-
Notifications
You must be signed in to change notification settings - Fork 4.1k
Pull requests: NVIDIA/Megatron-LM
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
[dev] bump emerging optimizers to v0.3.0
#5320
opened Jun 12, 2026 by
FDecaYed
Contributor
Loading…
6 tasks
fix: guard MoE metrics tensor allocation and ensure contiguous GPU tensors (closes #4660)
community-request
#5319
opened Jun 12, 2026 by
botbikamordehai2-sketch
Loading…
ci: Update test configurations to unify legacy scope names
complexity: medium
Run functional tests
#5316
opened Jun 12, 2026 by
balasaajay
Contributor
Loading…
6 tasks
Enable dbias_dprob triton kernel in TE GroupedLinear
#5315
opened Jun 12, 2026 by
vasunvidia
Contributor
•
Draft
6 tasks
chore: nightly sync main into dev (12_06_2026)
Run functional tests
Run MBridge tests
Attach this for testing this PR against MBridge main
#5314
opened Jun 12, 2026 by
svcnvidia-nemo-ci
•
Draft
Account for reasoning token stripping
complexity: low
Final Review
PR is in the "final review" stage
#5313
opened Jun 12, 2026 by
tdene
Contributor
Loading…
1 of 6 tasks
Update SFT dataset and loss calculation
#5311
opened Jun 12, 2026 by
parthmannan
Contributor
•
Draft
1 of 6 tasks
Support fused MLA QKV checkpoint reload
complexity: low
#5310
opened Jun 11, 2026 by
sraman-rgb
Contributor
Loading…
6 tasks
Keep DeepSeek V4 CSA compressor and indexer in high precision under FP8 training
community-request
#5308
opened Jun 11, 2026 by
ssam18
Loading…
Add dump_optimizer_parameters helper for capturing DTensor optimizer state
complexity: low
#5307
opened Jun 11, 2026 by
wujingyue
Contributor
Loading…
3 tasks done
Add RL rollout submission and consumption granularity controls
#5306
opened Jun 11, 2026 by
lauradang
Loading…
6 tasks done
Split dynamic decode bookkeeping for async scheduling
#5303
opened Jun 11, 2026 by
lmcafee-nvidia
Contributor
•
Draft
Fix crash due to tool call at sequence length
complexity: low
#5302
opened Jun 11, 2026 by
tdene
Contributor
Loading…
1 of 6 tasks
Allow for pre-bound socket to be passed in server
complexity: low
Final Review
PR is in the "final review" stage
#5301
opened Jun 11, 2026 by
tdene
Contributor
Loading…
1 of 6 tasks
Handle None values in sampling parameters
complexity: low
Final Review
PR is in the "final review" stage
#5300
opened Jun 11, 2026 by
tdene
Contributor
Loading…
1 of 6 tasks
Avoid X11 master port default
complexity: low
#5299
opened Jun 11, 2026 by
guihong-nv
Contributor
Loading…
Add code owners for optimizer-related files
complexity: low
#5297
opened Jun 11, 2026 by
janEbert
Contributor
Loading…
1 task done
feat(fusions): fused mRoPE for Qwen3.5-VL
complexity: high
#5294
opened Jun 11, 2026 by
wplf
Member
Loading…
feat(moe): discard-output recompute for shared experts
complexity: low
#5293
opened Jun 11, 2026 by
wplf
Member
Loading…
Previous Next
ProTip!
Add no:assignee to see everything that’s not assigned.