AbhishekGhosh1998 commited on
Commit
1c43af8
·
verified ·
1 Parent(s): 67fc4d2

granularity results granularity_next_b200/tiera_topk_trace @ 2026-09-21T06:02:36Z from sp-aiinfra-interactive-aghosh-8-mb99k-8xb200-moe-cg-0

Browse files
Files changed (28) hide show
  1. .gitattributes +2 -0
  2. granularity_next_b200/tiera_topk_trace/.completed/ultraep__tierbtrace__gbs16 +1 -0
  3. granularity_next_b200/tiera_topk_trace/STATUS.txt +3 -2
  4. granularity_next_b200/tiera_topk_trace/hf_watcher.log +3 -0
  5. granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/manifest.txt +19 -0
  6. granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/pip_freeze.txt +317 -0
  7. granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/provenance.txt +90 -0
  8. granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G2_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/G2_K2__moonep__matched__seq4096__gbs16__capnone__async.nsys-rep +3 -0
  9. granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G2_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/moonep_async_overlap.json +9 -0
  10. granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G2_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/moonep_async_overlap.txt +416 -0
  11. granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G2_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/moonep_async_overlap_summary.txt +1 -0
  12. granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G2_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/run.env +105 -0
  13. granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G2_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/train.log +1064 -0
  14. granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G2_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/verify_trace_ranks.log +6 -0
  15. granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/G3_K2__moonep__matched__seq4096__gbs16__capnone__async.nsys-rep +3 -0
  16. granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/moonep_async_overlap.json +9 -0
  17. granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/moonep_async_overlap.txt +416 -0
  18. granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/moonep_async_overlap_summary.txt +1 -0
  19. granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/run.env +105 -0
  20. granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/train.log +1194 -0
  21. granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/verify_trace_ranks.log +6 -0
  22. granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K8__moonep__matched__seq4096__gbs16__capnone__async/nsys/run.env +105 -0
  23. granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K8__moonep__matched__seq4096__gbs16__capnone__async/nsys/train.log +12 -0
  24. granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/results.tsv +1 -0
  25. granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/trace_extension_build.log +74 -0
  26. granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/trace_results.tsv +3 -0
  27. granularity_next_b200/tiera_topk_trace/steps.tsv +1 -0
  28. granularity_next_b200/tiera_topk_trace/study.log +41 -0
.gitattributes CHANGED
@@ -100,3 +100,5 @@ granularity_next_b200/tiera_topk_trace/ultraep/tierbtrace/tierbtrace_gbs16/G4_K1
100
  granularity_next_b200/tiera_topk_trace/ultraep/tierbtrace/tierbtrace_gbs16/G4_K16__ultraep__matched__seq4096__gbs16__cap1p08203125/nsys/G4_K16__ultraep__matched__seq4096__gbs16__cap1p08203125.nsys-rep filter=lfs diff=lfs merge=lfs -text
101
  granularity_next_b200/tiera_topk_trace/ultraep/tierbtrace/tierbtrace_gbs16/G4_K6__ultraep__matched__seq4096__gbs16__cap1p1354166666666667/nsys/G4_K6__ultraep__matched__seq4096__gbs16__cap1p1354166666666667.nsys-rep filter=lfs diff=lfs merge=lfs -text
102
  granularity_next_b200/tiera_topk_trace/ultraep/tierbtrace/tierbtrace_gbs16/G4_K8__ultraep__matched__seq4096__gbs16__cap1p109375/nsys/G4_K8__ultraep__matched__seq4096__gbs16__cap1p109375.nsys-rep filter=lfs diff=lfs merge=lfs -text
 
 
 
100
  granularity_next_b200/tiera_topk_trace/ultraep/tierbtrace/tierbtrace_gbs16/G4_K16__ultraep__matched__seq4096__gbs16__cap1p08203125/nsys/G4_K16__ultraep__matched__seq4096__gbs16__cap1p08203125.nsys-rep filter=lfs diff=lfs merge=lfs -text
101
  granularity_next_b200/tiera_topk_trace/ultraep/tierbtrace/tierbtrace_gbs16/G4_K6__ultraep__matched__seq4096__gbs16__cap1p1354166666666667/nsys/G4_K6__ultraep__matched__seq4096__gbs16__cap1p1354166666666667.nsys-rep filter=lfs diff=lfs merge=lfs -text
102
  granularity_next_b200/tiera_topk_trace/ultraep/tierbtrace/tierbtrace_gbs16/G4_K8__ultraep__matched__seq4096__gbs16__cap1p109375/nsys/G4_K8__ultraep__matched__seq4096__gbs16__cap1p109375.nsys-rep filter=lfs diff=lfs merge=lfs -text
103
+ granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G2_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/G2_K2__moonep__matched__seq4096__gbs16__capnone__async.nsys-rep filter=lfs diff=lfs merge=lfs -text
104
+ granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/G3_K2__moonep__matched__seq4096__gbs16__capnone__async.nsys-rep filter=lfs diff=lfs merge=lfs -text
granularity_next_b200/tiera_topk_trace/.completed/ultraep__tierbtrace__gbs16 ADDED
@@ -0,0 +1 @@
 
 
1
+ 2026-09-21T05:53:08+00:00
granularity_next_b200/tiera_topk_trace/STATUS.txt CHANGED
@@ -1,10 +1,11 @@
1
  study_root: /workspace/granularity_next_b200/tiera_topk_trace
2
  gbs: 16
3
- updated: 2026-09-21T05:22:22+00:00
4
- current: ultraep__tierbtrace__gbs16
5
 
6
  completed steps:
7
  echo__tierbtrace__gbs16
 
8
 
9
  failed steps:
10
  (none)
 
1
  study_root: /workspace/granularity_next_b200/tiera_topk_trace
2
  gbs: 16
3
+ updated: 2026-09-21T05:53:11+00:00
4
+ current: moonep__tierbtrace__gbs16
5
 
6
  completed steps:
7
  echo__tierbtrace__gbs16
8
+ ultraep__tierbtrace__gbs16
9
 
10
  failed steps:
11
  (none)
granularity_next_b200/tiera_topk_trace/hf_watcher.log CHANGED
@@ -45,3 +45,6 @@ https://huggingface.co/MoE-CUDA-Graph-Analysis/tierb-b200-gbs8/tree/main
45
  uploaded: /workspace/granularity_next_b200/tiera_topk_trace -> MoE-CUDA-Graph-Analysis/tierb-b200-gbs8/granularity_next_b200/tiera_topk_trace
46
  https://huggingface.co/MoE-CUDA-Graph-Analysis/tierb-b200-gbs8/tree/main
47
  [2026-09-21T05:52:31+00:00] backup pass
 
 
 
