AbhishekGhosh1998 commited on
Commit
bd80bfe
·
verified ·
1 Parent(s): 3f74d6a

granularity results tierb-fixed-moonep-full/moonep/tierbtrace @ 2026-08-30T01:06:13Z from sp-aiinfra-interactive-aghosh-7jtjl-8xb200-moe-cg-0

Browse files
Files changed (14) hide show
  1. .gitattributes +1 -0
  2. tierb-fixed-moonep-full/moonep/tierbtrace/manifest.txt +19 -0
  3. tierb-fixed-moonep-full/moonep/tierbtrace/pip_freeze.txt +317 -0
  4. tierb-fixed-moonep-full/moonep/tierbtrace/provenance.txt +44 -0
  5. tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/nsys/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async.nsys-rep +3 -0
  6. tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/nsys/moonep_async_overlap.json +9 -0
  7. tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/nsys/moonep_async_overlap.txt +416 -0
  8. tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/nsys/moonep_async_overlap_summary.txt +1 -0
  9. tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/nsys/run.env +73 -0
  10. tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/nsys/train.log +1326 -0
  11. tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/nsys/verify_trace_ranks.log +6 -0
  12. tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/results.tsv +1 -0
  13. tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/trace_extension_build.log +58 -0
  14. tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/trace_results.tsv +1 -0
.gitattributes CHANGED
@@ -63,3 +63,4 @@ tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__ultraep__
63
  tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/QWEN3_30B__ultraep__matched__seq4096__gbs8__cap1p11328125/nsys/QWEN3_30B__ultraep__matched__seq4096__gbs8__cap1p11328125.nsys-rep filter=lfs diff=lfs merge=lfs -text
64
  tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/DSV2_LITE_L24__ultraep__matched__seq8192__gbs8__cap1p0755208333333333/nsys/DSV2_LITE_L24__ultraep__matched__seq8192__gbs8__cap1p0755208333333333.nsys-rep filter=lfs diff=lfs merge=lfs -text
65
  tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/QWEN2_57B__ultraep__matched__seq4096__gbs8__cap1p08984375/nsys/QWEN2_57B__ultraep__matched__seq4096__gbs8__cap1p08984375.nsys-rep filter=lfs diff=lfs merge=lfs -text
 
 
63
  tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/QWEN3_30B__ultraep__matched__seq4096__gbs8__cap1p11328125/nsys/QWEN3_30B__ultraep__matched__seq4096__gbs8__cap1p11328125.nsys-rep filter=lfs diff=lfs merge=lfs -text
64
  tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/DSV2_LITE_L24__ultraep__matched__seq8192__gbs8__cap1p0755208333333333/nsys/DSV2_LITE_L24__ultraep__matched__seq8192__gbs8__cap1p0755208333333333.nsys-rep filter=lfs diff=lfs merge=lfs -text
65
  tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/QWEN2_57B__ultraep__matched__seq4096__gbs8__cap1p08984375/nsys/QWEN2_57B__ultraep__matched__seq4096__gbs8__cap1p08984375.nsys-rep filter=lfs diff=lfs merge=lfs -text
66
+ tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/nsys/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async.nsys-rep filter=lfs diff=lfs merge=lfs -text
tierb-fixed-moonep-full/moonep/tierbtrace/manifest.txt ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ timestamp: 2026-08-30T00:59:54+00:00
2
+ host: sp-aiinfra-interactive-aghosh-7jtjl-8xb200-moe-cg-0
3
+ gpu: NVIDIA B200 x8
4
+ driver: 580.95.05
5
+ study_scripts_branch: granularity-moonep
6
+ study_scripts_commit: eec9aee
7
+ campaign: moonep
8
+ backend: moonep
9
+ phase: tierbtrace
10
+ gbs: 8
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: 7bc235bf
tierb-fixed-moonep-full/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==1.29.0
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@7bc235bfc2a6d28129bf8df798ac4e667113af85#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==0.7.1
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
tierb-fixed-moonep-full/moonep/tierbtrace/provenance.txt CHANGED
@@ -46,3 +46,47 @@ command : run_ladder.sh --phase tierbtrace --gbs 8 --out /workspace/tierb-
46
  ---- toolchain -----------------------------------------------
47
  nvcc : release 12.9, V12.9.86
48
  gcc : gcc (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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: 1.29.0 [/usr/local/lib/python3.12/dist-packages/huggingface_hub/__init__.py]
60
+
61
+ ---- repositories --------------------------------------------
62
+ Megatron-MoE-Echo moonep_backend_async_grad_reduce 7bc235bfc2a6
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 eec9aee807fa
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=tierb-fixed-moonep-full
76
+ MOONEP_ASSERT_NO_PENDING_REDUCE=1
77
+ MOONEP_ASYNC_GRAD_REDUCE=1
78
+ NCCL_VERSION=2.27.3
79
+ NVSHMEM_VERSION=3.2.5
80
+ PYTORCH_BUILD_NUMBER=0
81
+ PYTORCH_BUILD_VERSION=2.8.0a0+5228986
82
+ PYTORCH_HOME=/opt/pytorch/pytorch
83
+ PYTORCH_VERSION=2.8.0a0+5228986
84
+ TORCH_ALLOW_TF32_CUBLAS_OVERRIDE=1
85
+ TORCH_CUDA_ARCH_LIST=7.5 8.0 8.6 9.0 10.0 12.0+PTX
86
+ TORCH_NCCL_USE_COMM_NONBLOCKING=0
87
+
88
+ ---- shared library resolution -------------------------------
89
+ libnvshmem_host.so: not in ldconfig
90
+ libnccl.so: /lib/x86_64-linux-gnu/libnccl.so.2
91
+ lrwxrwxrwx 1 root root 20 Aug 29 16:41 /usr/local/lib/python3.12/dist-packages/nvidia/nvshmem/lib/libnvshmem_host.so -> libnvshmem_host.so.3
92
+ -rw-r--r-- 1 root root 179045744 Aug 29 16:37 /usr/local/lib/python3.12/dist-packages/nvidia/nvshmem/lib/libnvshmem_host.so.3
tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/nsys/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async.nsys-rep ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:84bb6e84281049be49d947b37b07deb8d0d49d54de613f24bf17ea6cd5a04b60
3
+ size 7571093
tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/nsys/moonep_async_overlap.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "devices": 8,
3
+ "reduce_busy_ms_per_iteration": 5.61675,
4
+ "overlapped_ms_per_iteration": 0.0,
5
+ "exposed_ms_per_iteration": 5.61675,
6
+ "overlap_pct": 0.0,
7
+ "minimum_overlap_pct": 0.1,
8
+ "status": "FAIL"
9
+ }
tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/nsys/moonep_async_overlap.txt ADDED
@@ -0,0 +1,416 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+
2
+ === kernels on device 0, by total time ===
3
+ 75.174 ms elementwise_kernel
4
+ 73.755 ms device_kernel
5
+ 63.402 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
6
+ 61.319 ms multi_tensor_apply_kernel
7
+ 47.770 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o81924096i64i64_tensorptrbf16gmemalign16o5570564096i64i64_tensorptri32gmemalign16o655361_tensorptri32_0
8
+ 39.066 ms vectorized_elementwise_kernel
9
+ 35.485 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o5570564096i64i64_tensorptrbf16gmemalign16o81924096i64i64_tensorptri32gmemalign16o655361_tensorptri32gm_0
10
+ 34.071 ms cast_mxfp8_2D_kernel
11
+ 30.099 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
12
+ 21.224 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
13
+ 16.516 ms ncclDevKernel_AllGather_RING_LL
14
+ 9.980 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
15
+ 9.593 ms unrolled_elementwise_kernel
16
+ 9.244 ms nvjet_qqsss_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_NTT
17
+ 8.856 ms nvjet_qqtst_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_TNT
18
+ 8.786 ms nvjet_qqtst_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_NNT
19
+ 8.137 ms triton_poi_fused__to_copy_cat_0
20
+ REDUCE 8.023 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o3932164096i64i64_tensorptrf32gmemalign16o3932164096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
21
+ 6.379 ms setGroupedGemmWgradArguments
22
+ 4.042 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
23
+ 3.844 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o393216409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
24
+ REDUCE 3.761 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o5242881536i64i64_tensorptrf32gmemalign16o5242881536i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
25
+ 3.694 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
26
+ 3.612 ms CatArrayBatchedCopy_alignedK_contig
27
+ 3.370 ms gatherTopK
28
+ 3.353 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
29
+ 3.175 ms rmsnorm_bwd_general_kernel
30
+ 2.833 ms fused_rope_backward_kernel
31
+ 2.711 ms reduce_kernel
32
+ 2.666 ms triton_red_fused__to_copy_mul_silu_sum_1
33
+ 2.635 ms fused_rope_forward_kernel
34
+ 2.420 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o524288153611i640_CopyAtom_ThrID10_TVLayoutSrc_0
35
+ 2.229 ms triton_poi_fused__to_copy_mul_silu_0
36
+ 2.142 ms rmsnorm_bwd_tuned_kernel
37
+ 2.126 ms multi_tensor_swizzle_row_scaling_kernel
38
+ 2.113 ms rmsnorm_fwd_general_kernel
39
+ 1.971 ms compute_dot_do_o_specialized
40
+ 1.621 ms rmsnorm_bwd_finalize_general_kernel
41
+ 1.539 ms rmsnorm_fwd_tuned_kernel
42
+ 1.529 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_1281_tensorptri32_gmem_align16_o_655361_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_al_0
43
+
44
+ === QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async.sqlite (device 0, 2 iteration(s)) ===
45
+ reduce_grad kernels : 48 (24/iter)
46
+ reduce_grad sum : 5.892 ms/iter
47
+ reduce_grad busy (union) : 5.892 ms/iter
48
+ ... overlapped with compute: 0.000 ms/iter ( 0.0 %)
49
+ ... exposed on critical path: 5.892 ms/iter (100.0 %)
50
+ other GPU work (union) : 295.318 ms/iter
51
+ traced wall : 414.723 ms/iter
52
+ --> time hidden by overlap : 0.000 ms/iter
53
+
54
+ === kernels on device 1, by total time ===
55
+ 77.204 ms elementwise_kernel
56
+ 72.874 ms device_kernel
57
+ 63.529 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
58
+ 61.840 ms multi_tensor_apply_kernel
59
+ 47.737 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o81924096i64i64_tensorptrbf16gmemalign16o5570564096i64i64_tensorptri32gmemalign16o655361_tensorptri32_0
60
+ 40.712 ms vectorized_elementwise_kernel
61
+ 36.060 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o5570564096i64i64_tensorptrbf16gmemalign16o81924096i64i64_tensorptri32gmemalign16o655361_tensorptri32gm_0
62
+ 34.070 ms cast_mxfp8_2D_kernel
63
+ 29.504 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
64
+ 21.065 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
65
+ 15.240 ms ncclDevKernel_AllGather_RING_LL
66
+ 9.733 ms unrolled_elementwise_kernel
67
+ 9.223 ms nvjet_qqsss_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_NTT
68
+ 8.817 ms nvjet_qqtst_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_TNT
69
+ 8.686 ms nvjet_qqtst_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_NNT
70
+ 8.423 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
71
+ 8.147 ms triton_poi_fused__to_copy_cat_0
72
+ REDUCE 7.144 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o3932164096i64i64_tensorptrf32gmemalign16o3932164096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
73
+ 6.358 ms setGroupedGemmWgradArguments
74
+ 4.019 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
75
+ REDUCE 3.812 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o5242881536i64i64_tensorptrf32gmemalign16o5242881536i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
76
+ 3.698 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
77
+ 3.615 ms CatArrayBatchedCopy_alignedK_contig
78
+ 3.377 ms gatherTopK
79
+ 3.346 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
80
+ 3.223 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o393216409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
81
+ 3.223 ms rmsnorm_bwd_general_kernel
82
+ 2.843 ms fused_rope_backward_kernel
83
+ 2.707 ms reduce_kernel
84
+ 2.667 ms triton_red_fused__to_copy_mul_silu_sum_1
85
+ 2.642 ms fused_rope_forward_kernel
86
+ 2.391 ms ncclDevKernel_AllReduce_Sum_f32_RING_LL
87
+ 2.254 ms triton_poi_fused__to_copy_mul_silu_0
88
+ 2.163 ms rmsnorm_bwd_tuned_kernel
89
+ 2.143 ms multi_tensor_swizzle_row_scaling_kernel
90
+ 2.113 ms rmsnorm_fwd_general_kernel
91
+ 1.994 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o524288153611i640_CopyAtom_ThrID10_TVLayoutSrc_0
92
+ 1.971 ms compute_dot_do_o_specialized
93
