Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
126 commits
Select commit Hold shift + click to select a range
e1c4495
Support the MIMO cross-grid path in training loop (#5373)
yashaswikarnati Jun 22, 2026
6bd392f
Stabilize hybrid_2b GB200 perf test against run-to-run noise (#5364)
shanmugamr1992 Jun 22, 2026
b6b44a7
Consistent oncall schedule (#5404)
Phlip79 Jun 22, 2026
93a7642
Disag MR3: Add heterogeneous KV/Mamba reshard planners (#5188)
wdykas Jun 22, 2026
2a46893
Add RADIO vision encoder wrapper for MIMO example (#5397)
yashaswikarnati Jun 22, 2026
76f6ccc
Clean up MTP inference control flow (#5418)
santhnm2 Jun 22, 2026
f8170b4
Add MIMO dual gradient finalization (colocated + non-colocated) (#5286)
yashaswikarnati Jun 23, 2026
a58373f
Add RL rollout submission and consumption granularity controls (#5306)
lauradang Jun 23, 2026
f66c28f
Add --functional-test-name to trigger_internal_ci (#5449)
ko3n1g Jun 23, 2026
8fa1831
Rename CP batch helpers to describe balancing granularity (#5403)
deepakn94 Jun 23, 2026
06ae6a9
build: point flash_mla at the nv_dev branch (#5448)
ko3n1g Jun 23, 2026
a2bb5e5
Add logprobs_mode (raw/processed) to inference config (#5419)
tdene Jun 23, 2026
fcbb6ed
Support SWA and sink attention in dynamic inference (#5249)
cuichenx Jun 23, 2026
b1884d1
Add hetero grid args and MoE process groups for MIMO example (#5375)
yashaswikarnati Jun 24, 2026
b549290
ci: Set test_save_verify_integrity_manifest_directly as flaky (#5468)
chtruong814 Jun 24, 2026
47cb413
Remove DBuffer mesh axis validation (#5441)
wujingyue Jun 23, 2026
a27b040
feat(inference): default use_coordinator to True in high-level APIs (…
shanmugamr1992 Jun 24, 2026
811bd29
Support HybridModel feature specs in ModelOpt (#5354)
Phlip79 Jun 24, 2026
e7af860
Add experimental Megatron-FSDP fully_shard implementation (#5387)
wujingyue Jun 24, 2026
cc0c960
chore: rotate oncall schedule
github-actions[bot] Jun 24, 2026
4d44e37
Add inference functions to support MCore-/MBridge- training refactor …
shanmugamr1992 Jun 24, 2026
5a256f3
ci: launch GB200 unit tests via launch_on_gb200 marker (#5477)
ko3n1g Jun 24, 2026
82de1b8
build: install flash_mla from source in the CI image (#5481)
ko3n1g Jun 24, 2026
0b0d985
[split 2/4] Scale DSA indexer loss in pipeline schedules (#5244)
HollowMan6 Jun 24, 2026
0938eb7
ci: check megatron.training imports in installation test (#5458)
ko3n1g Jun 24, 2026
239959b
Fix merges_file kwarg name in HuggingFaceTokenizer (#5406)
muyihao Jun 24, 2026
3330d12
Automated community request assignment (#5147)
Phlip79 Jun 24, 2026
311416f
Clean up training.py module header (dedupe + reorganize imports/globa…
ilml Jun 24, 2026
9038381
Thread process groups through training checkpoint paths (#5486)
yashaswikarnati Jun 24, 2026
5863721
Narrow oncall responsibilities (#5490)
Phlip79 Jun 25, 2026
1c1d6b5
Add MIMO forward step and per-token loss for hetero training (#5376)
yashaswikarnati Jun 25, 2026
ea967a7
Add Nemotron6-MoE VLM model provider for MIMO example (#5374)
yashaswikarnati Jun 25, 2026
7168714
ci: auto-retry test-data download in container-build job (#5498)
ko3n1g Jun 25, 2026
3bfd87b
Force RL inference to CP=1 (#5423)
tdene Jun 25, 2026
2a43e0d
Merge cu_seqlens across micro-batch for THD attention (#5454)
deepakn94 Jun 25, 2026
da482cf