 
45
  uploaded: /workspace/granularity_next_b200/tiera_topk_trace -> MoE-CUDA-Graph-Analysis/tierb-b200-gbs8/granularity_next_b200/tiera_topk_trace
46
  https://huggingface.co/MoE-CUDA-Graph-Analysis/tierb-b200-gbs8/tree/main
47
  [2026-09-21T05:52:31+00:00] backup pass
48
+ uploaded: /workspace/granularity_next_b200/tiera_topk_trace -> MoE-CUDA-Graph-Analysis/tierb-b200-gbs8/granularity_next_b200/tiera_topk_trace
49
+ https://huggingface.co/MoE-CUDA-Graph-Analysis/tierb-b200-gbs8/tree/main
50
+ [2026-09-21T06:02:36+00:00] backup pass
granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/manifest.txt ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ timestamp: 2026-09-21T05:53:25+00:00
2
+ host: sp-aiinfra-interactive-aghosh-8-mb99k-8xb200-moe-cg-0
3
+ gpu: NVIDIA B200 x8
4
+ driver: 580.95.05
5
+ study_scripts_branch: granularity-moonep
6
+ study_scripts_commit: d29f657
7
+ campaign: moonep
8
+ backend: moonep
9
+ phase: tierbtrace
10
+ gbs: 16
11
+ seq: 4096
12
+ configs: G1 G2 G3 G4
13
+ capacity: 1.25
14
+ repeats: 3
15
+ nsys: 1
16
+ megatron_repo: /workspace/Megatron-MoE-Echo
17
+ megatron_branch_required: moonep_backend_async_grad_reduce
18
+ megatron_branch_actual: moonep_backend_async_grad_reduce
19
+ megatron_commit: 2e68c8ad
granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/pip_freeze.txt ADDED
@@ -0,0 +1,317 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ absl-py==2.3.0
2
+ aiohappyeyeballs @ file:///rapids/aiohappyeyeballs-2.6.1-py3-none-any.whl#sha256=f349ba8f4b75cb25c99c5c2d84e997e485204d2902a9597802b0371f09331fb8
3
+ aiohttp @ file:///rapids/aiohttp-3.12.7-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl#sha256=94f98e0e5a49f89b252e115844f756c04fc8050f38252a32a3dd994ce8121f10
4
+ aiosignal @ file:///rapids/aiosignal-1.3.2-py2.py3-none-any.whl#sha256=45cde58e409a301715980c2b01d0c28bdde3770d8290b5eb2173759d9acb31a5
5
+ annotated-types==0.7.0
6
+ anyio==4.9.0
7
+ apex @ file:///opt/pytorch/apex/dist/apex-0.1-cp312-cp312-linux_x86_64.whl#sha256=a1508e4a386362e9c8f46c92ccada9110e421212b964c694235d0fd5a224b170
8
+ argon2-cffi==25.1.0
9
+ argon2-cffi-bindings==21.2.0
10
+ arrow==1.3.0
11
+ asciitree @ file:///rapids/asciitree-0.3.3-py3-none-any.whl#sha256=3e507ccfaeb417823158091fdb9d26b0205f6a59ba680b479a71db65c73d01d4
12
+ asttokens==3.0.0
13
+ astunparse==1.6.3
14
+ async-lru==2.0.5
15
+ attrs==25.3.0
16
+ audioread==3.0.1
17
+ babel==2.17.0
18
+ beautifulsoup4==4.13.4
19
+ black==25.1.0
20
+ bleach==6.2.0
21
+ blis==0.7.11
22
+ cachetools @ file:///rapids/cachetools-6.0.0-py3-none-any.whl#sha256=82e73ba88f7b30228b5507dce1a1f878498fc669d972aef2dde4f3a3c24f103e
23
+ catalogue==2.0.10
24
+ certifi==2025.4.26
25
+ cffi==1.17.1
26
+ charset-normalizer==3.4.2
27
+ click==8.5.0
28
+ cloudpathlib==0.21.1
29
+ cloudpickle @ file:///rapids/cloudpickle-3.1.1-py3-none-any.whl#sha256=c8c5a44295039331ee9dad40ba100a9c7297b6f988e50e87ccdf3765a668350e
30
+ cmake==3.31.6
31
+ comm==0.2.2
32
+ confection==0.1.5
33
+ contourpy==1.3.2
34
+ cuda-bindings @ file:///rapids/cuda_bindings-12.9.0-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl#sha256=ff0e28d1e34758654b9c961e1f55e4786e49aee6a4dbceaf3cc24c46c672df7e
35
+ cuda-python @ file:///rapids/cuda_python-12.9.0-py3-none-any.whl#sha256=926acba49b2c0a0374c61b7c98f337c085199cf51cdfe4d6423c4129c20547a7
36
+ cudf @ file:///rapids/cudf-25.4.0-cp312-cp312-linux_x86_64.whl#sha256=b779a8e8c0e8577b9b560895958496da211bfac49cdc4f11a49e8d18cb82e348
37
+ cudf-polars @ file:///rapids/cudf_polars-25.4.0-py3-none-any.whl#sha256=227b6c5147e93ea86282c8f7219118f916769fa2ca0cdd931e97a148aa66b869
38
+ cugraph @ file:///rapids/cugraph-25.4.0-cp312-cp312-linux_x86_64.whl#sha256=b837c2dc2f7b235c702e4f8df64479b8fcbb342f848516d5faf0cdc4b5f31f29
39
+ cugraph-service-client @ file:///rapids/cugraph_service_client-25.4.0-py3-none-any.whl#sha256=7e568c3e41154b9752f5173555ffcc282ac52d72343feeb79653b2e2b1232470
40
+ cugraph-service-server @ file:///rapids/cugraph_service_server-25.4.0-py3-none-any.whl#sha256=a2830dcffafcccb16e12ebf85b98f8f247869d38c01ab4dd8664a1e7439d31be
41
+ cuml @ file:///rapids/cuml-25.4.0-cp312-cp312-linux_x86_64.whl#sha256=1e90a464a18b094e24797f1676688c5d7696c4cf97c934504214a53d5aa09aa5
42
+ cupy-cuda12x @ file:///rapids/cupy_cuda12x-13.3.0-cp312-cp312-linux_x86_64.whl#sha256=6a2dfa0aa1e8189a10d1534276c08f1d2f73b747fd28bfbe7b9f4ef66d9af73f
43
+ cuvs @ file:///rapids/cuvs-25.4.0-cp312-cp312-linux_x86_64.whl#sha256=b12f6b9f80e053a3b38542c2947d2516e468c02cad54a13d8b1a4a61d5aa7914
44
+ cycler==0.12.1
45
+ cymem==2.0.11
46
+ Cython==3.1.1
47
+ dask @ file:///rapids/dask-2025.2.0-py3-none-any.whl#sha256=f0fdeef6ceb0a06569d456c9e704f220f7f54e80f3a6ea42ab98cea6bc642b6e
48
+ dask-cuda @ file:///rapids/dask_cuda-25.4.0-py3-none-any.whl#sha256=442d87da0a95d00ca2a262b64c76e785a32ad905c82000fa2227932f81390e19
49
+ dask-cudf @ file:///rapids/dask_cudf-25.4.0-py3-none-any.whl#sha256=5e7cea0b26daea17da6e0bb457435e24460f76ba740358a3a18e239d6015c95c
50
+ datasets==5.0.1
51
+ debugpy==1.8.14
52
+ decorator==5.2.1
53
+ -e git+https://github.com/deepseek-ai/DeepEP.git@34152ae28f80bcc3ee38d7a12cb2ad87cfd4ea72#egg=deep_ep
54
+ defusedxml==0.7.1
55
+ dill==0.4.1
56
+ distributed @ file:///rapids/distributed-2025.2.0-py3-none-any.whl#sha256=368462084be47092d0a835f6c64c9b852da64dc88302b76d5f93bd5131228c67
57
+ distributed-ucxx @ file:///rapids/distributed_ucxx-0.43.0-py3-none-any.whl#sha256=94f60d0ccb0c94ba8dbfa1eaf455874d18aaab17e2ae218c38590e9fdc2d721d
58
+ dm-tree==0.1.9
59
+ einops==0.8.1
60
+ execnet==2.1.1
61
+ executing==2.2.0
62
+ expecttest==0.3.0
63
+ fasteners @ file:///rapids/fasteners-0.19-py3-none-any.whl#sha256=758819cb5d94cdedf4e836988b74de396ceacb8e2794d21f82d131fd9ee77237
64
+ fastjsonschema==2.21.1
65
+ fastrlock @ file:///rapids/fastrlock-0.8.3-cp312-cp312-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_28_x86_64.whl#sha256=5f13ec08f1adb1aa916c384b05ecb7dbebb8df9ea81abd045f60941c6283a670
66
+ filelock==3.18.0
67
+ flash_attn @ file:///opt/pytorch/flash_attn-2.7.4.post1-cp312-cp312-linux_x86_64.whl#sha256=a10455d6c56b387785b7413fd0bbcb392b42b3d4f4b88fda5d86cc8a9ce9182a
68
+ fonttools==4.58.1
69
+ fqdn==1.5.1
70
+ frozenlist @ file:///rapids/frozenlist-1.6.0-cp312-cp312-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_17_x86_64.manylinux2014_x86_64.whl#sha256=4e1be9111cb6756868ac242b3c2bd1f09d9aea09846e4f5c23715e7afb647103
71
+ fsspec==2025.5.1
72
+ gast==0.6.0
73
+ grpcio==1.62.1
74
+ h11==0.16.0
75
+ hf-xet==1.6.0
76
+ httpcore==1.0.9
77
+ httpx==0.28.1
78
+ huggingface_hub==1.32.0
79
+ hypothesis==6.130.8
80
+ idna==3.10
81
+ importlib_metadata==8.7.0
82
+ iniconfig==2.1.0
83
+ intel-openmp==2021.4.0
84
+ ipykernel==6.29.5
85
+ ipython==9.3.0
86
+ ipython_pygments_lexers==1.1.1
87
+ isoduration==20.11.0
88
+ isort==6.0.1
89
+ jedi==0.19.2
90
+ Jinja2==3.1.6
91
+ joblib==1.5.1
92
+ json5==0.12.0
93
+ jsonpointer==3.0.0
94
+ jsonschema==4.24.0
95
+ jsonschema-specifications==2025.4.1
96
+ jupyter-events==0.12.0
97
+ jupyter-lsp==2.2.5
98
+ jupyter_client==8.6.3
99
+ jupyter_core==5.8.1
100
+ jupyter_server==2.16.0
101
+ jupyter_server_terminals==0.5.3
102
+ jupyterlab==4.4.3
103
+ jupyterlab_code_formatter==3.0.2
104
+ jupyterlab_pygments==0.3.0
105
+ jupyterlab_server==2.27.3
106
+ jupyterlab_tensorboard_pro @ file:///builder/jupyterlab_tensorboard_pro-4.0.0-py2.py3-none-any.whl#sha256=6bfcb5ced3358eb72150f4b2ab91312faef2f0ca14d845e8379d8fa52df6902e
107
+ jupytext==1.17.2
108
+ kiwisolver==1.4.8
109
+ kvikio @ file:///rapids/kvikio-25.4.0-cp312-cp312-linux_x86_64.whl#sha256=8b40fd8ea9e4a2974c04b50b7fc26b06094ac807f9a233281d0bafbaa71b1d85
110
+ langcodes==3.5.0
111
+ language_data==1.3.0
112
+ lazy_loader==0.4
113
+ libcudf @ file:///rapids/libcudf-25.4.0-py3-none-linux_x86_64.whl#sha256=238320b4958c1737bba2a609673e2539a78216b811b5498a68e00395d9b021a4
114
+ libcugraph @ file:///rapids/libcugraph-25.4.0-py3-none-linux_x86_64.whl#sha256=29085e901b2f7569ff2653bb1355fe6fad20f2fcaa585f5649da9867d8df766d
115
+ libcuml @ file:///rapids/libcuml-25.4.0-py3-none-linux_x86_64.whl#sha256=8e2e444c72b0e995e15a48da37ff12fb410a47a2baa69ba86ccc9b1a3db91e8c
116
+ libcuvs @ file:///rapids/libcuvs-25.4.0-py3-none-linux_x86_64.whl#sha256=e2f8db61c92b769c590dd752f43439ddd62b2f6592a3c337362b6c1ef220c3b9
117
+ libkvikio @ file:///rapids/libkvikio-25.4.0-py3-none-linux_x86_64.whl#sha256=da8a5713b05fc478bf40529a3e0fb61a2d605bb5a70ca4d461f96a6005dbe8b7
118
+ libraft @ file:///rapids/libraft-25.4.0-py3-none-linux_x86_64.whl#sha256=b921dec005aa4a321dd156776a18376feb47c96a5b67eec125c8a8e19edd1e24
119
+ librmm @ file:///rapids/librmm-25.4.0-py3-none-linux_x86_64.whl#sha256=ff951811bda729d4fa1109ab906225ae1f489a6ec2eeab5d1b674cc34bd4422f
120
+ librmm-cu12 @ file:///rapids/librmm_cu12-25.4.0-py3-none-any.whl#sha256=29296ef2b0e90679a0db280e86015be12d6ab9f4e62a181aa085f62dbadd06b3
121
+ librosa==0.11.0
122
+ libucxx @ file:///rapids/libucxx-0.43.0-py3-none-linux_x86_64.whl#sha256=f15849d7ac2fb68f0fd8998838a13321ddcfb0c594f7e6cbc401a3bcc4750566
123
+ lightning-thunder @ file:///opt/pytorch/lightning-thunder
124
+ lightning-utilities==0.14.3
125
+ lintrunner==0.12.7
126
+ llvmlite==0.42.0
127
+ locket @ file:///rapids/locket-1.0.0-py2.py3-none-any.whl#sha256=b6c819a722f7b6bd955b80781788e4a66a55628b858d347536b7e81325a3a5e3
128
+ looseversion==1.3.0
129
+ marisa-trie==1.2.1
130
+ Markdown==3.8
131
+ markdown-it-py==3.0.0
132
+ MarkupSafe==3.0.2
133
+ matplotlib==3.10.3
134
+ matplotlib-inline==0.1.7
135
+ mdit-py-plugins==0.4.2
136
+ mdurl==0.1.2
137
+ -e git+ssh://git@github.com/sriram-17-17/my-Megatron-MoE-Echo.git@2e68c8ad6bfb5bac5d6eb962fa7d4fbf0a5fb095#egg=megatron_core
138
+ mistune==3.1.3
139
+ mkl==2021.1.1
140
+ mkl-devel==2021.1.1
141
+ mkl-include==2021.1.1
142
+ ml_dtypes==0.6.0
143
+ mock==5.2.0
144
+ -e git+ssh://git@github.com/sriram-17-17/MoonEP.git@d8bc6667fad7177e6e09b637d773fa67f08bae93#egg=moonep
145
+ mpmath==1.3.0
146
+ msgpack==1.1.0
147
+ multidict @ file:///rapids/multidict-6.4.4-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl#sha256=a920f9cf2abdf6e493c519492d892c362007f113c94da4c239ae88429835bad1
148
+ multiprocess==0.70.19
149
+ murmurhash==1.0.13
150
+ mypy_extensions==1.1.0
151
+ nbclient==0.10.2
152
+ nbconvert==7.16.6
153
+ nbformat==5.10.4
154
+ nest-asyncio==1.6.0
155
+ networkx==3.5
156
+ ninja==1.11.1.4
157
+ notebook==7.4.3
158
+ notebook_shim==0.2.4
159
+ numba==0.59.1
160
+ numba-cuda @ file:///rapids/numba_cuda-0.4.0-py3-none-any.whl#sha256=5b95a5f42068771e37142944d44fe844fa87f1b2f89051461a4cb0af91bcc28a
161
+ numcodecs @ file:///rapids/numcodecs-0.13.1-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl#sha256=da2230484e6102e5fa3cc1a5dd37ca1f92dfbd183d91662074d6f7574e3e8f53
162
+ numpy==1.26.4
163
+ nvdlfw_inspect @ git+https://github.com/NVIDIA/nvidia-dlfw-inspect.git@6b60eb0e675606fb2fbbfcb12667ecbec75eaf87
164
+ nvfuser==0.2.27a0+9bf5aca
165
+ nvidia-cuda-cccl-cu12==12.9.27
166
+ nvidia-cudnn-frontend @ file:///opt/pytorch/lightning-thunder/tmp/cudnn_frontend
167
+ nvidia-cutlass-dsl==4.4.2
168
+ nvidia-cutlass-dsl-libs-base==4.4.2
169
+ nvidia-dali-cuda120==1.50.0
170
+ nvidia-ml-py @ file:///rapids/nvidia_ml_py-12.575.51-py3-none-any.whl#sha256=eb8641800d98ce40a22f479873f34b482e214a7e80349c63be51c3919845446e
171
+ nvidia-modelopt==0.29.0
172
+ nvidia-modelopt-core==0.29.0
173
+ nvidia-nvcomp-cu12==4.2.0.14
174
+ nvidia-nvimgcodec-cu12==0.5.0.13
175
+ nvidia-nvjpeg-cu12==12.4.0.16
176
+ nvidia-nvjpeg2k-cu12==0.8.1.40
177
+ nvidia-nvshmem-cu12==3.7.2
178
+ nvidia-nvtiff-cu12==0.5.0.67
179
+ nvidia-resiliency-ext==0.4.0
180
+ nvtx @ file:///rapids/nvtx-0.2.11-cp312-cp312-linux_x86_64.whl#sha256=73521e37cba42b4db0b3cd68fc5eb814e55c500a61eeed6edcdf6746315ac53e
181
+ nx-cugraph @ file:///rapids/nx_cugraph-25.4.0-py3-none-any.whl#sha256=e625444486cc899a77bd48e7da76a2edd95c0006dd95fbf323745feae7f33111
182
+ onnx @ file:///opt/pytorch/pytorch/third_party/onnx
183
+ onnx-ir==1.0.0
184
+ onnxscript==0.7.2
185
+ opt_einsum==3.4.0
186
+ optree==0.16.0
187
+ overrides==7.7.0
188
+ packaging==23.2
189
+ pandas @ file:///rapids/pandas-2.2.3-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl#sha256=fffb8ae78d8af97f849404f21411c95062db1496aeb3e56f146f0355c9989319
190
+ pandocfilters==1.5.1
191
+ parso==0.8.4
192
+ partd @ file:///rapids/partd-1.4.2-py3-none-any.whl#sha256=978e4ac767ec4ba5b86c6eaa52e5a2a3bc748a2ca839e8cc798f1cc6ce6efb0f
193
+ pathspec==0.12.1
194
+ pexpect==4.9.0
195
+ pillow==11.2.1
196
+ platformdirs==4.3.8
197
+ pluggy==1.6.0
198
+ ply @ file:///rapids/ply-3.11-py2.py3-none-any.whl#sha256=096f9b8350b65ebd2fd1346b12452efe5b9607f7482813ffca50c22722a807ce
199
+ polars @ file:///rapids/polars-1.25.2-cp39-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl#sha256=f7fcbb4f476784384ccda48757fca4e8c2e2c5a0a3aef3717aaf56aee4e30e09
200
+ polygraphy==0.49.20
201
+ pooch==1.8.2
202
+ preshed==3.0.10
203
+ prometheus_client==0.22.1
204
+ prompt_toolkit==3.0.51
205
+ propcache @ file:///rapids/propcache-0.3.1-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl#sha256=563f9d8c03ad645597b8d010ef4e9eab359faeb11a0a2ac9f7b4bc8c28ebef25
206
+ protobuf==4.24.4
207
+ psutil==7.0.0
208
+ ptyprocess==0.7.0
209
+ PuLP==3.2.1
210
+ pure_eval==0.2.3
211
+ pyarrow==25.0.1
212
+ pybind11==2.13.6
213
+ pybind11_global==2.13.6
214
+ pycocotools @ git+https://github.com/nvidia/cocoapi.git@08350a157b41f0755a536441b5abec89ab934c5b#subdirectory=PythonAPI
215
+ pycparser==2.22
216
+ pydantic==2.11.5
217
+ pydantic_core==2.33.2
218
+ Pygments==2.19.1
219
+ pylibcudf @ file:///rapids/pylibcudf-25.4.0-cp312-cp312-linux_x86_64.whl#sha256=401b8bfd022d06f84e120779ca7c3d01a28b68762aaa8f7873b10d6b1975cf6a
220
+ pylibcugraph @ file:///rapids/pylibcugraph-25.4.0-cp312-cp312-linux_x86_64.whl#sha256=061c7389a88c815b54fd5d3a3584fb87bf693e5602f5be335260d0ba575ba557
221
+ pylibcugraphops @ file:///rapids/pylibcugraphops-25.4.0-cp312-cp312-linux_x86_64.whl#sha256=a25ef4b22166c80718d5c7e5bddb755e81a2fe340fa5becd0ac36ec3edfb208a
222
+ pylibraft @ file:///rapids/pylibraft-25.4.0-cp312-cp312-linux_x86_64.whl#sha256=f2ecd23fe51f9c2caa87bbbe85cb32fcaff4ecfb52ceb9a9991bc50f82782c2e
223
+ pylibwholegraph @ file:///rapids/pylibwholegraph-25.4.0-cp312-cp312-linux_x86_64.whl#sha256=b82913d1e38bda76d8c272fbd8cc9072d987209d3d47cc93a464a37c2925aca2
224
+ pynvjitlink @ file:///rapids/pynvjitlink-0.3.0-cp312-cp312-linux_x86_64.whl#sha256=066caf1314730643d3b48c9847bd65789defdd6a7c78993176eb3d8bbe3f4919
225
+ pynvml @ file:///rapids/pynvml-12.0.0-py3-none-any.whl#sha256=fdff84b62a27dbe98e08e1a647eb77342bef1aebe0878bcd15e99a83fcbecb9e
226
+ pyparsing==3.2.3
227
+ pytest==8.1.1
228
+ pytest-flakefinder==1.1.0
229
+ pytest-rerunfailures==15.1
230
+ pytest-shard==0.1.2
231
+ pytest-xdist==3.7.0
232
+ python-dateutil==2.9.0.post0
233
+ python-hostlist==2.2.1
234
+ python-json-logger==3.3.0
235
+ pytorch-triton @ file:///tmp/dist/pytorch_triton-3.3.0%2Bgit96316ce52.nvinternal-cp312-cp312-linux_x86_64.whl#sha256=6d6921772b9dfd0243c055e0d054de42a113e92cabcfdd1d3e80a5ad9c08cf7d
236
+ pytz @ file:///rapids/pytz-2023.4-py2.py3-none-any.whl#sha256=f90ef520d95e7c46951105338d918664ebfd6f1d995bd7d153127ce90efafa6a
237
+ PyYAML==6.0.2
238
+ pyzmq==26.4.0
239
+ raft-dask @ file:///rapids/raft_dask-25.4.0-cp312-cp312-linux_x86_64.whl#sha256=36cf37ba6ce35454c4729c2cb597a360f22780eefb11d514a16a07137da5eb6c
240
+ rapids-dask-dependency @ file:///rapids/rapids_dask_dependency-25.4.0a0-py3-none-any.whl#sha256=484faca626e7760112a7d8ad3426f5bdd720b609006d8dca87c2607076a348bd
241
+ rapids-logger @ file:///rapids/rapids_logger-0.1.18-py3-none-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl#sha256=3617b2d86291ac6cb957542895f20fc9737ace29773ca90337e6fb04c73bd120
242
+ referencing==0.36.2
243
+ regex==2024.11.6
244
+ requests==2.32.3
245
+ rfc3339-validator==0.1.4
246
+ rfc3986-validator==0.1.1
247
+ rich==14.0.0
248
+ rmm @ file:///rapids/rmm-25.4.0-cp312-cp312-linux_x86_64.whl#sha256=1824bcab3836296666980a997ddcb9c4bca24dfb873012cfa0a7ecea583ca712
249
+ rpds-py==0.25.1
250
+ safetensors==0.5.3
251
+ scikit-learn==1.6.1
252
+ scipy==1.15.3
253
+ Send2Trash==1.8.3
254
+ setuptools==78.1.1
255
+ shellingham==1.5.4
256
+ six==1.16.0
257
+ smart-open==7.1.0
258
+ sniffio==1.3.1
259
+ sortedcontainers==2.4.0
260
+ soundfile==0.13.1
261
+ soupsieve==2.7
262
+ soxr==0.5.0.post1
263
+ spacy==3.7.5
264
+ spacy-legacy==3.0.12
265
+ spacy-loggers==1.0.5
266
+ srsly==2.5.1
267
+ stack-data==0.6.3
268
+ sympy==1.14.0
269
+ tabulate==0.9.0
270
+ tbb==2021.13.1
271
+ tblib @ file:///rapids/tblib-3.1.0-py3-none-any.whl#sha256=670bb4582578134b3d81a84afa1b016128b429f3d48e6cbbaecc9d15675e984e
272
+ tensorboard==2.16.2
273
+ tensorboard-data-server==0.7.2
274
+ tensorrt @ file:///workspace/TensorRT-10.11.0.33/python/tensorrt-10.11.0.33-cp312-none-linux_x86_64.whl#sha256=c2f18f72758a3d54a71901495c8fda9e7094a8bcdd2a5b6b58967fbf05733b3a
275
+ terminado==0.18.1
276
+ thinc==8.2.5
277
+ threadpoolctl==3.6.0
278
+ thriftpy2 @ file:///rapids/thriftpy2-0.5.2-cp312-cp312-linux_x86_64.whl#sha256=c0c7d939bdc5ccf7739154959938445d5fdf72f1d6e12bd563947a36f67e2147
279
+ tinycss2==1.4.0
280
+ toolz @ file:///rapids/toolz-1.0.0-py3-none-any.whl#sha256=292c8f1c4e7516bf9086f8850935c799a874039c8bcf959d47b600e4c44a6236
281
+ torch @ file:///opt/transfer/torch-2.8.0a0%2B5228986c39.nv25.6-29946474-cp312-cp312-linux_x86_64.whl#sha256=7a3c9ff5288d8102a31296b3d62491f3f47c8820a5c9aaefc7dc0fe20e1e8ad5
282
+ torch_tensorrt @ file:///opt/pytorch/torch_tensorrt/dist/torch_tensorrt-2.8.0a0-cp312-cp312-linux_x86_64.whl#sha256=26f4d7a04e7e6e017a5b917b8b98b83720bb2bb0195f48d25f52d567d286bfe1
283
+ torchao @ file:///opt/pytorch/ao
284
+ torchprofile==0.0.4
285
+ torchvision @ file:///opt/transfer/torchvision-0.22.0a0%2B95f10a4e-29946474-cp312-cp312-manylinux_2_28_x86_64.whl#sha256=746a3c674cc61680a86f4a14b374b21af1b5c21f629c42765b40c9494f6b234d
286
+ tornado==6.5.1
287
+ tqdm==4.67.1
288
+ traitlets==5.14.3
289
+ transformer_engine @ file:///pvcdata/aiinfra/b-aghosh/wheels/transformer_engine-2.10.0.dev0%2B10f6128d-cp312-cp312-linux_x86_64.whl#sha256=35cf5f4b83b2baad864e12e512bb422e0929409ac05610812fa521d4942d0189
290
+ treelite @ file:///rapids/treelite-4.4.1-py3-none-linux_x86_64.whl#sha256=174741724205540a1a5b37da81483aeff55d89548f42b3461cf07acac7ae96eb
291
+ typer==0.16.0
292
+ types-dataclasses==0.6.6
293
+ types-python-dateutil==2.9.0.20250516
294
+ typing-inspection==0.4.1
295
+ typing_extensions==4.14.0
296
+ tzdata @ file:///rapids/tzdata-2025.2-py2.py3-none-any.whl#sha256=1a403fada01ff9221ca8044d701868fa132215d84beb92242d9acd2147f667a8
297
+ ucx-py @ file:///rapids/ucx_py-0.43.0-cp312-cp312-linux_x86_64.whl#sha256=4a452fd8d8c983283ada3d49086e3657dc2ee94fbad08efdd708d15ab6e0b73a
298
+ ucxx @ file:///rapids/ucxx-0.43.0-cp312-cp312-linux_x86_64.whl#sha256=f1b4e55cb45c66fe88a2d8f9c2b87751165f3cfec21fbdc9c47c4ddff75e8a5f
299
+ -e git+https://github.com/Dots-Infra/UltraEP.git@94cab099b44fffa99a82fea99e7c12d89cf65e4f#egg=ultra_ep
300
+ uri-template==1.3.0
301
+ urllib3==2.0.7
302
+ wasabi==1.1.3
303
+ wcwidth==0.2.13
304
+ weasel==0.4.1
305
+ webcolors==24.11.1
306
+ webencodings==0.5.1
307
+ websocket-client==1.8.0
308
+ Werkzeug==3.1.3
309
+ wheel==0.45.1
310
+ wrapt==1.17.2
311
+ xdoctest==1.0.2
312
+ xgboost @ file:///rapids/xgboost-2.1.4-py3-none-manylinux_2_39_x86_64.whl#sha256=25eeca27f4048fc491940b3a0b4b5782ee48c33cb3aec059cef71f8da542930b
313
+ xxhash==4.0.1
314
+ yarl @ file:///rapids/yarl-1.20.0-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl#sha256=bf099e2432131093cc611623e0b0bcc399b8cddd9a91eded8bfb50402ec35018
315
+ zarr @ file:///rapids/zarr-2.18.7-py3-none-any.whl#sha256=ac3dc4033e9ae4e9d7b5e27c97ea3eaf1003cc0a07f010bd83d5134bf8c4b223
316
+ zict @ file:///rapids/zict-3.0.0-py2.py3-none-any.whl#sha256=5796e36bd0e0cc8cf0fbc1ace6a68912611c1dbd74750a3f3026b9b9d6a327ae
317
+ zipp==3.22.0
granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/provenance.txt ADDED
@@ -0,0 +1,90 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ==============================================================
2
+ PROVENANCE: moonep__tierbtrace__gbs16
3
+ ==============================================================
4
+ captured_at : 2026-09-21T05:53:11+00:00
5
+ hostname : sp-aiinfra-interactive-aghosh-8-mb99k-8xb200-moe-cg-0
6
+ kernel : 6.8.0-1043-nvidia
7
+ cwd : /workspace/blackwell_run_scripts
8
+ command : run_ladder.sh --phase tierbtrace --gbs 16 --out /workspace/granularity_next_b200/tiera_topk_trace/moonep/tierbtrace --nsys
9
+
10
+ ---- GPU / driver --------------------------------------------
11
+ 0, NVIDIA B200, 10.0, 183359 MiB, 580.95.05
12
+ 1, NVIDIA B200, 10.0, 183359 MiB, 580.95.05
13
+ 2, NVIDIA B200, 10.0, 183359 MiB, 580.95.05
14
+ 3, NVIDIA B200, 10.0, 183359 MiB, 580.95.05
15
+ 4, NVIDIA B200, 10.0, 183359 MiB, 580.95.05
16
+ 5, NVIDIA B200, 10.0, 183359 MiB, 580.95.05
17
+ 6, NVIDIA B200, 10.0, 183359 MiB, 580.95.05
18
+ 7, NVIDIA B200, 10.0, 183359 MiB, 580.95.05
19
+ persistence/clocks:
20
+ 0, 120 MHz, 1965 MHz, 1000.00 W, 29
21
+ 1, 1965 MHz, 1965 MHz, 1000.00 W, 35
22
+ 2, 120 MHz, 1965 MHz, 1000.00 W, 34
23
+ 3, 120 MHz, 1965 MHz, 1000.00 W, 30
24
+ 4, 1965 MHz, 1965 MHz, 1000.00 W, 30
25
+ 5, 120 MHz, 1965 MHz, 1000.00 W, 33
26
+ 6, 120 MHz, 1965 MHz, 1000.00 W, 33
27
+ 7, 120 MHz, 1965 MHz, 1000.00 W, 29
28
+
29
+ NVLink topology:
30
+ GPU0 GPU1 GPU2 GPU3 GPU4 GPU5 GPU6 GPU7 NIC0 NIC1 NIC2 NIC3 NIC4 NIC5 NIC6 NIC7 NIC8 CPU Affinity NUMA Affinity GPU NUMA ID
31
+ GPU0 X NV18 NV18 NV18 NV18 NV18 NV18 NV18 PIX NODE SYS SYS SYS SYS SYS SYS NODE 0-27 0 N/A
32
+ GPU1 NV18 X NV18 NV18 NV18 NV18 NV18 NV18 NODE PIX SYS SYS SYS SYS SYS SYS NODE 0-27 0 N/A
33
+ GPU2 NV18 NV18 X NV18 NV18 NV18 NV18 NV18 SYS SYS PIX NODE SYS SYS SYS SYS SYS 28-55 1 N/A
34
+ GPU3 NV18 NV18 NV18 X NV18 NV18 NV18 NV18 SYS SYS NODE PIX SYS SYS SYS SYS SYS 28-55 1 N/A
35
+ GPU4 NV18 NV18 NV18 NV18 X NV18 NV18 NV18 SYS SYS SYS SYS PIX NODE SYS SYS SYS 56-83 2 N/A
36
+ GPU5 NV18 NV18 NV18 NV18 NV18 X NV18 NV18 SYS SYS SYS SYS NODE PIX SYS SYS SYS 56-83 2 N/A
37
+ GPU6 NV18 NV18 NV18 NV18 NV18 NV18 X NV18 SYS SYS SYS SYS SYS SYS PIX NODE SYS 84-111 3 N/A
38
+ GPU7 NV18 NV18 NV18 NV18 NV18 NV18 NV18 X SYS SYS SYS SYS SYS SYS NODE PIX SYS 84-111 3 N/A
39
+ NIC0 PIX NODE SYS SYS SYS SYS SYS SYS X NODE SYS SYS SYS SYS SYS SYS NODE
40
+ NIC1 NODE PIX SYS SYS SYS SYS SYS SYS NODE X SYS SYS SYS SYS SYS SYS NODE
41
+ NIC2 SYS SYS PIX NODE SYS SYS SYS SYS SYS SYS X NODE SYS SYS SYS SYS SYS
42
+ NIC3 SYS SYS NODE PIX SYS SYS SYS SYS SYS SYS NODE X SYS SYS SYS SYS SYS
43
+ NIC4 SYS SYS SYS SYS PIX NODE SYS SYS SYS SYS SYS SYS X NODE SYS SYS SYS
44
+ unavailable
45
+
46
+ ---- toolchain -----------------------------------------------
47
+ nvcc : release 12.9, V12.9.86
48
+ gcc : gcc (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0
49
+ python : 3.12.3
50
+ torch : 2.8.0a0+5228986c39.nv25.06 cuda=12.9 nccl=2.27.3
51
+ torch arch : ['sm_75', 'sm_80', 'sm_86', 'sm_90', 'sm_100', 'sm_120', 'compute_120']
52
+ transformer_engine: 2.10.0.dev0+10f6128d [/usr/local/lib/python3.12/dist-packages/transformer_engine/__init__.py]
53
+ deep_ep : 1.2.1+34152ae [/workspace/DeepEP/deep_ep/__init__.py]
54
+ ultra_ep : 1.0.0 [/workspace/UltraEP/ultra_ep/__init__.py]
55
+ moonep : 0.0.1 [/workspace/MoonEP/moonep/__init__.py]
56
+ megatron : 0.14.0rc5 [None]
57
+ flash_attn : 2.7.4.post1 [/usr/local/lib/python3.12/dist-packages/flash_attn/__init__.py]
58
+ numpy : 1.26.4 [/usr/local/lib/python3.12/dist-packages/numpy/__init__.py]
59
+ huggingface_hub: 1.32.0 [/usr/local/lib/python3.12/dist-packages/huggingface_hub/__init__.py]
60
+
61
+ ---- repositories --------------------------------------------
62
+ Megatron-MoE-Echo moonep_backend_async_grad_reduce 2e68c8ad6bfb
63
+ Megatron-UltraEP ultraep-bracing-regions-through-kernels 03d57ef7d476
64
+ UltraEP main 94cab099b44f
65
+ DeepEP HEAD 34152ae28f80 (+3 untracked: build output)
66
+ MoonEP megatron-ddp-symmetric-params d8bc6667fad7
67
+ blackwell_run_scripts granularity-moonep d29f6570cdce
68
+
69
+ ---- environment ---------------------------------------------
70
+ CUDA_ARCH_LIST=7.5 8.0 8.6 9.0 10.0 12.0
71
+ CUDA_DRIVER_VERSION=575.57.08
72
+ CUDA_HOME=/usr/local/cuda
73
+ CUDA_MODULE_LOADING=LAZY
74
+ CUDA_VERSION=12.9.1.010
75
+ HF_REMOTE_ROOT=granularity_next_b200/tiera_topk_trace
76
+ NCCL_VERSION=2.27.3
77
+ NVSHMEM_VERSION=3.2.5
78
+ PYTORCH_BUILD_NUMBER=0
79
+ PYTORCH_BUILD_VERSION=2.8.0a0+5228986
80
+ PYTORCH_HOME=/opt/pytorch/pytorch
81
+ PYTORCH_VERSION=2.8.0a0+5228986
82
+ TORCH_ALLOW_TF32_CUBLAS_OVERRIDE=1
83
+ TORCH_CUDA_ARCH_LIST=7.5 8.0 8.6 9.0 10.0 12.0+PTX
84
+ TORCH_NCCL_USE_COMM_NONBLOCKING=0
85
+
86
+ ---- shared library resolution -------------------------------
87
+ libnvshmem_host.so: not in ldconfig
88
+ libnccl.so: /lib/x86_64-linux-gnu/libnccl.so.2
89
+ lrwxrwxrwx 1 root root 20 Sep 21 04:57 /usr/local/lib/python3.12/dist-packages/nvidia/nvshmem/lib/libnvshmem_host.so -> libnvshmem_host.so.3
90
+ -rw-r--r-- 1 root root 179045744 Sep 21 04:54 /usr/local/lib/python3.12/dist-packages/nvidia/nvshmem/lib/libnvshmem_host.so.3
granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G2_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/G2_K2__moonep__matched__seq4096__gbs16__capnone__async.nsys-rep ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:46f5c5d8fdabb8e99368a93ce52f9f8828359b2abf7aac55b20926c4ea767f5f
3
+ size 10170747
granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G2_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/moonep_async_overlap.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "devices": 8,
3
+ "reduce_busy_ms_per_iteration": 32.253625,
4
+ "overlapped_ms_per_iteration": 31.28375,
5
+ "exposed_ms_per_iteration": 0.97,
6
+ "overlap_pct": 96.99297365799968,
7
+ "minimum_overlap_pct": 0.1,
8
+ "status": "PASS"
9
+ }
granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G2_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/moonep_async_overlap.txt ADDED
@@ -0,0 +1,416 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+
2
+ === kernels on device 0, by total time ===
3
+ 66.685 ms device_kernel
4
+ 48.197 ms multi_tensor_apply_kernel
5
+ REDUCE 44.275 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2293764096i64i64_tensorptrf32gmemalign16o2293764096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
6
+ 34.902 ms elementwise_kernel
7
+ 31.943 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o229376409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
8
+ 30.604 ms vectorized_elementwise_kernel
9
+ 28.261 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
10
+ 27.750 ms cast_mxfp8_2D_kernel
11
+ 27.285 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
12
+ 23.376 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o40964096i64i64_tensorptrbf16gmemalign16o737284096i64i64_tensorptri32gmemalign16o81921_tensorptri32gm_0
13
+ 22.885 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
14
+ 22.634 ms ncclDevKernel_AllGather_RING_LL
15
+ REDUCE 19.725 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o1310723584i64i64_tensorptrf32gmemalign16o1310723584i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
16
+ 16.416 ms unrolled_elementwise_kernel
17
+ 16.312 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o737284096i64i64_tensorptrbf16gmemalign16o40964096i64i64_tensorptri32gmemalign16o81921_tensorptri32gmem_0
18
+ 15.355 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o131072358411i640_CopyAtom_ThrID10_TVLayoutSrc_0
19
+ 9.074 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NNT
20
+ 8.242 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_TNT
21
+ 7.713 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
22
+ 6.895 ms triton_poi_fused__to_copy_cat_0
23
+ 5.397 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_badd_Avec32UE8M0_Bvec32UE8M0_NTT
24
+ 5.202 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NTT
25
+ 4.753 ms setGroupedGemmWgradArguments
26
+ 3.980 ms rmsnorm_bwd_tuned_kernel
27
+ 3.902 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
28
+ 3.812 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
29
+ 3.617 ms CatArrayBatchedCopy_alignedK_contig
30
+ 3.561 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_321_tensorptri32_gmem_align16_o_81921_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_alig_0
31
+ 3.447 ms reduce_kernel
32
+ 3.435 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
33
+ 3.048 ms fused_rope_backward_kernel
34
+ 2.912 ms _scatter_gather_elementwise_kernel
35
+ 2.550 ms triton_red_fused__to_copy_mul_silu_sum_1
36
+ 2.544 ms multi_tensor_swizzle_col_scaling_kernel
37
+ 2.472 ms multi_tensor_swizzle_row_scaling_kernel
38
+ 2.447 ms fused_rope_forward_kernel
39
+ 2.247 ms swizzle_col_scaling_kernel
40
+ 1.995 ms triton_poi_fused__to_copy_mul_silu_0
41
+ 1.863 ms rmsnorm_fwd_tuned_kernel
42
+ 1.793 ms triton_poi_fused_add_0
43
+
44
+ === G2_K2__moonep__matched__seq4096__gbs16__capnone__async.sqlite (device 0, 2 iteration(s)) ===
45
+ reduce_grad kernels : 128 (64/iter)
46
+ reduce_grad sum : 32.000 ms/iter
47
+ reduce_grad busy (union) : 32.000 ms/iter
48
+ ... overlapped with compute: 31.062 ms/iter ( 97.1 %)
49
+ ... exposed on critical path: 0.938 ms/iter ( 2.9 %)
50
+ other GPU work (union) : 246.977 ms/iter
51
+ traced wall : 271.346 ms/iter
52
+ --> time hidden by overlap : 31.062 ms/iter
53
+
54
+ === kernels on device 1, by total time ===
55
+ 66.141 ms device_kernel
56
+ 48.256 ms multi_tensor_apply_kernel
57
+ REDUCE 44.267 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2293764096i64i64_tensorptrf32gmemalign16o2293764096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
58
+ 34.960 ms elementwise_kernel
59
+ 32.852 ms vectorized_elementwise_kernel
60
+ 29.110 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
61
+ 28.210 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
62
+ 27.986 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o229376409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
63
+ 27.718 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
64
+ 27.690 ms cast_mxfp8_2D_kernel
65
+ 23.301 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o40964096i64i64_tensorptrbf16gmemalign16o737284096i64i64_tensorptri32gmemalign16o81921_tensorptri32gm_0
66
+ REDUCE 20.471 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o1310723584i64i64_tensorptrf32gmemalign16o1310723584i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
67
+ 16.502 ms unrolled_elementwise_kernel
68
+ 15.526 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o737284096i64i64_tensorptrbf16gmemalign16o40964096i64i64_tensorptri32gmemalign16o81921_tensorptri32gmem_0
69
+ 14.047 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o131072358411i640_CopyAtom_ThrID10_TVLayoutSrc_0
70
+ 9.085 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NNT
71
+ 8.706 ms ncclDevKernel_AllGather_RING_LL
72
+ 8.185 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_TNT
73
+ 7.625 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
74
+ 6.892 ms triton_poi_fused__to_copy_cat_0
75
+ 5.331 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_badd_Avec32UE8M0_Bvec32UE8M0_NTT
76
+ 5.155 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NTT
77
+ 4.907 ms setGroupedGemmWgradArguments
78
+ 3.946 ms rmsnorm_bwd_tuned_kernel
79
+ 3.795 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
80
+ 3.769 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
81
+ 3.665 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_321_tensorptri32_gmem_align16_o_81921_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_alig_0
82
+ 3.611 ms CatArrayBatchedCopy_alignedK_contig
83
+ 3.482 ms reduce_kernel
84
+ 3.373 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
85
+ 3.066 ms fused_rope_backward_kernel
86
+ 2.948 ms _scatter_gather_elementwise_kernel
87
+ 2.542 ms triton_red_fused__to_copy_mul_silu_sum_1
88
+ 2.541 ms multi_tensor_swizzle_col_scaling_kernel
89
+ 2.460 ms multi_tensor_swizzle_row_scaling_kernel
90
+ 2.452 ms fused_rope_forward_kernel
91
+ 2.274 ms swizzle_col_scaling_kernel
92
+ 2.161 ms ncclDevKernel_AllReduce_Sum_f32_RING_LL
93
+ 1.985 ms triton_poi_fused__to_copy_mul_silu_0
94
+ 1.864 ms rmsnorm_fwd_tuned_kernel
95
+
96
+ === G2_K2__moonep__matched__seq4096__gbs16__capnone__async.sqlite (device 1, 2 iteration(s)) ===
97
+ reduce_grad kernels : 128 (64/iter)
98
+ reduce_grad sum : 32.369 ms/iter
99
+ reduce_grad busy (union) : 32.369 ms/iter
100
+ ... overlapped with compute: 31.410 ms/iter ( 97.0 %)
101
+ ... exposed on critical path: 0.959 ms/iter ( 3.0 %)
102
+ other GPU work (union) : 241.764 ms/iter
103
+ traced wall : 312.471 ms/iter
104
+ --> time hidden by overlap : 31.410 ms/iter
105
+
106
+ === kernels on device 2, by total time ===
107
+ 67.181 ms device_kernel
108
+ 48.180 ms multi_tensor_apply_kernel
109
+ REDUCE 44.304 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2293764096i64i64_tensorptrf32gmemalign16o2293764096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
110
+ 34.903 ms elementwise_kernel
111
+ 32.616 ms vectorized_elementwise_kernel
112
+ 31.408 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o229376409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
113
+ 28.229 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
114
+ 27.708 ms cast_mxfp8_2D_kernel
115
+ 27.507 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
116
+ 23.516 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
117
+ 23.244 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o40964096i64i64_tensorptrbf16gmemalign16o737284096i64i64_tensorptri32gmemalign16o81921_tensorptri32gm_0
118
+ REDUCE 20.025 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o1310723584i64i64_tensorptrf32gmemalign16o1310723584i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
119
+ 16.447 ms unrolled_elementwise_kernel
120
+ 16.013 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o737284096i64i64_tensorptrbf16gmemalign16o40964096i64i64_tensorptri32gmemalign16o81921_tensorptri32gmem_0
121
+ 15.468 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o131072358411i640_CopyAtom_ThrID10_TVLayoutSrc_0
122
+ 10.800 ms ncclDevKernel_AllGather_RING_LL
123
+ 9.085 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NNT
124
+ 8.222 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_TNT
125
+ 7.656 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
126
+ 6.894 ms triton_poi_fused__to_copy_cat_0
127
+ 5.331 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_badd_Avec32UE8M0_Bvec32UE8M0_NTT
128
+ 5.151 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NTT
129
+ 4.644 ms setGroupedGemmWgradArguments
130
+ 3.964 ms rmsnorm_bwd_tuned_kernel
131
+ 3.903 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
132
+ 3.805 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
133
+ 3.632 ms CatArrayBatchedCopy_alignedK_contig
134
+ 3.631 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_321_tensorptri32_gmem_align16_o_81921_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_alig_0
135
+ 3.446 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
136
+ 3.440 ms reduce_kernel
137
+ 3.113 ms fused_rope_backward_kernel
138
+ 2.923 ms _scatter_gather_elementwise_kernel
139
+ 2.861 ms ncclDevKernel_AllReduce_Sum_f32_RING_LL
140
+ 2.546 ms multi_tensor_swizzle_col_scaling_kernel
141
+ 2.544 ms triton_red_fused__to_copy_mul_silu_sum_1
142
+ 2.458 ms multi_tensor_swizzle_row_scaling_kernel
143
+ 2.446 ms fused_rope_forward_kernel
144
+ 2.254 ms swizzle_col_scaling_kernel
145
+ 1.997 ms triton_poi_fused__to_copy_mul_silu_0
146
+ 1.864 ms rmsnorm_fwd_tuned_kernel
147
+
148
+ === G2_K2__moonep__matched__seq4096__gbs16__capnone__async.sqlite (device 2, 2 iteration(s)) ===
149
+ reduce_grad kernels : 128 (64/iter)
150
+ reduce_grad sum : 32.164 ms/iter
151
+ reduce_grad busy (union) : 32.164 ms/iter
152
+ ... overlapped with compute: 31.177 ms/iter ( 96.9 %)
153
+ ... exposed on critical path: 0.988 ms/iter ( 3.1 %)
154
+ other GPU work (union) : 243.149 ms/iter
155
+ traced wall : 309.441 ms/iter
156
+ --> time hidden by overlap : 31.177 ms/iter
157
+
158
+ === kernels on device 3, by total time ===
159
+ 68.456 ms device_kernel
160
+ 48.206 ms multi_tensor_apply_kernel
161
+ REDUCE 45.190 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2293764096i64i64_tensorptrf32gmemalign16o2293764096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
162
+ 36.199 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o229376409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
163
+ 34.980 ms elementwise_kernel
164
+ 30.497 ms vectorized_elementwise_kernel
165
+ 28.332 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
166
+ 27.735 ms cast_mxfp8_2D_kernel
167
+ 27.722 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
168
+ 23.231 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o40964096i64i64_tensorptrbf16gmemalign16o737284096i64i64_tensorptri32gmemalign16o81921_tensorptri32gm_0
169
+ REDUCE 19.310 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o1310723584i64i64_tensorptrf32gmemalign16o1310723584i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
170
+ 17.351 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o131072358411i640_CopyAtom_ThrID10_TVLayoutSrc_0
171
+ 16.458 ms unrolled_elementwise_kernel
172
+ 15.601 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o737284096i64i64_tensorptrbf16gmemalign16o40964096i64i64_tensorptri32gmemalign16o81921_tensorptri32gmem_0
173
+ 15.246 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
174
+ 11.119 ms ncclDevKernel_AllGather_RING_LL
175
+ 9.106 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NNT
176
+ 8.249 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_TNT
177
+ 7.734 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
178
+ 6.901 ms triton_poi_fused__to_copy_cat_0
179
+ 5.390 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_badd_Avec32UE8M0_Bvec32UE8M0_NTT
180
+ 5.212 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NTT
181
+ 4.673 ms setGroupedGemmWgradArguments
182
+ 3.969 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
183
+ 3.968 ms rmsnorm_bwd_tuned_kernel
184
+ 3.866 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
185
+ 3.653 ms CatArrayBatchedCopy_alignedK_contig
186
+ 3.635 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_321_tensorptri32_gmem_align16_o_81921_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_alig_0
187
+ 3.485 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
188
+ 3.450 ms reduce_kernel
189
+ 3.019 ms fused_rope_backward_kernel
190
+ 2.921 ms _scatter_gather_elementwise_kernel
191
+ 2.899 ms ncclDevKernel_AllReduce_Sum_f32_RING_LL
192
+ 2.548 ms triton_red_fused__to_copy_mul_silu_sum_1
193
+ 2.546 ms multi_tensor_swizzle_col_scaling_kernel
194
+ 2.464 ms multi_tensor_swizzle_row_scaling_kernel
195
+ 2.441 ms fused_rope_forward_kernel
196
+ 2.273 ms swizzle_col_scaling_kernel
197
+ 1.987 ms triton_poi_fused__to_copy_mul_silu_0
198
+ 1.866 ms rmsnorm_fwd_tuned_kernel
199
+
200
+ === G2_K2__moonep__matched__seq4096__gbs16__capnone__async.sqlite (device 3, 2 iteration(s)) ===
201
+ reduce_grad kernels : 128 (64/iter)
202
+ reduce_grad sum : 32.250 ms/iter
203
+ reduce_grad busy (union) : 32.250 ms/iter
204
+ ... overlapped with compute: 31.261 ms/iter ( 96.9 %)
205
+ ... exposed on critical path: 0.989 ms/iter ( 3.1 %)
206
+ other GPU work (union) : 242.329 ms/iter
207
+ traced wall : 271.577 ms/iter
208
+ --> time hidden by overlap : 31.261 ms/iter
209
+
210
+ === kernels on device 4, by total time ===
211
+ 66.606 ms device_kernel
212
+ 48.184 ms multi_tensor_apply_kernel
213
+ REDUCE 44.177 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2293764096i64i64_tensorptrf32gmemalign16o2293764096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
214
+ 34.918 ms elementwise_kernel
215
+ 30.654 ms vectorized_elementwise_kernel
216
+ 30.328 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o229376409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
217
+ 28.301 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
218
+ 27.759 ms cast_mxfp8_2D_kernel
219
+ 27.638 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
220
+ 25.485 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
221
+ 23.272 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o40964096i64i64_tensorptrbf16gmemalign16o737284096i64i64_tensorptri32gmemalign16o81921_tensorptri32gm_0
222
+ REDUCE 20.241 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o1310723584i64i64_tensorptrf32gmemalign16o1310723584i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
223
+ 16.449 ms unrolled_elementwise_kernel
224
+ 16.061 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o737284096i64i64_tensorptrbf16gmemalign16o40964096i64i64_tensorptri32gmemalign16o81921_tensorptri32gmem_0
225
+ 14.700 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o131072358411i640_CopyAtom_ThrID10_TVLayoutSrc_0
226
+ 10.604 ms ncclDevKernel_AllGather_RING_LL
227
+ 9.051 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NNT
228
+ 8.248 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_TNT
229
+ 7.720 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
230
+ 6.899 ms triton_poi_fused__to_copy_cat_0
231
+ 5.347 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_badd_Avec32UE8M0_Bvec32UE8M0_NTT
232
+ 5.165 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NTT
233
+ 4.673 ms setGroupedGemmWgradArguments
234
+ 3.961 ms rmsnorm_bwd_tuned_kernel
235
+ 3.860 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
236
+ 3.812 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
237
+ 3.667 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_321_tensorptri32_gmem_align16_o_81921_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_alig_0
238
+ 3.624 ms CatArrayBatchedCopy_alignedK_contig
239
+ 3.438 ms reduce_kernel
240
+ 3.413 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
241
+ 3.043 ms fused_rope_backward_kernel
242
+ 2.936 ms _scatter_gather_elementwise_kernel
243
+ 2.552 ms triton_red_fused__to_copy_mul_silu_sum_1
244
+ 2.537 ms multi_tensor_swizzle_col_scaling_kernel
245
+ 2.446 ms fused_rope_forward_kernel
246
+ 2.435 ms multi_tensor_swizzle_row_scaling_kernel
247
+ 2.252 ms swizzle_col_scaling_kernel
248
+ 1.999 ms triton_poi_fused__to_copy_mul_silu_0
249
+ 1.864 ms rmsnorm_fwd_tuned_kernel
250
+ 1.803 ms triton_poi_fused_add_0
251
+
252
+ === G2_K2__moonep__matched__seq4096__gbs16__capnone__async.sqlite (device 4, 2 iteration(s)) ===
253
+ reduce_grad kernels : 128 (64/iter)
254
+ reduce_grad sum : 32.209 ms/iter
255
+ reduce_grad busy (union) : 32.209 ms/iter
256
+ ... overlapped with compute: 31.222 ms/iter ( 96.9 %)
257
+ ... exposed on critical path: 0.987 ms/iter ( 3.1 %)
258
+ other GPU work (union) : 240.788 ms/iter
259
+ traced wall : 271.415 ms/iter
260
+ --> time hidden by overlap : 31.222 ms/iter
261
+
262
+ === kernels on device 5, by total time ===
263
+ 67.326 ms device_kernel
264
+ 48.194 ms multi_tensor_apply_kernel
265
+ REDUCE 44.518 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2293764096i64i64_tensorptrf32gmemalign16o2293764096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
266
+ 34.946 ms elementwise_kernel
267
+ 31.874 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o229376409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
268
+ 30.475 ms vectorized_elementwise_kernel
269
+ 28.357 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
270
+ 27.742 ms cast_mxfp8_2D_kernel
271
+ 27.614 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
272
+ 23.251 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o40964096i64i64_tensorptrbf16gmemalign16o737284096i64i64_tensorptri32gmemalign16o81921_tensorptri32gm_0
273
+ 22.813 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
274
+ REDUCE 20.448 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o1310723584i64i64_tensorptrf32gmemalign16o1310723584i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
275
+ 16.424 ms unrolled_elementwise_kernel
276
+ 15.624 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o737284096i64i64_tensorptrbf16gmemalign16o40964096i64i64_tensorptri32gmemalign16o81921_tensorptri32gmem_0
277
+ 15.251 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o131072358411i640_CopyAtom_ThrID10_TVLayoutSrc_0
278
+ 10.571 ms ncclDevKernel_AllGather_RING_LL
279
+ 9.085 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NNT
280
+ 8.232 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_TNT
281
+ 7.712 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
282
+ 6.901 ms triton_poi_fused__to_copy_cat_0
283
+ 5.361 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_badd_Avec32UE8M0_Bvec32UE8M0_NTT
284
+ 5.167 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NTT
285
+ 4.641 ms setGroupedGemmWgradArguments
286
+ 3.961 ms rmsnorm_bwd_tuned_kernel
287
+ 3.903 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
288
+ 3.809 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
289
+ 3.634 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_321_tensorptri32_gmem_align16_o_81921_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_alig_0
290
+ 3.628 ms CatArrayBatchedCopy_alignedK_contig
291
+ 3.443 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
292
+ 3.438 ms reduce_kernel
293
+ 3.055 ms fused_rope_backward_kernel
294
+ 2.935 ms _scatter_gather_elementwise_kernel
295
+ 2.928 ms ncclDevKernel_AllReduce_Sum_f32_RING_LL
296
+ 2.545 ms multi_tensor_swizzle_col_scaling_kernel
297
+ 2.545 ms triton_red_fused__to_copy_mul_silu_sum_1
298
+ 2.451 ms fused_rope_forward_kernel
299
+ 2.442 ms multi_tensor_swizzle_row_scaling_kernel
300
+ 2.259 ms swizzle_col_scaling_kernel
301
+ 1.995 ms triton_poi_fused__to_copy_mul_silu_0
302
+ 1.871 ms rmsnorm_fwd_tuned_kernel
303
+
304
+ === G2_K2__moonep__matched__seq4096__gbs16__capnone__async.sqlite (device 5, 2 iteration(s)) ===
305
+ reduce_grad kernels : 128 (64/iter)
306
+ reduce_grad sum : 32.483 ms/iter
307
+ reduce_grad busy (union) : 32.483 ms/iter
308
+ ... overlapped with compute: 31.488 ms/iter ( 96.9 %)
309
+ ... exposed on critical path: 0.995 ms/iter ( 3.1 %)
310
+ other GPU work (union) : 242.056 ms/iter
311
+ traced wall : 271.272 ms/iter
312
+ --> time hidden by overlap : 31.488 ms/iter
313
+
314
+ === kernels on device 6, by total time ===
315
+ 68.748 ms device_kernel
316
+ 48.208 ms multi_tensor_apply_kernel
317
+ REDUCE 44.146 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2293764096i64i64_tensorptrf32gmemalign16o2293764096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
318
+ 34.901 ms elementwise_kernel
319
+ 32.899 ms vectorized_elementwise_kernel
320
+ 29.421 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o229376409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
321
+ 28.526 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
322
+ 27.776 ms cast_mxfp8_2D_kernel
323
+ 27.139 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
324
+ 24.122 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
325
+ 23.299 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o40964096i64i64_tensorptrbf16gmemalign16o737284096i64i64_tensorptri32gmemalign16o81921_tensorptri32gm_0
326
+ REDUCE 20.192 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o1310723584i64i64_tensorptrf32gmemalign16o1310723584i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
327
+ 16.506 ms unrolled_elementwise_kernel
328
+ 16.123 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o737284096i64i64_tensorptrbf16gmemalign16o40964096i64i64_tensorptri32gmemalign16o81921_tensorptri32gmem_0
329
+ 14.071 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o131072358411i640_CopyAtom_ThrID10_TVLayoutSrc_0
330
+ 9.122 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NNT
331
+ 8.480 ms ncclDevKernel_AllGather_RING_LL
332
+ 8.285 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_TNT
333
+ 7.785 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
334
+ 6.900 ms triton_poi_fused__to_copy_cat_0
335
+ 5.398 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_badd_Avec32UE8M0_Bvec32UE8M0_NTT
336
+ 5.220 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NTT
337
+ 4.608 ms setGroupedGemmWgradArguments
338
+ 4.017 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
339
+ 3.952 ms rmsnorm_bwd_tuned_kernel
340
+ 3.860 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
341
+ 3.677 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_321_tensorptri32_gmem_align16_o_81921_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_alig_0
342
+ 3.616 ms CatArrayBatchedCopy_alignedK_contig
343
+ 3.471 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
344
+ 3.468 ms reduce_kernel
345
+ 3.014 ms fused_rope_backward_kernel
346
+ 2.923 ms _scatter_gather_elementwise_kernel
347
+ 2.690 ms ncclDevKernel_AllReduce_Sum_f32_RING_LL
348
+ 2.557 ms multi_tensor_swizzle_col_scaling_kernel
349
+ 2.548 ms triton_red_fused__to_copy_mul_silu_sum_1
350
+ 2.448 ms fused_rope_forward_kernel
351
+ 2.437 ms multi_tensor_swizzle_row_scaling_kernel
352
+ 2.267 ms swizzle_col_scaling_kernel
353
+ 1.984 ms triton_poi_fused__to_copy_mul_silu_0
354
+ 1.877 ms rmsnorm_fwd_tuned_kernel
355
+
356
+ === G2_K2__moonep__matched__seq4096__gbs16__capnone__async.sqlite (device 6, 2 iteration(s)) ===
357
+ reduce_grad kernels : 128 (64/iter)
358
+ reduce_grad sum : 32.169 ms/iter
359
+ reduce_grad busy (union) : 32.169 ms/iter
360
+ ... overlapped with compute: 31.230 ms/iter ( 97.1 %)
361
+ ... exposed on critical path: 0.939 ms/iter ( 2.9 %)
362
+ other GPU work (union) : 242.180 ms/iter
363
+ traced wall : 321.952 ms/iter
364
+ --> time hidden by overlap : 31.230 ms/iter
365
+
366
+ === kernels on device 7, by total time ===
367
+ 65.621 ms device_kernel
368
+ 48.219 ms multi_tensor_apply_kernel
369
+ REDUCE 45.002 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2293764096i64i64_tensorptrf32gmemalign16o2293764096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
370
+ 34.982 ms elementwise_kernel
371
+ 34.259 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o229376409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
372
+ 32.657 ms vectorized_elementwise_kernel
373
+ 28.118 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
374
+ 27.893 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
375
+ 27.759 ms cast_mxfp8_2D_kernel
376
+ 23.278 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o40964096i64i64_tensorptrbf16gmemalign16o737284096i64i64_tensorptri32gmemalign16o81921_tensorptri32gm_0
377
+ 20.417 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
378
+ REDUCE 19.768 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o1310723584i64i64_tensorptrf32gmemalign16o1310723584i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
379
+ 17.393 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o131072358411i640_CopyAtom_ThrID10_TVLayoutSrc_0
380
+ 16.446 ms unrolled_elementwise_kernel
381
+ 15.609 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o737284096i64i64_tensorptrbf16gmemalign16o40964096i64i64_tensorptri32gmemalign16o81921_tensorptri32gmem_0
382
+ 10.348 ms ncclDevKernel_AllGather_RING_LL
383
+ 9.089 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NNT
384
+ 8.190 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_TNT
385
+ 7.729 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
386
+ 6.879 ms triton_poi_fused__to_copy_cat_0
387
+ 5.365 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_badd_Avec32UE8M0_Bvec32UE8M0_NTT
388
+ 5.167 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NTT
389
+ 4.685 ms setGroupedGemmWgradArguments
390
+ 3.945 ms rmsnorm_bwd_tuned_kernel
391
+ 3.846 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
392
+ 3.823 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
393
+ 3.661 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_321_tensorptri32_gmem_align16_o_81921_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_alig_0
394
+ 3.631 ms CatArrayBatchedCopy_alignedK_contig
395
+ 3.472 ms reduce_kernel
396
+ 3.395 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
397
+ 2.986 ms fused_rope_backward_kernel
398
+ 2.949 ms _scatter_gather_elementwise_kernel
399
+ 2.586 ms multi_tensor_swizzle_row_scaling_kernel
400
+ 2.552 ms multi_tensor_swizzle_col_scaling_kernel
401
+ 2.541 ms triton_red_fused__to_copy_mul_silu_sum_1
402
+ 2.447 ms fused_rope_forward_kernel
403
+ 2.420 ms ncclDevKernel_AllReduce_Sum_f32_RING_LL
404
+ 2.284 ms swizzle_col_scaling_kernel
405
+ 1.988 ms triton_poi_fused__to_copy_mul_silu_0
406
+ 1.872 ms rmsnorm_fwd_tuned_kernel
407
+
408
+ === G2_K2__moonep__matched__seq4096__gbs16__capnone__async.sqlite (device 7, 2 iteration(s)) ===
409
+ reduce_grad kernels : 128 (64/iter)
410
+ reduce_grad sum : 32.385 ms/iter
411
+ reduce_grad busy (union) : 32.385 ms/iter
412
+ ... overlapped with compute: 31.420 ms/iter ( 97.0 %)
413
+ ... exposed on critical path: 0.965 ms/iter ( 3.0 %)
414
+ other GPU work (union) : 242.770 ms/iter
415
+ traced wall : 341.748 ms/iter
416
+ --> time hidden by overlap : 31.420 ms/iter
granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G2_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/moonep_async_overlap_summary.txt ADDED
@@ -0,0 +1 @@
 