+ 1.635 ms rmsnorm_bwd_finalize_general_kernel
94
+ 1.619 ms kernel_cutlass_kernel_moonepcombine_prologueCombinePrologueKernel_object_at__tensorptrbf16gmemalign16o696004096i64i64_tensorptri32gmemalign16o2088001_tensorptri32gmemalign16o696001_tensor_0
95
+
96
+ === QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async.sqlite (device 1, 2 iteration(s)) ===
97
+ reduce_grad kernels : 48 (24/iter)
98
+ reduce_grad sum : 5.478 ms/iter
99
+ reduce_grad busy (union) : 5.478 ms/iter
100
+ ... overlapped with compute: 0.000 ms/iter ( 0.0 %)
101
+ ... exposed on critical path: 5.478 ms/iter (100.0 %)
102
+ other GPU work (union) : 297.187 ms/iter
103
+ traced wall : 381.934 ms/iter
104
+ --> time hidden by overlap : 0.000 ms/iter
105
+
106
+ === kernels on device 2, by total time ===
107
+ 77.156 ms elementwise_kernel
108
+ 72.421 ms device_kernel
109
+ 62.858 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
110
+ 61.674 ms multi_tensor_apply_kernel
111
+ 47.734 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o81924096i64i64_tensorptrbf16gmemalign16o5570564096i64i64_tensorptri32gmemalign16o655361_tensorptri32_0
112
+ 37.832 ms vectorized_elementwise_kernel
113
+ 36.347 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o5570564096i64i64_tensorptrbf16gmemalign16o81924096i64i64_tensorptri32gmemalign16o655361_tensorptri32gm_0
114
+ 34.167 ms cast_mxfp8_2D_kernel
115
+ 30.871 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
116
+ 21.201 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
117
+ 15.404 ms ncclDevKernel_AllGather_RING_LL
118
+ 9.719 ms unrolled_elementwise_kernel
119
+ 9.124 ms nvjet_qqsss_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_NTT
120
+ 8.814 ms nvjet_qqtst_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_TNT
121
+ 8.715 ms nvjet_qqtst_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_NNT
122
+ 8.137 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
123
+ 8.133 ms triton_poi_fused__to_copy_cat_0
124
+ REDUCE 7.306 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o3932164096i64i64_tensorptrf32gmemalign16o3932164096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
125
+ 6.538 ms setGroupedGemmWgradArguments
126
+ 4.237 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o393216409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
127
+ 3.924 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
128
+ REDUCE 3.749 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o5242881536i64i64_tensorptrf32gmemalign16o5242881536i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
129
+ 3.698 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
130
+ 3.617 ms CatArrayBatchedCopy_alignedK_contig
131
+ 3.389 ms gatherTopK
132
+ 3.296 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
133
+ 3.221 ms rmsnorm_bwd_general_kernel
134
+ 2.842 ms fused_rope_backward_kernel
135
+ 2.713 ms reduce_kernel
136
+ 2.666 ms triton_red_fused__to_copy_mul_silu_sum_1
137
+ 2.647 ms fused_rope_forward_kernel
138
+ 2.487 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o524288153611i640_CopyAtom_ThrID10_TVLayoutSrc_0
139
+ 2.249 ms triton_poi_fused__to_copy_mul_silu_0
140
+ 2.168 ms multi_tensor_swizzle_row_scaling_kernel
141
+ 2.167 ms rmsnorm_bwd_tuned_kernel
142
+ 2.122 ms rmsnorm_fwd_general_kernel
143
+ 1.972 ms compute_dot_do_o_specialized
144
+ 1.609 ms rmsnorm_bwd_finalize_general_kernel
145
+ 1.562 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_1281_tensorptri32_gmem_align16_o_655361_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_al_0
146
+ 1.548 ms rmsnorm_fwd_tuned_kernel
147
+
148
+ === QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async.sqlite (device 2, 2 iteration(s)) ===
149
+ reduce_grad kernels : 48 (24/iter)
150
+ reduce_grad sum : 5.528 ms/iter
151
+ reduce_grad busy (union) : 5.528 ms/iter
152
+ ... overlapped with compute: 0.000 ms/iter ( 0.0 %)
153
+ ... exposed on critical path: 5.528 ms/iter (100.0 %)
154
+ other GPU work (union) : 295.416 ms/iter
155
+ traced wall : 337.823 ms/iter
156
+ --> time hidden by overlap : 0.000 ms/iter
157
+
158
+ === kernels on device 3, by total time ===
159
+ 77.129 ms elementwise_kernel
160
+ 73.509 ms device_kernel
161
+ 64.516 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
162
+ 61.742 ms multi_tensor_apply_kernel
163
+ 47.752 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o81924096i64i64_tensorptrbf16gmemalign16o5570564096i64i64_tensorptri32gmemalign16o655361_tensorptri32_0
164
+ 40.630 ms vectorized_elementwise_kernel
165
+ 36.253 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o5570564096i64i64_tensorptrbf16gmemalign16o81924096i64i64_tensorptri32gmemalign16o655361_tensorptri32gm_0
166
+ 34.204 ms cast_mxfp8_2D_kernel
167
+ 28.416 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
168
+ 21.405 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
169
+ 10.239 ms ncclDevKernel_AllGather_RING_LL
170
+ 9.734 ms unrolled_elementwise_kernel
171
+ 9.337 ms nvjet_qqsss_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_NTT
172
+ 8.880 ms nvjet_qqtst_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_TNT
173
+ 8.816 ms nvjet_qqtst_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_NNT
174
+ 8.162 ms triton_poi_fused__to_copy_cat_0
175
+ REDUCE 7.248 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o3932164096i64i64_tensorptrf32gmemalign16o3932164096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
176
+ 6.465 ms setGroupedGemmWgradArguments
177
+ 5.657 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
178
+ 4.065 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
179
+ REDUCE 3.773 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o5242881536i64i64_tensorptrf32gmemalign16o5242881536i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
180
+ 3.745 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
181
+ 3.668 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o393216409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
182
+ 3.614 ms CatArrayBatchedCopy_alignedK_contig
183
+ 3.379 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
184
+ 3.377 ms gatherTopK
185
+ 3.212 ms rmsnorm_bwd_general_kernel
186
+ 2.842 ms fused_rope_backward_kernel
187
+ 2.708 ms reduce_kernel
188
+ 2.687 ms ncclDevKernel_AllReduce_Sum_f32_RING_LL
189
+ 2.665 ms triton_red_fused__to_copy_mul_silu_sum_1
190
+ 2.644 ms fused_rope_forward_kernel
191
+ 2.255 ms triton_poi_fused__to_copy_mul_silu_0
192
+ 2.179 ms rmsnorm_bwd_tuned_kernel
193
+ 2.149 ms multi_tensor_swizzle_row_scaling_kernel
194
+ 2.120 ms rmsnorm_fwd_general_kernel
195
+ 1.970 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o524288153611i640_CopyAtom_ThrID10_TVLayoutSrc_0
196
+ 1.966 ms compute_dot_do_o_specialized
197
+ 1.589 ms rmsnorm_bwd_finalize_general_kernel
198
+ 1.572 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_1281_tensorptri32_gmem_align16_o_655361_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_al_0
199
+
200
+ === QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async.sqlite (device 3, 2 iteration(s)) ===
201
+ reduce_grad kernels : 48 (24/iter)
202
+ reduce_grad sum : 5.510 ms/iter
203
+ reduce_grad busy (union) : 5.510 ms/iter
204
+ ... overlapped with compute: 0.000 ms/iter ( 0.0 %)
205
+ ... exposed on critical path: 5.510 ms/iter (100.0 %)
206
+ other GPU work (union) : 294.831 ms/iter
207
+ traced wall : 379.421 ms/iter
208
+ --> time hidden by overlap : 0.000 ms/iter
209
+
210
+ === kernels on device 4, by total time ===
211
+ 77.208 ms elementwise_kernel
212
+ 74.034 ms device_kernel
213
+ 64.303 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
214
+ 61.646 ms multi_tensor_apply_kernel
215
+ 47.678 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o81924096i64i64_tensorptrbf16gmemalign16o5570564096i64i64_tensorptri32gmemalign16o655361_tensorptri32_0
216
+ 39.151 ms vectorized_elementwise_kernel
217
+ 36.494 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o5570564096i64i64_tensorptrbf16gmemalign16o81924096i64i64_tensorptri32gmemalign16o655361_tensorptri32gm_0
218
+ 34.190 ms cast_mxfp8_2D_kernel
219
+ 28.782 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
220
+ 21.334 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
221
+ 10.040 ms ncclDevKernel_AllGather_RING_LL
222
+ 9.724 ms unrolled_elementwise_kernel
223
+ 9.378 ms nvjet_qqsss_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_NTT
224
+ 8.857 ms nvjet_qqtst_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_NNT
225
+ 8.852 ms nvjet_qqtst_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_TNT
226
+ 8.170 ms triton_poi_fused__to_copy_cat_0
227
+ REDUCE 7.242 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o3932164096i64i64_tensorptrf32gmemalign16o3932164096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
228
+ 6.494 ms setGroupedGemmWgradArguments
229
+ 4.788 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
230
+ 4.087 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
231
+ 3.995 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o393216409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
232
+ REDUCE 3.737 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o5242881536i64i64_tensorptrf32gmemalign16o5242881536i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
233
+ 3.727 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
234
+ 3.615 ms CatArrayBatchedCopy_alignedK_contig
235
+ 3.384 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
236
+ 3.378 ms gatherTopK
237
+ 3.216 ms rmsnorm_bwd_general_kernel
238
+ 2.842 ms fused_rope_backward_kernel
239
+ 2.722 ms reduce_kernel
240
+ 2.664 ms triton_red_fused__to_copy_mul_silu_sum_1
241
+ 2.651 ms fused_rope_forward_kernel
242
+ 2.250 ms triton_poi_fused__to_copy_mul_silu_0
243
+ 2.186 ms rmsnorm_bwd_tuned_kernel
244
+ 2.158 ms multi_tensor_swizzle_row_scaling_kernel
245
+ 2.125 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o524288153611i640_CopyAtom_ThrID10_TVLayoutSrc_0
246
+ 2.121 ms rmsnorm_fwd_general_kernel
247
+ 1.975 ms compute_dot_do_o_specialized
248
+ 1.601 ms rmsnorm_bwd_finalize_general_kernel
249
+ 1.556 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_1281_tensorptri32_gmem_align16_o_655361_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_al_0
250
+ 1.554 ms rmsnorm_fwd_tuned_kernel
251
+
252
+ === QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async.sqlite (device 4, 2 iteration(s)) ===
253
+ reduce_grad kernels : 48 (24/iter)
254
+ reduce_grad sum : 5.490 ms/iter
255
+ reduce_grad busy (union) : 5.490 ms/iter
256
+ ... overlapped with compute: 0.000 ms/iter ( 0.0 %)
257
+ ... exposed on critical path: 5.490 ms/iter (100.0 %)
258
+ other GPU work (union) : 293.149 ms/iter
259
+ traced wall : 411.904 ms/iter
260
+ --> time hidden by overlap : 0.000 ms/iter
261
+
262
+ === kernels on device 5, by total time ===
263
+ 77.196 ms elementwise_kernel
264
+ 72.558 ms device_kernel
265
+ 62.998 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
266
+ 61.809 ms multi_tensor_apply_kernel
267
+ 47.715 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o81924096i64i64_tensorptrbf16gmemalign16o5570564096i64i64_tensorptri32gmemalign16o655361_tensorptri32_0
268
+ 40.425 ms vectorized_elementwise_kernel
269
+ 36.766 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o5570564096i64i64_tensorptrbf16gmemalign16o81924096i64i64_tensorptri32gmemalign16o655361_tensorptri32gm_0
270
+ 34.095 ms cast_mxfp8_2D_kernel
271
+ 30.677 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
272
+ 21.208 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
273
+ 14.930 ms ncclDevKernel_AllGather_RING_LL
274
+ 9.723 ms unrolled_elementwise_kernel
275
+ 9.134 ms nvjet_qqsss_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_NTT
276
+ 8.863 ms nvjet_qqtst_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_TNT
277
+ 8.668 ms nvjet_qqtst_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_NNT
278
+ 8.300 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
279
+ 8.130 ms triton_poi_fused__to_copy_cat_0
280
+ REDUCE 7.482 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o3932164096i64i64_tensorptrf32gmemalign16o3932164096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
