-
Notifications
You must be signed in to change notification settings - Fork 930
Pull requests: pytorch/torchtitan
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
[rl] Add a standalone weight-sync transport benchmark
ciflow/rl
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
#4083
opened Aug 7, 2026 by
JenniferWang
Contributor
•
3/3
•
Draft
[rl] Add weight-sync transport dial and a manual CPU-staging escape hatch
ciflow/rl
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
#4082
opened Aug 7, 2026 by
JenniferWang
Contributor
•
2/3
•
Draft
Use DTensor placements for distributed metrics
ciflow/rl
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
inference moe expert sp padding
ciflow/rl
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
#4080
opened Aug 7, 2026 by
acisseJZhong
Contributor
Loading…
[spmd_types] Kimi 2.7 enablement
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
#4079
opened Aug 6, 2026 by
pianpwk
Contributor
Loading…
Support oversharded QKV heads in DistributedMuon
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
Support explicit replicated Muon compute
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
Remove renderer from Rollouter group API
ciflow/rl
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
#4076
opened Aug 6, 2026 by
pzhan9
Contributor
Loading…
Mask loss at document boundaries
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
#4075
opened Aug 6, 2026 by
francesco-bertolotti
Contributor
Loading…
Run RolloutWorker in a Monarch actor mesh
ciflow/rl
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
#4074
opened Aug 5, 2026 by
pzhan9
Contributor
Loading…
Refactor rollout execution into RolloutWorker
ciflow/rl
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
#4073
opened Aug 5, 2026 by
pzhan9
Contributor
Loading…
Run InterGeneratorRouter inside a monarch actor
ciflow/rl
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
#4072
opened Aug 5, 2026 by
pzhan9
Contributor
Loading…
[do not review] inference moe padding
ciflow/rl
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
#4067
opened Aug 5, 2026 by
acisseJZhong
Contributor
•
Draft
[ROCm][CI] Skip cp_spmd_types on ROCm pending upstream spmd_types fix
ciflow/rocm
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
module: rocm
#4065
opened Aug 4, 2026 by
rishisinhanj
Contributor
Loading…
Add standalone distributed Muon optimizer
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
[docs]: document environment variables
CLA Signed
This label is managed by the Meta Open Source bot.
#4059
opened Aug 4, 2026 by
gtnv
Loading…
Add torch_checkpointing save config plumbing (#4058)
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
meta-exported
#4058
opened Aug 3, 2026 by
ivy-zhou
Loading…
[bugfix]: fix VLM MFU miscomputation
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
#4057
opened Aug 3, 2026 by
gtnv
Loading…
feat(checkpoint): support allow_partial_load_on_model
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
#4055
opened Aug 3, 2026 by
chango9543
Loading…
3 tasks done
fix async tp
ciflow/rl
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
#4045
opened Aug 1, 2026 by
acisseJZhong
Contributor
Loading…
[wip][not for review] Add eager dist GEMM attention override
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
#4044
opened Jul 31, 2026 by
anijain2305
Contributor
•
Draft
Revert "Always Pre-Split Microbatches for PP"
ci-no-td
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
#4042
opened Jul 31, 2026 by
wconstab
Contributor
Loading…
[ROCm][CI] Re-enable ROCm CI
ciflow/rocm
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
module: rocm
#4039
opened Jul 31, 2026 by
akashveramd
Contributor
•
Draft
Add standalone distributed Muon optimizer
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
graph_trainer: auto-size GPU_MAX_HW_QUEUES on ROCm
ciflow/rocm
ciflow/8gpu
CLA Signed
This label is managed by the Meta Open Source bot.
module: rocm
#4032
opened Jul 30, 2026 by
rishisinhanj
Contributor
Loading…
3 tasks done
Previous Next
ProTip!
Filter pull requests by the default branch with base:main.