 
1
+ devices=8 reduce=32.254 ms/iter overlap=31.284 ms/iter (97.0%) exposed=0.970 ms/iter status=PASS
granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G2_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/run.env ADDED
@@ -0,0 +1,105 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ backend=moonep
2
+ config=G2_K2
3
+ family=tierA
4
+ source_model=
5
+ standard_num_layers=
6
+ base_rung=G2
7
+ repo=/workspace/Megatron-MoE-Echo
8
+ branch=moonep_backend_async_grad_reduce
9
+ commit=2e68c8ad
10
+ moonep_source_dir=/workspace/MoonEP
11
+ moonep_source_branch=megatron-ddp-symmetric-params
12
+ moonep_source_commit=d8bc666
13
+ num_layers=16
14
+ hidden=4096
15
+ ffn_hidden=14336
16
+ moe_ffn_hidden=3584
17
+ num_experts=32
18
+ topk=2
19
+ experts_per_rank=4
20
+ expert_params=44040192
21
+ heads=32
22
+ kv_groups=8
23
+ kv_channels=
24
+ qk_layernorm=0
25
+ shared_ffn=0
26
+ shared_expert_overlap=0
27
+ moe_token_dispatcher=flex
28
+ mla=0
29
+ q_lora_rank=
30
+ kv_lora_rank=512
31
+ qk_head_dim=128
32
+ qk_pos_emb_head_dim=64
33
+ v_head_dim=128
34
+ dense_first_layers=0
35
+ moe_layers=16
36
+ rotary_base=1000000
37
+ rotary_percent=1.0
38
+ rotary_scaling=
39
+ mscale=
40
+ mscale_all_dim=
41
+ qkv_bias=0
42
+ router_balancing=aux_loss
43
+ aux_loss_coeff=1e-2
44
+ router_pre_softmax=0
45
+ router_fusion=0
46
+ router_dtype=fp32
47
+ router_groups=
48
+ router_group_topk=
49
+ router_score_function=softmax
50
+ router_expert_bias=0
51
+ router_bias_update_rate=
52
+ router_topk_scaling_factor=
53
+ shared_expert_gate=0
54
+ apply_layernorm_1p=0
55
+ attention_output_gate=0
56
+ linear_attention_type=
57
+ linear_attention_freq=
58
+ linear_conv_kernel_dim=
59
+ linear_key_head_dim=
60
+ linear_value_head_dim=
61
+ linear_num_key_heads=
62
+ linear_num_value_heads=
63
+ mtp_num_layers=0
64
+ mtp_loss_scaling_factor=0.1
65
+ no_weight_decay_cond_type=
66
+ max_position_embeddings=32768
67
+ init_method_std=0.01
68
+ device_initiated_grouped_gemm=1
69
+ norm_epsilon=1e-5
70
+ seq_length=4096
71
+ micro_batch_size=1
72
+ global_batch_size=16
73
+ num_microbatches=2
74
+ assignments_per_rank=8192
75
+ mean_tokens_per_expert=2048
76
+ slot_policy=matched
77
+ slots_per_rank=4
78
+ capacity_mode=none
79
+ capacity=none
80
+ cuda_graph=1
81
+ probe_mode=0
82
+ async_grad_reduce=1
83
+ branch_role=timing
84
+ memory_dump_mode=none
85
+ memory_snapshot_path=
86
+ cuda_graph_warmup_steps=3
87
+ recompute=none
88
+ force_balance=0
89
+ train_iters=10
90
+ ep_size=8
91
+ echo_dispatch_overlap=1
92
+ echo_dispatch_recompute=0
93
+ ultraep_quota_min_tokens=
94
+ ultraep_weight_sync_join=
95
+ profile_enable=1
96
+ kernel_trace=1
97
+ fp8_mode=mxfp8
98
+ fp8_recipe=mxfp8
99
+ training_seed=1234
100
+ learning_rate=1e-4
101
+ min_learning_rate=1.0e-5
102
+ lr_warmup_iters=5
103
+ state_dump_enable=0
104
+ state_dump_dir=
105
+ moonep_assert_no_pending_reduce=1
granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G2_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/train.log ADDED
@@ -0,0 +1,1064 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ==============================================================
2
+ G2_K2 / moonep (slots/rank=4 policy=matched)
3
+ E=32 K=2 H'=3584 m=4
4
+ L=16 seq=4096 gbs=16 (2 microbatches)
5
+ cap_mode=none cap=none cg=1 recompute=none
6
+ memory_dump=none snapshots=/workspace/granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G2_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/memory_snapshots/snapshot.pickle
7
+ repo=/workspace/Megatron-MoE-Echo @ 2e68c8ad
8
+ ==============================================================
9
+ W0921 05:57:41.171000 131368 torch/distributed/run.py:766]
10
+ W0921 05:57:41.171000 131368 torch/distributed/run.py:766] *****************************************
11
+ W0921 05:57:41.171000 131368 torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.
12
+ W0921 05:57:41.171000 131368 torch/distributed/run.py:766] *****************************************
13
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
14
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
15
+ using world size: 8, data-parallel size: 8, context-parallel size: 1, hierarchical context-parallel sizes: None, tensor-model-parallel size: 1, pipeline-model-parallel size: 1
16
+ WARNING: overriding default arguments for tokenizer_type:GPT2BPETokenizer with tokenizer_type:GPT2BPETokenizer
17
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
18
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
19
+ Number of virtual stages per pipeline stage: None
20
+ accumulate and all-reduce gradients in fp32 for bfloat16 data type.
21
+ using torch.bfloat16 for parameters ...
22
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
23
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
24
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
25
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
26
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
27
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
28
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
29
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
30
+ ------------------------ arguments ------------------------
31
+ account_for_embedding_in_pipeline_split ......... False
32
+ account_for_loss_in_pipeline_split .............. False
33
+ accumulate_allreduce_grads_in_fp32 .............. True
34
+ adam_beta1 ...................................... 0.9
35
+ adam_beta2 ...................................... 0.999
36
+ adam_eps ........................................ 1e-08
37
+ add_bias_linear ................................. False
38
+ add_position_embedding .......................... False
39
+ add_qkv_bias .................................... False
40
+ adlr_autoresume ................................. False
41
+ adlr_autoresume_interval ........................ 1000
42
+ align_grad_reduce ............................... True
43
+ align_param_gather .............................. False
44
+ app_tag_run_name ................................ None
45
+ app_tag_run_version ............................. 0.0.0
46
+ apply_layernorm_1p .............................. False
47
+ apply_query_key_layer_scaling ................... False
48
+ apply_residual_connection_post_layernorm ........ False
49
+ apply_rope_fusion ............................... True
50
+ async_save ...................................... None
51
+ async_tensor_model_parallel_allreduce ........... True
52
+ attention_backend ............................... AttnBackend.auto
53
+ attention_dropout ............................... 0.0
54
+ attention_softmax_in_fp32 ....................... False
55
+ auto_detect_ckpt_format ......................... False
56
+ barrier_with_L1_time ............................ True
57
+ bert_binary_head ................................ True
58
+ bert_embedder_type .............................. megatron
59
+ bert_load ....................................... None
60
+ bf16 ............................................ True
61
+ bias_dropout_fusion ............................. True
62
+ bias_gelu_fusion ................................ False
63
+ bias_swiglu_fusion .............................. True
64
+ biencoder_projection_dim ........................ 0
65
+ biencoder_shared_query_context_model ............ False
66
+ block_data_path ................................. None
67
+ calc_ft_timeouts ................................ False
68
+ calculate_per_token_loss ........................ False
69
+ check_for_large_grads ........................... False
70
+ check_for_nan_in_loss_and_grad .................. False
71
+ check_for_spiky_loss ............................ False
72
+ check_weight_hash_across_dp_replicas_interval ... None
73
+ ckpt_assume_constant_structure .................. False
74
+ ckpt_convert_format ............................. None
75
+ ckpt_convert_save ............................... None
76
+ ckpt_convert_update_legacy_dist_opt_format ...... False
77
+ ckpt_format ..................................... torch_dist
78
+ ckpt_fully_parallel_load ........................ False
79
+ ckpt_fully_parallel_save ........................ True
80
+ ckpt_fully_parallel_save_deprecated ............. False
81
+ ckpt_step ....................................... None
82
+ classes_fraction ................................ 1.0
83
+ clip_grad ....................................... 1.0
84
+ clone_scatter_output_in_embedding ............... True
85
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
86
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
87
+ config_logger_dir ...............................
88
+ consumed_train_samples .......................... 0
89
+ consumed_valid_samples .......................... 0
90
+ context_parallel_size ........................... 1
91
+ cp_comm_type .................................... ['p2p']
92
+ create_attention_mask_in_dataloader ............. True
93
+ cross_entropy_fusion_impl ....................... native
94
+ cross_entropy_loss_fusion ....................... False
95
+ cuda_graph_scope ................................ full_iteration
96
+ cuda_graph_warmup_steps ......................... 3
97
+ data_args_path .................................. None
98
+ data_cache_path ................................. None
99
+ data_parallel_random_init ....................... False
100
+ data_parallel_sharding_strategy ................. no_shard
101
+ data_parallel_size .............................. 8
102
+ data_path ....................................... ['/workspace/data/wikitext/wikitext103_gpt2_text_document']
103
+ data_per_class_fraction ......................... 1.0
104
+ data_sharding ................................... True
105
+ dataloader_type ................................. single
106
+ ddp_average_in_collective ....................... False
107
+ ddp_bucket_size ................................. None
108
+ ddp_num_buckets ................................. None
109
+ ddp_pad_buckets_for_high_nccl_busbw ............. False
110
+ decoder_first_pipeline_num_layers ............... None
111
+ decoder_last_pipeline_num_layers ................ None
112
+ decoder_num_layers .............................. None
113
+ decoder_seq_length .............................. None
114
+ decoupled_lr .................................... None
115
+ decoupled_min_lr ................................ None
116
+ decrease_batch_size_if_needed ................... False
117
+ defer_embedding_wgrad_compute ................... False
118
+ delay_wgrad_compute ............................. False
119
+ deprecated_use_mcore_models ..................... True
120
+ deterministic_mode .............................. False
121
+ dino_bottleneck_size ............................ 256
122
+ dino_freeze_last_layer .......................... 1
123
+ dino_head_hidden_size ........................... 2048
124
+ dino_local_crops_number ......................... 10
125
+ dino_local_img_size ............................. 96
126
+ dino_norm_last_layer ............................ False
127
+ dino_teacher_temp ............................... 0.07
128
+ dino_warmup_teacher_temp ........................ 0.04
129
+ dino_warmup_teacher_temp_epochs ................. 30
130
+ disable_bf16_reduced_precision_matmul ........... False
131
+ disable_mamba_mem_eff_path ...................... False
132
+ disable_straggler_on_startup .................... False
133
+ dist_ckpt_format_deprecated ..................... None
134
+ dist_ckpt_strictness ............................ assume_ok_unexpected
135
+ distribute_saved_activations .................... False
136
+ distributed_backend ............................. nccl
137
+ distributed_timeout_minutes ..................... 10
138
+ embedding_init_method_std ....................... None
139
+ embedding_path .................................. None
140
+ empty_unused_memory_level ....................... 0
141
+ enable_cuda_graph ............................... True
142
+ enable_experimental ............................. False
143
+ enable_ft_package ............................... False
144
+ enable_gloo_process_groups ...................... True
145
+ enable_msc ...................................... True
146
+ enable_one_logger ............................... True
147
+ encoder_num_layers .............................. 16
148
+ encoder_seq_length .............................. 4096
149
+ end_weight_decay ................................ 0.1
150
+ eod_mask_loss ................................... False
151
+ error_injection_rate ............................ 0
152
+ error_injection_type ............................ transient_error
153
+ eval_interval ................................... 1000000
154
+ eval_iters ...................................... 0
155
+ evidence_data_path .............................. None
156
+ exit_duration_in_mins ........................... None
157
+ exit_interval ................................... None
158
+ exit_on_missing_checkpoint ...................... False
159
+ exit_signal_handler ............................. False
160
+ exp_avg_dtype ................................... torch.float32
161
+ exp_avg_sq_dtype ................................ torch.float32
162
+ expert_model_parallel_size ...................... 8
163
+ expert_tensor_parallel_size ..................... 1
164
+ export_force_local_attention .................... False
165
+ export_kd_cfg ................................... None
166
+ export_kd_teacher_ckpt_format ................... None
167
+ export_kd_teacher_load .......................... None
168
+ export_kv_cache_quant ........................... False
169
+ export_legacy_megatron .......................... False
170
+ export_model_type ............................... GPTModel
171
+ export_moe_apply_probs_on_input ................. False
172
+ export_qk_l2_norm ............................... False
173
+ export_quant_cfg ................................ None
174
+ export_real_quant_cfg ........................... None
175
+ export_te_mcore_model ........................... False
176
+ external_cuda_graph ............................. False
177
+ ffn_hidden_size ................................. 14336
178
+ finetune ........................................ False
179
+ finetune_data_split ............................. train
180
+ finetune_hf_dataset ............................. None
181
+ first_last_layers_bf16 .......................... False
182
+ flash_decode .................................... False
183
+ fp16 ............................................ False
184
+ fp16_lm_cross_entropy ........................... False
185
+ fp32_residual_connection ........................ False
186
+ fp8 ............................................. e4m3
187
+ fp8_amax_compute_algo ........................... most_recent
188
+ fp8_amax_history_len ............................ 1
189
+ fp8_interval .................................... 1
190
+ fp8_margin ...................................... 0
191
+ fp8_param_gather ................................ False
192
+ fp8_recipe ...................................... mxfp8
193
+ fp8_wgrad ....................................... True
194
+ fsdp_double_buffer .............................. False
195
+ global_batch_size ............................... 16
196
+ grad_reduce_in_bf16 ............................. False
197
+ gradient_accumulation_fusion .................... True
198
+ gradient_reduce_div_fusion ...................... True
199
+ group_query_attention ........................... True
200
+ head_lr_mult .................................... 1.0
201
+ heterogeneous_layers_config_encoded_json ........ None
202
+ heterogeneous_layers_config_path ................ None
203
+ hidden_dropout .................................. 0.0
204
+ hidden_size ..................................... 4096
205
+ hierarchical_context_parallel_sizes ............. None
206
+ high_priority_stream_groups ..................... []
207
+ hybrid_attention_ratio .......................... 0.0
208
+ hybrid_mlp_ratio ................................ 0.0
209
+ hybrid_override_pattern ......................... None
210
+ hysteresis ...................................... 2
211
+ ict_head_size ................................... None
212
+ ict_load ........................................ None
213
+ img_h ........................................... 224
214
+ img_w ........................................... 224
215
+ indexer_batch_size .............................. 128
216
+ indexer_log_interval ............................ 1000
217
+ inference_batch_times_seqlen_threshold .......... -1
218
+ inference_dynamic_batching ...................... False
219
+ inference_dynamic_batching_buffer_guaranteed_fraction 0.2
220
+ inference_dynamic_batching_buffer_overflow_factor None
221
+ inference_dynamic_batching_buffer_size_gb ....... 40.0
222
+ inference_dynamic_batching_chunk_size ........... 256
223
+ inference_dynamic_batching_max_requests_override None
224
+ inference_dynamic_batching_max_tokens_override .. None
225
+ inference_dynamic_batching_num_cuda_graphs ...... 16
226
+ inference_max_batch_size ........................ 8
227
+ inference_max_seq_length ........................ 2560
228
+ inference_rng_tracker ........................... False
229
+ init_method_std ................................. 0.01
230
+ init_method_xavier_uniform ...................... False
231
+ init_model_with_meta_device ..................... False
232
+ initial_loss_scale .............................. 4294967296
233
+ inprocess_active_world_size ..................... 8
234
+ inprocess_barrier_timeout ....................... 120
235
+ inprocess_completion_timeout .................... 120
236
+ inprocess_empty_cuda_cache ...................... False
237
+ inprocess_granularity ........................... node
238
+ inprocess_hard_timeout .......................... 90
239
+ inprocess_heartbeat_interval .................... 30
240
+ inprocess_heartbeat_timeout ..................... 60
241
+ inprocess_last_call_wait ........................ 1
242
+ inprocess_max_iterations ........................ None
243
+ inprocess_monitor_process_interval .............. 1.0
244
+ inprocess_monitor_thread_interval ............... 1.0
245
+ inprocess_progress_watchdog_interval ............ 1.0
246
+ inprocess_restart ............................... False
247
+ inprocess_soft_timeout .......................... 60
248
+ inprocess_termination_grace_time ................ 1
249
+ is_hybrid_model ................................. False
250
+ iter_per_epoch .................................. 1250
251
+ iterations_to_skip .............................. []
252
+ keep_fp8_transpose_cache_when_using_custom_fsdp . False
253
+ kitchen_config_file ............................. None
254
+ kitchen_recipe_number ........................... None
255
+ kv_channels ..................................... 128
256
+ kv_lora_rank .................................... 32
257
+ lazy_mpu_init ................................... None
258
+ load ............................................ None
259
+ load_main_params_from_ckpt ...................... None
260
+ load_model_opt_format ........................... False
261
+ local_rank ...................................... 0
262
+ log_energy ...................................... False
263
+ log_interval .................................... 1
264
+ log_loss_scale_to_tensorboard ................... True
265
+ log_memory_to_tensorboard ....................... False
266
+ log_num_zeros_in_grad ........................... False
267
+ log_params_norm ................................. False
268
+ log_progress .................................... False
269
+ log_straggler ................................... False
270
+ log_throughput .................................. True
271
+ log_timers_to_tensorboard ....................... False
272
+ log_validation_ppl_to_tensorboard ............... False
273
+ log_world_size_to_tensorboard ................... False
274
+ logging_level ................................... None
275
+ loss_scale ...................................... None
276
+ loss_scale_window ............................... 1000
277
+ lr .............................................. 0.0001
278
+ lr_decay_iters .................................. 8
279
+ lr_decay_samples ................................ None
280
+ lr_decay_style .................................. cosine
281
+ lr_warmup_fraction .............................. None
282
+ lr_warmup_init .................................. 0.0
283
+ lr_warmup_iters ................................. 5
284
+ lr_warmup_samples ............................... 0
285
+ lr_wsd_decay_iters .............................. None
286
+ lr_wsd_decay_samples ............................ None
287
+ lr_wsd_decay_style .............................. exponential
288
+ main_grads_dtype ................................ torch.float32
289
+ main_params_dtype ............................... torch.float32
290
+ make_vocab_size_divisible_by .................... 128
291
+ mamba_head_dim .................................. 64
292
+ mamba_num_groups ................................ 8
293
+ mamba_num_heads ................................. None
294
+ mamba_state_dim ................................. 128
295
+ manual_gc ....................................... False
296
+ manual_gc_eval .................................. True
297
+ manual_gc_interval .............................. 0
298
+ mask_factor ..................................... 1.0
299
+ mask_prob ....................................... 0.15
300
+ mask_type ....................................... random
301
+ masked_softmax_fusion ........................... False
302
+ max_position_embeddings ......................... 32768
303
+ max_tokens_to_oom ............................... 12000
304
+ memory_snapshot_path ............................ snapshot.pickle
305
+ merge_file ...................................... /workspace/data/wikitext/gpt2-merges.txt
306
+ micro_batch_size ................................ 1
307
+ microbatch_group_size_per_vp_stage .............. None
308
+ mid_level_dataset_surplus ....................... 0.005
309
+ min_loss_scale .................................. 1.0
310
+ min_lr .......................................... 1e-05
311
+ mlp_chunks_for_prefill .......................... 1
312
+ mmap_bin_files .................................. True
313
+ mock_data ....................................... False
314
+ moe_apply_probs_on_input ........................ False
315
+ moe_aux_loss_coeff .............................. 0.01
316
+ moe_deepep_num_sms .............................. 20
317
+ moe_echo_enable_random_offloading ............... False
318
+ moe_echo_expert_dispatch_overlap ................ False
319
+ moe_echo_expert_dispatcher_type ................. hybridep
320
+ moe_echo_recompute_expert_dispatch .............. False
321
+ moe_enable_deepep ............................... False
322
+ moe_enable_echo ................................. False
323
+ moe_enable_hybridep ............................. False
324
+ moe_enable_moonep ............................... True
325
+ moe_expert_capacity_factor ...................... None
326
+ moe_extended_tp ................................. False
327
+ moe_ffn_hidden_size ............................. 3584
328
+ moe_grouped_gemm ................................ True
329
+ moe_hybridep_num_sms ............................ 16
330
+ moe_input_jitter_eps ............................ None
331
+ moe_layer_freq .................................. 1
332
+ moe_layer_recompute ............................. False
333
+ moe_moonep_async_grad_reduce .................... True
334
+ moe_moonep_num_sms .............................. None
335
+ moe_num_echo_experts ............................ None
336
+ moe_pad_expert_input_to_capacity ................ False
337
+ moe_per_layer_logging ........................... False
338
+ moe_permute_fusion .............................. False
339
+ moe_received_token_capacity ..................... None
340
+ moe_router_bias_update_rate ..................... 0.001
341
+ moe_router_dtype ................................ fp32
342
+ moe_router_enable_expert_bias ................... False
343
+ moe_router_force_load_balancing ................. False
344
+ moe_router_group_topk ........................... None
345
+ moe_router_load_balancing_type .................. aux_loss
346
+ moe_router_num_groups ........................... None
347
+ moe_router_padding_for_fp8 ...................... False
348
+ moe_router_pre_softmax .......................... False
349
+ moe_router_score_function ....................... softmax
350
+ moe_router_topk ................................. 2
351
+ moe_router_topk_scaling_factor .................. None
352
+ moe_shared_expert_intermediate_size ............. None
353
+ moe_shared_expert_overlap ....................... False
354
+ moe_token_dispatcher_type ....................... flex
355
+ moe_token_drop_policy ........................... probs
356
+ moe_upcycling_granularity ....................... 1
357
+ moe_use_device_initiated_grouped_gemm ........... True
358
+ moe_use_legacy_grouped_gemm ..................... False
359
+ moe_use_upcycling ............................... False
360
+ moe_z_loss_coeff ................................ None
361
+ mrope_section ................................... None
362
+ mscale .......................................... 1.0
363
+ mscale_all_dim .................................. 0.0
364
+ mtp_loss_scaling_factor ......................... 0.1
365
+ mtp_num_layers .................................. None
366
+ multi_latent_attention .......................... False
367
+ nccl_all_reduce_for_prefill ..................... False
368
+ nccl_communicator_config_path ................... None
369
+ nccl_ub ......................................... False
370
+ no_load_optim ................................... None
371
+ no_load_rng ..................................... None
372
+ no_persist_layer_norm ........................... False
373
+ no_rope_freq .................................... None
374
+ no_save_optim ................................... None
375
+ no_save_rng ..................................... None
376
+ non_persistent_ckpt_type ........................ None
377
+ non_persistent_global_ckpt_dir .................. None
378
+ non_persistent_local_ckpt_algo .................. fully_parallel
379
+ non_persistent_local_ckpt_dir ................... None
380
+ non_persistent_save_interval .................... None
381
+ norm_epsilon .................................... 1e-05
382
+ normalization ................................... RMSNorm
383
+ num_attention_heads ............................. 32
384
+ num_channels .................................... 3
385
+ num_classes ..................................... 1000
386
+ num_dataset_builder_threads ..................... 1
387
+ num_distributed_optimizer_instances ............. 1
388
+ num_experts ..................................... 32
389
+ num_layers ...................................... 16
390
+ num_layers_at_end_in_bf16 ....................... 1
391
+ num_layers_at_start_in_bf16 ..................... 1
392
+ num_layers_per_virtual_pipeline_stage ........... None
393
+ num_query_groups ................................ 8
394
+ num_virtual_stages_per_pipeline_rank ............ None
395
+ num_workers ..................................... 2
396
+ object_storage_cache_path ....................... None
397
+ one_logger_async ................................ False
398
+ one_logger_project .............................. megatron-lm
399
+ one_logger_run_name ............................. None
400
+ onnx_safe ....................................... None
401
+ openai_gelu ..................................... False
402
+ optimizer ....................................... adam
403
+ optimizer_cpu_offload ........................... False
404
+ optimizer_offload_fraction ...................... 1.0
405
+ output_bert_embeddings .......................... False
406
+ overlap_cpu_optimizer_d2h_h2d ................... False
407
+ overlap_grad_reduce ............................. True
408
+ overlap_moe_expert_parallel_comm ................ False
409
+ overlap_p2p_comm ................................ False
410
+ overlap_p2p_comm_warmup_flush ................... False
411
+ overlap_param_gather ............................ True
412
+ overlap_param_gather_with_optimizer_step ........ False
413
+ override_opt_param_scheduler .................... False
414
+ params_dtype .................................... torch.bfloat16
415
+ patch_dim ....................................... 16
416
+ per_split_data_args_path ........................ None
417
+ perform_initialization .......................... True
418
+ pin_cpu_grads ................................... True
419
+ pin_cpu_params .................................. True
420
+ pipeline_model_parallel_comm_backend ............ None
421
+ pipeline_model_parallel_layout .................. None
422
+ pipeline_model_parallel_size .................... 1
423
+ position_embedding_type ......................... rope
424
+ pretrained_checkpoint ........................... None
425
+ profile ......................................... True
426
+ profile_ranks ................................... [0, 1, 2, 3, 4, 5, 6, 7]
427
+ profile_step_end ................................ 8
428
+ profile_step_start .............................. 6
429
+ q_lora_rank ..................................... None
430
+ qk_head_dim ..................................... 128
431
+ qk_l2_norm ...................................... False
432
+ qk_layernorm .................................... False
433
+ qk_pos_emb_head_dim ............................. 64
434
+ query_in_block_prob ............................. 0.1
435
+ rampup_batch_size ............................... None
436
+ rank ............................................ 0
437
+ recompute_granularity ........................... None
438
+ recompute_method ................................ None
439
+ recompute_modules ............................... None
440
+ recompute_num_layers ............................ None
441
+ record_memory_history ........................... False
442
+ relative_attention_max_distance ................. 128
443
+ relative_attention_num_buckets .................. 32
444
+ replication ..................................... False
445
+ replication_factor .............................. 2
446
+ replication_jump ................................ None
447
+ rerun_mode ...................................... validate_results
448
+ reset_attention_mask ............................ False
449
+ reset_position_ids .............................. False
450
+ result_rejected_tracker_filename ................ None
451
+ retriever_report_topk_accuracies ................ []
452
+ retriever_score_scaling ......................... False
453
+ retriever_seq_length ............................ 256
454
+ retro_add_retriever ............................. False
455
+ retro_attention_gate ............................ 1
456
+ retro_cyclic_train_iters ........................ None
457
+ retro_encoder_attention_dropout ................. 0.1
458
+ retro_encoder_hidden_dropout .................... 0.1
459
+ retro_encoder_layers ............................ 2
460
+ retro_num_neighbors ............................. 2
461
+ retro_num_retrieved_chunks ...................... 2
462
+ retro_project_dir ............................... None
463
+ retro_verify_neighbor_count ..................... True
464
+ reuse_grad_buf_for_mxfp8_param_ag ............... False
465
+ rope_scaling_factor ............................. 8.0
466
+ rope_type ....................................... None
467
+ rotary_base ..................................... 1000000
468
+ rotary_interleaved .............................. False
469
+ rotary_percent .................................. 1.0
470
+ rotary_scaling_factor ........................... 1.0
471
+ rotary_seq_len_interpolation_factor ............. None
472
+ run_workload_inspector_server ................... False
473
+ sample_rate ..................................... 1.0
474
+ save ............................................ None
475
+ save_interval ................................... None
476
+ save_retain_interval ............................ None
477
+ scatter_gather_tensors_in_pipeline .............. True
478
+ seed ............................................ 1234
479
+ seq_length ...................................... 4096
480
+ sequence_parallel ............................... False
481
+ sft ............................................. False
482
+ sft_tokenizer_prompt_format ..................... nemotron-h-aligned
483
+ sgd_momentum .................................... 0.9
484
+ short_seq_prob .................................. 0.1
485
+ skip_train ...................................... False
486
+ skipped_train_samples ........................... 0
487
+ spec ............................................ None
488
+ split ........................................... 100,0,0
489
+ squared_relu .................................... False
490
+ start_weight_decay .............................. 0.1
491
+ straggler_ctrlr_port ............................ 65535
492
+ straggler_minmax_count .......................... 1
493
+ suggested_communication_unit_size ............... None
494
+ swiglu .......................................... True/workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
495
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
496
+
497
+ swin_backbone_type .............................. tiny
498
+ symmetric_ar_type ............................... None
499
+ te_rng_tracker .................................. True
500
+ tensor_model_parallel_size ...................... 1
501
+ tensorboard_dir ................................. None
502
+ tensorboard_log_interval ........................ 1
503
+ tensorboard_queue_size .......................... 1000
504
+ test_data_path .................................. None
505
+ test_mode ....................................... False
506
+ tiktoken_num_special_tokens ..................... 1000
507
+ tiktoken_pattern ................................ None
508
+ tiktoken_special_tokens ......................... None
509
+ timing_log_level ................................ 0
510
+ timing_log_option ............................... minmax
511
+ titles_data_path ................................ None
512
+ tokenizer_model ................................. None
513
+ tokenizer_type .................................. GPT2BPETokenizer
514
+ torch_fsdp2_reshard_after_forward ............... True
515
+ tp_comm_bootstrap_backend ....................... nccl
516
+ tp_comm_bulk_dgrad .............................. True
517
+ tp_comm_bulk_wgrad .............................. True
518
+ tp_comm_overlap ................................. False
519
+ tp_comm_overlap_ag .............................. True
520
+ tp_comm_overlap_cfg ............................. None
521
+ tp_comm_overlap_rs .............................. True
522
+ tp_comm_overlap_rs_dgrad ........................ False
523
+ tp_comm_split_ag ................................ True
524
+ tp_comm_split_rs ................................ True
525
+ train_data_path ................................. None
526
+ train_iters ..................................... 10
527
+ train_samples ................................... None
528
+ train_sync_interval ............................. None
529
+ transformer_impl ................................ transformer_engine
530
+ transformer_pipeline_model_parallel_size ........ 1
531
+ untie_embeddings_and_output_weights ............. True
532
+ use_checkpoint_args ............................. False
533
+ use_checkpoint_opt_param_scheduler .............. False
534
+ use_cpu_initialization .......................... None
535
+ use_custom_fsdp ................................. False
536
+ use_dist_ckpt ................................... True
537
+ use_dist_ckpt_deprecated ........................ False
538
+ use_distributed_optimizer ....................... True
539
+ use_flash_attn .................................. False
540
+ use_legacy_models ............................... False
541
+ use_mp_args_from_checkpoint_args ................ False
542
+ use_one_sent_docs ............................... False
543
+ use_persistent_ckpt_worker ...................... False
544
+ use_precision_aware_optimizer ................... False
545
+ use_pytorch_profiler ............................ False
546
+ use_ring_exchange_p2p ........................... False
547
+ use_rope_scaling ................................ False
548
+ use_rotary_position_embeddings .................. False
549
+ use_sharp ....................................... False
550
+ use_tokenizer_model_from_checkpoint_args ........ True
551
+ use_torch_fsdp2 ................................. False
552
+ use_torch_optimizer_for_cpu_offload ............. False
553
+ use_tp_pp_dp_mapping ............................ False
554
+ v_head_dim ...................................... 128
555
+ valid_data_path ................................. None
556
+ variable_seq_lengths ............................ False
557
+ virtual_pipeline_model_parallel_size ............ None
558
+ vision_backbone_type ............................ vit
559
+ vision_pretraining .............................. False
560
+ vision_pretraining_type ......................... classify
561
+ vocab_extra_ids ................................. 0
562
+ vocab_file ...................................... /workspace/data/wikitext/gpt2-vocab.json
563
+ vocab_size ...................................... None
564
+ wandb_exp_name ..................................
565
+ wandb_project ...................................
566
+ wandb_save_dir ..................................
567
+ weight_decay .................................... 0.1
568
+ weight_decay_incr_style ......................... constant
569
+ wgrad_deferral_limit ............................ 0
570
+ world_size ...................................... 8
571
+ yaml_cfg ........................................ None
572
+ -------------------- end of arguments ---------------------
573
+ INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 2
574
+ > building GPT2BPETokenizer tokenizer ...
575
+ WARNING: one_logger package is required to enable e2e metrics tracking. please go to https://confluence.nvidia.com/display/MLWFO/Package+Repositories for details to install it
576
+ > padded vocab (size: 50257) with 47 dummy tokens (new size: 50304)
577
+ WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.VALIDATE_RESULTS
578
+ > initializing torch distributed ...
579
+ [Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
580
+ [Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
581
+ [Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
582
+ [Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
583
+ [Gloo] Rank 5 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
584
+ [Gloo] Rank 6 is connected to 7 peer ranks. Expected number of connected peer ranks is : [Gloo] Rank 7
585
+ 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
586
+ [Gloo] Rank 7 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
587
+ [Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
588
+ [Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
589
+ [Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
590
+ [Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
591
+ [Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
592
+ [Gloo] Rank 5 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
593
+ [Gloo] Rank [Gloo] Rank 7 is connected to 7 peer ranks. Expected number of connected peer ranks is : 67 is connected to 7 peer ranks.
594
+ Expected number of connected peer ranks is : 7
595
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
596
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0[Gloo] Rank
597
+ 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
598
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
599
+ > initialized tensor model parallel with size 1
600
+ > initialized pipeline model parallel with size 1
601
+ > setting random seeds to 1234 ...
602
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
603
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
604
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
605
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
606
+ [rank5]:[W921 05:57:54.620750318 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
607
+ [rank4]:[W921 05:57:54.620834186 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
608
+ [rank7]:[W921 05:57:54.620883557 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
609
+ [rank3]:[W921 05:57:54.621006260 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
610
+ [rank6]:[W921 05:57:54.621079946 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
611
+ [rank2]:[W921 05:57:54.621113053 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
612
+ [rank1]:[W921 05:57:54.622303082 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
613
+ > compiling dataset index builder ...
614
+ make: Entering directory '/workspace/Megatron-MoE-Echo/megatron/core/datasets'
615
+ make: Nothing to be done for 'default'.
616
+ make: Leaving directory '/workspace/Megatron-MoE-Echo/megatron/core/datasets'
617
+ >>> done with dataset index builder. Compilation time: 0.427 seconds
618
+ WARNING: constraints for invoking optimized fused softmax kernel are not met. We default back to unfused kernel invocations.
619
+ > compiling and loading fused kernels ...
620
+ [rank0]:[W921 05:57:55.221921556 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
621
+ >>> done with compiling and loading fused kernels. Compilation time: 3.966 seconds
622
+ time to initialize megatron (seconds): 7.995
623
+ [after megatron is initialized] datetime: 2026-09-21 05:58:01
624
+ building GPT model ...
625
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
626
+ warnings.warn(
627
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
628
+ warnings.warn(
629
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
630
+ warnings.warn(
631
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
632
+ warnings.warn(
633
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
634
+ warnings.warn(
635
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
636
+ warnings.warn(
637
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
638
+ warnings.warn(
639
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
640
+ warnings.warn(
641
+ > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 3903983616
642
+ INFO:megatron.core.distributed.distributed_data_parallel:Setting up DistributedDataParallel with config DistributedDataParallelConfig(grad_reduce_in_fp32=True, overlap_grad_reduce=True, overlap_param_gather=True, align_param_gather=False, use_distributed_optimizer=True, num_distributed_optimizer_instances=1, check_for_nan_in_grad=False, check_for_large_grads=False, bucket_size=40000000, pad_buckets_for_high_nccl_busbw=False, average_in_collective=False, fp8_param_gather=False, reuse_grad_buf_for_mxfp8_param_ag=False, use_custom_fsdp=False, data_parallel_sharding_strategy='no_shard', gradient_reduce_div_fusion=True, suggested_communication_unit_size=None, preserve_fp32_weights=True, keep_fp8_transpose_cache_when_using_custom_fsdp=False, nccl_ub=False, fsdp_double_buffer=False)
643
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 18
644
+ Params for bucket 1 (206045184 elements, 206045184 padded size):
645
+ module.output_layer.weight
646
+ Params for bucket 2 (42086400 elements, 42086400 padded size):
647
+ module.decoder.final_layernorm.weight
648
+ module.decoder.layers.15.self_attention.linear_qkv.layer_norm_weight
649
+ module.decoder.layers.15.mlp.router.weight
650
+ module.decoder.layers.15.self_attention.linear_proj.weight
651
+ module.decoder.layers.15.self_attention.linear_qkv.weight
652
+ module.decoder.layers.15.pre_mlp_layernorm.weight
653
+ Params for bucket 3 (42082304 elements, 42082304 padded size):
654
+ module.decoder.layers.14.mlp.router.weight
655
+ module.decoder.layers.14.self_attention.linear_proj.weight
656
+ module.decoder.layers.14.self_attention.linear_qkv.layer_norm_weight
657
+ module.decoder.layers.14.self_attention.linear_qkv.weight
658
+ module.decoder.layers.14.pre_mlp_layernorm.weight
659
+ Params for bucket 4 (42082304 elements, 42082304 padded size):
660
+ module.decoder.layers.13.mlp.router.weight
661
+ module.decoder.layers.13.self_attention.linear_proj.weight
662
+ module.decoder.layers.13.self_attention.linear_qkv.layer_norm_weight
663
+ module.decoder.layers.13.pre_mlp_layernorm.weight
664
+ module.decoder.layers.13.self_attention.linear_qkv.weight
665
+ Params for bucket 5 (42082304 elements, 42082304 padded size):
666
+ module.decoder.layers.12.self_attention.linear_qkv.weight
667
+ module.decoder.layers.12.pre_mlp_layernorm.weight
668
+ module.decoder.layers.12.mlp.router.weight
669
+ module.decoder.layers.12.self_attention.linear_qkv.layer_norm_weight
670
+ module.decoder.layers.12.self_attention.linear_proj.weight
671
+ Params for bucket 6 (42082304 elements, 42082304 padded size):
672
+ module.decoder.layers.11.pre_mlp_layernorm.weight
673
+ module.decoder.layers.11.self_attention.linear_qkv.weight
674
+ module.decoder.layers.11.mlp.router.weight
675
+ module.decoder.layers.11.self_attention.linear_proj.weight
676
+ module.decoder.layers.11.self_attention.linear_qkv.layer_norm_weight
677
+ Params for bucket 7 (42082304 elements, 42082304 padded size):
678
+ module.decoder.layers.10.mlp.router.weight
679
+ module.decoder.layers.10.self_attention.linear_proj.weight
680
+ module.decoder.layers.10.self_attention.linear_qkv.layer_norm_weight
681
+ module.decoder.layers.10.self_attention.linear_qkv.weight
682
+ module.decoder.layers.10.pre_mlp_layernorm.weight
683
+ Params for bucket 8 (42082304 elements, 42082304 padded size):
684
+ module.decoder.layers.9.pre_mlp_layernorm.weight
685
+ module.decoder.layers.9.self_attention.linear_qkv.weight
686
+ module.decoder.layers.9.mlp.router.weight
687
+ module.decoder.layers.9.self_attention.linear_proj.weight
688
+ module.decoder.layers.9.self_attention.linear_qkv.layer_norm_weight
689
+ Params for bucket 9 (42082304 elements, 42082304 padded size):
690
+ module.decoder.layers.8.mlp.router.weight
691
+ module.decoder.layers.8.self_attention.linear_proj.weight
692
+ module.decoder.layers.8.self_attention.linear_qkv.layer_norm_weight
693
+ module.decoder.layers.8.pre_mlp_layernorm.weight
694
+ module.decoder.layers.8.self_attention.linear_qkv.weight
695
+ Params for bucket 10 (42082304 elements, 42082304 padded size):
696
+ module.decoder.layers.7.self_attention.linear_qkv.weight
697
+ module.decoder.layers.7.pre_mlp_layernorm.weight
698
+ module.decoder.layers.7.mlp.router.weight
699
+ module.decoder.layers.7.self_attention.linear_proj.weight
700
+ module.decoder.layers.7.self_attention.linear_qkv.layer_norm_weight
701
+ Params for bucket 11 (42082304 elements, 42082304 padded size):
702
+ module.decoder.layers.6.self_attention.linear_qkv.weight
703
+ module.decoder.layers.6.pre_mlp_layernorm.weight
704
+ module.decoder.layers.6.mlp.router.weight
705
+ module.decoder.layers.6.self_attention.linear_proj.weight
706
+ module.decoder.layers.6.self_attention.linear_qkv.layer_norm_weight
707
+ Params for bucket 12 (42082304 elements, 42082304 padded size):
708
+ module.decoder.layers.5.mlp.router.weight
709
+ module.decoder.layers.5.self_attention.linear_qkv.layer_norm_weight
710
+ module.decoder.layers.5.self_attention.linear_qkv.weight
711
+ module.decoder.layers.5.self_attention.linear_proj.weight
712
+ module.decoder.layers.5.pre_mlp_layernorm.weight
713
+ Params for bucket 13 (42082304 elements, 42082304 padded size):
714
+ module.decoder.layers.4.pre_mlp_layernorm.weight
715
+ module.decoder.layers.4.mlp.router.weight
716
+ module.decoder.layers.4.self_attention.linear_qkv.weight
717
+ module.decoder.layers.4.self_attention.linear_qkv.layer_norm_weight
718
+ module.decoder.layers.4.self_attention.linear_proj.weight
719
+ Params for bucket 14 (42082304 elements, 42082304 padded size):
720
+ module.decoder.layers.3.self_attention.linear_proj.weight
721
+ module.decoder.layers.3.self_attention.linear_qkv.layer_norm_weight
722
+ module.decoder.layers.3.mlp.router.weight
723
+ module.decoder.layers.3.self_attention.linear_qkv.weight
724
+ module.decoder.layers.3.pre_mlp_layernorm.weight
725
+ Params for bucket 15 (42082304 elements, 42082304 padded size):
726
+ module.decoder.layers.2.pre_mlp_layernorm.weight
727
+ module.decoder.layers.2.mlp.router.weight
728
+ module.decoder.layers.2.self_attention.linear_qkv.weight
729
+ module.decoder.layers.2.self_attention.linear_qkv.layer_norm_weight
730
+ module.decoder.layers.2.self_attention.linear_proj.weight
731
+ Params for bucket 16 (42082304 elements, 42082304 padded size):
732
+ module.decoder.layers.1.mlp.router.weight
733
+ module.decoder.layers.1.self_attention.linear_proj.weight
734
+ module.decoder.layers.1.self_attention.linear_qkv.layer_norm_weight
735
+ module.decoder.layers.1.pre_mlp_layernorm.weight
736
+ module.decoder.layers.1.self_attention.linear_qkv.weight
737
+ Params for bucket 17 (42082304 elements, 42082304 padded size):
738
+ module.decoder.layers.0.self_attention.linear_qkv.weight
739
+ module.decoder.layers.0.self_attention.linear_qkv.layer_norm_weight
740
+ module.decoder.layers.0.mlp.router.weight
741
+ module.decoder.layers.0.pre_mlp_layernorm.weight
742
+ module.decoder.layers.0.self_attention.linear_proj.weight
743
+ Params for bucket 18 (206045184 elements, 206045184 padded size):
744
+ module.embedding.word_embeddings.weight
745
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 56
746
+ Params for bucket 1 (44040192 elements, 44040192 padded size):
747
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight0
748
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight2
749
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight1
750
+ Params for bucket 2 (44040192 elements, 44040192 padded size):
751
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight3
752
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight0
753
+ Params for bucket 3 (58720256 elements, 58720256 padded size):
754
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight1
755
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight2
756
+ Params for bucket 4 (44040192 elements, 44040192 padded size):
757
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight3
758
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight0
759
+ Params for bucket 5 (44040192 elements, 44040192 padded size):
760
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight1
761
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight3
762
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight2
763
+ Params for bucket 6 (58720256 elements, 58720256 padded size):
764
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight0
765
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight1
766
+ Params for bucket 7 (58720256 elements, 58720256 padded size):
767
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight2
768
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight3
769
+ Params for bucket 8 (44040192 elements, 44040192 padded size):
770
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight0
771
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight1
772
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight2
773
+ Params for bucket 9 (44040192 elements, 44040192 padded size):
774
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight3
775
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight0
776
+ Params for bucket 10 (58720256 elements, 58720256 padded size):
777
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight1
778
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight2
779
+ Params for bucket 11 (44040192 elements, 44040192 padded size):
780
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight3
781
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight0