281
+ 6.451 ms setGroupedGemmWgradArguments
282
+ 4.568 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o393216409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
283
+ 3.941 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
284
+ 3.712 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
285
+ REDUCE 3.700 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o5242881536i64i64_tensorptrf32gmemalign16o5242881536i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
286
+ 3.612 ms CatArrayBatchedCopy_alignedK_contig
287
+ 3.374 ms gatherTopK
288
+ 3.326 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
289
+ 3.230 ms rmsnorm_bwd_general_kernel
290
+ 2.840 ms fused_rope_backward_kernel
291
+ 2.708 ms reduce_kernel
292
+ 2.666 ms triton_red_fused__to_copy_mul_silu_sum_1
293
+ 2.645 ms fused_rope_forward_kernel
294
+ 2.308 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o524288153611i640_CopyAtom_ThrID10_TVLayoutSrc_0
295
+ 2.253 ms triton_poi_fused__to_copy_mul_silu_0
296
+ 2.153 ms rmsnorm_bwd_tuned_kernel
297
+ 2.152 ms multi_tensor_swizzle_row_scaling_kernel
298
+ 2.121 ms rmsnorm_fwd_general_kernel
299
+ 2.049 ms ncclDevKernel_AllReduce_Sum_f32_RING_LL
300
+ 1.963 ms compute_dot_do_o_specialized
301
+ 1.634 ms rmsnorm_bwd_finalize_general_kernel
302
+ 1.548 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_1281_tensorptri32_gmem_align16_o_655361_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_al_0
303
+
304
+ === QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async.sqlite (device 5, 2 iteration(s)) ===
305
+ reduce_grad kernels : 48 (24/iter)
306
+ reduce_grad sum : 5.591 ms/iter
307
+ reduce_grad busy (union) : 5.591 ms/iter
308
+ ... overlapped with compute: 0.000 ms/iter ( 0.0 %)
309
+ ... exposed on critical path: 5.591 ms/iter (100.0 %)
310
+ other GPU work (union) : 296.768 ms/iter
311
+ traced wall : 407.038 ms/iter
312
+ --> time hidden by overlap : 0.000 ms/iter
313
+
314
+ === kernels on device 6, by total time ===
315
+ 77.247 ms elementwise_kernel
316
+ 72.289 ms device_kernel
317
+ 63.308 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
318
+ 61.737 ms multi_tensor_apply_kernel
319
+ 47.707 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o81924096i64i64_tensorptrbf16gmemalign16o5570564096i64i64_tensorptri32gmemalign16o655361_tensorptri32_0
320
+ 37.725 ms vectorized_elementwise_kernel
321
+ 35.975 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o5570564096i64i64_tensorptrbf16gmemalign16o81924096i64i64_tensorptri32gmemalign16o655361_tensorptri32gm_0
322
+ 34.153 ms cast_mxfp8_2D_kernel
323
+ 30.437 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
324
+ 21.279 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
325
+ 16.308 ms ncclDevKernel_AllGather_RING_LL
326
+ 9.710 ms unrolled_elementwise_kernel
327
+ 9.159 ms nvjet_qqsss_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_NTT
328
+ 8.842 ms nvjet_qqtst_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_TNT
329
+ 8.656 ms nvjet_qqtst_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_NNT
330
+ 8.457 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
331
+ 8.147 ms triton_poi_fused__to_copy_cat_0
332
+ REDUCE 7.667 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o3932164096i64i64_tensorptrf32gmemalign16o3932164096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
333
+ 6.522 ms setGroupedGemmWgradArguments
334
+ 3.974 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
335
+ 3.692 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
336
+ REDUCE 3.685 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o5242881536i64i64_tensorptrf32gmemalign16o5242881536i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
337
+ 3.610 ms CatArrayBatchedCopy_alignedK_contig
338
+ 3.541 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o393216409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
339
+ 3.374 ms gatherTopK
340
+ 3.327 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
341
+ 3.202 ms rmsnorm_bwd_general_kernel
342
+ 2.965 ms ncclDevKernel_AllReduce_Sum_f32_RING_LL
343
+ 2.842 ms fused_rope_backward_kernel
344
+ 2.698 ms reduce_kernel
345
+ 2.663 ms triton_red_fused__to_copy_mul_silu_sum_1
346
+ 2.648 ms fused_rope_forward_kernel
347
+ 2.433 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o524288153611i640_CopyAtom_ThrID10_TVLayoutSrc_0
348
+ 2.256 ms triton_poi_fused__to_copy_mul_silu_0
349
+ 2.158 ms multi_tensor_swizzle_row_scaling_kernel
350
+ 2.157 ms rmsnorm_bwd_tuned_kernel
351
+ 2.115 ms rmsnorm_fwd_general_kernel
352
+ 1.966 ms compute_dot_do_o_specialized
353
+ 1.615 ms rmsnorm_bwd_finalize_general_kernel
354
+ 1.554 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_1281_tensorptri32_gmem_align16_o_655361_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_al_0
355
+
356
+ === QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async.sqlite (device 6, 2 iteration(s)) ===
357
+ reduce_grad kernels : 48 (24/iter)
358
+ reduce_grad sum : 5.676 ms/iter
359
+ reduce_grad busy (union) : 5.676 ms/iter
360
+ ... overlapped with compute: 0.000 ms/iter ( 0.0 %)
361
+ ... exposed on critical path: 5.676 ms/iter (100.0 %)
362
+ other GPU work (union) : 296.438 ms/iter
363
+ traced wall : 337.788 ms/iter
364
+ --> time hidden by overlap : 0.000 ms/iter
365
+
366
+ === kernels on device 7, by total time ===
367
+ 77.202 ms elementwise_kernel
368
+ 72.774 ms device_kernel
369
+ 62.854 ms cudnn_generated_fort_native_sdpa_sm100_flash_bprop_f16_knob_31_128x128x128_1x4x1_kernel0_0
370
+ 61.727 ms multi_tensor_apply_kernel
371
+ 47.735 ms kernel_cutlass_kernel_moonepdispatchDispatchKernel_object_at__tensorptrbf16gmemalign16o81924096i64i64_tensorptrbf16gmemalign16o5570564096i64i64_tensorptri32gmemalign16o655361_tensorptri32_0
372
+ 37.952 ms vectorized_elementwise_kernel
373
+ 35.603 ms kernel_cutlass_kernel_moonepcombineCombineKernel_object_at__tensorptrbf16gmemalign16o5570564096i64i64_tensorptrbf16gmemalign16o81924096i64i64_tensorptri32gmemalign16o655361_tensorptri32gm_0
374
+ 34.140 ms cast_mxfp8_2D_kernel
375
+ 30.667 ms ncclDevKernel_ReduceScatter_Sum_f32_RING_LL
376
+ 21.145 ms cudnn_generated_fort_native_sdpa_sm100_flash_fprop_f16_knob_7_128x128x128_4x1x1_kernel0_0
377
+ 15.950 ms ncclDevKernel_AllGather_RING_LL
378
+ 9.731 ms unrolled_elementwise_kernel
379
+ 9.153 ms nvjet_qqsss_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_NTT
380
+ 8.838 ms nvjet_qqtst_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_TNT
381
+ 8.767 ms kernel_cutlass_kernel_moonepinter_rank_syncInterRankSyncKernel_object_at__tensorptri32gmemalign16o41943041_tensorptri32gmemalign16o10___0
382
+ 8.694 ms nvjet_qqtst_128x256_128x6_2x1_2cta_v_bz_Avec32UE8M0_Bvec32UE8M0_NNT
383
+ 8.126 ms triton_poi_fused__to_copy_cat_0
384
+ REDUCE 7.789 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o3932164096i64i64_tensorptrf32gmemalign16o3932164096i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
385
+ 6.502 ms setGroupedGemmWgradArguments
386
+ 3.947 ms nvjet_tss_128x256_64x6_2x1_2cta_v_badd_NTT
387
+ REDUCE 3.749 ms kernel_cutlass_kernel_moonepgrad_reduceGradReduceKernel_object_at__tensorptrf32gmemalign16o5242881536i64i64_tensorptrf32gmemalign16o5242881536i64i64_CopyAtom_ThrID10_TVLayoutSrc11638401_T_0
388
+ 3.688 ms nvjet_tst_256x256_64x4_2x2_2cta_h_bz_NNT
389
+ 3.615 ms CatArrayBatchedCopy_alignedK_contig
390
+ 3.550 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o393216409611i640_CopyAtom_ThrID10_TVLayoutSrc_0
391
+ 3.346 ms gatherTopK
392
+ 3.331 ms nvjet_tst_128x256_64x6_2x1_2cta_v_bz_TNT
393
+ 3.231 ms rmsnorm_bwd_general_kernel
394
+ 2.846 ms fused_rope_backward_kernel
395
+ 2.704 ms reduce_kernel
396
+ 2.667 ms triton_red_fused__to_copy_mul_silu_sum_1
397
+ 2.642 ms fused_rope_forward_kernel
398
+ 2.267 ms kernel_cutlass_kernel_moonepprefetchPrefetchKernel_object_at__CopyAtom_ThrID10_TVLayoutSrc11638401_TVLayoutDst11638401_Valuetypebf16_tensor00o524288153611i640_CopyAtom_ThrID10_TVLayoutSrc_0
399
+ 2.248 ms triton_poi_fused__to_copy_mul_silu_0
400
+ 2.159 ms rmsnorm_bwd_tuned_kernel
401
+ 2.154 ms multi_tensor_swizzle_row_scaling_kernel
402
+ 2.110 ms rmsnorm_fwd_general_kernel
403
+ 1.970 ms compute_dot_do_o_specialized
404
+ 1.621 ms rmsnorm_bwd_finalize_general_kernel
405
+ 1.563 ms kernel_cutlass_kernel_moonepplanningPlanningKernel_object_at__tensorptri32_gmem_align16_o_1281_tensorptri32_gmem_align16_o_655361_tensorptri32_gmem_align16_o_41943041_tensorptri32_gmem_al_0
406
+ 1.558 ms rmsnorm_fwd_tuned_kernel
407
+
408
+ === QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async.sqlite (device 7, 2 iteration(s)) ===
409
+ reduce_grad kernels : 48 (24/iter)
410
+ reduce_grad sum : 5.769 ms/iter
411
+ reduce_grad busy (union) : 5.769 ms/iter
412
+ ... overlapped with compute: 0.000 ms/iter ( 0.0 %)
413
+ ... exposed on critical path: 5.769 ms/iter (100.0 %)
414
+ other GPU work (union) : 295.378 ms/iter
415
+ traced wall : 337.756 ms/iter
416
+ --> time hidden by overlap : 0.000 ms/iter
tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/nsys/moonep_async_overlap_summary.txt ADDED
@@ -0,0 +1 @@
 
 
1
+ devices=8 reduce=5.617 ms/iter overlap=0.000 ms/iter (0.0%) exposed=5.617 ms/iter status=FAIL
tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/nsys/run.env ADDED
@@ -0,0 +1,73 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ backend=moonep
2
+ config=QWEN3_235B
3
+ family=tierB
4
+ repo=/workspace/Megatron-MoE-Echo
5
+ branch=moonep_backend_async_grad_reduce
6
+ commit=7bc235bf
7
+ moonep_source_dir=/workspace/MoonEP
8
+ moonep_source_branch=megatron-ddp-symmetric-params
9
+ moonep_source_commit=d8bc666
10
+ num_layers=12
11
+ hidden=4096
12
+ ffn_hidden=12288
13
+ moe_ffn_hidden=1536
14
+ num_experts=128
15
+ topk=8
16
+ experts_per_rank=16
17
+ expert_params=18874368
18
+ heads=64
19
+ kv_groups=4
20
+ kv_channels=128
21
+ qk_layernorm=1
22
+ shared_ffn=0
23
+ shared_expert_overlap=0
24
+ moe_token_dispatcher=flex
25
+ mla=0
26
+ dense_first_layers=0
27
+ moe_layers=12
28
+ rotary_base=1000000
29
+ rotary_scaling=
30
+ qkv_bias=0
31
+ router_balancing=aux_loss
32
+ aux_loss_coeff=1e-3
33
+ router_pre_softmax=0
34
+ router_fusion=0
35
+ router_dtype=fp32
36
+ device_initiated_grouped_gemm=1
37
+ norm_epsilon=1e-6
38
+ seq_length=8192
39
+ micro_batch_size=1
40
+ global_batch_size=8
41
+ num_microbatches=1
42
+ assignments_per_rank=65536
43
+ mean_tokens_per_expert=4096
44
+ slot_policy=matched
45
+ slots_per_rank=16
46
+ capacity_mode=none
47
+ capacity=none
48
+ cuda_graph=1
49
+ probe_mode=0
50
+ async_grad_reduce=1
51
+ branch_role=timing
52
+ memory_dump_mode=none
53
+ memory_snapshot_path=
54
+ cuda_graph_warmup_steps=3
55
+ recompute=none
56
+ force_balance=0
57
+ train_iters=10
58
+ ep_size=8
59
+ echo_dispatch_overlap=1
60
+ echo_dispatch_recompute=0
61
+ ultraep_quota_min_tokens=
62
+ ultraep_weight_sync_join=
63
+ profile_enable=1
64
+ kernel_trace=1
65
+ fp8_mode=mxfp8
66
+ fp8_recipe=mxfp8
67
+ training_seed=1234
68
+ learning_rate=1e-4
69
+ min_learning_rate=1.0e-5
70
+ lr_warmup_iters=5
71
+ state_dump_enable=0
72
+ state_dump_dir=
73
+ moonep_assert_no_pending_reduce=1
tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/nsys/train.log ADDED
@@ -0,0 +1,1326 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [01:04:00] WARN: moonep_backend_async_grad_reduce lacks --moe-router-fusion; QWEN3_235B runs unfused (timing only, routing identical)
2
+ ==============================================================
3
+ QWEN3_235B / moonep (slots/rank=16 policy=matched)
4
+ E=128 K=8 H'=1536 m=16
5
+ L=12 seq=8192 gbs=8 (1 microbatches)
6
+ cap_mode=none cap=none cg=1 recompute=none
7
+ memory_dump=none snapshots=/workspace/tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/nsys/memory_snapshots/snapshot.pickle
8
+ repo=/workspace/Megatron-MoE-Echo @ 7bc235bf
9
+ ==============================================================
10
+ W0830 01:04:02.188000 363852 torch/distributed/run.py:766]
11
+ W0830 01:04:02.188000 363852 torch/distributed/run.py:766] *****************************************
12
+ W0830 01:04:02.188000 363852 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.