[split 4/4] Enable DSA CP and THD hooks (#5246)
HollowMan6 Jun 25, 2026
e1b8454
Fix fused MLA down projection with tensor parallelism (#5383)
sraman-rgb Jun 25, 2026
8bafe7c
Fix NameError in is_flashinfer_min_version when check_equality=False …
adityasingh2400 Jun 26, 2026
da42015
Add hybrid FSDP unit module support (#4329)
Phlip79 Jun 26, 2026
476228d
fix: set DATA_PATH for moe-dynamic-inference recipe (#5506)
ko3n1g Jun 26, 2026
c0d7848
Add --qad-train-target {base|mtp|both} for QAD / MTP QAT (frozen-base…
yeyu-nvidia Jun 26, 2026
0552f29
[Main] Generalized fix for mxfp8 param gather (#5236)
zhongbozhu Jun 26, 2026
5949478
Add CUDA graph training iteration test (#5417)
wujingyue Jun 26, 2026
847de23
test: restore G/G + lag=19 for gpt_grpo_tp4_pp1_dp2_8b throughput tes…
lauradang Jun 26, 2026
990ced9
Update copy-pr-bot.yaml [skip ci]
github-actions[bot] Jun 27, 2026
cbaa6eb
ci: cache-from a single coherent buildcache donor (#5509)
ko3n1g Jun 26, 2026
ed59a0a
ci: Use GB300 for Github CI tests (#5520)
chtruong814 Jun 27, 2026
0ff7226
ci: pin HF_HUB_CACHE to bind-mounted cache for gpt-oss-20b inference …
ko3n1g Jun 27, 2026
f88b85f
Add inter-document attention masking to GPTDataset (#5298)
deepakn94 Jun 28, 2026
25f6a09
Update copy-pr-bot.yaml [skip ci]
github-actions[bot] Jun 29, 2026
522a9dd
[CI] Fix `gpt_dynamic_inference_tp2_pp2_ep2_gptoss_20b_swa` tests (#5…
asolergi-nv Jun 29, 2026
3c327f3
Implement async scheduling for dynamic inference (#5453)
lmcafee-nvidia Jun 29, 2026
b9f7fb6
build: bump transformer-engine to release_v2.16.post (#5517)
ko3n1g Jun 29, 2026
819912c
Fix `isort` target Python version (#5567)
janEbert Jun 30, 2026
7b6eb02
Fix PR template typo (#5566)
janEbert Jun 30, 2026
872442a
Deduplicate tensor-splitting utility (#5545)
anlthms Jun 30, 2026
99b56a7
Add CI duties to oncall (#5510)
Phlip79 Jun 30, 2026
817c1d5
Thread dp_cp/expt_dp process groups through checkpoint load path (#5579)
yashaswikarnati Jun 30, 2026
f285ea5
Fix TEGroupedMLP pre-backward unshard in fine-grained FSDP hooks for …
rapatel Jun 30, 2026
223e244
[training migration] Finish ModelBuilder integration (#5516)
maanug-nv Jun 30, 2026
4849143
Update copy-pr-bot.yaml [skip ci]
github-actions[bot] Jul 1, 2026
a3d761b
Use NVIDIA inference credentials for Claude actions (#5589)
Phlip79 Jul 1, 2026
36d11ce
Update PR instructions (#5592)
Phlip79 Jul 1, 2026
f972da7
Update mcore skill owners (#5586)
Phlip79 Jul 1, 2026
8b3d8a5
chore: rotate oncall schedule
github-actions[bot] Jul 1, 2026
3c08255
Add /claude fix workflow for on-demand PR fixes (#4862)
Phlip79 Jul 1, 2026
4c4a8ee
Update copy-pr-bot.yaml [skip ci]
github-actions[bot] Jul 2, 2026
adfe9e1
[Megatron-FSDP] MaxPoolAllocator for double-buffering hybrid architec…
cspades Jul 2, 2026
60f338a
fix(tensor_parallel): _reduce returns unreduced tensor for non-contig…
Pearblossom-M Jul 2, 2026
2b551c6
Add Auto Quantize in ModelOpt quantize example (#4821)
jenchen13 Jul 2, 2026
69c4868
[Main][feat] Support CUDA Graph capture offloading modules (#3697)
lhb8125 Jul 2, 2026
0522099
E2E heterogenous non colocated MiMo training (#5602)
yashaswikarnati Jul 2, 2026
5e4fe9b
Optimize memory usage of partial CUDA graphs (#5451)
jiemingz Jul 2, 2026
ee623a9
Document stacked dependent PR handling in split PR skill (#5496)