782
+ Params for bucket 12 (44040192 elements, 44040192 padded size):
783
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight1
784
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight3
785
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight2
786
+ Params for bucket 13 (58720256 elements, 58720256 padded size):
787
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight0
788
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight1
789
+ Params for bucket 14 (58720256 elements, 58720256 padded size):
790
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight2
791
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight3
792
+ Params for bucket 15 (44040192 elements, 44040192 padded size):
793
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight0
794
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight2
795
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight1
796
+ Params for bucket 16 (44040192 elements, 44040192 padded size):
797
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight3
798
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight0
799
+ Params for bucket 17 (58720256 elements, 58720256 padded size):
800
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight1
801
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight2
802
+ Params for bucket 18 (44040192 elements, 44040192 padded size):
803
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight3
804
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight0
805
+ Params for bucket 19 (44040192 elements, 44040192 padded size):
806
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight1
807
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight3
808
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight2
809
+ Params for bucket 20 (58720256 elements, 58720256 padded size):
810
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight0
811
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight1
812
+ Params for bucket 21 (58720256 elements, 58720256 padded size):
813
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight2
814
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight3
815
+ Params for bucket 22 (44040192 elements, 44040192 padded size):
816
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight0
817
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight1
818
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight2
819
+ Params for bucket 23 (44040192 elements, 44040192 padded size):
820
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight3
821
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight0
822
+ Params for bucket 24 (58720256 elements, 58720256 padded size):
823
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight1
824
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight2
825
+ Params for bucket 25 (44040192 elements, 44040192 padded size):
826
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight3
827
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight0
828
+ Params for bucket 26 (44040192 elements, 44040192 padded size):
829
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight1
830
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight2
831
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight3
832
+ Params for bucket 27 (58720256 elements, 58720256 padded size):
833
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight0
834
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight1
835
+ Params for bucket 28 (58720256 elements, 58720256 padded size):
836
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight2
837
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight3
838
+ Params for bucket 29 (44040192 elements, 44040192 padded size):
839
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight0
840
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight2
841
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight1
842
+ Params for bucket 30 (44040192 elements, 44040192 padded size):
843
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight3
844
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight0
845
+ Params for bucket 31 (58720256 elements, 58720256 padded size):
846
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight1
847
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight2
848
+ Params for bucket 32 (44040192 elements, 44040192 padded size):
849
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight3
850
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight0
851
+ Params for bucket 33 (44040192 elements, 44040192 padded size):
852
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight1
853
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight3
854
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight2
855
+ Params for bucket 34 (58720256 elements, 58720256 padded size):
856
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight0
857
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight1
858
+ Params for bucket 35 (58720256 elements, 58720256 padded size):
859
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight2
860
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight3
861
+ Params for bucket 36 (44040192 elements, 44040192 padded size):
862
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight0
863
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight1
864
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight2
865
+ Params for bucket 37 (44040192 elements, 44040192 padded size):
866
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight3
867
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight0
868
+ Params for bucket 38 (58720256 elements, 58720256 padded size):
869
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight1
870
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight2
871
+ Params for bucket 39 (44040192 elements, 44040192 padded size):
872
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight3
873
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight0
874
+ Params for bucket 40 (44040192 elements, 44040192 padded size):
875
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight1
876
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight2
877
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight3
878
+ Params for bucket 41 (58720256 elements, 58720256 padded size):
879
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight0
880
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight1
881
+ Params for bucket 42 (58720256 elements, 58720256 padded size):
882
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight2
883
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight3
884
+ Params for bucket 43 (44040192 elements, 44040192 padded size):
885
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight0
886
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight1
887
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight2
888
+ Params for bucket 44 (44040192 elements, 44040192 padded size):
889
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight3
890
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight0
891
+ Params for bucket 45 (58720256 elements, 58720256 padded size):
892
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight1
893
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight2
894
+ Params for bucket 46 (44040192 elements, 44040192 padded size):
895
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight3
896
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight0
897
+ Params for bucket 47 (44040192 elements, 44040192 padded size):
898
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight1
899
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight2
900
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight3
901
+ Params for bucket 48 (58720256 elements, 58720256 padded size):
902
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight0
903
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight1
904
+ Params for bucket 49 (58720256 elements, 58720256 padded size):
905
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight2
906
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight3
907
+ Params for bucket 50 (44040192 elements, 44040192 padded size):
908
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight0
909
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight1
910
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight2
911
+ Params for bucket 51 (44040192 elements, 44040192 padded size):
912
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight3
913
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight0
914
+ Params for bucket 52 (58720256 elements, 58720256 padded size):
915
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight1
916
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight2
917
+ Params for bucket 53 (44040192 elements, 44040192 padded size):
918
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight3
919
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight0
920
+ Params for bucket 54 (44040192 elements, 44040192 padded size):
921
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight1
922
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight3
923
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight2
924
+ Params for bucket 55 (58720256 elements, 58720256 padded size):
925
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight0
926
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight1
927
+ Params for bucket 56 (58720256 elements, 58720256 padded size):
928
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight2
929
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight3
930
+ INFO:megatron.core.optimizer:Setting up optimizer with config OptimizerConfig(optimizer='adam', lr=0.0001, min_lr=1e-05, decoupled_lr=None, decoupled_min_lr=None, weight_decay=0.1, fp8_recipe='mxfp8', fp16=False, bf16=True, reuse_grad_buf_for_mxfp8_param_ag=False, params_dtype=torch.bfloat16, use_precision_aware_optimizer=False, store_param_remainders=True, main_grads_dtype=torch.float32, main_params_dtype=torch.float32, exp_avg_dtype=torch.float32, exp_avg_sq_dtype=torch.float32, loss_scale=None, initial_loss_scale=4294967296, min_loss_scale=1.0, loss_scale_window=1000, hysteresis=2, adam_beta1=0.9, adam_beta2=0.999, adam_eps=1e-08, sgd_momentum=0.9, use_distributed_optimizer=True, overlap_param_gather_with_optimizer_step=False, optimizer_cpu_offload=False, optimizer_offload_fraction=1.0, use_torch_optimizer_for_cpu_offload=False, overlap_cpu_optimizer_d2h_h2d=False, pin_cpu_grads=True, pin_cpu_params=True, clip_grad=1.0, log_num_zeros_in_grad=False, barrier_with_L1_time=True, timers=<megatron.core.timers.Timers object at 0x728717c1b050>, config_logger_dir='')
931
+ INFO:megatron.core.optimizer_param_scheduler:> learning rate decay style: cosine
932
+ [after model, optimizer, and learning rate scheduler are built] datetime: 2026-09-21 05:58:04
933
+ > building train, validation, and test datasets ...
934
+ > datasets target sizes (minimum size):
935
+ train: 160
936
+ validation: 0
937
+ test: 0
938
+ INFO:megatron.core.datasets.blended_megatron_dataset_config:Let split_matrix = [(0, 1.0), None, None]
939
+ > building train, validation, and test datasets for GPT ...
940
+ INFO:megatron.core.datasets.blended_megatron_dataset_builder:Building GPTDataset splits with sizes=(160, 0, 0) and config=GPTDatasetConfig(random_seed=1234, sequence_length=4096, blend=(['/workspace/data/wikitext/wikitext103_gpt2_text_document'], None), blend_per_split=None, split='100,0,0', split_matrix=[(0, 1.0), None, None], num_dataset_builder_threads=1, path_to_cache=None, mmap_bin_files=True, mock=False, tokenizer=<megatron.training.tokenizer.tokenizer._GPT2BPETokenizer object at 0x728717b78590>, mid_level_dataset_surplus=0.005, reset_position_ids=False, reset_attention_mask=False, eod_mask_loss=False, create_attention_mask=True, drop_last_partial_validation_sequence=True, add_extra_token_to_sequence=True, object_storage_cache_path=None)
941
+ INFO:megatron.core.datasets.indexed_dataset:Load the _IndexReader from /workspace/data/wikitext/wikitext103_gpt2_text_document.idx
942
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence lengths
943
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence pointers
944
+ INFO:megatron.core.datasets.indexed_dataset: Extract the document indices
945
+ INFO:megatron.core.datasets.indexed_dataset:> total number of sequences: 1165029
946
+ INFO:megatron.core.datasets.indexed_dataset:> total number of documents: 1801350
947
+ INFO:megatron.core.datasets.gpt_dataset:Load the GPTDataset train indices
948
+ INFO:megatron.core.datasets.gpt_dataset: Load the document index from 68684a515e118253a8572c4517cf2ffe-GPTDataset-train-document_index.npy
949
+ INFO:megatron.core.datasets.gpt_dataset: Load the sample index from 68684a515e118253a8572c4517cf2ffe-GPTDataset-train-sample_index.npy
950
+ INFO:megatron.core.datasets.gpt_dataset: Load the shuffle index from 68684a515e118253a8572c4517cf2ffe-GPTDataset-train-shuffle_index.npy
951
+ INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 29073
952
+ > finished creating GPT datasets ...
953
+ [after dataloaders are built] datetime: 2026-09-21 05:58:04
954
+ done with setup ...
955
+ training ...
956
+ (min, max) time across ranks (ms):
957
+ model-and-optimizer-setup ......................: (2880.96, 2885.64)
958
+ train/valid/test-data-iterators-setup ..........: (90.28, 111.15)
959
+ Overwriting rerun_state_machine.current_iteration from -1 to 0...
960
+ [before the start of training step] datetime: 2026-09-21 05:58:04
961
+ INFO:moonep.api:MoonEP comm buffer sizes:
962
+ config : R=8, S=4096, H=4096, E=32, K=2, NvS=9208, NvS_padded=9216
963
+ sm_count : num_sms=32, num_sms_dedup=148
964
+ hidden_buffer_size : 72.00 MiB
965
+ meta_buffer_size : 2.00 MiB
966
+ local_temp_buffer_size : 513.79 KiB
967
+ total_size : 74.50 MiB
968
+ [MoonEP selfcheck 1/4] dropped groups empty OK | covered rows [0, 1, 2, 3, 32, 33, 34, 35] sizes [128, 6400, 1024, 512, 384, 0, 0, 0] | duplicated experts [[23, -1, -1, -1], [1, -1, -1, -1], [1, -1, -1, -1], [31, -1, -1, -1], [22, -1, -1, -1], [-1, -1, -1, -1], [22, -1, -1, -1], [23, -1, -1, -1]]
969
+ [MoonEP selfcheck 2/4] dropped groups empty OK | covered rows [0, 1, 2, 3, 32, 33, 34, 35] sizes [1024, 128, 640, 128, 6656, 0, 0, 0] | duplicated experts [[27, -1, -1, -1], [27, -1, -1, -1], [28, -1, -1, -1], [19, -1, -1, -1], [28, -1, -1, -1], [19, -1, -1, -1], [28, -1, -1, -1], [-1, -1, -1, -1]]
970
+ [MoonEP selfcheck 3/4] dropped groups empty OK | covered rows [0, 1, 2, 3, 32, 33, 34, 35] sizes [1536, 512, 128, 2944, 3328, 0, 0, 0] | duplicated experts [[18, -1, -1, -1], [-1, -1, -1, -1], [6, -1, -1, -1], [6, -1, -1, -1], [6, -1, -1, -1], [6, -1, -1, -1], [17, -1, -1, -1], [17, -1, -1, -1]]
971
+ [MoonEP selfcheck 4/4] dropped groups empty OK | covered rows [0, 1, 2, 3, 32, 33, 34, 35] sizes [2816, 3072, 0, 2048, 512, 0, 0, 0] | duplicated experts [[12, -1, -1, -1], [12, -1, -1, -1], [3, -1, -1, -1], [-1, -1, -1, -1], [12, -1, -1, -1], [0, -1, -1, -1], [12, -1, -1, -1], [12, -1, -1, -1]]
972
+ Number of parameters in transformer block in billions: 23.22 [2026-09-21 05:58:27] iteration 1/ 10 | consumed samples: 16 | elapsed time per iteration (ms): 22625.5 | throughput per GPU (TFLOP/s/GPU): 5.6 | learning rate: 2.000000E-05 | global batch size: 16 | lm loss: 1.102902E+01 | loss scale: 1.0 | grad norm: 44.186 | number of skipped iterations: 0 | number of nan iterations: 0 |
973
+
974
+ Number of parameters in embedding layers in billions: 0.41
975
+ Total number of parameters in billions: 23.63
976
+ Number of parameters in most loaded shard in billions: 23.6320
977
+ Theoretical memory footprints: weight and optimizer=169029.43 MB
978
+ [Rank 0] (after 1 iterations) memory (MB) | allocated: 53085.119140625 | max allocated: 53085.11962890625 | reserved: 57302.0 | max reserved: 57302.0
979
+ [2026-09-21 05:58:27] iteration 2/ 10 | consumed samples: 32 | elapsed time per iteration (ms): 480.7 | throughput per GPU (TFLOP/s/GPU): 261.2 | learning rate: 4.000000E-05 | global batch size: 16 | lm loss: 1.101823E+01 | loss scale: 1.0 | grad norm: 45.730 | number of skipped iterations: 0 | number of nan iterations: 0 |
980
+ [2026-09-21 05:58:28] iteration 3/ 10 | consumed samples: 48 | elapsed time per iteration (ms): 400.2 | throughput per GPU (TFLOP/s/GPU): 313.8 | learning rate: 6.000000E-05 | global batch size: 16 | lm loss: 9.971961E+00 | loss scale: 1.0 | grad norm: 157.264 | number of skipped iterations: 0 | number of nan iterations: 0 |
981
+ Capture CUDA graph for training!!!
982
+ CUDA graph capture done!!!
983
+ [2026-09-21 05:58:29] iteration 4/ 10 | consumed samples: 64 | elapsed time per iteration (ms): 1150.4 | throughput per GPU (TFLOP/s/GPU): 109.2 | learning rate: 8.000000E-05 | global batch size: 16 | lm loss: 1.311218E+01 | loss scale: 1.0 | grad norm: 41.289 | number of skipped iterations: 0 | number of nan iterations: 0 |
984
+ [2026-09-21 05:58:29] iteration 5/ 10 | consumed samples: 80 | elapsed time per iteration (ms): 256.9 | throughput per GPU (TFLOP/s/GPU): 488.8 | learning rate: 1.000000E-04 | global batch size: 16 | lm loss: 1.040744E+01 | loss scale: 1.0 | grad norm: 25.682 | number of skipped iterations: 0 | number of nan iterations: 0 |
985
+ [2026-09-21 05:58:30] iteration 6/ 10 | consumed samples: 96 | elapsed time per iteration (ms): 257.0 | throughput per GPU (TFLOP/s/GPU): 488.5 | learning rate: 7.750000E-05 | global batch size: 16 | lm loss: 1.025443E+01 | loss scale: 1.0 | grad norm: 19.004 | number of skipped iterations: 0 | number of nan iterations: 0 |
986
+ [2026-09-21 05:58:33] iteration 7/ 10 | consumed samples: 112 | elapsed time per iteration (ms): 3149.2 | throughput per GPU (TFLOP/s/GPU): 39.9 | learning rate: 3.250000E-05 | global batch size: 16 | lm loss: 1.002115E+01 | loss scale: 1.0 | grad norm: 8.675 | number of skipped iterations: 0 | number of nan iterations: 0 |
987
+ [2026-09-21 05:58:33] iteration 8/ 10 | consumed samples: 128 | elapsed time per iteration (ms): 264.3 | throughput per GPU (TFLOP/s/GPU): 475.2 | learning rate: 1.000000E-05 | global batch size: 16 | lm loss: 9.458863E+00 | loss scale: 1.0 | grad norm: 8.073 | number of skipped iterations: 0 | number of nan iterations: 0 |
988
+ Capture range started in the application.
989
+ Capture range ended in the application.
990
+ Generating '/tmp/nsys-report-574d.qdstrm'
991
+
992
+
993
+ [2026-09-21 05:59:16] iteration 9/ 10 | consumed samples: 144 | elapsed time per iteration (ms): 42958.5 | throughput per GPU (TFLOP/s/GPU): 2.9 | learning rate: 1.000000E-05 | global batch size: 16 | lm loss: 9.209682E+00 | loss scale: 1.0 | grad norm: 7.234 | number of skipped iterations: 0 | number of nan iterations: 0 |
994
+ [2026-09-21 05:59:16] iteration 10/ 10 | consumed samples: 160 | elapsed time per iteration (ms): 282.5 | throughput per GPU (TFLOP/s/GPU): 444.5 | learning rate: 1.000000E-05 | global batch size: 16 | lm loss: 9.049531E+00 | loss scale: 1.0 | grad norm: 6.792 | number of skipped iterations: 0 | number of nan iterations: 0 |
995
+ [after training is done] datetime: 2026-09-21 05:59:16
996
+ [rank0]:[W921 05:59:18.030874754 ProcessGroupNCCL.cpp:1505] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
997
+ [rank1]:[W921 05:59:19.430528208 ProcessGroupNCCL.cpp:1505] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
998
+ [rank6]:[W921 05:59:19.431292971 ProcessGroupNCCL.cpp:1505] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
999
+ [rank2]:[W921 05:59:19.433226738 ProcessGroupNCCL.cpp:1505] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
1000
+ [rank3]:[W921 05:59:19.567742354 ProcessGroupNCCL.cpp:1505] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
1001
+ [rank5]:[W921 05:59:19.725135561 ProcessGroupNCCL.cpp:1505] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
1002
+ [rank7]:[W921 05:59:19.729759524 ProcessGroupNCCL.cpp:1505] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
1003
+ [rank4]:[W921 05:59:20.983853426 ProcessGroupNCCL.cpp:1505] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
1004
+ Collection stop is not allowed in this state.
1005
+ [rank5]:[W921 05:59:23.124197647 observer.cpp:116] Warning: Leaked callback handle: 1 (function operator())
1006
+ pure virtual method called
1007
+ terminate called without an active exception
1008
+ [rank6]:[W921 05:59:23.267073184 observer.cpp:116] Warning: Leaked callback handle: 1 (function operator())
1009
+ pure virtual method called
1010
+ terminate called without an active exception
1011
+ [rank2]:[W921 05:59:23.310011577 observer.cpp:116] Warning: Leaked callback handle: 1 (function operator())
1012
+ pure virtual method called
1013
+ terminate called without an active exception
1014
+ [rank3]:[W921 05:59:23.396476879 observer.cpp:116] Warning: Leaked callback handle: 1 (function operator())
1015
+ pure virtual method called
1016
+ terminate called without an active exception
1017
+ [rank1]:[W921 05:59:23.474878515 observer.cpp:116] Warning: Leaked callback handle: 1 (function operator())
1018
+ [rank0]:[W921 05:59:23.479024046 observer.cpp:116] Warning: Leaked callback handle: 1 (function operator())
1019
+ pure virtual method called
1020
+ terminate called without an active exception
1021
+ pure virtual method called
1022
+ terminate called without an active exception
1023
+ W0921 05:59:23.862000 131368 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 131577 closing signal SIGTERM
1024
+ W0921 05:59:23.863000 131368 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 131578 closing signal SIGTERM
1025
+ W0921 05:59:23.863000 131368 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 131579 closing signal SIGTERM
1026
+ W0921 05:59:23.863000 131368 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 131580 closing signal SIGTERM
1027
+ W0921 05:59:23.864000 131368 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 131581 closing signal SIGTERM
1028
+ W0921 05:59:23.864000 131368 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 131583 closing signal SIGTERM
1029
+ W0921 05:59:23.865000 131368 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 131584 closing signal SIGTERM
1030
+ E0921 05:59:25.509000 131368 torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: -6) local_rank: 5 (pid: 131582) of binary: /usr/bin/python
1031
+ Traceback (most recent call last):
1032
+ File "/usr/local/bin/torchrun", line 8, in <module>
1033
+ sys.exit(main())
1034
+ ^^^^^^
1035
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper
1036
+ return f(*args, **kwargs)
1037
+ ^^^^^^^^^^^^^^^^^^
1038
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 892, in main
1039
+ run(args)
1040
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 883, in run
1041
+ elastic_launch(
1042
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 139, in __call__
1043
+ return launch_agent(self._config, self._entrypoint, list(args))
1044
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1045
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 270, in launch_agent
1046
+ raise ChildFailedError(
1047
+ torch.distributed.elastic.multiprocessing.errors.ChildFailedError:
1048
+ ============================================================
1049
+ pretrain_gpt.py FAILED
1050
+ ------------------------------------------------------------
1051
+ Failures:
1052
+ <NO_OTHER_FAILURES>
1053
+ ------------------------------------------------------------
1054
+ Root Cause (first observed failure):
1055
+ [0]:
1056
+ time : 2026-09-21_05:59:23
1057
+ host : sp-aiinfra-interactive-aghosh-8-mb99k-8xb200-moe-cg-0
1058
+ rank : 5 (local_rank: 5)
1059
+ exitcode : -6 (pid: 131582)
1060
+ error_file: <N/A>
1061
+ traceback : Signal 6 (SIGABRT) received by PID 131582
1062
+ ============================================================
1063
+ Generated:
1064
+ /workspace/granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G2_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/G2_K2__moonep__matched__seq4096__gbs16__capnone__async.nsys-rep
granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G2_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/verify_trace_ranks.log ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ report : /workspace/granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G2_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/G2_K2__moonep__matched__seq4096__gbs16__capnone__async.nsys-rep
2
+ kernels : 133202
3
+ devices : 8 of 8 -> [0, 1, 2, 3, 4, 5, 6, 7]
4
+ processes: 1
5
+ markers : 17408 moonep_trace__ kernels on devices [0, 1, 2, 3, 4, 5, 6, 7]
6
+ PASS: all ranks present and backend markers emitted
granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/G3_K2__moonep__matched__seq4096__gbs16__capnone__async.nsys-rep ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6be2b7c114c9bf88589c1b87d926122b3858cada3093d0adf06883c33689e92e
3
+ size 11277816
granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/moonep_async_overlap.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "devices": 8,
3
+ "reduce_busy_ms_per_iteration": 18.631125,
4
+ "overlapped_ms_per_iteration": 17.887,
5
+ "exposed_ms_per_iteration": 0.744125,
6
+ "overlap_pct": 96.00601144590034,
7
+ "minimum_overlap_pct": 0.1,
8
+ "status": "PASS"
9
+ }
granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/moonep_async_overlap.txt ADDED
@@ -0,0 +1,416 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+
2
+ === kernels on device 0, by total time ===
3
+ 48.109 ms multi_tensor_apply_kernel
4
+ 41.667 ms device_kernel
5
+ 37.504 ms elementwise_kernel
6
+ 31.520 ms vectorized_elementwise_kernel
7
+ 28.167 ms cast_mxfp8_2D_kernel
8
+ REDUCE 24.975 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2293764096i64i64_tensorptrf32gmemalign16o2293764096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
9
+ 24.957 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
10
+ 24.321 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o40964096i64i64_tensorptrbf16gmemalign16o819204096i64i64_tensorptri32gmemalign16o81921_tensorptri32gm_0
11
+ 21.357 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
12
+ 19.303 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o229376409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
13
+ 16.542 ms unrolled_elementwise_kernel
14
+ 16.298 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o819204096i64i64_tensorptrbf16gmemalign16o40964096i64i64_tensorptri32gmemalign16o81921_tensorptri32gmem_0
15
+ REDUCE 11.973 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2621441792i64i64_tensorptrf32gmemalign16o2621441792i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
16
+ 11.672 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
17
+ 9.198 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o262144179211i640_CopyAtom_ThrID10_TVLayoutSrc_0
18
+ 8.871 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NNT
19
+ 8.708 ms ncclDevKernel_AllGather_RING_LL
20
+ 8.265 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_TNT
21
+ 7.748 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
22
+ 5.468 ms setGroupedGemmWgradArguments
23
+ 5.260 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_badd_Avec32UE8M0_Bvec32UE8M0_NTT
24
+ 5.033 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NTT
25
+ 3.941 ms triton_poi_fused__to_copy_cat_0
26
+ 3.907 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
27
+ 3.902 ms rmsnorm_bwd_tuned_kernel
28
+ 3.806 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
29
+ 3.587 ms reduce_kernel
30
+ 3.543 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_641_tensorptri32_gmem_align16_o_81921_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_alig_0
31
+ 3.462 ms CatArrayBatchedCopy_alignedK_contig
32
+ 3.427 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
33
+ 2.890 ms _scatter_gather_elementwise_kernel
34
+ 2.634 ms fused_rope_backward_kernel
35
+ 2.520 ms multi_tensor_swizzle_col_scaling_kernel
36
+ 2.469 ms multi_tensor_swizzle_row_scaling_kernel
37
+ 2.451 ms fused_rope_forward_kernel
38
+ 2.413 ms gatherTopK
39
+ 2.275 ms swizzle_col_scaling_kernel
40
+ 1.860 ms rmsnorm_fwd_tuned_kernel
41
+ 1.774 ms triton_poi_fused_add_0
42
+ 1.605 ms compute_dot_do_o_specialized
43
+
44
+ === G3_K2__moonep__matched__seq4096__gbs16__capnone__async.sqlite (device 0, 2 iteration(s)) ===
45
+ reduce_grad kernels : 128 (64/iter)
46
+ reduce_grad sum : 18.474 ms/iter
47
+ reduce_grad busy (union) : 18.474 ms/iter
48
+ ... overlapped with compute: 17.717 ms/iter ( 95.9 %)
49
+ ... exposed on critical path: 0.757 ms/iter ( 4.1 %)
50
+ other GPU work (union) : 211.392 ms/iter
51
+ traced wall : 240.832 ms/iter
52
+ --> time hidden by overlap : 17.717 ms/iter
53
+
54
+ === kernels on device 1, by total time ===
55
+ 48.141 ms multi_tensor_apply_kernel
56
+ 41.319 ms device_kernel
57
+ 37.525 ms elementwise_kernel
58
+ 31.593 ms vectorized_elementwise_kernel
59
+ 28.246 ms cast_mxfp8_2D_kernel
60
+ REDUCE 25.283 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2293764096i64i64_tensorptrf32gmemalign16o2293764096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
61
+ 24.844 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
62
+ 24.243 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o40964096i64i64_tensorptrbf16gmemalign16o819204096i64i64_tensorptri32gmemalign16o81921_tensorptri32gm_0
63
+ 21.827 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
64
+ 18.031 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o229376409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
65
+ 16.548 ms unrolled_elementwise_kernel
66
+ 15.754 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o819204096i64i64_tensorptrbf16gmemalign16o40964096i64i64_tensorptri32gmemalign16o81921_tensorptri32gmem_0
67
+ 13.763 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
68
+ REDUCE 12.141 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2621441792i64i64_tensorptrf32gmemalign16o2621441792i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
69
+ 10.947 ms ncclDevKernel_AllGather_RING_LL
70
+ 9.592 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o262144179211i640_CopyAtom_ThrID10_TVLayoutSrc_0
71
+ 8.867 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NNT
72
+ 8.214 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_TNT
73
+ 7.642 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
74
+ 5.448 ms setGroupedGemmWgradArguments
75
+ 5.278 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_badd_Avec32UE8M0_Bvec32UE8M0_NTT
76
+ 5.054 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NTT
77
+ 4.222 ms ncclDevKernel_AllReduce_Sum_f32_RING_LL
78
+ 3.930 ms triton_poi_fused__to_copy_cat_0
79
+ 3.917 ms rmsnorm_bwd_tuned_kernel
80
+ 3.796 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
81
+ 3.769 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
82
+ 3.638 ms reduce_kernel
83
+ 3.619 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_641_tensorptri32_gmem_align16_o_81921_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_alig_0
84
+ 3.466 ms CatArrayBatchedCopy_alignedK_contig
85
+ 3.364 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
86
+ 2.927 ms _scatter_gather_elementwise_kernel
87
+ 2.630 ms fused_rope_backward_kernel
88
+ 2.529 ms multi_tensor_swizzle_col_scaling_kernel
89
+ 2.466 ms multi_tensor_swizzle_row_scaling_kernel
90
+ 2.452 ms fused_rope_forward_kernel
91
+ 2.403 ms gatherTopK
92
+ 2.295 ms swizzle_col_scaling_kernel
93
+ 1.868 ms rmsnorm_fwd_tuned_kernel
94
+ 1.778 ms triton_poi_fused_add_0
95
+
96
+ === G3_K2__moonep__matched__seq4096__gbs16__capnone__async.sqlite (device 1, 2 iteration(s)) ===
97
+ reduce_grad kernels : 128 (64/iter)
98
+ reduce_grad sum : 18.712 ms/iter
99
+ reduce_grad busy (union) : 18.712 ms/iter
100
+ ... overlapped with compute: 17.983 ms/iter ( 96.1 %)
101
+ ... exposed on critical path: 0.729 ms/iter ( 3.9 %)
102
+ other GPU work (union) : 214.308 ms/iter
103
+ traced wall : 241.721 ms/iter
104
+ --> time hidden by overlap : 17.983 ms/iter
105
+
106
+ === kernels on device 2, by total time ===
107
+ 48.121 ms multi_tensor_apply_kernel
108
+ 41.314 ms device_kernel
109
+ 37.490 ms elementwise_kernel
110
+ 31.398 ms vectorized_elementwise_kernel
111
+ 28.169 ms cast_mxfp8_2D_kernel
112
+ REDUCE 25.331 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2293764096i64i64_tensorptrf32gmemalign16o2293764096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
113
+ 24.921 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
114
+ 24.212 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o40964096i64i64_tensorptrbf16gmemalign16o819204096i64i64_tensorptri32gmemalign16o81921_tensorptri32gm_0
115
+ 21.739 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
116
+ 19.581 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o229376409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
117
+ 16.519 ms unrolled_elementwise_kernel
118
+ 15.943 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o819204096i64i64_tensorptrbf16gmemalign16o40964096i64i64_tensorptri32gmemalign16o81921_tensorptri32gmem_0
119
+ 12.182 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
120
+ REDUCE 11.996 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2621441792i64i64_tensorptrf32gmemalign16o2621441792i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
121
+ 10.636 ms ncclDevKernel_AllGather_RING_LL
122
+ 9.486 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o262144179211i640_CopyAtom_ThrID10_TVLayoutSrc_0
123
+ 8.913 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NNT
124
+ 8.249 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_TNT
125
+ 7.709 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
126
+ 5.489 ms setGroupedGemmWgradArguments
127
+ 5.292 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_badd_Avec32UE8M0_Bvec32UE8M0_NTT
128
+ 5.062 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NTT
129
+ 3.931 ms triton_poi_fused__to_copy_cat_0
130
+ 3.920 ms rmsnorm_bwd_tuned_kernel
131
+ 3.895 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
132
+ 3.798 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
133
+ 3.607 ms reduce_kernel
134
+ 3.578 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_641_tensorptri32_gmem_align16_o_81921_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_alig_0
135
+ 3.449 ms CatArrayBatchedCopy_alignedK_contig
136
+ 3.427 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
137
+ 3.129 ms ncclDevKernel_AllReduce_Sum_f32_RING_LL
138
+ 2.907 ms _scatter_gather_elementwise_kernel
139
+ 2.629 ms fused_rope_backward_kernel
140
+ 2.511 ms multi_tensor_swizzle_col_scaling_kernel
141
+ 2.475 ms multi_tensor_swizzle_row_scaling_kernel
142
+ 2.454 ms fused_rope_forward_kernel
143
+ 2.400 ms gatherTopK
144
+ 2.276 ms swizzle_col_scaling_kernel
145
+ 1.861 ms rmsnorm_fwd_tuned_kernel
146
+ 1.778 ms triton_poi_fused_add_0
147
+
148
+ === G3_K2__moonep__matched__seq4096__gbs16__capnone__async.sqlite (device 2, 2 iteration(s)) ===
149
+ reduce_grad kernels : 128 (64/iter)
150
+ reduce_grad sum : 18.664 ms/iter
151
+ reduce_grad busy (union) : 18.664 ms/iter
152
+ ... overlapped with compute: 17.915 ms/iter ( 96.0 %)
153
+ ... exposed on critical path: 0.749 ms/iter ( 4.0 %)
154
+ other GPU work (union) : 213.630 ms/iter
155
+ traced wall : 241.650 ms/iter
156
+ --> time hidden by overlap : 17.915 ms/iter
157
+
158
+ === kernels on device 3, by total time ===
159
+ 48.144 ms multi_tensor_apply_kernel
160
+ 41.457 ms device_kernel
161
+ 37.525 ms elementwise_kernel
162
+ 31.394 ms vectorized_elementwise_kernel
163
+ 28.252 ms cast_mxfp8_2D_kernel
164
+ REDUCE 25.330 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2293764096i64i64_tensorptrf32gmemalign16o2293764096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
165
+ 25.095 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
166
+ 24.216 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o40964096i64i64_tensorptrbf16gmemalign16o819204096i64i64_tensorptri32gmemalign16o81921_tensorptri32gm_0
167
+ 21.529 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
168
+ 18.122 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o229376409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
169
+ 16.557 ms unrolled_elementwise_kernel
170
+ 15.710 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o819204096i64i64_tensorptrbf16gmemalign16o40964096i64i64_tensorptri32gmemalign16o81921_tensorptri32gmem_0
171
+ 13.996 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
172
+ REDUCE 12.139 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2621441792i64i64_tensorptrf32gmemalign16o2621441792i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
173
+ 10.914 ms ncclDevKernel_AllGather_RING_LL
174
+ 9.002 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NNT
175
+ 8.461 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o262144179211i640_CopyAtom_ThrID10_TVLayoutSrc_0
176
+ 8.305 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_TNT
177
+ 7.796 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
178
+ 5.532 ms setGroupedGemmWgradArguments
179
+ 5.367 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_badd_Avec32UE8M0_Bvec32UE8M0_NTT
180
+ 5.130 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NTT
181
+ 3.976 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
182
+ 3.975 ms ncclDevKernel_AllReduce_Sum_f32_RING_LL
183
+ 3.937 ms triton_poi_fused__to_copy_cat_0
184
+ 3.931 ms rmsnorm_bwd_tuned_kernel
185
+ 3.837 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
186
+ 3.601 ms reduce_kernel
187
+ 3.567 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_641_tensorptri32_gmem_align16_o_81921_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_alig_0
188
+ 3.466 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
189
+ 3.451 ms CatArrayBatchedCopy_alignedK_contig
190
+ 2.900 ms _scatter_gather_elementwise_kernel
191
+ 2.634 ms fused_rope_backward_kernel
192
+ 2.511 ms multi_tensor_swizzle_col_scaling_kernel
193
+ 2.460 ms fused_rope_forward_kernel
194
+ 2.434 ms multi_tensor_swizzle_row_scaling_kernel
195
+ 2.410 ms gatherTopK
196
+ 2.284 ms swizzle_col_scaling_kernel
197
+ 1.866 ms rmsnorm_fwd_tuned_kernel
198
+ 1.788 ms triton_poi_fused_add_0
199
+
200
+ === G3_K2__moonep__matched__seq4096__gbs16__capnone__async.sqlite (device 3, 2 iteration(s)) ===
201
+ reduce_grad kernels : 128 (64/iter)
202
+ reduce_grad sum : 18.734 ms/iter
203
+ reduce_grad busy (union) : 18.734 ms/iter
204
+ ... overlapped with compute: 17.986 ms/iter ( 96.0 %)
205
+ ... exposed on critical path: 0.748 ms/iter ( 4.0 %)
206
+ other GPU work (union) : 214.138 ms/iter
207
+ traced wall : 241.756 ms/iter
208
+ --> time hidden by overlap : 17.986 ms/iter
209
+
210
+ === kernels on device 4, by total time ===
211
+ 48.112 ms multi_tensor_apply_kernel
212
+ 42.490 ms device_kernel
213
+ 37.515 ms elementwise_kernel
214
+ 33.584 ms vectorized_elementwise_kernel
215
+ 28.231 ms cast_mxfp8_2D_kernel
216
+ REDUCE 25.185 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2293764096i64i64_tensorptrf32gmemalign16o2293764096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
217
+ 24.974 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
218
+ 24.214 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o40964096i64i64_tensorptrbf16gmemalign16o819204096i64i64_tensorptri32gmemalign16o81921_tensorptri32gm_0
219
+ 21.850 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
220
+ 18.332 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o229376409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
221
+ 16.497 ms unrolled_elementwise_kernel
222
+ 15.995 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o819204096i64i64_tensorptrbf16gmemalign16o40964096i64i64_tensorptri32gmemalign16o81921_tensorptri32gmem_0
223
+ 12.697 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
224
+ REDUCE 12.060 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2621441792i64i64_tensorptrf32gmemalign16o2621441792i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
225
+ 9.316 ms ncclDevKernel_AllGather_RING_LL
226
+ 8.905 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NNT
227
+ 8.637 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o262144179211i640_CopyAtom_ThrID10_TVLayoutSrc_0
228
+ 8.285 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_TNT
229
+ 7.759 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
230
+ 5.364 ms setGroupedGemmWgradArguments
231
+ 5.287 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_badd_Avec32UE8M0_Bvec32UE8M0_NTT
232
+ 5.090 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NTT
233
+ 3.939 ms triton_poi_fused__to_copy_cat_0
234
+ 3.908 ms rmsnorm_bwd_tuned_kernel
235
+ 3.854 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
236
+ 3.825 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
237
+ 3.616 ms reduce_kernel
238
+ 3.607 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_641_tensorptri32_gmem_align16_o_81921_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_alig_0
239
+ 3.459 ms CatArrayBatchedCopy_alignedK_contig
240
+ 3.401 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
241
+ 2.916 ms _scatter_gather_elementwise_kernel
242
+ 2.634 ms fused_rope_backward_kernel
243
+ 2.516 ms multi_tensor_swizzle_col_scaling_kernel
244
+ 2.461 ms fused_rope_forward_kernel
245
+ 2.436 ms multi_tensor_swizzle_row_scaling_kernel
246
+ 2.414 ms gatherTopK
247
+ 2.272 ms swizzle_col_scaling_kernel
248
+ 1.865 ms rmsnorm_fwd_tuned_kernel
249
+ 1.783 ms triton_poi_fused_add_0
250
+ 1.613 ms compute_dot_do_o_specialized
251
+
252
+ === G3_K2__moonep__matched__seq4096__gbs16__capnone__async.sqlite (device 4, 2 iteration(s)) ===
253
+ reduce_grad kernels : 128 (64/iter)
254
+ reduce_grad sum : 18.623 ms/iter
255
+ reduce_grad busy (union) : 18.623 ms/iter
256
+ ... overlapped with compute: 17.866 ms/iter ( 95.9 %)
257
+ ... exposed on critical path: 0.757 ms/iter ( 4.1 %)
258
+ other GPU work (union) : 212.934 ms/iter
259
+ traced wall : 291.658 ms/iter
260
+ --> time hidden by overlap : 17.866 ms/iter
261
+
262
+ === kernels on device 5, by total time ===
263
+ 48.173 ms multi_tensor_apply_kernel
264
+ 40.620 ms device_kernel
265
+ 37.513 ms elementwise_kernel
266
+ 33.536 ms vectorized_elementwise_kernel
267
+ 28.218 ms cast_mxfp8_2D_kernel
268
+ REDUCE 25.145 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2293764096i64i64_tensorptrf32gmemalign16o2293764096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
269
+ 24.959 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
270
+ 24.237 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o40964096i64i64_tensorptrbf16gmemalign16o819204096i64i64_tensorptri32gmemalign16o81921_tensorptri32gm_0
271
+ 21.674 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
272
+ 17.538 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o229376409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
273
+ 16.586 ms unrolled_elementwise_kernel
274
+ 15.848 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o819204096i64i64_tensorptrbf16gmemalign16o40964096i64i64_tensorptri32gmemalign16o81921_tensorptri32gmem_0
275
+ 15.561 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
276
+ REDUCE 12.163 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2621441792i64i64_tensorptrf32gmemalign16o2621441792i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
277
+ 11.068 ms ncclDevKernel_AllGather_RING_LL
278
+ 8.900 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NNT
279
+ 8.413 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o262144179211i640_CopyAtom_ThrID10_TVLayoutSrc_0
280
+ 8.270 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_TNT
281
+ 7.754 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
282
+ 5.556 ms setGroupedGemmWgradArguments
283
+ 5.298 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_badd_Avec32UE8M0_Bvec32UE8M0_NTT
284
+ 5.102 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NTT
285
+ 3.939 ms triton_poi_fused__to_copy_cat_0
286
+ 3.918 ms rmsnorm_bwd_tuned_kernel
287
+ 3.893 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
288
+ 3.804 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
289
+ 3.608 ms reduce_kernel
290
+ 3.587 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_641_tensorptri32_gmem_align16_o_81921_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_alig_0
291
+ 3.455 ms CatArrayBatchedCopy_alignedK_contig
292
+ 3.428 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
293
+ 3.284 ms ncclDevKernel_AllReduce_Sum_f32_RING_LL
294
+ 2.925 ms _scatter_gather_elementwise_kernel
295
+ 2.628 ms fused_rope_backward_kernel
296
+ 2.512 ms multi_tensor_swizzle_col_scaling_kernel
297
+ 2.458 ms fused_rope_forward_kernel
298
+ 2.424 ms multi_tensor_swizzle_row_scaling_kernel
299
+ 2.414 ms gatherTopK
300
+ 2.280 ms swizzle_col_scaling_kernel
301
+ 1.866 ms rmsnorm_fwd_tuned_kernel
302
+ 1.786 ms triton_poi_fused_add_0
303
+
304
+ === G3_K2__moonep__matched__seq4096__gbs16__capnone__async.sqlite (device 5, 2 iteration(s)) ===
305
+ reduce_grad kernels : 128 (64/iter)
306
+ reduce_grad sum : 18.654 ms/iter
307
+ reduce_grad busy (union) : 18.654 ms/iter
308
+ ... overlapped with compute: 17.898 ms/iter ( 95.9 %)
309
+ ... exposed on critical path: 0.756 ms/iter ( 4.1 %)
310
+ other GPU work (union) : 214.871 ms/iter
311
+ traced wall : 289.794 ms/iter
312
+ --> time hidden by overlap : 17.898 ms/iter
313
+
314
+ === kernels on device 6, by total time ===
315
+ 48.068 ms multi_tensor_apply_kernel
316
+ 42.203 ms device_kernel
317
+ 37.506 ms elementwise_kernel
318
+ 31.623 ms vectorized_elementwise_kernel
319
+ 28.314 ms cast_mxfp8_2D_kernel
320
+ 25.109 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
321
+ REDUCE 25.067 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2293764096i64i64_tensorptrf32gmemalign16o2293764096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
322
+ 24.195 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o40964096i64i64_tensorptrbf16gmemalign16o819204096i64i64_tensorptri32gmemalign16o81921_tensorptri32gm_0
323
+ 21.257 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
324
+ 16.860 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o229376409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
325
+ 16.545 ms unrolled_elementwise_kernel
326
+ 16.045 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o819204096i64i64_tensorptrbf16gmemalign16o40964096i64i64_tensorptri32gmemalign16o81921_tensorptri32gmem_0
327
+ 14.588 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
328
+ REDUCE 12.055 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2621441792i64i64_tensorptrf32gmemalign16o2621441792i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
329
+ 9.094 ms ncclDevKernel_AllGather_RING_LL
330
+ 9.032 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NNT
331
+ 8.315 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_TNT
332
+ 7.817 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
333
+ 7.610 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o262144179211i640_CopyAtom_ThrID10_TVLayoutSrc_0
334
+ 5.434 ms setGroupedGemmWgradArguments
335
+ 5.361 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_badd_Avec32UE8M0_Bvec32UE8M0_NTT
336
+ 5.104 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NTT
337
+ 4.012 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
338
+ 3.937 ms triton_poi_fused__to_copy_cat_0
339
+ 3.915 ms rmsnorm_bwd_tuned_kernel
340
+ 3.832 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
341
+ 3.643 ms reduce_kernel
342
+ 3.629 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_641_tensorptri32_gmem_align16_o_81921_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_alig_0
343
+ 3.460 ms CatArrayBatchedCopy_alignedK_contig
344
+ 3.452 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
345
+ 3.051 ms ncclDevKernel_AllReduce_Sum_f32_RING_LL
346
+ 2.911 ms _scatter_gather_elementwise_kernel
347
+ 2.632 ms fused_rope_backward_kernel
348
+ 2.534 ms multi_tensor_swizzle_col_scaling_kernel
349
+ 2.468 ms fused_rope_forward_kernel
350
+ 2.410 ms gatherTopK
351
+ 2.409 ms multi_tensor_swizzle_row_scaling_kernel
352
+ 2.287 ms swizzle_col_scaling_kernel
353
+ 1.869 ms rmsnorm_fwd_tuned_kernel
354
+ 1.782 ms triton_poi_fused_add_0
355
+
356
+ === G3_K2__moonep__matched__seq4096__gbs16__capnone__async.sqlite (device 6, 2 iteration(s)) ===
357
+ reduce_grad kernels : 128 (64/iter)
358
+ reduce_grad sum : 18.561 ms/iter
359
+ reduce_grad busy (union) : 18.561 ms/iter
360
+ ... overlapped with compute: 17.841 ms/iter ( 96.1 %)
361
+ ... exposed on critical path: 0.720 ms/iter ( 3.9 %)
362
+ other GPU work (union) : 212.924 ms/iter
363
+ traced wall : 241.175 ms/iter
364
+ --> time hidden by overlap : 17.841 ms/iter
365
+
366
+ === kernels on device 7, by total time ===
367
+ 48.111 ms multi_tensor_apply_kernel
368
+ 41.269 ms device_kernel
369
+ 37.509 ms elementwise_kernel
370
+ 33.627 ms vectorized_elementwise_kernel
371
+ 28.293 ms cast_mxfp8_2D_kernel
372
+ REDUCE 25.317 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2293764096i64i64_tensorptrf32gmemalign16o2293764096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
373
+ 24.758 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
374
+ 24.193 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o40964096i64i64_tensorptrbf16gmemalign16o819204096i64i64_tensorptri32gmemalign16o81921_tensorptri32gm_0
375
+ 21.609 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
376
+ 18.592 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o229376409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
377
+ 16.566 ms unrolled_elementwise_kernel
378
+ 15.906 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o819204096i64i64_tensorptrbf16gmemalign16o40964096i64i64_tensorptri32gmemalign16o81921_tensorptri32gmem_0
379
+ 12.867 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
380
+ REDUCE 11.937 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o2621441792i64i64_tensorptrf32gmemalign16o2621441792i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
381
+ 9.282 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o262144179211i640_CopyAtom_ThrID10_TVLayoutSrc_0
382
+ 8.907 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NNT
383
+ 8.674 ms ncclDevKernel_AllGather_RING_LL
384
+ 8.223 ms nvjet_qqtst_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_TNT
385
+ 7.793 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
386
+ 5.490 ms setGroupedGemmWgradArguments
387
+ 5.309 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_badd_Avec32UE8M0_Bvec32UE8M0_NTT
388
+ 5.061 ms nvjet_qqsss_128x256_128x6_2x2_2cta_h_bz_Avec32UE8M0_Bvec32UE8M0_NTT
389
+ 3.921 ms triton_poi_fused__to_copy_cat_0
390
+ 3.917 ms rmsnorm_bwd_tuned_kernel
391
+ 3.847 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
392
+ 3.838 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
393
+ 3.630 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_641_tensorptri32_gmem_align16_o_81921_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_alig_0
394
+ 3.620 ms reduce_kernel
395
+ 3.463 ms CatArrayBatchedCopy_alignedK_contig
396
+ 3.387 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
397
+ 2.947 ms _scatter_gather_elementwise_kernel
398
+ 2.624 ms fused_rope_backward_kernel
399
+ 2.571 ms multi_tensor_swizzle_row_scaling_kernel
400
+ 2.538 ms multi_tensor_swizzle_col_scaling_kernel
401
+ 2.454 ms fused_rope_forward_kernel
402
+ 2.403 ms gatherTopK
403
+ 2.307 ms swizzle_col_scaling_kernel
404
+ 2.192 ms ncclDevKernel_AllReduce_Sum_f32_RING_LL
405
+ 1.874 ms rmsnorm_fwd_tuned_kernel
406
+ 1.770 ms triton_poi_fused_add_0
407
+
408
+ === G3_K2__moonep__matched__seq4096__gbs16__capnone__async.sqlite (device 7, 2 iteration(s)) ===
409
+ reduce_grad kernels : 128 (64/iter)
410
+ reduce_grad sum : 18.627 ms/iter
411
+ reduce_grad busy (union) : 18.627 ms/iter
412
+ ... overlapped with compute: 17.890 ms/iter ( 96.0 %)
413
+ ... exposed on critical path: 0.737 ms/iter ( 4.0 %)
414
+ other GPU work (union) : 213.239 ms/iter
415
+ traced wall : 291.904 ms/iter
416
+ --> time hidden by overlap : 17.890 ms/iter
granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/moonep_async_overlap_summary.txt ADDED
@@ -0,0 +1 @@
 