13
+ W0830 01:04:02.188000 363852 torch/distributed/run.py:766] *****************************************
14
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
15
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
16
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
17
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
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
+ 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
23
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
24
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
25
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
26
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
27
+ 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
28
+ WARNING: overriding default arguments for tokenizer_type:GPT2BPETokenizer with tokenizer_type:GPT2BPETokenizer
29
+ Number of virtual stages per pipeline stage: None
30
+ accumulate and all-reduce gradients in fp32 for bfloat16 data type.
31
+ using torch.bfloat16 for parameters ...
32
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
33
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
34
+ ------------------------ arguments ------------------------
35
+ account_for_embedding_in_pipeline_split ......... False
36
+ account_for_loss_in_pipeline_split .............. False
37
+ accumulate_allreduce_grads_in_fp32 .............. True
38
+ adam_beta1 ...................................... 0.9
39
+ adam_beta2 ...................................... 0.999
40
+ adam_eps ........................................ 1e-08
41
+ add_bias_linear ................................. False
42
+ add_position_embedding .......................... False
43
+ add_qkv_bias .................................... False
44
+ adlr_autoresume ................................. False
45
+ adlr_autoresume_interval ........................ 1000
46
+ align_grad_reduce ............................... True
47
+ align_param_gather .............................. False
48
+ app_tag_run_name ................................ None
49
+ app_tag_run_version ............................. 0.0.0
50
+ apply_layernorm_1p .............................. False
51
+ apply_query_key_layer_scaling ................... False
52
+ apply_residual_connection_post_layernorm ........ False
53
+ apply_rope_fusion ............................... True
54
+ async_save ...................................... None
55
+ async_tensor_model_parallel_allreduce ........... True
56
+ attention_backend ............................... AttnBackend.auto
57
+ attention_dropout ............................... 0.0
58
+ attention_softmax_in_fp32 ....................... False
59
+ auto_detect_ckpt_format ......................... False
60
+ barrier_with_L1_time ............................ True
61
+ bert_binary_head ................................ True
62
+ bert_embedder_type .............................. megatron
63
+ bert_load ....................................... None
64
+ bf16 ............................................ True
65
+ bias_dropout_fusion ............................. True
66
+ bias_gelu_fusion ................................ False
67
+ bias_swiglu_fusion .............................. True
68
+ biencoder_projection_dim ........................ 0
69
+ biencoder_shared_query_context_model ............ False
70
+ block_data_path ................................. None
71
+ calc_ft_timeouts ................................ False
72
+ calculate_per_token_loss ........................ False
73
+ check_for_large_grads ........................... False
74
+ check_for_nan_in_loss_and_grad .................. False
75
+ check_for_spiky_loss ............................ False
76
+ check_weight_hash_across_dp_replicas_interval ... None
77
+ ckpt_assume_constant_structure .................. False
78
+ ckpt_convert_format ............................. None
79
+ ckpt_convert_save ............................... None
80
+ ckpt_convert_update_legacy_dist_opt_format ...... False
81
+ ckpt_format ..................................... torch_dist
82
+ ckpt_fully_parallel_load ........................ False
83
+ ckpt_fully_parallel_save ........................ True
84
+ ckpt_fully_parallel_save_deprecated ............. False
85
+ ckpt_step ....................................... None
86
+ classes_fraction ................................ 1.0
87
+ clip_grad ....................................... 1.0
88
+ clone_scatter_output_in_embedding ............... True
89
+ config_logger_dir ...............................
90
+ consumed_train_samples .......................... 0
91
+ consumed_valid_samples .......................... 0
92
+ context_parallel_size ........................... 1
93
+ cp_comm_type .................................... ['p2p']
94
+ create_attention_mask_in_dataloader ............. True
95
+ cross_entropy_fusion_impl ....................... native
96
+ cross_entropy_loss_fusion ....................... False
97
+ cuda_graph_scope ................................ full_iteration
98
+ cuda_graph_warmup_steps ......................... 3
99
+ data_args_path .................................. None
100
+ data_cache_path ................................. None
101
+ data_parallel_random_init ....................... False
102
+ data_parallel_sharding_strategy ................. no_shard
103
+ data_parallel_size .............................. 8
104
+ data_path ....................................... ['/workspace/data/wikitext/wikitext103_gpt2_text_document']
105
+ data_per_class_fraction ......................... 1.0
106
+ data_sharding ................................... True
107
+ dataloader_type ................................. single
108
+ ddp_average_in_collective ....................... False
109
+ ddp_bucket_size ................................. None
110
+ ddp_num_buckets ................................. None
111
+ ddp_pad_buckets_for_high_nccl_busbw ............. False
112
+ decoder_first_pipeline_num_layers ............... None
113
+ decoder_last_pipeline_num_layers ................ None
114
+ decoder_num_layers .............................. None
115
+ decoder_seq_length .............................. None
116
+ decoupled_lr .................................... None
117
+ decoupled_min_lr ................................ None
118
+ decrease_batch_size_if_needed ................... False
119
+ defer_embedding_wgrad_compute ................... False
120
+ delay_wgrad_compute ............................. False
121
+ deprecated_use_mcore_models ..................... True
122
+ deterministic_mode .............................. False
123
+ dino_bottleneck_size ............................ 256
124
+ dino_freeze_last_layer .......................... 1
125
+ dino_head_hidden_size ........................... 2048
126
+ dino_local_crops_number ......................... 10
127
+ dino_local_img_size ............................. 96
128
+ dino_norm_last_layer ............................ False
129
+ dino_teacher_temp ............................... 0.07
130
+ dino_warmup_teacher_temp ........................ 0.04
131
+ dino_warmup_teacher_temp_epochs ................. 30
132
+ disable_bf16_reduced_precision_matmul ........... False
133
+ disable_mamba_mem_eff_path ...................... False
134
+ disable_straggler_on_startup .................... False
135
+ dist_ckpt_format_deprecated ..................... None
136
+ dist_ckpt_strictness ............................ assume_ok_unexpected
137
+ distribute_saved_activations .................... False
138
+ distributed_backend ............................. nccl
139
+ distributed_timeout_minutes ..................... 10
140
+ embedding_init_method_std ....................... None
141
+ embedding_path .................................. None
142
+ empty_unused_memory_level ....................... 0
143
+ enable_cuda_graph ............................... True
144
+ enable_experimental ............................. False
145
+ enable_ft_package ............................... False
146
+ enable_gloo_process_groups ...................... True
147
+ enable_msc ...................................... True
148
+ enable_one_logger ............................... True
149
+ encoder_num_layers .............................. 12
150
+ encoder_seq_length .............................. 8192
151
+ end_weight_decay ................................ 0.1
152
+ eod_mask_loss ................................... False
153
+ error_injection_rate ............................ 0
154
+ error_injection_type ............................ transient_error
155
+ eval_interval ................................... 1000000
156
+ eval_iters ...................................... 0
157
+ evidence_data_path .............................. None
158
+ exit_duration_in_mins ........................... None
159
+ exit_interval ................................... None
160
+ exit_on_missing_checkpoint ...................... False
161
+ exit_signal_handler ............................. False
162
+ exp_avg_dtype ................................... torch.float32
163
+ exp_avg_sq_dtype ................................ torch.float32
164
+ expert_model_parallel_size ...................... 8
165
+ expert_tensor_parallel_size ..................... 1
166
+ export_force_local_attention .................... False
167
+ export_kd_cfg ................................... None
168
+ export_kd_teacher_ckpt_format ................... None
169
+ export_kd_teacher_load .......................... None
170
+ export_kv_cache_quant ........................... False
171
+ export_legacy_megatron .......................... False
172
+ export_model_type ............................... GPTModel
173
+ export_moe_apply_probs_on_input ................. False
174
+ export_qk_l2_norm ............................... False
175
+ export_quant_cfg ................................ None
176
+ export_real_quant_cfg ........................... None
177
+ export_te_mcore_model ........................... False
178
+ external_cuda_graph ............................. False
179
+ ffn_hidden_size ................................. 12288
180
+ finetune ........................................ False
181
+ finetune_data_split ............................. train
182
+ finetune_hf_dataset ............................. None
183
+ first_last_layers_bf16 .......................... False
184
+ flash_decode .................................... False
185
+ fp16 ............................................ False
186
+ fp16_lm_cross_entropy ........................... False
187
+ fp32_residual_connection ........................ False
188
+ fp8 ............................................. e4m3
189
+ fp8_amax_compute_algo ........................... most_recent
190
+ fp8_amax_history_len ............................ 1
191
+ fp8_interval .................................... 1
192
+ fp8_margin ...................................... 0
193
+ fp8_param_gather ................................ False
194
+ fp8_recipe ...................................... mxfp8
195
+ fp8_wgrad ....................................... True
196
+ fsdp_double_buffer .............................. False
197
+ global_batch_size ............................... 8
198
+ grad_reduce_in_bf16 ............................. False
199
+ gradient_accumulation_fusion .................... True
200
+ gradient_reduce_div_fusion ...................... True
201
+ group_query_attention ........................... True
202
+ head_lr_mult .................................... 1.0
203
+ heterogeneous_layers_config_encoded_json ........ None
204
+ heterogeneous_layers_config_path ................ None
205
+ hidden_dropout .................................. 0.0
206
+ hidden_size ..................................... 4096
207
+ hierarchical_context_parallel_sizes ............. None
208
+ high_priority_stream_groups ..................... []
209
+ hybrid_attention_ratio .......................... 0.0
210
+ hybrid_mlp_ratio ................................ 0.0
211
+ hybrid_override_pattern ......................... None
212
+ hysteresis ...................................... 2
213
+ ict_head_size ................................... None
214
+ ict_load ........................................ None
215
+ img_h ........................................... 224
216
+ img_w ........................................... 224
217
+ indexer_batch_size .............................. 128
218
+ indexer_log_interval ............................ 1000
219
+ inference_batch_times_seqlen_threshold .......... -1
220
+ inference_dynamic_batching ...................... False
221
+ inference_dynamic_batching_buffer_guaranteed_fraction 0.2
222
+ inference_dynamic_batching_buffer_overflow_factor None
223
+ inference_dynamic_batching_buffer_size_gb ....... 40.0
224
+ inference_dynamic_batching_chunk_size ........... 256
225
+ inference_dynamic_batching_max_requests_override None
226
+ inference_dynamic_batching_max_tokens_override .. None
227
+ inference_dynamic_batching_num_cuda_graphs ...... 16
228
+ inference_max_batch_size ........................ 8
229
+ inference_max_seq_length ........................ 2560
230
+ inference_rng_tracker ........................... False
231
+ init_method_std ................................. 0.01
232
+ init_method_xavier_uniform ...................... False
233
+ init_model_with_meta_device ..................... False
234
+ initial_loss_scale .............................. 4294967296
235
+ inprocess_active_world_size ..................... 8
236
+ inprocess_barrier_timeout ....................... 120
237
+ inprocess_completion_timeout .................... 120
238
+ inprocess_empty_cuda_cache ...................... False
239
+ inprocess_granularity ........................... node
240
+ inprocess_hard_timeout .......................... 90
241
+ inprocess_heartbeat_interval .................... 30
242
+ inprocess_heartbeat_timeout ..................... 60
243
+ inprocess_last_call_wait ........................ 1
244
+ inprocess_max_iterations ........................ None
245
+ inprocess_monitor_process_interval .............. 1.0
246
+ inprocess_monitor_thread_interval ............... 1.0
247
+ inprocess_progress_watchdog_interval ............ 1.0
248
+ inprocess_restart ............................... False
249
+ inprocess_soft_timeout .......................... 60
250
+ inprocess_termination_grace_time ................ 1
251
+ is_hybrid_model ................................. False
252
+ iter_per_epoch .................................. 1250
253
+ iterations_to_skip .............................. []
254
+ keep_fp8_transpose_cache_when_using_custom_fsdp . False
255
+ kitchen_config_file ............................. None
256
+ kitchen_recipe_number ........................... None
257