wujingyue Jul 2, 2026
060371c
Fix Claude reaction permissions (#5613)
Phlip79 Jul 2, 2026
6608a05
Fix smoke BERT/T5 test failures (#5629)
balasaajay Jul 2, 2026
25f6117
Add NCCL symmetric-memory staging to experimental FSDP (#5440)
wujingyue Jul 2, 2026
d89aae5
Add smoke test notification functionality and update notify script (#…
balasaajay Jul 2, 2026
4828d65
Update golden value files for GPT-3 weekly (#5459)
balasaajay Jul 2, 2026
06b07a1
Update copy-pr-bot.yaml [skip ci]
github-actions[bot] Jul 4, 2026
0823c73
Ignore contributor DCO failures in Claude fix (#5625)
Phlip79 Jul 5, 2026
6856424
chore(beep boop 🤖): Bump (main) (2026-07-06)
github-actions[bot] Jul 6, 2026
959f3f1
Pre-size the all-gather buffer for inference to max capacity (#5546)
santhnm2 Jul 6, 2026
7ee524e
Scatter embeddings for sequence parallelism in standalone LM forwards…
kevalmorabia97 Jul 6, 2026
bf32f44
Fix inter-document masking crash and NaNs with TP > 1 and micro_batch…
deepakn94 Jul 6, 2026
1bb5ff5
add safe version of numpy.load (#5500)
dimapihtar Jul 7, 2026
115ce7f
Fuse shared expert MLP with grouped GEMM (#5604)
sraman-rgb Jul 7, 2026
c1560d9
MoE routing analysis and metrics capture (#5220)
mathemakitten Jul 7, 2026
c797a5e
Add cspades to oncall rotation (#5695)
Phlip79 Jul 7, 2026
5dbea46
Add microbatch context helper (#5652)
wujingyue Jul 7, 2026
68f4c62
ci: Update test configurations to unify legacy scope names (#5316)
balasaajay Jul 7, 2026
6ee356a
Fix Torch FSDP2 crash: add force_all_reduce kwarg to base finish_grad…
factnn Jul 8, 2026
72a78d6
Separate mFSDP v2 unit tests (#5640)
wujingyue Jul 8, 2026
465264b
chore: rotate oncall schedule
github-actions[bot] Jul 8, 2026
3ab71ee
test(determinism): add determinism tests (#5041)
ZhiyuLi-Nvidia Jul 8, 2026
40b1fd3
deprecate common strategy (#5160)
dimapihtar Jul 8, 2026
509efe6
ci: revert unify legacy scope names (#5316) (#5709)
ko3n1g Jul 8, 2026
65d6c23
Normalize CRLF in Claude fix commands (#5712)
Phlip79 Jul 8, 2026
2a21b8e
Remove some barriers in save_checkpoint_and_time (#5557)
shurkat-nvidia Jul 8, 2026
91fcdfe
Add FSDP NVTX annotations (#5704)
wujingyue Jul 8, 2026
2d7060f
NCCL EP support (#5129)
YangFei1990 Jul 8, 2026
64bee49
Update base image to nvcr.io/nvidia/pytorch:26.06-py3 (#5632)
balasaajay Jul 8, 2026
027fa4a
Refactor RL rollout pipeline (#5491)
lauradang Jul 8, 2026
d1cab16
[2/2] Wiring cuDNN fused DSA kernels support with THD, CP and IndexSh…
HollowMan6 Jul 8, 2026
f29c747
Triton kernels - avoid recompilation and autotuning in prod (#5608)
sidsingh-nvidia Jul 8, 2026
cf2f07d
Add NeMo Transformer audio encoder model (#5565)
yqwangustc Jul 9, 2026
a2496aa
remove deprecated modules from core/dist_checkpointing (#5134)
dimapihtar Jul 9, 2026
328a77f
fix(fsdp): import os in safe_get_rank fallback (#4959)
fallintoplace Jul 9, 2026
ce8865c
Add forward all-gather overlap (#5513)
wujingyue Jul 9, 2026
e86c262
Fix seq_load_balancing loss with inter-document masking and MBS > 1 (…
deepakn94 Jul 10, 2026
779c5b7
Avoid X11 master port default (#5299)
guihong-nv Jul 10, 2026
1aa880d
Fix infinite recursion in abstract tokenizer special-id property alia…
asadbekXodjayev Jul 10, 2026
2c579b4
Set Bert TE spec q/k_layernorm to None (#5687)
bbuschkaemper Jul 10, 2026
5389d14
Set is_first_microbatch when quant_recipe is configured (#5642)