 
1
+ devices=8 reduce=18.631 ms/iter overlap=17.887 ms/iter (96.0%) exposed=0.744 ms/iter status=PASS
granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/run.env ADDED
@@ -0,0 +1,105 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ backend=moonep
2
+ config=G3_K2
3
+ family=tierA
4
+ source_model=
5
+ standard_num_layers=
6
+ base_rung=G3
7
+ repo=/workspace/Megatron-MoE-Echo
8
+ branch=moonep_backend_async_grad_reduce
9
+ commit=2e68c8ad
10
+ moonep_source_dir=/workspace/MoonEP
11
+ moonep_source_branch=megatron-ddp-symmetric-params
12
+ moonep_source_commit=d8bc666
13
+ num_layers=16
14
+ hidden=4096
15
+ ffn_hidden=14336
16
+ moe_ffn_hidden=1792
17
+ num_experts=64
18
+ topk=2
19
+ experts_per_rank=8
20
+ expert_params=22020096
21
+ heads=32
22
+ kv_groups=8
23
+ kv_channels=
24
+ qk_layernorm=0
25
+ shared_ffn=0
26
+ shared_expert_overlap=0
27
+ moe_token_dispatcher=flex
28
+ mla=0
29
+ q_lora_rank=
30
+ kv_lora_rank=512
31
+ qk_head_dim=128
32
+ qk_pos_emb_head_dim=64
33
+ v_head_dim=128
34
+ dense_first_layers=0
35
+ moe_layers=16
36
+ rotary_base=1000000
37
+ rotary_percent=1.0
38
+ rotary_scaling=
39
+ mscale=
40
+ mscale_all_dim=
41
+ qkv_bias=0
42
+ router_balancing=aux_loss
43
+ aux_loss_coeff=1e-2
44
+ router_pre_softmax=0
45
+ router_fusion=0
46
+ router_dtype=fp32
47
+ router_groups=
48
+ router_group_topk=
49
+ router_score_function=softmax
50
+ router_expert_bias=0
51
+ router_bias_update_rate=
52
+ router_topk_scaling_factor=
53
+ shared_expert_gate=0
54
+ apply_layernorm_1p=0
55
+ attention_output_gate=0
56
+ linear_attention_type=
57
+ linear_attention_freq=
58
+ linear_conv_kernel_dim=
59
+ linear_key_head_dim=
60
+ linear_value_head_dim=
61
+ linear_num_key_heads=
62
+ linear_num_value_heads=
63
+ mtp_num_layers=0
64
+ mtp_loss_scaling_factor=0.1
65
+ no_weight_decay_cond_type=
66
+ max_position_embeddings=32768
67
+ init_method_std=0.01
68
+ device_initiated_grouped_gemm=1
69
+ norm_epsilon=1e-5
70
+ seq_length=4096
71
+ micro_batch_size=1
72
+ global_batch_size=16
73
+ num_microbatches=2
74
+ assignments_per_rank=8192
75
+ mean_tokens_per_expert=1024
76
+ slot_policy=matched
77
+ slots_per_rank=8
78
+ capacity_mode=none
79
+ capacity=none
80
+ cuda_graph=1
81
+ probe_mode=0
82
+ async_grad_reduce=1
83
+ branch_role=timing
84
+ memory_dump_mode=none
85
+ memory_snapshot_path=
86
+ cuda_graph_warmup_steps=3
87
+ recompute=none
88
+ force_balance=0
89
+ train_iters=10
90
+ ep_size=8
91
+ echo_dispatch_overlap=1
92
+ echo_dispatch_recompute=0
93
+ ultraep_quota_min_tokens=
94
+ ultraep_weight_sync_join=
95
+ profile_enable=1
96
+ kernel_trace=1
97
+ fp8_mode=mxfp8
98
+ fp8_recipe=mxfp8
99
+ training_seed=1234
100
+ learning_rate=1e-4
101
+ min_learning_rate=1.0e-5
102
+ lr_warmup_iters=5
103
+ state_dump_enable=0
104
+ state_dump_dir=
105
+ moonep_assert_no_pending_reduce=1
granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/train.log ADDED
@@ -0,0 +1,1194 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ==============================================================
2
+ G3_K2 / moonep (slots/rank=8 policy=matched)
3
+ E=64 K=2 H'=1792 m=8
4
+ L=16 seq=4096 gbs=16 (2 microbatches)
5
+ cap_mode=none cap=none cg=1 recompute=none
6
+ memory_dump=none snapshots=/workspace/granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/memory_snapshots/snapshot.pickle
7
+ repo=/workspace/Megatron-MoE-Echo @ 2e68c8ad
8
+ ==============================================================
9
+ W0921 06:00:04.170000 142863 torch/distributed/run.py:766]
10
+ W0921 06:00:04.170000 142863 torch/distributed/run.py:766] *****************************************
11
+ W0921 06:00:04.170000 142863 torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.
12
+ W0921 06:00:04.170000 142863 torch/distributed/run.py:766] *****************************************
13
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
14
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
15
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
16
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
17
+ WARNING: one_logger package is required to enable e2e metrics tracking. please go to https://confluence.nvidia.com/display/MLWFO/Package+Repositories for details to install it
18
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
19
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
20
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
21
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
22
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
23
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
24
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
25
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
26
+ using world size: 8, data-parallel size: 8, context-parallel size: 1, hierarchical context-parallel sizes: None, tensor-model-parallel size: 1, pipeline-model-parallel size: 1
27
+ WARNING: overriding default arguments for tokenizer_type:GPT2BPETokenizer with tokenizer_type:GPT2BPETokenizer
28
+ Number of virtual stages per pipeline stage: None
29
+ accumulate and all-reduce gradients in fp32 for bfloat16 data type.
30
+ using torch.bfloat16 for parameters ...
31
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
32
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
33
+ ------------------------ arguments ------------------------
34
+ account_for_embedding_in_pipeline_split ......... False
35
+ account_for_loss_in_pipeline_split .............. False
36
+ accumulate_allreduce_grads_in_fp32 .............. True
37
+ adam_beta1 ...................................... 0.9
38
+ adam_beta2 ...................................... 0.999
39
+ adam_eps ........................................ 1e-08
40
+ add_bias_linear ................................. False
41
+ add_position_embedding .......................... False
42
+ add_qkv_bias .................................... False
43
+ adlr_autoresume ................................. False
44
+ adlr_autoresume_interval ........................ 1000
45
+ align_grad_reduce ............................... True
46
+ align_param_gather .............................. False
47
+ app_tag_run_name ................................ None
48
+ app_tag_run_version ............................. 0.0.0
49
+ apply_layernorm_1p .............................. False
50
+ apply_query_key_layer_scaling ................... False
51
+ apply_residual_connection_post_layernorm ........ False
52
+ apply_rope_fusion ............................... True
53
+ async_save ...................................... None
54
+ async_tensor_model_parallel_allreduce ........... True
55
+ attention_backend ............................... AttnBackend.auto
56
+ attention_dropout ............................... 0.0
57
+ attention_softmax_in_fp32 ....................... False
58
+ auto_detect_ckpt_format ......................... False
59
+ barrier_with_L1_time ............................ True
60
+ bert_binary_head ................................ True
61
+ bert_embedder_type .............................. megatron
62
+ bert_load ....................................... None
63
+ bf16 ............................................ True
64
+ bias_dropout_fusion ............................. True
65
+ bias_gelu_fusion ................................ False
66
+ bias_swiglu_fusion .............................. True
67
+ biencoder_projection_dim ........................ 0
68
+ biencoder_shared_query_context_model ............ False
69
+ block_data_path ................................. None
70
+ calc_ft_timeouts ................................ False
71
+ calculate_per_token_loss ........................ False
72
+ check_for_large_grads ........................... False
73
+ check_for_nan_in_loss_and_grad .................. False
74
+ check_for_spiky_loss ............................ False
75
+ check_weight_hash_across_dp_replicas_interval ... None
76
+ ckpt_assume_constant_structure .................. False
77
+ ckpt_convert_format ............................. None
78
+ ckpt_convert_save ............................... None
79
+ ckpt_convert_update_legacy_dist_opt_format ...... False
80
+ ckpt_format ..................................... torch_dist
81
+ ckpt_fully_parallel_load ........................ False
82
+ ckpt_fully_parallel_save ........................ True
83
+ ckpt_fully_parallel_save_deprecated ............. False
84
+ ckpt_step ....................................... None
85
+ classes_fraction ................................ 1.0
86
+ clip_grad ....................................... 1.0
87
+ clone_scatter_output_in_embedding ............... True
88
+ config_logger_dir ...............................
89
+ consumed_train_samples .......................... 0
90
+ consumed_valid_samples .......................... 0
91
+ context_parallel_size ........................... 1
92
+ cp_comm_type .................................... ['p2p']
93
+ create_attention_mask_in_dataloader ............. True
94
+ cross_entropy_fusion_impl ....................... native
95
+ cross_entropy_loss_fusion ....................... False
96
+ cuda_graph_scope ................................ full_iteration
97
+ cuda_graph_warmup_steps ......................... 3
98
+ data_args_path .................................. None
99
+ data_cache_path ................................. None
100
+ data_parallel_random_init ....................... False
101
+ data_parallel_sharding_strategy ................. no_shard
102
+ data_parallel_size .............................. 8
103
+ data_path ....................................... ['/workspace/data/wikitext/wikitext103_gpt2_text_document']
104
+ data_per_class_fraction ......................... 1.0
105
+ data_sharding ................................... True
106
+ dataloader_type ................................. single
107
+ ddp_average_in_collective ....................... False
108
+ ddp_bucket_size ................................. None
109
+ ddp_num_buckets ................................. None
110
+ ddp_pad_buckets_for_high_nccl_busbw ............. False
111
+ decoder_first_pipeline_num_layers ............... None
112
+ decoder_last_pipeline_num_layers ................ None
113
+ decoder_num_layers .............................. None
114
+ decoder_seq_length .............................. None
115
+ decoupled_lr .................................... None
116
+ decoupled_min_lr ................................ None
117
+ decrease_batch_size_if_needed ................... False
118
+ defer_embedding_wgrad_compute ................... False
119
+ delay_wgrad_compute ............................. False
120
+ deprecated_use_mcore_models ..................... True
121
+ deterministic_mode .............................. False
122
+ dino_bottleneck_size ............................ 256
123
+ dino_freeze_last_layer .......................... 1
124
+ dino_head_hidden_size ........................... 2048
125
+ dino_local_crops_number ......................... 10
126
+ dino_local_img_size ............................. 96
127
+ dino_norm_last_layer ............................ False
128
+ dino_teacher_temp ............................... 0.07
129
+ dino_warmup_teacher_temp ........................ 0.04
130
+ dino_warmup_teacher_temp_epochs ................. 30
131
+ disable_bf16_reduced_precision_matmul ........... False
132
+ disable_mamba_mem_eff_path ...................... False
133
+ disable_straggler_on_startup .................... False
134
+ dist_ckpt_format_deprecated ..................... None
135
+ dist_ckpt_strictness ............................ assume_ok_unexpected
136
+ distribute_saved_activations .................... False
137
+ distributed_backend ............................. nccl
138
+ distributed_timeout_minutes ..................... 10
139
+ embedding_init_method_std ....................... None
140
+ embedding_path .................................. None
141
+ empty_unused_memory_level ....................... 0
142
+ enable_cuda_graph ............................... True
143
+ enable_experimental ............................. False
144
+ enable_ft_package ............................... False
145
+ enable_gloo_process_groups ...................... True
146
+ enable_msc ...................................... True
147
+ enable_one_logger ............................... True
148
+ encoder_num_layers .............................. 16
149
+ encoder_seq_length .............................. 4096
150
+ end_weight_decay ................................ 0.1
151
+ eod_mask_loss ................................... False
152
+ error_injection_rate ............................ 0
153
+ error_injection_type ............................ transient_error
154
+ eval_interval ................................... 1000000
155
+ eval_iters ...................................... 0
156
+ evidence_data_path .............................. None
157
+ exit_duration_in_mins ........................... None
158
+ exit_interval ................................... None
159
+ exit_on_missing_checkpoint ...................... False
160
+ exit_signal_handler ............................. False
161
+ exp_avg_dtype ................................... torch.float32
162
+ exp_avg_sq_dtype ................................ torch.float32
163
+ expert_model_parallel_size ...................... 8
164
+ expert_tensor_parallel_size ..................... 1
165
+ export_force_local_attention .................... False
166
+ export_kd_cfg ................................... None
167
+ export_kd_teacher_ckpt_format ................... None
168
+ export_kd_teacher_load .......................... None
169
+ export_kv_cache_quant ........................... False
170
+ export_legacy_megatron .......................... False
171
+ export_model_type ............................... GPTModel
172
+ export_moe_apply_probs_on_input ................. False
173
+ export_qk_l2_norm ............................... False
174
+ export_quant_cfg ................................ None
175
+ export_real_quant_cfg ........................... None
176
+ export_te_mcore_model ........................... False
177
+ external_cuda_graph ............................. False
178
+ ffn_hidden_size ................................. 14336
179
+ finetune ........................................ False
180
+ finetune_data_split ............................. train
181
+ finetune_hf_dataset ............................. None
182
+ first_last_layers_bf16 .......................... False
183
+ flash_decode .................................... False
184
+ fp16 ............................................ False
185
+ fp16_lm_cross_entropy ........................... False
186
+ fp32_residual_connection ........................ False
187
+ fp8 ............................................. e4m3
188
+ fp8_amax_compute_algo ........................... most_recent
189
+ fp8_amax_history_len ............................ 1
190
+ fp8_interval .................................... 1
191
+ fp8_margin ...................................... 0
192
+ fp8_param_gather ................................ False
193
+ fp8_recipe ...................................... mxfp8
194
+ fp8_wgrad ....................................... True
195
+ fsdp_double_buffer .............................. False
196
+ global_batch_size ............................... 16
197
+ grad_reduce_in_bf16 ............................. False
198
+ gradient_accumulation_fusion .................... True
199
+ gradient_reduce_div_fusion ...................... True
200
+ group_query_attention ........................... True
201
+ head_lr_mult .................................... 1.0
202
+ heterogeneous_layers_config_encoded_json ........ None
203
+ heterogeneous_layers_config_path ................ None
204
+ hidden_dropout .................................. 0.0
205
+ hidden_size ..................................... 4096
206
+ hierarchical_context_parallel_sizes ............. None
207
+ high_priority_stream_groups ..................... []
208
+ hybrid_attention_ratio .......................... 0.0
209
+ hybrid_mlp_ratio ................................ 0.0
210
+ hybrid_override_pattern ......................... None
211
+ hysteresis ...................................... 2
212
+ ict_head_size ................................... None
213
+ ict_load ........................................ None
214
+ img_h ........................................... 224
215
+ img_w ........................................... 224
216
+ indexer_batch_size .............................. 128
217
+ indexer_log_interval ............................ 1000
218
+ inference_batch_times_seqlen_threshold .......... -1
219
+ inference_dynamic_batching ...................... False
220
+ inference_dynamic_batching_buffer_guaranteed_fraction 0.2
221
+ inference_dynamic_batching_buffer_overflow_factor None
222
+ inference_dynamic_batching_buffer_size_gb ....... 40.0
223
+ inference_dynamic_batching_chunk_size ........... 256
224
+ inference_dynamic_batching_max_requests_override None
225
+ inference_dynamic_batching_max_tokens_override .. None
226
+ inference_dynamic_batching_num_cuda_graphs ...... 16
227
+ inference_max_batch_size ........................ 8
228
+ inference_max_seq_length ........................ 2560
229
+ inference_rng_tracker ........................... False
230
+ init_method_std ................................. 0.01
231
+ init_method_xavier_uniform ...................... False
232
+ init_model_with_meta_device ..................... False
233
+ initial_loss_scale .............................. 4294967296
234
+ inprocess_active_world_size ..................... 8
235
+ inprocess_barrier_timeout ....................... 120
236
+ inprocess_completion_timeout .................... 120
237
+ inprocess_empty_cuda_cache ...................... False
238
+ inprocess_granularity ........................... node
239
+ inprocess_hard_timeout .......................... 90
240
+ inprocess_heartbeat_interval .................... 30
241
+ inprocess_heartbeat_timeout ..................... 60
242
+ inprocess_last_call_wait ........................ 1
243
+ inprocess_max_iterations ........................ None
244
+ inprocess_monitor_process_interval .............. 1.0
245
+ inprocess_monitor_thread_interval ............... 1.0
246
+ inprocess_progress_watchdog_interval ............ 1.0
247
+ inprocess_restart ............................... False
248
+ inprocess_soft_timeout .......................... 60
249
+ inprocess_termination_grace_time ................ 1
250
+ is_hybrid_model ................................. False
251
+ iter_per_epoch .................................. 1250
252
+ iterations_to_skip .............................. []
253
+ keep_fp8_transpose_cache_when_using_custom_fsdp . False
254
+ kitchen_config_file ............................. None
255
+ kitchen_recipe_number ........................... None
256
+ kv_channels ..................................... 128
257
+ kv_lora_rank .................................... 32
258
+ lazy_mpu_init ................................... None
259
+ load ............................................ None
260
+ load_main_params_from_ckpt ...................... None
261
+ load_model_opt_format ........................... False
262
+ local_rank ...................................... 0
263
+ log_energy ...................................... False
264
+ log_interval .................................... 1
265
+ log_loss_scale_to_tensorboard ................... True
266
+ log_memory_to_tensorboard ....................... False
267
+ log_num_zeros_in_grad ........................... False
268
+ log_params_norm ................................. False
269
+ log_progress .................................... False
270
+ log_straggler ................................... False
271
+ log_throughput .................................. True
272
+ log_timers_to_tensorboard ....................... False
273
+ log_validation_ppl_to_tensorboard ............... False
274
+ log_world_size_to_tensorboard ................... False
275
+ logging_level ................................... None
276
+ loss_scale ...................................... None
277
+ loss_scale_window ............................... 1000
278
+ lr .............................................. 0.0001
279
+ lr_decay_iters .................................. 8
280
+ lr_decay_samples ................................ None
281
+ lr_decay_style .................................. cosine
282
+ lr_warmup_fraction .............................. None
283
+ lr_warmup_init .................................. 0.0
284
+ lr_warmup_iters ................................. 5
285
+ lr_warmup_samples ............................... 0
286
+ lr_wsd_decay_iters .............................. None
287
+ lr_wsd_decay_samples ............................ None
288
+ lr_wsd_decay_style .............................. exponential
289
+ main_grads_dtype ................................ torch.float32
290
+ main_params_dtype ............................... torch.float32
291
+ make_vocab_size_divisible_by .................... 128
292
+ mamba_head_dim .................................. 64
293
+ mamba_num_groups ................................ 8
294
+ mamba_num_heads ................................. None
295
+ mamba_state_dim ................................. 128
296
+ manual_gc ....................................... False
297
+ manual_gc_eval .................................. True
298
+ manual_gc_interval .............................. 0
299
+ mask_factor ..................................... 1.0
300
+ mask_prob ....................................... 0.15
301
+ mask_type ....................................... random
302
+ masked_softmax_fusion ........................... False
303
+ max_position_embeddings ......................... 32768
304
+ max_tokens_to_oom ............................... 12000
305
+ memory_snapshot_path ............................ snapshot.pickle
306
+ merge_file ...................................... /workspace/data/wikitext/gpt2-merges.txt
307
+ micro_batch_size ................................ 1
308
+ microbatch_group_size_per_vp_stage .............. None
309
+ mid_level_dataset_surplus ....................... 0.005
310
+ min_loss_scale .................................. 1.0
311
+ min_lr .......................................... 1e-05
312
+ mlp_chunks_for_prefill .......................... 1
313
+ mmap_bin_files .................................. True
314
+ mock_data ....................................... False
315
+ moe_apply_probs_on_input ........................ False
316
+ moe_aux_loss_coeff .............................. 0.01
317
+ moe_deepep_num_sms .............................. 20
318
+ moe_echo_enable_random_offloading ............... False
319
+ moe_echo_expert_dispatch_overlap ................ False
320
+ moe_echo_expert_dispatcher_type ................. hybridep
321
+ moe_echo_recompute_expert_dispatch .............. False
322
+ moe_enable_deepep ............................... False
323
+ moe_enable_echo ................................. False
324
+ moe_enable_hybridep ............................. False
325
+ moe_enable_moonep ............................... True
326
+ moe_expert_capacity_factor ...................... None
327
+ moe_extended_tp ................................. False
328
+ moe_ffn_hidden_size ............................. 1792
329
+ moe_grouped_gemm ................................ True
330
+ moe_hybridep_num_sms ............................ 16
331
+ moe_input_jitter_eps ............................ None
332
+ moe_layer_freq .................................. 1
333
+ moe_layer_recompute ............................. False
334
+ moe_moonep_async_grad_reduce .................... True
335
+ moe_moonep_num_sms .............................. None
336
+ moe_num_echo_experts ............................ None
337
+ moe_pad_expert_input_to_capacity ................ False
338
+ moe_per_layer_logging ........................... False
339
+ moe_permute_fusion .............................. False
340
+ moe_received_token_capacity ..................... None
341
+ moe_router_bias_update_rate ..................... 0.001
342
+ moe_router_dtype ................................ fp32
343
+ moe_router_enable_expert_bias ................... False
344
+ moe_router_force_load_balancing ................. False
345
+ moe_router_group_topk ........................... None
346
+ moe_router_load_balancing_type .................. aux_loss
347
+ moe_router_num_groups ........................... None
348
+ moe_router_padding_for_fp8 ...................... False
349
+ moe_router_pre_softmax .......................... False
350
+ moe_router_score_function ....................... softmax
351
+ moe_router_topk ................................. 2
352
+ moe_router_topk_scaling_factor .................. None
353
+ moe_shared_expert_intermediate_size ............. None
354
+ moe_shared_expert_overlap ....................... False
355
+ moe_token_dispatcher_type ....................... flex
356
+ moe_token_drop_policy ........................... probs
357
+ moe_upcycling_granularity ....................... 1
358
+ moe_use_device_initiated_grouped_gemm ........... True
359
+ moe_use_legacy_grouped_gemm ..................... False
360
+ moe_use_upcycling ............................... False
361
+ moe_z_loss_coeff ................................ None
362
+ mrope_section ................................... None
363
+ mscale .......................................... 1.0
364
+ mscale_all_dim .................................. 0.0
365
+ mtp_loss_scaling_factor ......................... 0.1
366
+ mtp_num_layers .................................. None
367
+ multi_latent_attention .......................... False
368
+ nccl_all_reduce_for_prefill ..................... False
369
+ nccl_communicator_config_path ................... None
370
+ nccl_ub ......................................... False
371
+ no_load_optim ................................... None
372
+ no_load_rng ..................................... None
373
+ no_persist_layer_norm ........................... False
374
+ no_rope_freq .................................... None
375
+ no_save_optim ................................... None
376
+ no_save_rng ..................................... None
377
+ non_persistent_ckpt_type ........................ None
378
+ non_persistent_global_ckpt_dir .................. None
379
+ non_persistent_local_ckpt_algo .................. fully_parallel
380
+ non_persistent_local_ckpt_dir ................... None
381
+ non_persistent_save_interval .................... None
382
+ norm_epsilon .................................... 1e-05
383
+ normalization ................................... RMSNorm
384
+ num_attention_heads ............................. 32
385
+ num_channels .................................... 3
386
+ num_classes ..................................... 1000
387
+ num_dataset_builder_threads ..................... 1
388
+ num_distributed_optimizer_instances ............. 1
389
+ num_experts ..................................... 64
390
+ num_layers ...................................... 16
391
+ num_layers_at_end_in_bf16 ....................... 1
392
+ num_layers_at_start_in_bf16 ..................... 1
393
+ num_layers_per_virtual_pipeline_stage ........... None
394
+ num_query_groups ................................ 8
395
+ num_virtual_stages_per_pipeline_rank ............ None
396
+ num_workers ..................................... 2
397
+ object_storage_cache_path ....................... None
398
+ one_logger_async ................................ False
399
+ one_logger_project .............................. megatron-lm
400
+ one_logger_run_name ............................. None
401
+ onnx_safe ....................................... None
402
+ openai_gelu ..................................... False
403
+ optimizer ....................................... adam
404
+ optimizer_cpu_offload ........................... False
405
+ optimizer_offload_fraction ...................... 1.0
406
+ output_bert_embeddings .......................... False
407
+ overlap_cpu_optimizer_d2h_h2d ................... False
408
+ overlap_grad_reduce ............................. True
409
+ overlap_moe_expert_parallel_comm ................ False
410
+ overlap_p2p_comm ................................ False
411
+ overlap_p2p_comm_warmup_flush ................... False
412
+ overlap_param_gather ............................ True
413
+ overlap_param_gather_with_optimizer_step ........ False
414
+ override_opt_param_scheduler .................... False
415
+ params_dtype .................................... torch.bfloat16
416
+ patch_dim ....................................... 16
417
+ per_split_data_args_path ........................ None
418
+ perform_initialization .......................... True
419
+ pin_cpu_grads ................................... True
420
+ pin_cpu_params .................................. True
421
+ pipeline_model_parallel_comm_backend ............ None
422
+ pipeline_model_parallel_layout .................. None
423
+ pipeline_model_parallel_size .................... 1
424
+ position_embedding_type ......................... rope
425
+ pretrained_checkpoint ........................... None
426
+ profile ......................................... True
427
+ profile_ranks ................................... [0, 1, 2, 3, 4, 5, 6, 7]
428
+ profile_step_end ................................ 8
429
+ profile_step_start .............................. 6
430
+ q_lora_rank ..................................... None
431
+ qk_head_dim ..................................... 128
432
+ qk_l2_norm ...................................... False
433
+ qk_layernorm .................................... False
434
+ qk_pos_emb_head_dim ............................. 64
435
+ query_in_block_prob ............................. 0.1
436
+ rampup_batch_size ............................... None
437
+ rank ............................................ 0
438
+ recompute_granularity ........................... None
439
+ recompute_method ................................ None
440
+ recompute_modules ............................... None
441
+ recompute_num_layers ............................ None
442
+ record_memory_history ........................... False
443
+ relative_attention_max_distance ................. 128
444
+ relative_attention_num_buckets .................. 32
445
+ replication ..................................... False
446
+ replication_factor .............................. 2
447
+ replication_jump ................................ None
448
+ rerun_mode ...................................... validate_results
449
+ reset_attention_mask ............................ False
450
+ reset_position_ids .............................. False
451
+ result_rejected_tracker_filename ................ None
452
+ retriever_report_topk_accuracies ................ []
453
+ retriever_score_scaling ......................... False
454
+ retriever_seq_length ............................ 256
455
+ retro_add_retriever ............................. False
456
+ retro_attention_gate ............................ 1
457
+ retro_cyclic_train_iters ........................ None
458
+ retro_encoder_attention_dropout ................. 0.1
459
+ retro_encoder_hidden_dropout .................... 0.1
460
+ retro_encoder_layers ............................ 2
461
+ retro_num_neighbors ............................. 2
462
+ retro_num_retrieved_chunks ...................... 2
463
+ retro_project_dir ............................... None
464
+ retro_verify_neighbor_count ..................... True
465
+ reuse_grad_buf_for_mxfp8_param_ag ............... False
466
+ rope_scaling_factor ............................. 8.0
467
+ rope_type ....................................... None
468
+ rotary_base ..................................... 1000000
469
+ rotary_interleaved .............................. False
470
+ rotary_percent .................................. 1.0
471
+ rotary_scaling_factor ........................... 1.0
472
+ rotary_seq_len_interpolation_factor ............. None
473
+ run_workload_inspector_server ................... False
474
+ sample_rate ..................................... 1.0
475
+ save ............................................ None
476
+ save_interval ................................... None
477
+ save_retain_interval ............................ None
478
+ scatter_gather_tensors_in_pipeline .............. True
479
+ seed ............................................ 1234
480
+ seq_length ...................................... 4096
481
+ sequence_parallel ............................... False
482
+ sft ............................................. False
483
+ sft_tokenizer_prompt_format ..................... nemotron-h-aligned
484
+ sgd_momentum .................................... 0.9
485
+ short_seq_prob .................................. 0.1
486
+ skip_train ...................................... False
487
+ skipped_train_samples ........................... 0
488
+ spec ............................................ None
489
+ split ........................................... 100,0,0
490
+ squared_relu .................................... False
491
+ start_weight_decay .............................. 0.1
492
+ straggler_ctrlr_port ............................ 65535
493
+ straggler_minmax_count .......................... 1
494
+ suggested_communication_unit_size ............... None
495
+ swiglu .......................................... True
496
+ swin_backbone_type .............................. tiny
497
+ symmetric_ar_type ............................... None
498
+ te_rng_tracker .................................. True
499
+ tensor_model_parallel_size ...................... 1
500
+ tensorboard_dir ................................. None
501
+ tensorboard_log_interval ........................ 1
502
+ tensorboard_queue_size .......................... 1000
503
+ test_data_path .................................. None
504
+ test_mode ....................................... False
505
+ tiktoken_num_special_tokens ..................... 1000
506
+ tiktoken_pattern ................................ None
507
+ tiktoken_special_tokens ......................... None
508
+ timing_log_level ................................ 0
509
+ timing_log_option ............................... minmax
510
+ titles_data_path ................................ None
511
+ tokenizer_model ................................. None
512
+ tokenizer_type .................................. GPT2BPETokenizer
513
+ torch_fsdp2_reshard_after_forward ............... True
514
+ tp_comm_bootstrap_backend ....................... nccl
515
+ tp_comm_bulk_dgrad .............................. True
516
+ tp_comm_bulk_wgrad .............................. True
517
+ tp_comm_overlap ................................. False
518
+ tp_comm_overlap_ag .............................. True
519
+ tp_comm_overlap_cfg ............................. None
520
+ tp_comm_overlap_rs .............................. True
521
+ tp_comm_overlap_rs_dgrad ........................ False
522
+ tp_comm_split_ag ................................ True
523
+ tp_comm_split_rs ................................ True
524
+ train_data_path ................................. None
525
+ train_iters ..................................... 10
526
+ train_samples ................................... None
527
+ train_sync_interval ............................. None
528
+ transformer_impl ................................ transformer_engine
529
+ transformer_pipeline_model_parallel_size ........ 1
530
+ untie_embeddings_and_output_weights ............. True
531
+ use_checkpoint_args ............................. False
532
+ use_checkpoint_opt_param_scheduler .............. False
533
+ use_cpu_initialization .......................... None
534
+ use_custom_fsdp ................................. False
535
+ use_dist_ckpt ................................... True
536
+ use_dist_ckpt_deprecated ........................ False
537
+ use_distributed_optimizer ....................... True
538
+ use_flash_attn .................................. False
539
+ use_legacy_models ............................... False
540
+ use_mp_args_from_checkpoint_args ................ False
541
+ use_one_sent_docs ............................... False
542
+ use_persistent_ckpt_worker ...................... False
543
+ use_precision_aware_optimizer ................... False
544
+ use_pytorch_profiler ............................ False
545
+ use_ring_exchange_p2p ........................... False
546
+ use_rope_scaling ................................ False
547
+ use_rotary_position_embeddings .................. False
548
+ use_sharp ....................................... False
549
+ use_tokenizer_model_from_checkpoint_args ........ True
550
+ use_torch_fsdp2 ................................. False
551
+ use_torch_optimizer_for_cpu_offload ............. False
552
+ use_tp_pp_dp_mapping ............................ False
553
+ v_head_dim ...................................... 128
554
+ valid_data_path ................................. None
555
+ variable_seq_lengths ............................ False
556
+ virtual_pipeline_model_parallel_size ............ None
557
+ vision_backbone_type ............................ vit
558
+ vision_pretraining .............................. False
559
+ vision_pretraining_type ......................... classify
560
+ vocab_extra_ids ................................. 0
561
+ vocab_file ...................................... /workspace/data/wikitext/gpt2-vocab.json
562
+ vocab_size ...................................... None
563
+ wandb_exp_name ..................................
564
+ wandb_project ...................................
565
+ wandb_save_dir ..................................
566
+ weight_decay .................................... 0.1
567
+ weight_decay_incr_style ......................... constant
568
+ wgrad_deferral_limit ............................ 0
569
+ world_size ...................................... 8
570
+ yaml_cfg ........................................ None
571
+ -------------------- end of arguments ---------------------
572
+ INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 2
573
+ > building GPT2BPETokenizer tokenizer ...
574
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
575
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
576
+ > padded vocab (size: 50257) with 47 dummy tokens (new size: 50304)
577
+ WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.VALIDATE_RESULTS
578
+ > initializing torch distributed ...
579
+ [Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
580
+ [Gloo] Rank 5 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
581
+ [Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
582
+ [Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
583
+ [Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
584
+ [Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
585
+ [Gloo] Rank [Gloo] Rank 6 is connected to 7 peer ranks. 7Expected number of connected peer ranks is : is connected to 77 peer ranks. Expected number of connected peer ranks is :
586
+ 7
587
+ [Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
588
+ [Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
589
+ [Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
590
+ [Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
591
+ [Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
592
+ [Gloo] Rank 5 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
593
+ [Gloo] Rank 7 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
594
+ [Gloo] Rank 6 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
595
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
596
+ > initialized tensor model parallel with size 1
597
+ > initialized pipeline model parallel with size 1
598
+ > setting random seeds to 1234 ...
599
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
600
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
601
+ [Gloo] Rank 0 is connected to 0[Gloo] Rank peer ranks. Expected number of connected peer ranks is : 00
602
+ is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
603
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
604
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
605
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
606
+ [rank4]:[W921 06:00:17.114091013 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
607
+ [rank1]:[W921 06:00:17.114371390 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
608
+ [rank5]:[W921 06:00:17.114931094 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
609
+ [rank7]:[W921 06:00:17.114948922 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
610
+ [rank6]:[W921 06:00:17.115143652 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
611
+ [rank3]:[W921 06:00:17.115192315 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
612
+ [rank2]:[W921 06:00:17.115274188 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
613
+ > compiling dataset index builder ...
614
+ make: Entering directory '/workspace/Megatron-MoE-Echo/megatron/core/datasets'
615
+ make: Nothing to be done for 'default'.
616
+ make: Leaving directory '/workspace/Megatron-MoE-Echo/megatron/core/datasets'
617
+ >>> done with dataset index builder. Compilation time: 0.429 seconds
618
+ WARNING: constraints for invoking optimized fused softmax kernel are not met. We default back to unfused kernel invocations.
619
+ > compiling and loading fused kernels ...
620
+ [rank0]:[W921 06:00:17.721691351 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
621
+ >>> done with compiling and loading fused kernels. Compilation time: 3.571 seconds
622
+ time to initialize megatron (seconds): 9.560
623
+ [after megatron is initialized] datetime: 2026-09-21 06:00:23
624
+ building GPT model ...
625
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
626
+ warnings.warn(
627
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
628
+ warnings.warn(
629
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
630
+ warnings.warn(
631
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
632
+ warnings.warn(
633
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
634
+ warnings.warn(
635
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
636
+ warnings.warn(
637
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
638
+ warnings.warn(
639
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
640
+ warnings.warn(
641
+ > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 3906080768
642
+ INFO:megatron.core.distributed.distributed_data_parallel:Setting up DistributedDataParallel with config DistributedDataParallelConfig(grad_reduce_in_fp32=True, overlap_grad_reduce=True, overlap_param_gather=True, align_param_gather=False, use_distributed_optimizer=True, num_distributed_optimizer_instances=1, check_for_nan_in_grad=False, check_for_large_grads=False, bucket_size=40000000, pad_buckets_for_high_nccl_busbw=False, average_in_collective=False, fp8_param_gather=False, reuse_grad_buf_for_mxfp8_param_ag=False, use_custom_fsdp=False, data_parallel_sharding_strategy='no_shard', gradient_reduce_div_fusion=True, suggested_communication_unit_size=None, preserve_fp32_weights=True, keep_fp8_transpose_cache_when_using_custom_fsdp=False, nccl_ub=False, fsdp_double_buffer=False)
643
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 18
644
+ Params for bucket 1 (206045184 elements, 206045184 padded size):
645
+ module.output_layer.weight
646
+ Params for bucket 2 (42217472 elements, 42217472 padded size):
647
+ module.decoder.final_layernorm.weight
648
+ module.decoder.layers.15.self_attention.linear_qkv.weight
649
+ module.decoder.layers.15.pre_mlp_layernorm.weight
650
+ module.decoder.layers.15.self_attention.linear_qkv.layer_norm_weight
651
+ module.decoder.layers.15.mlp.router.weight
652
+ module.decoder.layers.15.self_attention.linear_proj.weight
653
+ Params for bucket 3 (42213376 elements, 42213376 padded size):
654
+ module.decoder.layers.14.pre_mlp_layernorm.weight
655
+ module.decoder.layers.14.self_attention.linear_qkv.weight
656
+ module.decoder.layers.14.mlp.router.weight
657
+ module.decoder.layers.14.self_attention.linear_proj.weight
658
+ module.decoder.layers.14.self_attention.linear_qkv.layer_norm_weight
659
+ Params for bucket 4 (42213376 elements, 42213376 padded size):
660
+ module.decoder.layers.13.pre_mlp_layernorm.weight
661
+ module.decoder.layers.13.self_attention.linear_qkv.weight
662
+ module.decoder.layers.13.mlp.router.weight
663
+ module.decoder.layers.13.self_attention.linear_proj.weight
664
+ module.decoder.layers.13.self_attention.linear_qkv.layer_norm_weight
665
+ Params for bucket 5 (42213376 elements, 42213376 padded size):