+ kv_channels ..................................... 128
258
+ kv_lora_rank .................................... 32
259
+ lazy_mpu_init ................................... None
260
+ load ............................................ None
261
+ load_main_params_from_ckpt ...................... None
262
+ load_model_opt_format ........................... False
263
+ local_rank ...................................... 0
264
+ log_energy ...................................... False
265
+ log_interval .................................... 1
266
+ log_loss_scale_to_tensorboard ................... True
267
+ log_memory_to_tensorboard ....................... False
268
+ log_num_zeros_in_grad ........................... False
269
+ log_params_norm ................................. False
270
+ log_progress .................................... False
271
+ log_straggler ................................... False
272
+ log_throughput .................................. True
273
+ log_timers_to_tensorboard ....................... False
274
+ log_validation_ppl_to_tensorboard ............... False
275
+ log_world_size_to_tensorboard ................... False
276
+ logging_level ................................... None
277
+ loss_scale ...................................... None
278
+ loss_scale_window ............................... 1000
279
+ lr .............................................. 0.0001
280
+ lr_decay_iters .................................. 8
281
+ lr_decay_samples ................................ None
282
+ lr_decay_style .................................. cosine
283
+ lr_warmup_fraction .............................. None
284
+ lr_warmup_init .................................. 0.0
285
+ lr_warmup_iters ................................. 5
286
+ lr_warmup_samples ............................... 0
287
+ lr_wsd_decay_iters .............................. None
288
+ lr_wsd_decay_samples ............................ None
289
+ lr_wsd_decay_style .............................. exponential
290
+ main_grads_dtype ................................ torch.float32
291
+ main_params_dtype ............................... torch.float32
292
+ make_vocab_size_divisible_by .................... 128
293
+ mamba_head_dim .................................. 64
294
+ mamba_num_groups ................................ 8
295
+ mamba_num_heads ................................. None
296
+ mamba_state_dim ................................. 128
297
+ manual_gc ....................................... False
298
+ manual_gc_eval .................................. True
299
+ manual_gc_interval .............................. 0
300
+ mask_factor ..................................... 1.0
301
+ mask_prob ....................................... 0.15
302
+ mask_type ....................................... random
303
+ masked_softmax_fusion ........................... False
304
+ max_position_embeddings ......................... 32768
305
+ max_tokens_to_oom ............................... 12000
306
+ memory_snapshot_path ............................ snapshot.pickle
307
+ merge_file ...................................... /workspace/data/wikitext/gpt2-merges.txt
308
+ micro_batch_size ................................ 1
309
+ microbatch_group_size_per_vp_stage .............. None
310
+ mid_level_dataset_surplus ....................... 0.005
311
+ min_loss_scale .................................. 1.0
312
+ min_lr .......................................... 1e-05
313
+ mlp_chunks_for_prefill .......................... 1
314
+ mmap_bin_files .................................. True
315
+ mock_data ....................................... False
316
+ moe_apply_probs_on_input ........................ False
317
+ moe_aux_loss_coeff .............................. 0.001
318
+ moe_deepep_num_sms .............................. 20
319
+ moe_echo_enable_random_offloading ............... False
320
+ moe_echo_expert_dispatch_overlap ................ False
321
+ moe_echo_expert_dispatcher_type ................. hybridep
322
+ moe_echo_recompute_expert_dispatch .............. False
323
+ moe_enable_deepep ............................... False
324
+ moe_enable_echo ................................. False
325
+ moe_enable_hybridep ............................. False
326
+ moe_enable_moonep ............................... True
327
+ moe_expert_capacity_factor ...................... None
328
+ moe_extended_tp ................................. False
329
+ moe_ffn_hidden_size ............................. 1536
330
+ moe_grouped_gemm ................................ True
331
+ moe_hybridep_num_sms ............................ 16
332
+ moe_input_jitter_eps ............................ None
333
+ moe_layer_freq .................................. 1
334
+ moe_layer_recompute ............................. False
335
+ moe_moonep_async_grad_reduce .................... True
336
+ moe_moonep_num_sms .............................. None
337
+ moe_num_echo_experts ............................ None
338
+ moe_pad_expert_input_to_capacity ................ False
339
+ moe_per_layer_logging ........................... False
340
+ moe_permute_fusion .............................. False
341
+ moe_received_token_capacity ..................... None
342
+ moe_router_bias_update_rate ..................... 0.001
343
+ moe_router_dtype ................................ fp32
344
+ moe_router_enable_expert_bias ................... False
345
+ moe_router_force_load_balancing ................. False
346
+ moe_router_group_topk ........................... None
347
+ moe_router_load_balancing_type .................. aux_loss
348
+ moe_router_num_groups ........................... None
349
+ moe_router_padding_for_fp8 ...................... False
350
+ moe_router_pre_softmax .......................... False
351
+ moe_router_score_function ....................... softmax
352
+ moe_router_topk ................................. 8
353
+ moe_router_topk_scaling_factor .................. None
354
+ moe_shared_expert_intermediate_size ............. None
355
+ moe_shared_expert_overlap ....................... False
356
+ moe_token_dispatcher_type ....................... flex
357
+ moe_token_drop_policy ........................... probs
358
+ moe_upcycling_granularity ....................... 1
359
+ moe_use_device_initiated_grouped_gemm ........... True
360
+ moe_use_legacy_grouped_gemm ..................... False
361
+ moe_use_upcycling ............................... False
362
+ moe_z_loss_coeff ................................ None
363
+ mrope_section ................................... None
364
+ mscale .......................................... 1.0
365
+ mscale_all_dim .................................. 0.0
366
+ mtp_loss_scaling_factor ......................... 0.1
367
+ mtp_num_layers .................................. None
368
+ multi_latent_attention .......................... False
369
+ nccl_all_reduce_for_prefill ..................... False
370
+ nccl_communicator_config_path ................... None
371
+ nccl_ub ......................................... False
372
+ no_load_optim ................................... None
373
+ no_load_rng ..................................... None
374
+ no_persist_layer_norm ........................... False
375
+ no_rope_freq .................................... None
376
+ no_save_optim ................................... None
377
+ no_save_rng ..................................... None
378
+ non_persistent_ckpt_type ........................ None
379
+ non_persistent_global_ckpt_dir .................. None
380
+ non_persistent_local_ckpt_algo .................. fully_parallel
381
+ non_persistent_local_ckpt_dir ................... None
382
+ non_persistent_save_interval .................... None
383
+ norm_epsilon .................................... 1e-06
384
+ normalization ................................... RMSNorm
385
+ num_attention_heads ............................. 64
386
+ num_channels .................................... 3
387
+ num_classes ..................................... 1000
388
+ num_dataset_builder_threads ..................... 1
389
+ num_distributed_optimizer_instances ............. 1
390
+ num_experts ..................................... 128
391
+ num_layers ...................................... 12
392
+ num_layers_at_end_in_bf16 ....................... 1
393
+ num_layers_at_start_in_bf16 ..................... 1
394
+ num_layers_per_virtual_pipeline_stage ........... None
395
+ num_query_groups ................................ 4
396
+ num_virtual_stages_per_pipeline_rank ............ None
397
+ num_workers ..................................... 2
398
+ object_storage_cache_path ....................... None
399
+ one_logger_async ................................ False
400
+ one_logger_project .............................. megatron-lm
401
+ one_logger_run_name ............................. None
402
+ onnx_safe ....................................... None
403
+ openai_gelu ..................................... False
404
+ optimizer ....................................... adam
405
+ optimizer_cpu_offload ........................... False
406
+ optimizer_offload_fraction ...................... 1.0
407
+ output_bert_embeddings .......................... False
408
+ overlap_cpu_optimizer_d2h_h2d ................... False
409
+ overlap_grad_reduce ............................. True
410
+ overlap_moe_expert_parallel_comm ................ False
411
+ overlap_p2p_comm ................................ False
412
+ overlap_p2p_comm_warmup_flush ................... False
413
+ overlap_param_gather ............................ True
414
+ overlap_param_gather_with_optimizer_step ........ False
415
+ override_opt_param_scheduler .................... False
416
+ params_dtype .................................... torch.bfloat16
417
+ patch_dim ....................................... 16
418
+ per_split_data_args_path ........................ None
419
+ perform_initialization .......................... True
420
+ pin_cpu_grads ................................... True
421
+ pin_cpu_params .................................. True
422
+ pipeline_model_parallel_comm_backend ............ None
423
+ pipeline_model_parallel_layout .................. None
424
+ pipeline_model_parallel_size .................... 1
425
+ position_embedding_type ......................... rope
426
+ pretrained_checkpoint ........................... None
427
+ profile ......................................... True
428
+ profile_ranks ................................... [0, 1, 2, 3, 4, 5, 6, 7]
429
+ profile_step_end ................................ 8
430
+ profile_step_start .............................. 6
431
+ q_lora_rank ..................................... None
432
+ qk_head_dim ..................................... 128
433
+ qk_l2_norm ...................................... False
434
+ qk_layernorm .................................... True
435
+ qk_pos_emb_head_dim ............................. 64
436
+ query_in_block_prob ............................. 0.1
437
+ rampup_batch_size ............................... None
438
+ rank ............................................ 0
439
+ recompute_granularity ........................... None
440
+ recompute_method ................................ None
441
+ recompute_modules ............................... None
442
+ recompute_num_layers ............................ None
443
+ record_memory_history ........................... False
444
+ relative_attention_max_distance ................. 128
445
+ relative_attention_num_buckets .................. 32
446
+ replication ..................................... False
447
+ replication_factor .............................. 2
448
+ replication_jump ................................ None
449
+ rerun_mode ...................................... validate_results
450
+ reset_attention_mask ............................ False
451
+ reset_position_ids .............................. False
452
+ result_rejected_tracker_filename ................ None
453
+ retriever_report_topk_accuracies ................ []
454
+ retriever_score_scaling ......................... False
455
+ retriever_seq_length ............................ 256
456
+ retro_add_retriever ............................. False
457
+ retro_attention_gate ............................ 1
458
+ retro_cyclic_train_iters ........................ None
459
+ retro_encoder_attention_dropout ................. 0.1
460
+ retro_encoder_hidden_dropout .................... 0.1
461
+ retro_encoder_layers ............................ 2
462
+ retro_num_neighbors ............................. 2
463
+ retro_num_retrieved_chunks ...................... 2
464
+ retro_project_dir ............................... None
465
+ retro_verify_neighbor_count ..................... True
466
+ reuse_grad_buf_for_mxfp8_param_ag ............... False
467
+ rope_scaling_factor ............................. 8.0
468
+ rope_type ....................................... None
469
+ rotary_base ..................................... 1000000
470
+ rotary_interleaved .............................. False
471
+ rotary_percent .................................. 1.0
472
+ rotary_scaling_factor ........................... 1.0
473
+ rotary_seq_len_interpolation_factor ............. None
474
+ run_workload_inspector_server ................... False
475
+ sample_rate ..................................... 1.0
476
+ save ............................................ None
477
+ save_interval ................................... None
478
+ save_retain_interval ............................ None
479
+ scatter_gather_tensors_in_pipeline .............. True
480
+ seed ............................................ 1234
481
+ seq_length ...................................... 8192
482
+ sequence_parallel ............................... False
483
+ sft ............................................. False
484
+ sft_tokenizer_prompt_format ..................... nemotron-h-aligned
485
+ sgd_momentum .................................... 0.9
486
+ short_seq_prob .................................. 0.1
487
+ skip_train ...................................... False
488
+ skipped_train_samples ........................... 0
489
+ spec ............................................ None
490
+ split ........................................... 100,0,0
491
+ squared_relu .................................... False
492
+ start_weight_decay .............................. 0.1
493
+ straggler_ctrlr_port ............................ 65535
494
+ straggler_minmax_count .......................... 1
495
+ suggested_communication_unit_size ............... None
496
+ swiglu .......................................... True
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 1
574
+ > building GPT2BPETokenizer tokenizer ...
