-
Notifications
You must be signed in to change notification settings - Fork 593
Pull requests: AI-Hypercomputer/maxtext
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
Format and lint across MoE and decoder files.
#5062
opened Aug 29, 2026 by
copybara-service
Bot
Loading…
Expose enable_continuous_profiling CLI flag in MaxText RL
#5061
opened Aug 29, 2026 by
youchunni
Loading…
3 of 4 tasks
Match Tunix peft_trainer_v2 performance in MaxTextTrainingEngine
#5060
opened Aug 29, 2026 by
NuojCheng
Collaborator
Loading…
4 tasks done
Multimodal evaluation runner and end-to-end pipeline for omni POC
#5059
opened Aug 29, 2026 by
subawocit
Collaborator
Loading…
4 tasks done
MaxText/Tunix: Expose enable_continuous_profiling CLI flag in MaxText and Tunix profiler to allow saving >2GB profile as xplane.riegeli
#5058
opened Aug 28, 2026 by
copybara-service
Bot
Loading…
Add a remat checkpoint after the MoE expert-parallel combine
#5057
opened Aug 28, 2026 by
NuojCheng
Collaborator
Loading…
4 tasks done
test(fp8): add PyTorch reference parity suite and logit checker metrics
#5055
opened Aug 28, 2026 by
snehalv2002
Collaborator
•
Draft
3 of 4 tasks
feat(models): add llama3.1-8b-fp8 model config and HuggingFace mapping
#5054
opened Aug 28, 2026 by
snehalv2002
Collaborator
•
Draft
3 of 4 tasks
feat(layers): implement FP8 weight-only dynamic dequantization engine and types
#5053
opened Aug 28, 2026 by
snehalv2002
Collaborator
•
Draft
3 of 4 tasks
feat(checkpoint_conversion): support direct FP8 and scale tensor ingestion in to_maxtext
#5052
opened Aug 28, 2026 by
snehalv2002
Collaborator
•
Draft
3 of 4 tasks
feat(checkpointing): add Mode C dequantize-on-load parameter restoration
#5051
opened Aug 28, 2026 by
snehalv2002
Collaborator
•
Draft
3 of 4 tasks
Run pre-commit to fix MaxText linter
pull ready
#5049
opened Aug 28, 2026 by
bvandermoon
Collaborator
Loading…
4 tasks done
Support 2D transposed shape alignment in model_creation_utils.
#5048
opened Aug 28, 2026 by
copybara-service
Bot
Loading…
RL logprob parity scripts: MaxText trainer vs vLLM sampler (Qwen3.5 35B/397B)
#5047
opened Aug 28, 2026 by
wenxindongwork
Collaborator
•
Draft
1 of 4 tasks
[Maxtext][Tunix] Fix Weight Conversion for Qwen3.5-35B
#5045
opened Aug 28, 2026 by
YixuanWang-99
Collaborator
Loading…
4 tasks done
Add context parallelism to the Gated Delta Net (Qwen3-Next / Qwen3.5)
#5040
opened Aug 28, 2026 by
NuojCheng
Collaborator
Loading…
4 tasks done
Allocate the gradient accumulation carry in config.grad_dtype
#5039
opened Aug 28, 2026 by
NuojCheng
Collaborator
Loading…
4 tasks done
Refactor multi-domain PyGrain dataset mixtures to use MapDataset.mix
#5037
opened Aug 28, 2026 by
snehalv2002
Collaborator
•
Draft
4 tasks
Add Qwen3.5 weight mapping for tpu-inference's vLLM (torchax) path
#5034
opened Aug 27, 2026 by
wenxindongwork
Collaborator
•
Draft
Add Indexer Loss [Deepseek v4]
#5033
opened Aug 27, 2026 by
octatrifan
Collaborator
Loading…
4 tasks done
Improve testing for Muon and the sharded Muon variant.
#5029
opened Aug 27, 2026 by
copybara-service
Bot
Loading…
Fix and generalize Muon weight dimension extraction in MaxText.
#5028
opened Aug 27, 2026 by
copybara-service
Bot
Loading…
Previous Next
ProTip!
Exclude everything labeled
bug with -label:bug.