666
+ module.decoder.layers.12.mlp.router.weight
667
+ module.decoder.layers.12.self_attention.linear_qkv.layer_norm_weight
668
+ module.decoder.layers.12.self_attention.linear_proj.weight
669
+ module.decoder.layers.12.pre_mlp_layernorm.weight
670
+ module.decoder.layers.12.self_attention.linear_qkv.weight
671
+ Params for bucket 6 (42213376 elements, 42213376 padded size):
672
+ module.decoder.layers.11.mlp.router.weight
673
+ module.decoder.layers.11.self_attention.linear_proj.weight
674
+ module.decoder.layers.11.self_attention.linear_qkv.layer_norm_weight
675
+ module.decoder.layers.11.self_attention.linear_qkv.weight
676
+ module.decoder.layers.11.pre_mlp_layernorm.weight
677
+ Params for bucket 7 (42213376 elements, 42213376 padded size):
678
+ module.decoder.layers.10.self_attention.linear_qkv.weight
679
+ module.decoder.layers.10.pre_mlp_layernorm.weight
680
+ module.decoder.layers.10.mlp.router.weight
681
+ module.decoder.layers.10.self_attention.linear_proj.weight
682
+ module.decoder.layers.10.self_attention.linear_qkv.layer_norm_weight
683
+ Params for bucket 8 (42213376 elements, 42213376 padded size):
684
+ module.decoder.layers.9.mlp.router.weight
685
+ module.decoder.layers.9.self_attention.linear_proj.weight
686
+ module.decoder.layers.9.self_attention.linear_qkv.layer_norm_weight
687
+ module.decoder.layers.9.pre_mlp_layernorm.weight
688
+ module.decoder.layers.9.self_attention.linear_qkv.weight
689
+ Params for bucket 9 (42213376 elements, 42213376 padded size):
690
+ module.decoder.layers.8.self_attention.linear_qkv.weight
691
+ module.decoder.layers.8.pre_mlp_layernorm.weight
692
+ module.decoder.layers.8.mlp.router.weight
693
+ module.decoder.layers.8.self_attention.linear_proj.weight
694
+ module.decoder.layers.8.self_attention.linear_qkv.layer_norm_weight
695
+ Params for bucket 10 (42213376 elements, 42213376 padded size):
696
+ module.decoder.layers.7.self_attention.linear_qkv.weight
697
+ module.decoder.layers.7.pre_mlp_layernorm.weight
698
+ module.decoder.layers.7.mlp.router.weight
699
+ module.decoder.layers.7.self_attention.linear_proj.weight
700
+ module.decoder.layers.7.self_attention.linear_qkv.layer_norm_weight
701
+ Params for bucket 11 (42213376 elements, 42213376 padded size):
702
+ module.decoder.layers.6.mlp.router.weight
703
+ module.decoder.layers.6.self_attention.linear_proj.weight
704
+ module.decoder.layers.6.self_attention.linear_qkv.layer_norm_weight
705
+ module.decoder.layers.6.self_attention.linear_qkv.weight
706
+ module.decoder.layers.6.pre_mlp_layernorm.weight
707
+ Params for bucket 12 (42213376 elements, 42213376 padded size):
708
+ module.decoder.layers.5.mlp.router.weight
709
+ module.decoder.layers.5.self_attention.linear_qkv.layer_norm_weight
710
+ module.decoder.layers.5.self_attention.linear_qkv.weight
711
+ module.decoder.layers.5.self_attention.linear_proj.weight
712
+ module.decoder.layers.5.pre_mlp_layernorm.weight
713
+ Params for bucket 13 (42213376 elements, 42213376 padded size):
714
+ module.decoder.layers.4.self_attention.linear_qkv.weight
715
+ module.decoder.layers.4.pre_mlp_layernorm.weight
716
+ module.decoder.layers.4.mlp.router.weight
717
+ module.decoder.layers.4.self_attention.linear_qkv.layer_norm_weight
718
+ module.decoder.layers.4.self_attention.linear_proj.weight
719
+ Params for bucket 14 (42213376 elements, 42213376 padded size):
720
+ module.decoder.layers.3.self_attention.linear_qkv.weight
721
+ module.decoder.layers.3.pre_mlp_layernorm.weight
722
+ module.decoder.layers.3.mlp.router.weight
723
+ module.decoder.layers.3.self_attention.linear_qkv.layer_norm_weight
724
+ module.decoder.layers.3.self_attention.linear_proj.weight
725
+ Params for bucket 15 (42213376 elements, 42213376 padded size):
726
+ module.decoder.layers.2.pre_mlp_layernorm.weight
727
+ module.decoder.layers.2.self_attention.linear_qkv.weight
728
+ module.decoder.layers.2.mlp.router.weight
729
+ module.decoder.layers.2.self_attention.linear_proj.weight
730
+ module.decoder.layers.2.self_attention.linear_qkv.layer_norm_weight
731
+ Params for bucket 16 (42213376 elements, 42213376 padded size):
732
+ module.decoder.layers.1.mlp.router.weight
733
+ module.decoder.layers.1.self_attention.linear_proj.weight
734
+ module.decoder.layers.1.self_attention.linear_qkv.layer_norm_weight
735
+ module.decoder.layers.1.pre_mlp_layernorm.weight
736
+ module.decoder.layers.1.self_attention.linear_qkv.weight
737
+ Params for bucket 17 (42213376 elements, 42213376 padded size):
738
+ module.decoder.layers.0.mlp.router.weight
739
+ module.decoder.layers.0.pre_mlp_layernorm.weight
740
+ module.decoder.layers.0.self_attention.linear_qkv.layer_norm_weight
741
+ module.decoder.layers.0.self_attention.linear_proj.weight
742
+ module.decoder.layers.0.self_attention.linear_qkv.weight
743
+ Params for bucket 18 (206045184 elements, 206045184 padded size):
744
+ module.embedding.word_embeddings.weight
745
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 64
746
+ Params for bucket 1 (44040192 elements, 44040192 padded size):
747
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight1
748
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight3
749
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight5
750
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight0
751
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight4
752
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight2
753
+ Params for bucket 2 (44040192 elements, 44040192 padded size):
754
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight6
755
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight7
756
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight0
757
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight1
758
+ Params for bucket 3 (44040192 elements, 44040192 padded size):
759
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight2
760
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight3
761
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight4
762
+ Params for bucket 4 (44040192 elements, 44040192 padded size):
763
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight5
764
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight6
765
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight7
766
+ Params for bucket 5 (44040192 elements, 44040192 padded size):
767
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight0
768
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight2
769
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight4
770
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight3
771
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight1
772
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight5
773
+ Params for bucket 6 (44040192 elements, 44040192 padded size):
774
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight6
775
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight0
776
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight7
777
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight1
778
+ Params for bucket 7 (44040192 elements, 44040192 padded size):
779
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight2
780
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight3
781
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight4
782
+ Params for bucket 8 (44040192 elements, 44040192 padded size):
783
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight5
784
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight6
785
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight7
786
+ Params for bucket 9 (44040192 elements, 44040192 padded size):
787
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight3
788
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight1
789
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight5
790
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight0
791
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight4
792
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight2
793
+ Params for bucket 10 (44040192 elements, 44040192 padded size):
794
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight6
795
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight0
796
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight1
797
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight7
798
+ Params for bucket 11 (44040192 elements, 44040192 padded size):
799
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight2
800
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight4
801
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight3
802
+ Params for bucket 12 (44040192 elements, 44040192 padded size):
803
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight5
804
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight7
805
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight6
806
+ Params for bucket 13 (44040192 elements, 44040192 padded size):
807
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight0
808
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight4
809
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight2
810
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight1
811
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight3
812
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight5
813
+ Params for bucket 14 (44040192 elements, 44040192 padded size):
814
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight6
815
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight0
816
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight7
817
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight1
818
+ Params for bucket 15 (44040192 elements, 44040192 padded size):
819
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight2
820
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight3
821
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight4
822
+ Params for bucket 16 (44040192 elements, 44040192 padded size):
823
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight5
824
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight7
825
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight6
826
+ Params for bucket 17 (44040192 elements, 44040192 padded size):
827
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight3
828
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight1
829
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight5
830
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight0
831
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight4
832
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight2
833
+ Params for bucket 18 (44040192 elements, 44040192 padded size):
834
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight6
835
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight0
836
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight7
837
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight1
838
+ Params for bucket 19 (44040192 elements, 44040192 padded size):
839
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight2
840
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight4
841
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight3
842
+ Params for bucket 20 (44040192 elements, 44040192 padded size):
843
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight5
844
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight7
845
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight6
846
+ Params for bucket 21 (44040192 elements, 44040192 padded size):
847
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight3
848
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight1
849
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight5
850
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight0
851
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight4
852
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight2
853
+ Params for bucket 22 (44040192 elements, 44040192 padded size):
854
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight6
855
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight0
856
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight7
857
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight1
858
+ Params for bucket 23 (44040192 elements, 44040192 padded size):
859
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight2
860
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight3
861
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight4
862
+ Params for bucket 24 (44040192 elements, 44040192 padded size):
863
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight5
864
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight6
865
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight7
866
+ Params for bucket 25 (44040192 elements, 44040192 padded size):
867
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight0
868
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight2
869
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight4
870
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight3
871
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight1
872
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight5
873
+ Params for bucket 26 (44040192 elements, 44040192 padded size):
874
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight6
875
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight0
876
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight7
877
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight1
878
+ Params for bucket 27 (44040192 elements, 44040192 padded size):
879
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight2
880
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight3
881
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight4
882
+ Params for bucket 28 (44040192 elements, 44040192 padded size):
883
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight5
884
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight6
885
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight7
886
+ Params for bucket 29 (44040192 elements, 44040192 padded size):
887
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight3
888
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight1
889
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight5
890
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight0
891
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight4
892
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight2
893
+ Params for bucket 30 (44040192 elements, 44040192 padded size):
894
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight6
895
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight0
896
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight7
897
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight1
898
+ Params for bucket 31 (44040192 elements, 44040192 padded size):
899
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight2
900
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight3
901
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight4
902
+ Params for bucket 32 (44040192 elements, 44040192 padded size):
903
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight5
904
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight6
905
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight7
906
+ Params for bucket 33 (44040192 elements, 44040192 padded size):
907
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight3
908
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight1
909
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight5
910
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight0
911
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight2
912
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight4
913
+ Params for bucket 34 (44040192 elements, 44040192 padded size):
914
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight6
915
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight0
916
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight7
917
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight1
918
+ Params for bucket 35 (44040192 elements, 44040192 padded size):
919
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight2
920
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight3
921
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight4
922
+ Params for bucket 36 (44040192 elements, 44040192 padded size):
923
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight5
924
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight6
925
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight7
926
+ Params for bucket 37 (44040192 elements, 44040192 padded size):
927
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight0
928
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight4
929
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight2
930
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight3
931
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight1
932
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight5
933
+ Params for bucket 38 (44040192 elements, 44040192 padded size):
934
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight6
935
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight0
936
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight7
937
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight1
938
+ Params for bucket 39 (44040192 elements, 44040192 padded size):
939
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight2
940
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight3
941
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight4
942
+ Params for bucket 40 (44040192 elements, 44040192 padded size):
943
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight5
944
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight6
945
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight7
946
+ Params for bucket 41 (44040192 elements, 44040192 padded size):
947
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight3
948
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight1
949
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight5
950
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight0
951
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight4
952
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight2
953
+ Params for bucket 42 (44040192 elements, 44040192 padded size):
954
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight6
955
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight0
956
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight7
957
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight1
958
+ Params for bucket 43 (44040192 elements, 44040192 padded size):
959
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight2
960
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight3
961
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight4
962
+ Params for bucket 44 (44040192 elements, 44040192 padded size):
963
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight5
964
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight6
965
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight7
966
+ Params for bucket 45 (44040192 elements, 44040192 padded size):
967
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight0
968
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight4
969
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight2
970
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight3
971
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight1
972
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight5
973
+ Params for bucket 46 (44040192 elements, 44040192 padded size):
974
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight6
975
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight7
976
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight0
977
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight1
978
+ Params for bucket 47 (44040192 elements, 44040192 padded size):
979
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight2
980
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight3
981
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight4
982
+ Params for bucket 48 (44040192 elements, 44040192 padded size):
983
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight5
984
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight6
985
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight7
986
+ Params for bucket 49 (44040192 elements, 44040192 padded size):
987
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight0
988
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight4
989
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight2
990
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight3
991
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight1
992
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight5
993
+ Params for bucket 50 (44040192 elements, 44040192 padded size):
994
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight6
995
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight1
996
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight0
997
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight7
998
+ Params for bucket 51 (44040192 elements, 44040192 padded size):
999
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight2
1000
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight3
1001
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight4
1002
+ Params for bucket 52 (44040192 elements, 44040192 padded size):
1003
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight5
1004
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight6
1005
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight7
1006
+ Params for bucket 53 (44040192 elements, 44040192 padded size):
1007
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight0
1008
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight2
1009
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight4
1010
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight1
1011
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight3
1012
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight5
1013
+ Params for bucket 54 (44040192 elements, 44040192 padded size):
1014
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight6
1015
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight0
1016
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight1
1017
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight7
1018
+ Params for bucket 55 (44040192 elements, 44040192 padded size):
1019
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight2
1020
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight4
1021
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight3
1022
+ Params for bucket 56 (44040192 elements, 44040192 padded size):
1023
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight5
1024
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight7
1025
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight6
1026
+ Params for bucket 57 (44040192 elements, 44040192 padded size):
1027
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight0
1028
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight4
1029
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight2
1030
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight1
1031
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight3
1032
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight5
1033
+ Params for bucket 58 (44040192 elements, 44040192 padded size):
1034
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight6
1035
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight0
1036
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight7
1037
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight1
1038
+ Params for bucket 59 (44040192 elements, 44040192 padded size):
1039
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight2
1040
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight3
1041
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight4
1042
+ Params for bucket 60 (44040192 elements, 44040192 padded size):
1043
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight5
1044
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight7
1045
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight6
1046
+ Params for bucket 61 (44040192 elements, 44040192 padded size):
1047
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight0
1048
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight2
1049
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight4
1050
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight3
1051
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight1
1052
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight5
1053
+ Params for bucket 62 (44040192 elements, 44040192 padded size):
1054
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight6
1055
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight1
1056
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight0
1057
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight7
1058
+ Params for bucket 63 (44040192 elements, 44040192 padded size):
1059
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight2
1060
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight3
1061
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight4
1062
+ Params for bucket 64 (44040192 elements, 44040192 padded size):
1063
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight5
1064
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight6
1065
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight7
1066
+ INFO:megatron.core.optimizer:Setting up optimizer with config OptimizerConfig(optimizer='adam', lr=0.0001, min_lr=1e-05, decoupled_lr=None, decoupled_min_lr=None, weight_decay=0.1, fp8_recipe='mxfp8', fp16=False, bf16=True, reuse_grad_buf_for_mxfp8_param_ag=False, params_dtype=torch.bfloat16, use_precision_aware_optimizer=False, store_param_remainders=True, main_grads_dtype=torch.float32, main_params_dtype=torch.float32, exp_avg_dtype=torch.float32, exp_avg_sq_dtype=torch.float32, loss_scale=None, initial_loss_scale=4294967296, min_loss_scale=1.0, loss_scale_window=1000, hysteresis=2, adam_beta1=0.9, adam_beta2=0.999, adam_eps=1e-08, sgd_momentum=0.9, use_distributed_optimizer=True, overlap_param_gather_with_optimizer_step=False, optimizer_cpu_offload=False, optimizer_offload_fraction=1.0, use_torch_optimizer_for_cpu_offload=False, overlap_cpu_optimizer_d2h_h2d=False, pin_cpu_grads=True, pin_cpu_params=True, clip_grad=1.0, log_num_zeros_in_grad=False, barrier_with_L1_time=True, timers=<megatron.core.timers.Timers object at 0x70dedbfee9f0>, config_logger_dir='')
1067
+ INFO:megatron.core.optimizer_param_scheduler:> learning rate decay style: cosine
1068
+ [after model, optimizer, and learning rate scheduler are built] datetime: 2026-09-21 06:00:26
1069
+ > building train, validation, and test datasets ...
1070
+ > datasets target sizes (minimum size):
1071
+ train: 160
1072
+ validation: 0
1073
+ test: 0
1074
+ INFO:megatron.core.datasets.blended_megatron_dataset_config:Let split_matrix = [(0, 1.0), None, None]
1075
+ > building train, validation, and test datasets for GPT ...
1076
+ INFO:megatron.core.datasets.blended_megatron_dataset_builder:Building GPTDataset splits with sizes=(160, 0, 0) and config=GPTDatasetConfig(random_seed=1234, sequence_length=4096, blend=(['/workspace/data/wikitext/wikitext103_gpt2_text_document'], None), blend_per_split=None, split='100,0,0', split_matrix=[(0, 1.0), None, None], num_dataset_builder_threads=1, path_to_cache=None, mmap_bin_files=True, mock=False, tokenizer=<megatron.training.tokenizer.tokenizer._GPT2BPETokenizer object at 0x70dee2fd8380>, mid_level_dataset_surplus=0.005, reset_position_ids=False, reset_attention_mask=False, eod_mask_loss=False, create_attention_mask=True, drop_last_partial_validation_sequence=True, add_extra_token_to_sequence=True, object_storage_cache_path=None)
1077
+ INFO:megatron.core.datasets.indexed_dataset:Load the _IndexReader from /workspace/data/wikitext/wikitext103_gpt2_text_document.idx
1078
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence lengths
1079
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence pointers
1080
+ INFO:megatron.core.datasets.indexed_dataset: Extract the document indices
1081
+ INFO:megatron.core.datasets.indexed_dataset:> total number of sequences: 1165029
1082
+ INFO:megatron.core.datasets.indexed_dataset:> total number of documents: 1801350
1083
+ INFO:megatron.core.datasets.gpt_dataset:Load the GPTDataset train indices
1084
+ INFO:megatron.core.datasets.gpt_dataset: Load the document index from 68684a515e118253a8572c4517cf2ffe-GPTDataset-train-document_index.npy
1085
+ INFO:megatron.core.datasets.gpt_dataset: Load the sample index from 68684a515e118253a8572c4517cf2ffe-GPTDataset-train-sample_index.npy
1086
+ INFO:megatron.core.datasets.gpt_dataset: Load the shuffle index from 68684a515e118253a8572c4517cf2ffe-GPTDataset-train-shuffle_index.npy
1087
+ INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 29073
1088
+ > finished creating GPT datasets ...
1089
+ [after dataloaders are built] datetime: 2026-09-21 06:00:27
1090
+ done with setup ...
1091
+ training ...
1092
+ Overwriting rerun_state_machine.current_iteration from -1 to 0...
1093
+ (min, max) time across ranks (ms):
1094
+ model-and-optimizer-setup ......................: (3102.00, 3115.71)
1095
+ train/valid/test-data-iterators-setup ..........: (93.14, 98.19)
1096
+ [before the start of training step] datetime: 2026-09-21 06:00:27
1097
+ INFO:moonep.api:MoonEP comm buffer sizes:
1098
+ config : R=8, S=4096, H=4096, E=64, K=2, NvS=10224, NvS_padded=10240
1099
+ sm_count : num_sms=32, num_sms_dedup=148
1100
+ hidden_buffer_size : 80.00 MiB
1101
+ meta_buffer_size : 2.00 MiB
1102
+ local_temp_buffer_size : 515.29 KiB
1103
+ total_size : 82.50 MiB
1104
+ [MoonEP selfcheck 1/4] dropped groups empty OK | covered rows [0, 1, 2, 3, 4, 5, 6, 7, 64, 65, 66, 67, 68, 69, 70, 71] sizes [128, 4736, 384, 128, 1280, 256, 1536, 256, 0, 0, 0, 0, 0, 0, 0, 0] | duplicated experts [[-1, -1, -1, -1, -1, -1, -1, -1], [40, -1, -1, -1, -1, -1, -1, -1], [1, -1, -1, -1, -1, -1, -1, -1], [1, -1, -1, -1, -1, -1, -1, -1], [31, -1, -1, -1, -1, -1, -1, -1], [31, -1, -1, -1, -1, -1, -1, -1], [32, -1, -1, -1, -1, -1, -1, -1], [22, -1, -1, -1, -1, -1, -1, -1]]
1105
+ [MoonEP selfcheck 2/4] dropped groups empty OK | covered rows [0, 1, 2, 3, 4, 5, 6, 7, 64, 65, 66, 67, 68, 69, 70, 71] sizes [2816, 128, 128, 768, 256, 128, 4608, 128, 128, 0, 0, 0, 0, 0, 0, 0] | duplicated experts [[28, -1, -1, -1, -1, -1, -1, -1], [6, -1, -1, -1, -1, -1, -1, -1], [6, -1, -1, -1, -1, -1, -1, -1], [-1, -1, -1, -1, -1, -1, -1, -1], [28, -1, -1, -1, -1, -1, -1, -1], [6, -1, -1, -1, -1, -1, -1, -1], [17, -1, -1, -1, -1, -1, -1, -1], [17, -1, -1, -1, -1, -1, -1, -1]]
1106
+ [MoonEP selfcheck 3/4] dropped groups empty OK | covered rows [0, 1, 2, 3, 4, 5, 6, 7, 64, 65, 66, 67, 68, 69, 70, 71] sizes [768, 128, 4992, 128, 384, 0, 256, 0, 2048, 0, 0, 0, 0, 0, 0, 0] | duplicated experts [[34, -1, -1, -1, -1, -1, -1, -1], [7, 2, -1, -1, -1, -1, -1, -1], [57, -1, -1, -1, -1, -1, -1, -1], [43, -1, -1, -1, -1, -1, -1, -1], [43, -1, -1, -1, -1, -1, -1, -1], [-1, -1, -1, -1, -1, -1, -1, -1], [57, 60, -1, -1, -1, -1, -1, -1], [30, -1, -1, -1, -1, -1, -1, -1]]
1107
+ [MoonEP selfcheck 4/4] dropped groups empty OK | covered rows [0, 1, 2, 3, 4, 5, 6, 7, 64, 65, 66, 67, 68, 69, 70, 71] sizes [256, 128, 5760, 768, 0, 128, 128, 1024, 640, 0, 0, 0, 0, 0, 0, 0] | duplicated experts [[43, -1, -1, -1, -1, -1, -1, -1], [56, -1, -1, -1, -1, -1, -1, -1], [40, -1, -1, -1, -1, -1, -1, -1], [2, -1, -1, -1, -1, -1, -1, -1], [56, -1, -1, -1, -1, -1, -1, -1], [-1, -1, -1, -1, -1, -1, -1, -1], [56, -1, -1, -1, -1, -1, -1, -1], [40, -1, -1, -1, -1, -1, -1, -1]]
1108
+ [2026-09-21 06:00:49] iteration 1/ 10 | consumed samples: 16 | elapsed time per iteration (ms): 21865.4 | throughput per GPU (TFLOP/s/GPU): 4.2 | learning rate: 2.000000E-05 | global batch size: 16 | lm loss: 1.102587E+01 | loss scale: 1.0 | grad norm: 46.864 | number of skipped iterations: 0 | number of nan iterations: 0 |
1109
+ Number of parameters in transformer block in billions: 23.22
1110
+ Number of parameters in embedding layers in billions: 0.41
1111
+ Total number of parameters in billions: 23.63
1112
+ Number of parameters in most loaded shard in billions: 23.6320
1113
+ Theoretical memory footprints: weight and optimizer=169029.43 MB
1114
+ [Rank 0] (after 1 iterations) memory (MB) | allocated: 53065.47216796875 | max allocated: 53065.47265625 | reserved: 56294.0 | max reserved: 56294.0
1115
+ [2026-09-21 06:00:49] iteration 2/ 10 | consumed samples: 32 | elapsed time per iteration (ms): 449.3 | throughput per GPU (TFLOP/s/GPU): 202.4 | learning rate: 4.000000E-05 | global batch size: 16 | lm loss: 1.101135E+01 | loss scale: 1.0 | grad norm: 47.657 | number of skipped iterations: 0 | number of nan iterations: 0 |
1116
+ [2026-09-21 06:00:49] iteration 3/ 10 | consumed samples: 48 | elapsed time per iteration (ms): 399.7 | throughput per GPU (TFLOP/s/GPU): 227.5 | learning rate: 6.000000E-05 | global batch size: 16 | lm loss: 1.024838E+01 | loss scale: 1.0 | grad norm: 187.881 | number of skipped iterations: 0 | number of nan iterations: 0 |
1117
+ Capture CUDA graph for training!!!
1118
+ CUDA graph capture done!!!
1119
+ [2026-09-21 06:00:51] iteration 4/ 10 | consumed samples: 64 | elapsed time per iteration (ms): 1151.1 | throughput per GPU (TFLOP/s/GPU): 79.0 | learning rate: 8.000000E-05 | global batch size: 16 | lm loss: 1.308662E+01 | loss scale: 1.0 | grad norm: 44.335 | number of skipped iterations: 0 | number of nan iterations: 0 |
1120
+ [2026-09-21 06:00:51] iteration 5/ 10 | consumed samples: 80 | elapsed time per iteration (ms): 231.1 | throughput per GPU (TFLOP/s/GPU): 393.5 | learning rate: 1.000000E-04 | global batch size: 16 | lm loss: 1.040516E+01 | loss scale: 1.0 | grad norm: 20.433 | number of skipped iterations: 0 | number of nan iterations: 0 |
1121
+ [2026-09-21 06:00:51] iteration 6/ 10 | consumed samples: 96 | elapsed time per iteration (ms): 229.3 | throughput per GPU (TFLOP/s/GPU): 396.7 | learning rate: 7.750000E-05 | global batch size: 16 | lm loss: 1.022641E+01 | loss scale: 1.0 | grad norm: 31.705 | number of skipped iterations: 0 | number of nan iterations: 0 |
1122
+ [2026-09-21 06:00:54] iteration 7/ 10 | consumed samples: 112 | elapsed time per iteration (ms): 3159.3 | throughput per GPU (TFLOP/s/GPU): 28.8 | learning rate: 3.250000E-05 | global batch size: 16 | lm loss: 9.915960E+00 | loss scale: 1.0 | grad norm: 9.769 | number of skipped iterations: 0 | number of nan iterations: 0 |
1123
+ [2026-09-21 06:00:54] iteration 8/ 10 | consumed samples: 128 | elapsed time per iteration (ms): 240.4 | throughput per GPU (TFLOP/s/GPU): 378.3 | learning rate: 1.000000E-05 | global batch size: 16 | lm loss: 9.462394E+00 | loss scale: 1.0 | grad norm: 8.361 | number of skipped iterations: 0 | number of nan iterations: 0 |
1124
+ Capture range started in the application.
1125
+ Capture range ended in the application.
1126
+ Generating '/tmp/nsys-report-971e.qdstrm'
1127
+
1128
+
1129
+ [2026-09-21 06:01:38] iteration 9/ 10 | consumed samples: 144 | elapsed time per iteration (ms): 43212.2 | throughput per GPU (TFLOP/s/GPU): 2.1 | learning rate: 1.000000E-05 | global batch size: 16 | lm loss: 9.254077E+00 | loss scale: 1.0 | grad norm: 7.822 | number of skipped iterations: 0 | number of nan iterations: 0 |
1130
+ [2026-09-21 06:01:38] iteration 10/ 10 | consumed samples: 160 | elapsed time per iteration (ms): 252.7 | throughput per GPU (TFLOP/s/GPU): 359.9 | learning rate: 1.000000E-05 | global batch size: 16 | lm loss: 9.098603E+00 | loss scale: 1.0 | grad norm: 7.308 | number of skipped iterations: 0 | number of nan iterations: 0 |
1131
+ [after training is done] datetime: 2026-09-21 06:01:38
1132
+ [rank0]:[W921 06:01:39.757173273 ProcessGroupNCCL.cpp:1505] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
1133
+ [rank2]:[W921 06:01:41.006759775 ProcessGroupNCCL.cpp:1505] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
1134
+ [rank3]:[W921 06:01:41.007832771 ProcessGroupNCCL.cpp:1505] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
1135
+ [rank7]:[W921 06:01:41.013820410 ProcessGroupNCCL.cpp:1505] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
1136
+ [rank5]:[W921 06:01:41.058733522 ProcessGroupNCCL.cpp:1505] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
1137
+ [rank4]:[W921 06:01:41.354105238 ProcessGroupNCCL.cpp:1505] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
1138
+ [rank6]:[W921 06:01:41.575027704 ProcessGroupNCCL.cpp:1505] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
1139
+ [rank1]:[W921 06:01:41.586023040 ProcessGroupNCCL.cpp:1505] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
1140
+ Collection stop is not allowed in this state.
1141
+ [rank2]:[W921 06:01:44.565137282 observer.cpp:116] Warning: Leaked callback handle: 1 (function operator())
1142
+ pure virtual method called
1143
+ terminate called without an active exception
1144
+ [rank0]:[W921 06:01:44.607116968 observer.cpp:116] Warning: Leaked callback handle: 1 (function operator())
1145
+ pure virtual method called
1146
+ terminate called without an active exception
1147
+ [rank7]:[W921 06:01:44.752292148 observer.cpp:116] Warning: Leaked callback handle: 1 (function operator())
1148
+ pure virtual method called
1149
+ terminate called without an active exception
1150
+ [rank5]:[W921 06:01:45.959850729 observer.cpp:116] Warning: Leaked callback handle: 1 (function operator())
1151
+ pure virtual method called
1152
+ terminate called without an active exception
1153
+ W0921 06:01:45.360000 142863 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 143072 closing signal SIGTERM
1154
+ W0921 06:01:45.361000 142863 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 143073 closing signal SIGTERM
1155
+ W0921 06:01:45.361000 142863 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 143075 closing signal SIGTERM
1156
+ W0921 06:01:45.362000 142863 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 143076 closing signal SIGTERM
1157
+ W0921 06:01:45.362000 142863 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 143077 closing signal SIGTERM
1158
+ W0921 06:01:45.363000 142863 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 143078 closing signal SIGTERM
1159
+ W0921 06:01:45.363000 142863 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 143079 closing signal SIGTERM
1160
+ E0921 06:01:47.245000 142863 torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: -6) local_rank: 2 (pid: 143074) of binary: /usr/bin/python
1161
+ Traceback (most recent call last):
1162
+ File "/usr/local/bin/torchrun", line 8, in <module>
1163
+ sys.exit(main())
1164
+ ^^^^^^
1165
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper
1166
+ return f(*args, **kwargs)
1167
+ ^^^^^^^^^^^^^^^^^^
1168
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 892, in main
1169
+ run(args)
1170
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 883, in run
1171
+ elastic_launch(
1172
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 139, in __call__
1173
+ return launch_agent(self._config, self._entrypoint, list(args))
1174
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1175
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 270, in launch_agent
1176
+ raise ChildFailedError(
1177
+ torch.distributed.elastic.multiprocessing.errors.ChildFailedError:
1178
+ ============================================================
1179
+ pretrain_gpt.py FAILED
1180
+ ------------------------------------------------------------
1181
+ Failures:
1182
+ <NO_OTHER_FAILURES>
1183
+ ------------------------------------------------------------
1184
+ Root Cause (first observed failure):
1185
+ [0]:
1186
+ time : 2026-09-21_06:01:45
1187
+ host : sp-aiinfra-interactive-aghosh-8-mb99k-8xb200-moe-cg-0
1188
+ rank : 2 (local_rank: 2)
1189
+ exitcode : -6 (pid: 143074)
1190
+ error_file: <N/A>
1191
+ traceback : Signal 6 (SIGABRT) received by PID 143074
1192
+ ============================================================
1193
+ Generated:
1194
+ /workspace/granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/G3_K2__moonep__matched__seq4096__gbs16__capnone__async.nsys-rep
granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/verify_trace_ranks.log ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ report : /workspace/granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/G3_K2__moonep__matched__seq4096__gbs16__capnone__async.nsys-rep
2
+ kernels : 141507
3
+ devices : 8 of 8 -> [0, 1, 2, 3, 4, 5, 6, 7]
4
+ processes: 1
5
+ markers : 17408 moonep_trace__ kernels on devices [0, 1, 2, 3, 4, 5, 6, 7]
6
+ PASS: all ranks present and backend markers emitted
granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K8__moonep__matched__seq4096__gbs16__capnone__async/nsys/run.env ADDED
@@ -0,0 +1,105 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ backend=moonep
2
+ config=G3_K8
3
+ family=tierA
4
+ source_model=
5
+ standard_num_layers=
6
+ base_rung=G3
7
+ repo=/workspace/Megatron-MoE-Echo
8
+ branch=moonep_backend_async_grad_reduce
9
+ commit=2e68c8ad
10
+ moonep_source_dir=/workspace/MoonEP
11
+ moonep_source_branch=megatron-ddp-symmetric-params
12
+ moonep_source_commit=d8bc666
13
+ num_layers=16
14
+ hidden=4096
15
+ ffn_hidden=14336
16
+ moe_ffn_hidden=1792
17
+ num_experts=64
18
+ topk=8
19
+ experts_per_rank=8
20
+ expert_params=22020096
21
+ heads=32
22
+ kv_groups=8
23
+ kv_channels=
24
+ qk_layernorm=0
25
+ shared_ffn=0
26
+ shared_expert_overlap=0
27
+ moe_token_dispatcher=flex
28
+ mla=0
29
+ q_lora_rank=
30
+ kv_lora_rank=512
31
+ qk_head_dim=128
32
+ qk_pos_emb_head_dim=64
33
+ v_head_dim=128
34
+ dense_first_layers=0
35
+ moe_layers=16
36
+ rotary_base=1000000
37
+ rotary_percent=1.0
38
+ rotary_scaling=
39
+ mscale=
40
+ mscale_all_dim=
41
+ qkv_bias=0
42
+ router_balancing=aux_loss
43
+ aux_loss_coeff=1e-2
44
+ router_pre_softmax=0
45
+ router_fusion=0
46
+ router_dtype=fp32
47
+ router_groups=
48
+ router_group_topk=
49
+ router_score_function=softmax
50
+ router_expert_bias=0
51
+ router_bias_update_rate=
52
+ router_topk_scaling_factor=
53
+ shared_expert_gate=0
54
+ apply_layernorm_1p=0
55
+ attention_output_gate=0
56
+ linear_attention_type=
57
+ linear_attention_freq=
58
+ linear_conv_kernel_dim=
59
+ linear_key_head_dim=
60
+ linear_value_head_dim=
61
+ linear_num_key_heads=
62
+ linear_num_value_heads=
63
+ mtp_num_layers=0
64
+ mtp_loss_scaling_factor=0.1
65
+ no_weight_decay_cond_type=
66
+ max_position_embeddings=32768
67
+ init_method_std=0.01
68
+ device_initiated_grouped_gemm=1
69
+ norm_epsilon=1e-5
70
+ seq_length=4096
71
+ micro_batch_size=1
72
+ global_batch_size=16
73
+ num_microbatches=2
74
+ assignments_per_rank=32768
75
+ mean_tokens_per_expert=4096
76
+ slot_policy=matched
77
+ slots_per_rank=8
78
+ capacity_mode=none
79
+ capacity=none
80
+ cuda_graph=1
81
+ probe_mode=0
82
+ async_grad_reduce=1
83
+ branch_role=timing
84
+ memory_dump_mode=none
85
+ memory_snapshot_path=
86
+ cuda_graph_warmup_steps=3
87
+ recompute=none
88
+ force_balance=0
89
+ train_iters=10
90
+ ep_size=8
91
+ echo_dispatch_overlap=1
92
+ echo_dispatch_recompute=0
93
+ ultraep_quota_min_tokens=
94
+ ultraep_weight_sync_join=
95
+ profile_enable=1
96
+ kernel_trace=1
97
+ fp8_mode=mxfp8
98
+ fp8_recipe=mxfp8
99
+ training_seed=1234
100
+ learning_rate=1e-4
101
+ min_learning_rate=1.0e-5
102
+ lr_warmup_iters=5
103
+ state_dump_enable=0
104
+ state_dump_dir=
105
+ moonep_assert_no_pending_reduce=1
granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K8__moonep__matched__seq4096__gbs16__capnone__async/nsys/train.log ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ==============================================================
2
+ G3_K8 / moonep (slots/rank=8 policy=matched)
3
+ E=64 K=8 H'=1792 m=8
4
+ L=16 seq=4096 gbs=16 (2 microbatches)
5
+ cap_mode=none cap=none cg=1 recompute=none
6
+ memory_dump=none snapshots=/workspace/granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K8__moonep__matched__seq4096__gbs16__capnone__async/nsys/memory_snapshots/snapshot.pickle
7
+ repo=/workspace/Megatron-MoE-Echo @ 2e68c8ad
8
+ ==============================================================
9
+ W0921 06:02:27.205000 147335 torch/distributed/run.py:766]
10
+ W0921 06:02:27.205000 147335 torch/distributed/run.py:766] *****************************************
11
+ W0921 06:02:27.205000 147335 torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.
12
+ W0921 06:02:27.205000 147335 torch/distributed/run.py:766] *****************************************
granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/results.tsv ADDED
@@ -0,0 +1 @@
 