575
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
576
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
577
+ > padded vocab (size: 50257) with 47 dummy tokens (new size: 50304)
578
+ WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.VALIDATE_RESULTS
579
+ > initializing torch distributed ...
580
+ [Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
581
+ [Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
582
+ [Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
583
+ [Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
584
+ [Gloo] Rank 5 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
585
+ [Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
586
+ [Gloo] Rank [Gloo] Rank 67 is connected to is connected to 77 peer ranks. peer ranks. Expected number of connected peer ranks is : Expected number of connected peer ranks is : 77
587
+
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 0 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 7 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
593
+ [Gloo] Rank [Gloo] Rank 5 is connected to 7 peer ranks. 6Expected number of connected peer ranks is : [Gloo] Rank is connected to 774 peer ranks. is connected to Expected number of connected peer ranks is :
594
+ 77 peer ranks. Expected number of connected peer ranks is :
595
+ 7
596
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
597
+ [Gloo] Rank 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
+ [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[Gloo] Rank
601
+ 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
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
+ > initialized tensor model parallel with size 1
605
+ > initialized pipeline model parallel with size 1
606
+ > setting random seeds to 1234 ...
607
+ [rank5]:[W830 01:04:16.307997743 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.
608
+ [rank7]:[W830 01:04:16.308011893 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
+ [rank6]:[W830 01:04:16.308065594 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.
610
+ [rank3]:[W830 01:04:16.308160662 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.
611
+ [rank1]:[W830 01:04:16.308376944 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.
612
+ [rank2]:[W830 01:04:16.310729701 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
+ [rank4]:[W830 01:04:16.310917340 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.
614
+ > compiling dataset index builder ...
615
+ make: Entering directory '/workspace/Megatron-MoE-Echo/megatron/core/datasets'
616
+ make: Nothing to be done for 'default'.
617
+ make: Leaving directory '/workspace/Megatron-MoE-Echo/megatron/core/datasets'
618
+ >>> done with dataset index builder. Compilation time: 0.424 seconds
619
+ WARNING: constraints for invoking optimized fused softmax kernel are not met. We default back to unfused kernel invocations.
620
+ > compiling and loading fused kernels ...
621
+ [rank0]:[W830 01:04:16.939116966 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.
622
+ >>> done with compiling and loading fused kernels. Compilation time: 3.726 seconds
623
+ time to initialize megatron (seconds): 9.407
624
+ [after megatron is initialized] datetime: 2026-08-30 01:04:22
625
+ building GPT model ...
626
+ /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.
627
+ warnings.warn(
628
+ /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.
629
+ warnings.warn(
630
+ /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.
631
+ warnings.warn(
632
+ /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.
633
+ warnings.warn(
634
+ /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.
635
+ warnings.warn(
636
+ /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.
637
+ warnings.warn(
638
+ /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.
639
+ warnings.warn(
640
+ /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.
641
+ warnings.warn(
642
+ > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 4898003968
643
+ 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)
644
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 14
645
+ Params for bucket 1 (206045184 elements, 206045184 padded size):
646
+ module.output_layer.weight
647
+ Params for bucket 2 (71840000 elements, 71840000 padded size):
648
+ module.decoder.layers.11.self_attention.q_layernorm.weight
649
+ module.decoder.layers.11.pre_mlp_layernorm.weight
650
+ module.decoder.layers.11.mlp.router.weight
651
+ module.decoder.layers.11.self_attention.k_layernorm.weight
652
+ module.decoder.layers.11.self_attention.linear_qkv.weight
653
+ module.decoder.layers.11.self_attention.linear_qkv.layer_norm_weight
654
+ module.decoder.layers.11.self_attention.linear_proj.weight
655
+ module.decoder.final_layernorm.weight
656
+ Params for bucket 3 (71835904 elements, 71835904 padded size):
657
+ module.decoder.layers.10.pre_mlp_layernorm.weight
658
+ module.decoder.layers.10.self_attention.q_layernorm.weight
659
+ module.decoder.layers.10.self_attention.linear_qkv.weight
660
+ module.decoder.layers.10.self_attention.linear_qkv.layer_norm_weight
661
+ module.decoder.layers.10.mlp.router.weight
662
+ module.decoder.layers.10.self_attention.k_layernorm.weight
663
+ module.decoder.layers.10.self_attention.linear_proj.weight
664
+ Params for bucket 4 (71835904 elements, 71835904 padded size):
665
+ module.decoder.layers.9.self_attention.q_layernorm.weight
666
+ module.decoder.layers.9.self_attention.linear_proj.weight
667
+ module.decoder.layers.9.mlp.router.weight
668
+ module.decoder.layers.9.self_attention.linear_qkv.weight
669
+ module.decoder.layers.9.pre_mlp_layernorm.weight
670
+ module.decoder.layers.9.self_attention.k_layernorm.weight
671
+ module.decoder.layers.9.self_attention.linear_qkv.layer_norm_weight
672
+ Params for bucket 5 (71835904 elements, 71835904 padded size):
673
+ module.decoder.layers.8.self_attention.k_layernorm.weight
674
+ module.decoder.layers.8.mlp.router.weight
675
+ module.decoder.layers.8.pre_mlp_layernorm.weight
676
+ module.decoder.layers.8.self_attention.linear_qkv.weight
677
+ module.decoder.layers.8.self_attention.q_layernorm.weight
678
+ module.decoder.layers.8.self_attention.linear_qkv.layer_norm_weight
679
+ module.decoder.layers.8.self_attention.linear_proj.weight
680
+ Params for bucket 6 (71835904 elements, 71835904 padded size):
681
+ module.decoder.layers.7.mlp.router.weight
682
+ module.decoder.layers.7.pre_mlp_layernorm.weight
683
+ module.decoder.layers.7.self_attention.k_layernorm.weight
684
+ module.decoder.layers.7.self_attention.linear_qkv.weight
685
+ module.decoder.layers.7.self_attention.q_layernorm.weight
686
+ module.decoder.layers.7.self_attention.linear_qkv.layer_norm_weight
687
+ module.decoder.layers.7.self_attention.linear_proj.weight
688
+ Params for bucket 7 (71835904 elements, 71835904 padded size):
689
+ module.decoder.layers.6.mlp.router.weight
690
+ module.decoder.layers.6.self_attention.q_layernorm.weight
691
+ module.decoder.layers.6.self_attention.linear_qkv.weight
692
+ module.decoder.layers.6.pre_mlp_layernorm.weight
693
+ module.decoder.layers.6.self_attention.k_layernorm.weight
694
+ module.decoder.layers.6.self_attention.linear_qkv.layer_norm_weight
695
+ module.decoder.layers.6.self_attention.linear_proj.weight
696
+ Params for bucket 8 (71835904 elements, 71835904 padded size):
697
+ module.decoder.layers.5.self_attention.q_layernorm.weight
698
+ module.decoder.layers.5.pre_mlp_layernorm.weight
699
+ module.decoder.layers.5.self_attention.linear_qkv.layer_norm_weight
700
+ module.decoder.layers.5.mlp.router.weight
701
+ module.decoder.layers.5.self_attention.linear_qkv.weight
702
+ module.decoder.layers.5.self_attention.k_layernorm.weight
703
+ module.decoder.layers.5.self_attention.linear_proj.weight
704
+ Params for bucket 9 (71835904 elements, 71835904 padded size):
705
+ module.decoder.layers.4.pre_mlp_layernorm.weight
706
+ module.decoder.layers.4.self_attention.q_layernorm.weight
707
+ module.decoder.layers.4.self_attention.linear_qkv.weight
708
+ module.decoder.layers.4.self_attention.linear_qkv.layer_norm_weight
709
+ module.decoder.layers.4.mlp.router.weight
710
+ module.decoder.layers.4.self_attention.k_layernorm.weight
711
+ module.decoder.layers.4.self_attention.linear_proj.weight
712
+ Params for bucket 10 (71835904 elements, 71835904 padded size):
713
+ module.decoder.layers.3.mlp.router.weight
714
+ module.decoder.layers.3.pre_mlp_layernorm.weight
715
+ module.decoder.layers.3.self_attention.k_layernorm.weight
716
+ module.decoder.layers.3.self_attention.q_layernorm.weight
717
+ module.decoder.layers.3.self_attention.linear_qkv.weight
718
+ module.decoder.layers.3.self_attention.linear_qkv.layer_norm_weight
719
+ module.decoder.layers.3.self_attention.linear_proj.weight
720
+ Params for bucket 11 (71835904 elements, 71835904 padded size):
721
+ module.decoder.layers.2.mlp.router.weight
722
+ module.decoder.layers.2.self_attention.linear_qkv.weight
723
+ module.decoder.layers.2.pre_mlp_layernorm.weight
724
+ module.decoder.layers.2.self_attention.k_layernorm.weight
725
+ module.decoder.layers.2.self_attention.q_layernorm.weight
726
+ module.decoder.layers.2.self_attention.linear_qkv.layer_norm_weight
727
+ module.decoder.layers.2.self_attention.linear_proj.weight
728
+ Params for bucket 12 (71835904 elements, 71835904 padded size):
729
+ module.decoder.layers.1.mlp.router.weight
730
+ module.decoder.layers.1.self_attention.linear_qkv.weight
731
+ module.decoder.layers.1.self_attention.k_layernorm.weight
732
+ module.decoder.layers.1.pre_mlp_layernorm.weight
733
+ module.decoder.layers.1.self_attention.linear_qkv.layer_norm_weight
734
+ module.decoder.layers.1.self_attention.linear_proj.weight
735
+ module.decoder.layers.1.self_attention.q_layernorm.weight
736
+ Params for bucket 13 (71835904 elements, 71835904 padded size):
737
+ module.decoder.layers.0.mlp.router.weight
738
+ module.decoder.layers.0.pre_mlp_layernorm.weight
739
+ module.decoder.layers.0.self_attention.linear_qkv.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.q_layernorm.weight
743
+ module.decoder.layers.0.self_attention.k_layernorm.weight
744
+ Params for bucket 14 (206045184 elements, 206045184 padded size):
745
+ module.embedding.word_embeddings.weight
746
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 76
747
+ Params for bucket 1 (44040192 elements, 44040192 padded size):
748
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight0
749
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight2
750
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight4
751
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight6
752
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight3
753
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight1
754
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight5
755
+ Params for bucket 2 (44040192 elements, 44040192 padded size):
756
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight10
757
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight8
758
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight12
759
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight7
760
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight9
761
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight11
762
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight13
763
+ Params for bucket 3 (50331648 elements, 50331648 padded size):
764
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight14
765
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight1
766
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight15
767
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight0
768
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight2
769
+ Params for bucket 4 (50331648 elements, 50331648 padded size):
770
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight3
771
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight5
772
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight6
773
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight4
774
+ Params for bucket 5 (50331648 elements, 50331648 padded size):
775
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight7
776
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight9
777
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight8
778
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight10
779
+ Params for bucket 6 (50331648 elements, 50331648 padded size):
780
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight11
781
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight12
782
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight13
783
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight14
784
+ Params for bucket 7 (44040192 elements, 44040192 padded size):
785
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight15
786
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight0
787
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight2
788
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight4
789
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight1
790
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight3
791
+ Params for bucket 8 (44040192 elements, 44040192 padded size):
792
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight8
793
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight6
794
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight10
795
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight5
796
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight7
797
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight9
798
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight11
799
+ Params for bucket 9 (50331648 elements, 50331648 padded size):
800
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight12
801
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight0
802
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight14
803
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight13
804
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight15
805
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight1
806
+ Params for bucket 10 (50331648 elements, 50331648 padded size):
807
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight2
808
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight3
809
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight4
810
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight5
811
+ Params for bucket 11 (50331648 elements, 50331648 padded size):
812
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight6
813
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight9
814
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight7
815
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight8
816
+ Params for bucket 12 (50331648 elements, 50331648 padded size):
817
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight10
818
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight12
819
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight13
820
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight11
821
+ Params for bucket 13 (44040192 elements, 44040192 padded size):
822
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight14
823
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight2
824
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight0
825
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight1
826
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight15
827
+ Params for bucket 14 (44040192 elements, 44040192 padded size):
828
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight6
829
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight4
830
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight8
831
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight3
832
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight5
833
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight7
834
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight9
835
+ Params for bucket 15 (50331648 elements, 50331648 padded size):
836
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight10
837
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight12
838
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight14
839
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight13
840
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight11
841
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight15
842
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight0
843
+ Params for bucket 16 (50331648 elements, 50331648 padded size):
844
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight1
845
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight2
846
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight3
847
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight4
848
+ Params for bucket 17 (50331648 elements, 50331648 padded size):
849
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight5
850
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight8
851
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight6
852
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight7
853
+ Params for bucket 18 (50331648 elements, 50331648 padded size):
854
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight9
855
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight11
856
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight12
857
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight10
858
+ Params for bucket 19 (44040192 elements, 44040192 padded size):
859
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight13
860
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight15
861
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight0
862
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight14
863
+ Params for bucket 20 (44040192 elements, 44040192 padded size):
864
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight1
865
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight5
866
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight3
867
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight7
868
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight4
869
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight2
870
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight6
871
+ Params for bucket 21 (44040192 elements, 44040192 padded size):
872
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight11
873
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight9
874
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight13
875
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight8
876
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight10
877
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight12
878
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight14
879
+ Params for bucket 22 (44040192 elements, 44040192 padded size):
880
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight15
881
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight2
882
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight1
883
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight0
884
+ Params for bucket 23 (50331648 elements, 50331648 padded size):
885
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight3
886
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight5
887
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight4
888
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight6
889
+ Params for bucket 24 (50331648 elements, 50331648 padded size):
890
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight7
891
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight8
892
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight9
893
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight10
894
+ Params for bucket 25 (50331648 elements, 50331648 padded size):
895
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight11
896
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight14
897
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight12
898
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight13
899
+ Params for bucket 26 (44040192 elements, 44040192 padded size):
900
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight3