yezhengmao1 Jul 10, 2026
75a2132
Assign BERT CODEOWNERS to GPT team (#5746)
Phlip79 Jul 10, 2026
48a887f
Remove use of exec_module (#5744)
jon-barker Jul 10, 2026
a5e420a
Merge remote-tracking branch 'origin/main' into main2dev/13_07_2026
github-actions[bot] Jul 13, 2026
3f0539f
fix: post-CI corrections (unused imports + AbsorbedMLA submodule fields)
github-actions[bot] Jul 13, 2026
e348ed1
fix: forward padding mask in fine-grained MoE routing
svcnvidia-nemo-ci Jul 15, 2026
dc693da
fix: align pretrain provider call order
svcnvidia-nemo-ci Jul 15, 2026
015de62
fix: restore dev training-loop behavior
svcnvidia-nemo-ci Jul 15, 2026
48ca9de
fix: restore MLA TP fusion and DeepEP v2 SM config
FDecaYed Jul 17, 2026
775f4e3
Merge remote-tracking branch 'origin/dev' into main2dev/13_07_2026
FDecaYed Jul 17, 2026
1de26bc
fix: address post-merge unit test regressions
FDecaYed Jul 17, 2026
a77ba35
fix: resolve callable hybrid specs
FDecaYed Jul 17, 2026
63754d9
fix: preserve static hybrid module specs
Phlip79 Jul 17, 2026
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
The table of contents is too big for display.
Diff view
Diff view
  •  
  •  
  •  
2 changes: 1 addition & 1 deletion .github/copy-pr-bot.yaml
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
enabled: true
auto_sync_draft: false
auto_sync_ready: true
trustees_override: ["AAnoosheh", "ArEsKay3", "Autumn1998", "BestJuly", "BoxiangW", "CarlosGomes98", "ChenhanYu", "Connor-XY", "FDecaYed", "HaochenYuan", "ISEEKYAN", "JRD971000", "Mellonta", "Phlip79", "QiZhangNV", "RPrenger", "ShriyaRishab", "Victarry", "WanZzzzzz", "Wohox", "YangFei1990", "ZhiyuLi-Nvidia", "adistomar", "ahmadki", "aklife97", "alokpathy", "ananthsub", "anlthms", "aroshanghias-nvd", "ashehper", "asolergi-nv", "athitten", "balasaajay", "buptzyb", "chtruong814", "cjld", "cspades", "cuichenx", "deepakn94", "dimapihtar", "dingqingy-nv", "duncanriach", "erhoo82", "ericharper", "fanshiqing", "faradawn", "fitsumreda", "frsun-nvda", "gautham-kollu", "gdengk", "guihong-nv", "guyueh1", "hexinw-nvidia", "huvunvidia", "hxbai", "ilml", "jalbericiola", "janEbert", "jaredcasper", "jenchen13", "jiemingz", "jingqiny-99", "jkamalu", "jon-barker", "jstjohn", "kajalj22", "kamran-nvidia", "kevalmorabia97", "ko3n1g", "ksivaman", "kunlunl", "kvareddy", "kwyss-nvidia", "layalir", "lhb8125", "liding-nv", "lmcafee-nvidia", "maanug-nv", "macandro96", "mathemakitten", "matthieule", "mchrzanowski", "mehraakash", "minitu", "mkhona-nvidia", "nanz-nv", "ntajbakhsh", "parthmannan", "philipcmonk", "prajwal1210", "pthombre", "rapatel", "rhewett-nv", "rogerwaleffe", "sajadn", "sanandaraj5597", "sancha", "santhnm2", "sbak5", "shanmugamr1992", "sharathts", "sheliang-nv", "shengf-nv", "shifangx", "shjwudp", "sidsingh-nvidia", "skyw", "sraman-rgb", "sudhakarsingh27", "tdene", "theothermike", "thomasdhc", "tomlifu", "trintamaki", "tylerpoon", "wdykas", "wplf", "wujingyue", "xiaoyao0115", "xuantengh", "xuwchen", "yaox12", "yaoyu-33", "yashaswikarnati", "yeyu-nvidia", "yobibyte", "youngeunkwon0405", "yueshen2016", "yuzhongw-nvidia", "zhongbozhu"]