 
1
+ config backend policy slots seq gbs cap recompute rep status iter_ms_mean iter_ms_min peak_mib run_dir
granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/trace_extension_build.log ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Could not find an up-to-date installation of `packaging`. License expressions might not be validated. To enforce validation, please install `packaging>=24.2`.
2
+ /usr/local/lib/python3.12/dist-packages/setuptools/config/_apply_pyprojecttoml.py:82: SetuptoolsDeprecationWarning: `project.license` as a TOML table is deprecated
3
+ !!
4
+
5
+ ********************************************************************************
6
+ Please use a simple string containing a SPDX expression for `project.license`. You can also use `project.license-files`. (Both options available on setuptools>=77.0.0).
7
+
8
+ This deprecation is overdue, please update your project and remove deprecated
9
+ calls to avoid build errors in the future.
10
+
11
+ See https://packaging.python.org/en/latest/guides/writing-pyproject-toml/#license for details.
12
+ ********************************************************************************
13
+
14
+ !!
15
+ corresp(dist, value, root_dir)
16
+ /usr/local/lib/python3.12/dist-packages/setuptools/config/_apply_pyprojecttoml.py:61: SetuptoolsDeprecationWarning: License classifiers are deprecated.
17
+ !!
18
+
19
+ ********************************************************************************
20
+ Please consider removing the following classifiers in favor of a SPDX license expression:
21
+
22
+ License :: OSI Approved :: BSD License
23
+
24
+ See https://packaging.python.org/en/latest/guides/writing-pyproject-toml/#license for details.
25
+ ********************************************************************************
26
+
27
+ !!
28
+ dist._finalize_license_expression()
29
+ /usr/local/lib/python3.12/dist-packages/setuptools/dist.py:759: SetuptoolsDeprecationWarning: License classifiers are deprecated.
30
+ !!
31
+
32
+ ********************************************************************************
33
+ Please consider removing the following classifiers in favor of a SPDX license expression:
34
+
35
+ License :: OSI Approved :: BSD License
36
+
37
+ See https://packaging.python.org/en/latest/guides/writing-pyproject-toml/#license for details.
38
+ ********************************************************************************
39
+
40
+ !!
41
+ self._finalize_license_expression()
42
+ running build_ext
43
+ building 'megatron.core.datasets.helpers_cpp' extension
44
+ creating /workspace/Megatron-MoE-Echo/build/temp.linux-x86_64-cpython-312/megatron/core/datasets
45
+ Emitting ninja build file /workspace/Megatron-MoE-Echo/build/temp.linux-x86_64-cpython-312/build.ninja...
46
+ Compiling objects...
47
+ Allowing ninja to set a default number of workers... (overridable by setting the environment variable MAX_JOBS=N)
48
+ [1/1] c++ -MMD -MF /workspace/Megatron-MoE-Echo/build/temp.linux-x86_64-cpython-312/megatron/core/datasets/helpers.o.d -fno-strict-overflow -Wsign-compare -DNDEBUG -g -O2 -Wall -fPIC -I/usr/include/python3.12 -c -c /workspace/Megatron-MoE-Echo/megatron/core/datasets/helpers.cpp -o /workspace/Megatron-MoE-Echo/build/temp.linux-x86_64-cpython-312/megatron/core/datasets/helpers.o -I/usr/include/python3.12 -I/usr/local/lib/python3.12/dist-packages/pybind11/include -O3 -Wall -std=c++17 -DTORCH_API_INCLUDE_EXTENSION_H '-DPYBIND11_COMPILER_TYPE="_gcc"' '-DPYBIND11_STDLIB="_libstdcpp"' '-DPYBIND11_BUILD_ABI="_cxxabi1016"' -DTORCH_EXTENSION_NAME=helpers_cpp
49
+ In file included from /workspace/Megatron-MoE-Echo/megatron/core/datasets/helpers.cpp:12:
50
+ In function ‘pybind11::ssize_t pybind11::detail::byte_offset_unsafe(const Strides&, pybind11::ssize_t, Ix ...) [with long int Dim = 0; Strides = std::array<long int, 1>; Ix = {}]’,
51
+ inlined from ‘const T& pybind11::detail::unchecked_reference<T, Dims>::operator()(Ix ...) const [with Ix = {long int}; T = long int; long int Dims = 1]’ at /usr/local/lib/python3.12/dist-packages/pybind11/include/pybind11/numpy.h:545:65,
52
+ inlined from ‘const T& pybind11::detail::unchecked_reference<T, Dims>::operator[](pybind11::ssize_t) const [with long int D = 1; <template-parameter-2-2> = void; T = long int; long int Dims = 1]’ at /usr/local/lib/python3.12/dist-packages/pybind11/include/pybind11/numpy.h:553:26,
53
+ inlined from ‘void build_exhaustive_blending_indices(pybind11::array_t<short int>&, pybind11::array_t<long int>&, const pybind11::array_t<long int>&, int32_t)’ at /workspace/Megatron-MoE-Echo/megatron/core/datasets/helpers.cpp:67:31:
54
+ /usr/local/lib/python3.12/dist-packages/pybind11/include/pybind11/numpy.h:494:14: warning: ‘error_argmax’ may be used uninitialized [-Wmaybe-uninitialized]
55
+ 494 | return i * strides[Dim] + byte_offset_unsafe<Dim + 1>(strides, index...);
56
+ | ~~^~~~~~~~~~
57
+ /workspace/Megatron-MoE-Echo/megatron/core/datasets/helpers.cpp: In function ‘void build_exhaustive_blending_indices(pybind11::array_t<short int>&, pybind11::array_t<long int>&, const pybind11::array_t<long int>&, int32_t)’:
58
+ /workspace/Megatron-MoE-Echo/megatron/core/datasets/helpers.cpp:49:13: note: ‘error_argmax’ was declared here
59
+ 49 | int64_t error_argmax;
60
+ | ^~~~~~~~~~~~
61
+ creating build/lib.linux-x86_64-cpython-312/megatron/core/datasets
62
+ x86_64-linux-gnu-g++ -fno-strict-overflow -Wsign-compare -DNDEBUG -g -O2 -Wall -shared -Wl,-O1 -Wl,-Bsymbolic-functions /workspace/Megatron-MoE-Echo/build/temp.linux-x86_64-cpython-312/megatron/core/datasets/helpers.o -L/usr/lib/x86_64-linux-gnu -o build/lib.linux-x86_64-cpython-312/megatron/core/datasets/helpers_cpp.cpython-312-x86_64-linux-gnu.so
63
+ building 'megatron.core.transformer.moe._moonep_trace_C' extension
64
+ creating /workspace/Megatron-MoE-Echo/build/temp.linux-x86_64-cpython-312/megatron/core/transformer/moe
65
+ Emitting ninja build file /workspace/Megatron-MoE-Echo/build/temp.linux-x86_64-cpython-312/build.ninja...
66
+ Compiling objects...
67
+ Allowing ninja to set a default number of workers... (overridable by setting the environment variable MAX_JOBS=N)
68
+ [1/2] c++ -MMD -MF /workspace/Megatron-MoE-Echo/build/temp.linux-x86_64-cpython-312/megatron/core/transformer/moe/moonep_trace_ops.o.d -fno-strict-overflow -Wsign-compare -DNDEBUG -g -O2 -Wall -fPIC -I/usr/local/lib/python3.12/dist-packages/torch/include -I/usr/local/lib/python3.12/dist-packages/torch/include/torch/csrc/api/include -I/usr/local/cuda/include -I/usr/include/python3.12 -c -c /workspace/Megatron-MoE-Echo/megatron/core/transformer/moe/moonep_trace_ops.cpp -o /workspace/Megatron-MoE-Echo/build/temp.linux-x86_64-cpython-312/megatron/core/transformer/moe/moonep_trace_ops.o -O3 -std=c++17 -DTORCH_API_INCLUDE_EXTENSION_H '-DPYBIND11_COMPILER_TYPE="_gcc"' '-DPYBIND11_STDLIB="_libstdcpp"' '-DPYBIND11_BUILD_ABI="_cxxabi1016"' -DTORCH_EXTENSION_NAME=_moonep_trace_C
69
+ [2/2] /usr/local/cuda/bin/nvcc --generate-dependencies-with-compile --dependency-output /workspace/Megatron-MoE-Echo/build/temp.linux-x86_64-cpython-312/megatron/core/transformer/moe/moonep_trace_kernels.o.d -I/usr/local/lib/python3.12/dist-packages/torch/include -I/usr/local/lib/python3.12/dist-packages/torch/include/torch/csrc/api/include -I/usr/local/cuda/include -I/usr/include/python3.12 -c -c /workspace/Megatron-MoE-Echo/megatron/core/transformer/moe/moonep_trace_kernels.cu -o /workspace/Megatron-MoE-Echo/build/temp.linux-x86_64-cpython-312/megatron/core/transformer/moe/moonep_trace_kernels.o -D__CUDA_NO_HALF_OPERATORS__ -D__CUDA_NO_HALF_CONVERSIONS__ -D__CUDA_NO_BFLOAT16_CONVERSIONS__ -D__CUDA_NO_HALF2_OPERATORS__ --expt-relaxed-constexpr --compiler-options ''"'"'-fPIC'"'"'' -O3 -std=c++17 -lineinfo -DTORCH_API_INCLUDE_EXTENSION_H '-DPYBIND11_COMPILER_TYPE="_gcc"' '-DPYBIND11_STDLIB="_libstdcpp"' '-DPYBIND11_BUILD_ABI="_cxxabi1016"' -DTORCH_EXTENSION_NAME=_moonep_trace_C -gencode=arch=compute_100,code=sm_100 -gencode=arch=compute_120,code=compute_120 -gencode=arch=compute_120,code=sm_120 -gencode=arch=compute_75,code=sm_75 -gencode=arch=compute_80,code=sm_80 -gencode=arch=compute_86,code=sm_86 -gencode=arch=compute_90,code=sm_90
70
+ creating build/lib.linux-x86_64-cpython-312/megatron/core/transformer/moe
71
+ x86_64-linux-gnu-g++ -fno-strict-overflow -Wsign-compare -DNDEBUG -g -O2 -Wall -shared -Wl,-O1 -Wl,-Bsymbolic-functions /workspace/Megatron-MoE-Echo/build/temp.linux-x86_64-cpython-312/megatron/core/transformer/moe/moonep_trace_kernels.o /workspace/Megatron-MoE-Echo/build/temp.linux-x86_64-cpython-312/megatron/core/transformer/moe/moonep_trace_ops.o -L/usr/local/lib/python3.12/dist-packages/torch/lib -L/usr/local/cuda/lib64 -L/usr/lib/x86_64-linux-gnu -lc10 -ltorch -ltorch_cpu -ltorch_python -lcudart -lc10_cuda -ltorch_cuda -o build/lib.linux-x86_64-cpython-312/megatron/core/transformer/moe/_moonep_trace_C.cpython-312-x86_64-linux-gnu.so
72
+ copying build/lib.linux-x86_64-cpython-312/megatron/core/datasets/helpers_cpp.cpython-312-x86_64-linux-gnu.so -> megatron/core/datasets
73
+ copying build/lib.linux-x86_64-cpython-312/megatron/core/transformer/moe/_moonep_trace_C.cpython-312-x86_64-linux-gnu.so -> megatron/core/transformer/moe
74
+ load check passed: /workspace/Megatron-MoE-Echo/megatron/core/transformer/moe/_moonep_trace_C.cpython-312-x86_64-linux-gnu.so -> torch.ops.megatron_moonep_trace.fwd_moonep_forward_begin
granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/trace_results.tsv ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ model backend capacity status report async_overlap_json
2
+ G2_K2 moonep none OK /workspace/granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G2_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/G2_K2__moonep__matched__seq4096__gbs16__capnone__async.nsys-rep /workspace/granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G2_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/moonep_async_overlap.json
3
+ G3_K2 moonep none OK /workspace/granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/G3_K2__moonep__matched__seq4096__gbs16__capnone__async.nsys-rep /workspace/granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/G3_K2__moonep__matched__seq4096__gbs16__capnone__async/nsys/moonep_async_overlap.json
granularity_next_b200/tiera_topk_trace/steps.tsv CHANGED
@@ -1,2 +1,3 @@
1
  status campaign phase gbs started ended result_dir
2
  OK echo tierbtrace 16 2026-09-17T01:54:35+00:00 2026-09-17T03:22:23+00:00 /workspace/granularity_next_b200/tiera_topk_trace/echo/tierbtrace
 
