request#6051synchronizebycopybara-service[bot]Bot
test_1009954644
Oct 30, 202650m 9s

Navigation Menu

MaxText Package Tests

2500+ workflow runs

Gemma4 26B MoE on TPU v6e-128: combined optimization stack (#5572 + #5573 + #5608) & recipes

MaxText Package Tests

build(deps): move diffusers to shared base requirements

Add Halo Sliding-Window Context Parallelism and Per-Layer CP Strategy

Add Halo Sliding-Window Context Parallelism and Per-Layer CP Strategy

Add Halo Sliding-Window Context Parallelism and Per-Layer CP Strategy

OLMoE3: nanochat init, muP and CompletedP parametrizations (OLMo-core base-model mode)

OLMoE3: log the OLMo-core hero's metrics, fix z-loss under GA, and add OLMo-core's perplexity evals

Fix `compile_xla_flags` being discarded before training step 0

PR #6017: Training engine: buffer metrics on the host

MoE: add moe_late_tp_combine so routed experts take the tensor axis as expert parallelism

mesh: add mesh_ring_axis to give a mesh axis physical rings on a torus without wraparound

OLMoE3: nanochat init, muP and CompletedP parametrizations (OLMo-core base-model mode)

Add Halo Sliding-Window Context Parallelism and Per-Layer CP Strategy

Gemma4 26B MoE on TPU v6e-128: combined optimization stack (#5572 + #5573 + #5608) & recipes

Gemma4 26B MoE on TPU v6e-128: combined optimization stack (#5572 + #5573 + #5608) & recipes

PR #6017: Training engine: buffer metrics on the host

Training engine: buffer metrics on the host

PR #6017: Training engine: buffer metrics on the host

Training engine: buffer metrics on the host

Training engine: buffer metrics on the host

[GPU] DSv4 Support for TE MoE Block

Scaffolding Diffusers in Maxtext-M3 (for Weaver) via WAN [4/4]

Fix Engram's n-gram hashing and ShortConv under context-parallel load balancing

PR #6017: Training engine: buffer metrics on the host

Workflow runs for MaxText Package Tests

2500+ workflow runs