trustees_override: ["AAnoosheh", "ArEsKay3", "Autumn1998", "BestJuly", "BoxiangW", "CarlosGomes98", "ChenhanYu", "Connor-XY", "FDecaYed", "HaochenYuan", "HollowMan6", "ISEEKYAN", "JRD971000", "Leili", "Mellonta", "Phlip79", "QiZhangNV", "RPrenger", "ShriyaRishab", "Victarry", "WanZzzzzz", "Wohox", "YangFei1990", "ZhiyuLi-Nvidia", "adistomar", "ahmadki", "aklife97", "alokpathy", "ananthsub", "anlthms", "aroshanghias-nvd", "ashehper", "asolergi-nv", "athitten", "balasaajay", "buptzyb", "chtruong814", "cjld", "cspades", "cuichenx", "deepakn94", "desh2608", "dimapihtar", "dingqingy-nv", "duncanriach", "erhoo82", "ericharper", "fanshiqing", "faradawn", "fitsumreda", "freewym", "frsun-nvda", "gautham-kollu", "gdengk", "goelarushi", "guihong-nv", "guyueh1", "hexinw-nvidia", "huvunvidia", "hxbai", "ilml", "jalbericiola", "janEbert", "jaredcasper", "jenchen13", "jiaji-huang", "jiemingz", "jingqiny-99", "jkamalu", "jon-barker", "jstjohn", "kajalj22", "kamran-nvidia", "kevalmorabia97", "kingformatty", "ko3n1g", "ksivaman", "kunlunl", "kvareddy", "kwyss-nvidia", "lauradang", "layalir", "lhb8125", "liding-nv", "lmcafee-nvidia", "maanug-nv", "macandro96", "mathemakitten", "matthieule", "mchrzanowski", "mehraakash", "minitu", "mkhona-nvidia", "nanz-nv", "ntajbakhsh", "parthmannan", "philipcmonk", "prajwal1210", "pthombre", "rapatel", "rhewett-nv", "rogerwaleffe", "sajadn", "sanandaraj5597", "sancha", "santhnm2", "sbak5", "shanmugamr1992", "sharathts", "sheliang-nv", "shengf-nv", "shifangx", "shjwudp", "sidsingh-nvidia", "skyw", "sraman-rgb", "sudhakarsingh27", "svcnemo-autobot", "tdene", "theothermike", "thomasdhc", "tomlifu", "trintamaki", "tylerpoon", "wdykas", "wplf", "wujingyue", "xiaoyao0115", "xuantengh", "xuwchen", "yaox12", "yaoyu-33", "yashaswikarnati", "yeyu-nvidia", "yobibyte", "youngeunkwon0405", "yqwangustc", "yueshen2016", "yuzhongw-nvidia", "zhehuaichen", "zhongbozhu"]
42 changes: 21 additions & 21 deletions .github/oncall_schedule.json
Original file line number Diff line number Diff line change
@@ -1,50 +1,50 @@
[
{
"user": "Phlip79",
"date": "2026-06-17"
},
{
"user": "asolergi-nv",
"date": "2026-06-24"
},
{
"user": "maanug-nv",
"date": "2026-07-01"
},
{
"user": "wujingyue",
"user": "cspades",
"date": "2026-07-08"
},
{
"user": "Connor-XY",
"user": "dimapihtar",
"date": "2026-07-15"
},
{
"user": "Phlip79",
"user": "guihong-nv",
"date": "2026-07-22"
},
{
"user": "YangFei1990",
"user": "ilml",
"date": "2026-07-29"
},
{
"user": "asolergi-nv",
"user": "janEbert",
"date": "2026-08-05"
},
{
"user": "dimapihtar",
"user": "maanug-nv",
"date": "2026-08-12"
},
{
"user": "guihong-nv",
"user": "Phlip79",
"date": "2026-08-19"
},
{
"user": "ilml",
"user": "wujingyue",
"date": "2026-08-26"
},
{
"user": "janEbert",
"user": "YangFei1990",
"date": "2026-09-02"
},
{
"user": "asolergi-nv",
"date": "2026-09-09"
},
{
"user": "Connor-XY",
"date": "2026-09-16"
},
{
"user": "cspades",
"date": "2026-09-23"
}
]
2 changes: 1 addition & 1 deletion .github/pull_request_template.md
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
- [ ] I, the PR author, have personally reviewed every line of this PR.

# What does this PR do ?
# What does this PR do?
<!-- Add a one line overview of what this PR aims to accomplish. -->

:warning: For major changes (either in lines of code or in its impact), please make sure to first share a design doc with the team. If you're unsure what's the best way to do so, contact @NVIDIA/mcore-oncall.
Expand Down
Loading
Loading