 
1
  status campaign phase gbs started ended result_dir
2
  OK echo tierbtrace 16 2026-09-17T01:54:35+00:00 2026-09-17T03:22:23+00:00 /workspace/granularity_next_b200/tiera_topk_trace/echo/tierbtrace
3
+ OK ultraep tierbtrace 16 2026-09-21T05:22:22+00:00 2026-09-21T05:53:08+00:00 /workspace/granularity_next_b200/tiera_topk_trace/ultraep/tierbtrace
granularity_next_b200/tiera_topk_trace/study.log CHANGED
@@ -209,3 +209,44 @@ driver_repo: /workspace/blackwell_run_scripts @ b8bd100
209
  [guard] iter 8/10 (80s)
210
  [guard] no progress for 60s at iter 8/10 (past profile window); treating as done
211
  [guard] nsys stop (flushing report)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
209
  [guard] iter 8/10 (80s)
210
  [guard] no progress for 60s at iter 8/10 (past profile window); treating as done
211
  [guard] nsys stop (flushing report)
212
+ [guard] tearing down process group (outcome=STALLED)
213
+ [05:53:08] trace verified -> G4_K16__ultraep__matched__seq4096__gbs16__cap1p08203125.nsys-rep (15M, all 8 GPUs + ultraep markers)
214
+ --------------------------------------------------------------------------
215
+
216
+ ==============================================================
217
+ SWEEP COMPLETE ok=9 failed=0
218
+ ==============================================================
219
+ config backend policy slots seq gbs cap recompute rep status iter_ms_mean iter_ms_min peak_mib run_dir
220
+
221
+ Results: /workspace/granularity_next_b200/tiera_topk_trace/ultraep/tierbtrace/tierbtrace_gbs16/results.tsv
222
+ Roll up: python3 /workspace/blackwell_run_scripts/granularity/tools/collect.py /workspace/granularity_next_b200/tiera_topk_trace/ultraep/tierbtrace/tierbtrace_gbs16
223
+ ==============================================================
224
+
225
+ Campaign results: /workspace/granularity_next_b200/tiera_topk_trace/ultraep/tierbtrace
226
+ Roll up: python3 /workspace/blackwell_run_scripts/granularity/tools/collect.py /workspace/granularity_next_b200/tiera_topk_trace/ultraep/tierbtrace --compare
227
+ Already logged in to Hugging Face as: AbhishekGhosh1998
228
+ uploaded: /workspace/granularity_next_b200/tiera_topk_trace/ultraep/tierbtrace -> MoE-CUDA-Graph-Analysis/tierb-b200-gbs8/granularity_next_b200/tiera_topk_trace/ultraep/tierbtrace
229
+ https://huggingface.co/MoE-CUDA-Graph-Analysis/tierb-b200-gbs8/tree/main
230
+
231
+ === STEP moonep__tierbtrace__gbs16
232
+ ==============================================================
233
+ CAMPAIGN: moonep backend=moonep phase=tierbtrace
234
+ Megatron: /workspace/Megatron-MoE-Echo @ moonep_backend_async_grad_reduce
235
+ GBS=16 seq=4096 repeats=3 nsys=1
236
+ results -> /workspace/granularity_next_b200/tiera_topk_trace/moonep/tierbtrace
237
+ ==============================================================
238
+ [05:53:26] building moonep marker extension (log: /workspace/granularity_next_b200/tiera_topk_trace/moonep/tierbtrace/tierbtrace_gbs16/trace_extension_build.log)
239
+ [05:57:30] marker extension ready: _moonep_trace_C.cpython-312-x86_64-linux-gnu.so
240
+ [05:57:30] TRACE-ONLY G2_K2__moonep__matched__seq4096__gbs16__capnone__async (ranks: 0 1 2 3 4 5 6 7)
241
+ [guard] iter 6/10 (60s)
242
+ [guard] iter 8/10 (70s)
243
+ [guard] iter 10/10 (110s)
244
+ [guard] nsys stop (flushing report)
245
+ [05:59:52] trace verified -> G2_K2__moonep__matched__seq4096__gbs16__capnone__async.nsys-rep (9.8M, all 8 GPUs + moonep markers)
246
+ [05:59:53] TRACE-ONLY G3_K2__moonep__matched__seq4096__gbs16__capnone__async (ranks: 0 1 2 3 4 5 6 7)
247
+ [guard] iter 6/10 (60s)
248
+ [guard] iter 8/10 (70s)
249
+ [guard] iter 10/10 (110s)
250
+ [guard] nsys stop (flushing report)
251
+ [06:02:15] trace verified -> G3_K2__moonep__matched__seq4096__gbs16__capnone__async.nsys-rep (11M, all 8 GPUs + moonep markers)
252
+ [06:02:16] TRACE-ONLY G3_K8__moonep__matched__seq4096__gbs16__capnone__async (ranks: 0 1 2 3 4 5 6 7)