901
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight1
902
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight15
903
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight0
904
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight2
905
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight4
906
+ Params for bucket 27 (44040192 elements, 44040192 padded size):
907
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight5
908
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight9
909
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight7
910
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight11
911
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight6
912
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight8
913
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight10
914
+ Params for bucket 28 (50331648 elements, 50331648 padded size):
915
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight15
916
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight13
917
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight1
918
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight12
919
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight0
920
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight14
921
+ Params for bucket 29 (50331648 elements, 50331648 padded size):
922
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight2
923
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight5
924
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight3
925
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight4
926
+ Params for bucket 30 (50331648 elements, 50331648 padded size):
927
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight6
928
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight9
929
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight8
930
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight7
931
+ Params for bucket 31 (50331648 elements, 50331648 padded size):
932
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight10
933
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight12
934
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight11
935
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight13
936
+ Params for bucket 32 (44040192 elements, 44040192 padded size):
937
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight15
938
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight1
939
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight14
940
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight2
941
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight0
942
+ Params for bucket 33 (44040192 elements, 44040192 padded size):
943
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight3
944
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight5
945
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight7
946
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight9
947
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight4
948
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight6
949
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight8
950
+ Params for bucket 34 (50331648 elements, 50331648 padded size):
951
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight11
952
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight13
953
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight15
954
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight10
955
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight14
956
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight12
957
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight0
958
+ Params for bucket 35 (50331648 elements, 50331648 padded size):
959
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight1
960
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight3
961
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight4
962
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight2
963
+ Params for bucket 36 (50331648 elements, 50331648 padded size):
964
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight5
965
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight6
966
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight7
967
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight8
968
+ Params for bucket 37 (50331648 elements, 50331648 padded size):
969
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight9
970
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight10
971
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight12
972
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight11
973
+ Params for bucket 38 (44040192 elements, 44040192 padded size):
974
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight13
975
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight15
976
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight14
977
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight0
978
+ Params for bucket 39 (44040192 elements, 44040192 padded size):
979
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight2
980
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight4
981
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight6
982
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight1
983
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight5
984
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight3
985
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight7
986
+ Params for bucket 40 (44040192 elements, 44040192 padded size):
987
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight8
988
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight12
989
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight10
990
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight14
991
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight11
992
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight9
993
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight13
994
+ Params for bucket 41 (44040192 elements, 44040192 padded size):
995
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight15
996
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight1
997
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight2
998
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight0
999
+ Params for bucket 42 (50331648 elements, 50331648 padded size):
1000
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight3
1001
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight5
1002
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight6
1003
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight4
1004
+ Params for bucket 43 (50331648 elements, 50331648 padded size):
1005
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight7
1006
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight8
1007
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight9
1008
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight10
1009
+ Params for bucket 44 (50331648 elements, 50331648 padded size):
1010
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight11
1011
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight14
1012
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight12
1013
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight13
1014
+ Params for bucket 45 (44040192 elements, 44040192 padded size):
1015
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight15
1016
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight1
1017
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight3
1018
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight0
1019
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight2
1020
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight4
1021
+ Params for bucket 46 (44040192 elements, 44040192 padded size):
1022
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight5
1023
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight9
1024
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight7
1025
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight11
1026
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight6
1027
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight8
1028
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight10
1029
+ Params for bucket 47 (50331648 elements, 50331648 padded size):
1030
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight13
1031
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight0
1032
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight15
1033
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight12
1034
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight14
1035
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight1
1036
+ Params for bucket 48 (50331648 elements, 50331648 padded size):
1037
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight2
1038
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight3
1039
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight4
1040
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight5
1041
+ Params for bucket 49 (50331648 elements, 50331648 padded size):
1042
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight6
1043
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight9
1044
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight7
1045
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight8
1046
+ Params for bucket 50 (50331648 elements, 50331648 padded size):
1047
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight10
1048
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight12
1049
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight13
1050
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight11
1051
+ Params for bucket 51 (44040192 elements, 44040192 padded size):
1052
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight14
1053
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight2
1054
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight0
1055
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight15
1056
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight1
1057
+ Params for bucket 52 (44040192 elements, 44040192 padded size):
1058
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight4
1059
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight6
1060
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight8
1061
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight3
1062
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight5
1063
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight7
1064
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight9
1065
+ Params for bucket 53 (50331648 elements, 50331648 padded size):
1066
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight10
1067
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight12
1068
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight14
1069
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight11
1070
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight13
1071
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight15
1072
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight0
1073
+ Params for bucket 54 (50331648 elements, 50331648 padded size):
1074
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight1
1075
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight2
1076
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight3
1077
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight4
1078
+ Params for bucket 55 (50331648 elements, 50331648 padded size):
1079
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight5
1080
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight6
1081
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight8
1082
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight7
1083
+ Params for bucket 56 (50331648 elements, 50331648 padded size):
1084
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight9
1085
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight12
1086
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight11
1087
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight10
1088
+ Params for bucket 57 (44040192 elements, 44040192 padded size):
1089
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight13
1090
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight15
1091
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight0
1092
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight14
1093
+ Params for bucket 58 (44040192 elements, 44040192 padded size):
1094
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight4
1095
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight2
1096
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight6
1097
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight1
1098
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight3
1099
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight5
1100
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight7
1101
+ Params for bucket 59 (44040192 elements, 44040192 padded size):
1102
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight8
1103
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight10
1104
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight12
1105
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight14
1106
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight9
1107
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight11
1108
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight13
1109
+ Params for bucket 60 (44040192 elements, 44040192 padded size):
1110
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight15
1111
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight1
1112
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight2
1113
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight0
1114
+ Params for bucket 61 (50331648 elements, 50331648 padded size):
1115
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight3
1116
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight4
1117
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight5
1118
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight6
1119
+ Params for bucket 62 (50331648 elements, 50331648 padded size):
1120
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight7
1121
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight10
1122
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight8
1123
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight9
1124
+ Params for bucket 63 (50331648 elements, 50331648 padded size):
1125
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight11
1126
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight13
1127
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight14
1128
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight12
1129
+ Params for bucket 64 (44040192 elements, 44040192 padded size):
1130
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight15
1131
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight2
1132
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight0
1133
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight4
1134
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight3
1135
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight1
1136
+ Params for bucket 65 (44040192 elements, 44040192 padded size):
1137
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight8
1138
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight6
1139
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight10
1140
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight5
1141
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight9
1142
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight7
1143
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight11
1144
+ Params for bucket 66 (50331648 elements, 50331648 padded size):
1145
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight12
1146
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight14
1147
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight1
1148
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight15
1149
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight0
1150
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight13
1151
+ Params for bucket 67 (50331648 elements, 50331648 padded size):
1152
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight2
1153
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight4
1154
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight5
1155
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight3
1156
+ Params for bucket 68 (50331648 elements, 50331648 padded size):
1157
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight6
1158
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight7
1159
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight8
1160
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight9
1161
+ Params for bucket 69 (50331648 elements, 50331648 padded size):
1162
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight10
1163
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight11
1164
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight12
1165
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight13
1166
+ Params for bucket 70 (44040192 elements, 44040192 padded size):
1167
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight15
1168
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight2
1169
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight0
1170
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight14
1171
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight1
1172
+ Params for bucket 71 (44040192 elements, 44040192 padded size):
1173
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight4
1174
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight6
1175
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight8
1176
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight3
1177
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight5
1178
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight7
1179
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight9
1180
+ Params for bucket 72 (50331648 elements, 50331648 padded size):
1181
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight10
1182
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight14
1183
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight12
1184
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight11
1185
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight13
1186
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight15
1187
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight0
1188
+ Params for bucket 73 (50331648 elements, 50331648 padded size):
1189
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight1
1190
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight3
1191
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight2
1192
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight4
1193
+ Params for bucket 74 (50331648 elements, 50331648 padded size):
1194
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight5
1195
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight6
1196
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight7
1197
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight8
1198
+ Params for bucket 75 (50331648 elements, 50331648 padded size):
1199
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight9
1200
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight12
1201
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight10
1202
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight11
1203
+ Params for bucket 76 (37748736 elements, 37748736 padded size):
1204
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight13
1205
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight15
1206
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight14
1207
+ 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 0x79bb88c11490>, config_logger_dir='')
1208
+ INFO:megatron.core.optimizer_param_scheduler:> learning rate decay style: cosine
1209
+ [after model, optimizer, and learning rate scheduler are built] datetime: 2026-08-30 01:04:26
1210
+ > building train, validation, and test datasets ...
1211
+ > datasets target sizes (minimum size):
1212
+ train: 80
1213
+ validation: 0
1214
+ test: 0
1215
+ INFO:megatron.core.datasets.blended_megatron_dataset_config:Let split_matrix = [(0, 1.0), None, None]
1216
+ > building train, validation, and test datasets for GPT ...
1217
+ INFO:megatron.core.datasets.blended_megatron_dataset_builder:Building GPTDataset splits with sizes=(80, 0, 0) and config=GPTDatasetConfig(random_seed=1234, sequence_length=8192, 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 0x79bb5ecc7830>, 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)
1218
+ INFO:megatron.core.datasets.indexed_dataset:Load the _IndexReader from /workspace/data/wikitext/wikitext103_gpt2_text_document.idx
1219
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence lengths
1220
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence pointers
1221
+ INFO:megatron.core.datasets.indexed_dataset: Extract the document indices
1222
+ INFO:megatron.core.datasets.indexed_dataset:> total number of sequences: 1165029
1223
+ INFO:megatron.core.datasets.indexed_dataset:> total number of documents: 1801350
1224
+ INFO:megatron.core.datasets.gpt_dataset:Load the GPTDataset train indices
1225
+ INFO:megatron.core.datasets.gpt_dataset: Load the document index from 715d60356d3f5be048391704c2aaf5b2-GPTDataset-train-document_index.npy
1226
+ INFO:megatron.core.datasets.gpt_dataset: Load the sample index from 715d60356d3f5be048391704c2aaf5b2-GPTDataset-train-sample_index.npy
1227
+ INFO:megatron.core.datasets.gpt_dataset: Load the shuffle index from 715d60356d3f5be048391704c2aaf5b2-GPTDataset-train-shuffle_index.npy
1228
+ INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 14536
1229
+ > finished creating GPT datasets ...
1230
+ [after dataloaders are built] datetime: 2026-08-30 01:04:26
1231
+ done with setup ...
1232
+ training ...
1233
+ Overwriting rerun_state_machine.current_iteration from -1 to 0...
1234
+ (min, max) time across ranks (ms):
1235
+ model-and-optimizer-setup ......................: (3495.92, 3530.17)
1236
+ train/valid/test-data-iterators-setup ..........: (88.12, 122.20)
1237
+ [before the start of training step] datetime: 2026-08-30 01:04:26
1238
+ INFO:moonep.api:MoonEP comm buffer sizes:
1239
+ config : R=8, S=8192, H=4096, E=128, K=8, NvS=69600, NvS_padded=69632
1240
+ sm_count : num_sms=32, num_sms_dedup=148
1241
+ hidden_buffer_size : 544.00 MiB
1242
+ meta_buffer_size : 2.00 MiB
1243
+ local_temp_buffer_size : 2.52 MiB
1244
+ total_size : 548.52 MiB
1245
+ [MoonEP selfcheck 1/4] dropped groups empty OK | covered rows [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143] sizes [5376, 640, 256, 2816, 1280, 5632, 256, 256, 768, 1920, 14208, 4992, 14848, 896, 1664, 3584, 7296, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0] | duplicated experts [[127, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [43, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [127, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [64, 74, 68, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [0, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [107, 99, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [29, 17, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1]]
1246
+ [MoonEP selfcheck 2/4] dropped groups empty OK | covered rows [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143] sizes [9344, 2688, 128, 24192, 128, 0, 128, 256, 128, 768, 10496, 128, 128, 2432, 128, 1536, 14208, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0] | duplicated experts [[69, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [88, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [55, 52, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [63, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [124, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [10, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [58, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1]]
1247
+ [MoonEP selfcheck 3/4] dropped groups empty OK | covered rows [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143] sizes [128, 128, 128, 0, 128, 44416, 128, 128, 128, 6912, 128, 256, 128, 128, 128, 128, 13824, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0] | duplicated experts [[78, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [111, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [62, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [57, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [106, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [121, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [57, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1]]
1248
+ [MoonEP selfcheck 4/4] dropped groups empty OK | covered rows [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143] sizes [2304, 128, 128, 512, 128, 12416, 128, 1152, 1280, 16768, 640, 29056, 128, 128, 128, 1664, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0] | duplicated experts [[-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [118, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [11, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [118, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [39, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [11, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [73, 79, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1], [86, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1]]
1249
+ [2026-08-30 01:04:47] iteration 1/ 10 | consumed samples: 8 | elapsed time per iteration (ms): 21074.9 | throughput per GPU (TFLOP/s/GPU): 8.6 | learning rate: 2.000000E-05 | global batch size: 8 | lm loss: 1.115993E+01 | loss scale: 1.0 | grad norm: 44.807 | number of skipped iterations: 0 | number of nan iterations: 0 |
1250
+ Number of parameters in transformer block in billions: 29.85
1251
+ Number of parameters in embedding layers in billions: 0.41
1252
+ Total number of parameters in billions: 30.26
1253
+ Number of parameters in most loaded shard in billions: 30.2590
1254
+ Theoretical memory footprints: weight and optimizer=216428.96 MB
1255
+ [Rank 0] (after 1 iterations) memory (MB) | allocated: 67239.55517578125 | max allocated: 67239.5556640625 | reserved: 68548.0 | max reserved: 68548.0
1256
+ [2026-08-30 01:04:48] iteration 2/ 10 | consumed samples: 16 | elapsed time per iteration (ms): 374.7 | throughput per GPU (TFLOP/s/GPU): 482.5 | learning rate: 4.000000E-05 | global batch size: 8 | lm loss: 1.117600E+01 | loss scale: 1.0 | grad norm: 44.336 | number of skipped iterations: 0 | number of nan iterations: 0 |
1257
+ [2026-08-30 01:04:48] iteration 3/ 10 | consumed samples: 24 | elapsed time per iteration (ms): 333.6 | throughput per GPU (TFLOP/s/GPU): 542.1 | learning rate: 6.000000E-05 | global batch size: 8 | lm loss: 9.686143E+00 | loss scale: 1.0 | grad norm: 62.903 | number of skipped iterations: 0 | number of nan iterations: 0 |
1258
+ Capture CUDA graph for training!!!
1259
+ CUDA graph capture done!!!
1260
+ [2026-08-30 01:04:49] iteration 4/ 10 | consumed samples: 32 | elapsed time per iteration (ms): 992.1 | throughput per GPU (TFLOP/s/GPU): 182.3 | learning rate: 8.000000E-05 | global batch size: 8 | lm loss: 1.338079E+01 | loss scale: 1.0 | grad norm: 20.299 | number of skipped iterations: 0 | number of nan iterations: 0 |
1261
+ [2026-08-30 01:04:49] iteration 5/ 10 | consumed samples: 40 | elapsed time per iteration (ms): 324.1 | throughput per GPU (TFLOP/s/GPU): 557.9 | learning rate: 1.000000E-04 | global batch size: 8 | lm loss: 1.049441E+01 | loss scale: 1.0 | grad norm: 95.083 | number of skipped iterations: 0 | number of nan iterations: 0 |
1262
+ [2026-08-30 01:04:50] iteration 6/ 10 | consumed samples: 48 | elapsed time per iteration (ms): 325.5 | throughput per GPU (TFLOP/s/GPU): 555.6 | learning rate: 7.750000E-05 | global batch size: 8 | lm loss: 1.045251E+01 | loss scale: 1.0 | grad norm: 52.433 | number of skipped iterations: 0 | number of nan iterations: 0 |
1263
+ [2026-08-30 01:04:53] iteration 7/ 10 | consumed samples: 56 | elapsed time per iteration (ms): 3278.5 | throughput per GPU (TFLOP/s/GPU): 55.2 | learning rate: 3.250000E-05 | global batch size: 8 | lm loss: 1.027084E+01 | loss scale: 1.0 | grad norm: 6.326 | number of skipped iterations: 0 | number of nan iterations: 0 |
1264
+ [2026-08-30 01:04:53] iteration 8/ 10 | consumed samples: 64 | elapsed time per iteration (ms): 335.6 | throughput per GPU (TFLOP/s/GPU): 538.9 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 9.808949E+00 | loss scale: 1.0 | grad norm: 5.540 | number of skipped iterations: 0 | number of nan iterations: 0 |
1265
+ Capture range started in the application.
1266
+ Capture range ended in the application.
1267
+ Generating '/tmp/nsys-report-a863.qdstrm'
1268
+
1269
+
1270
+ [2026-08-30 01:05:35] iteration 9/ 10 | consumed samples: 72 | elapsed time per iteration (ms): 41630.6 | throughput per GPU (TFLOP/s/GPU): 4.3 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 9.545780E+00 | loss scale: 1.0 | grad norm: 5.279 | number of skipped iterations: 0 | number of nan iterations: 0 |
1271
+ [2026-08-30 01:05:35] iteration 10/ 10 | consumed samples: 80 | elapsed time per iteration (ms): 331.7 | throughput per GPU (TFLOP/s/GPU): 545.1 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 9.447124E+00 | loss scale: 1.0 | grad norm: 5.006 | number of skipped iterations: 0 | number of nan iterations: 0 |
1272
+ [after training is done] datetime: 2026-08-30 01:05:35
1273
+ [rank0]:[W830 01:05:37.366784257 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())
1274
+ [rank1]:[W830 01:05:38.395523828 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())
1275
+ [rank3]:[W830 01:05:38.598871266 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())
1276
+ [rank7]:[W830 01:05:38.614812968 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())
1277
+ [rank6]:[W830 01:05:38.621632404 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())
1278
+ [rank4]:[W830 01:05:38.717526439 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())
1279
+ [rank5]:[W830 01:05:38.887874062 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())
1280
+ [rank2]:[W830 01:05:38.949321875 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())
1281
+ [rank1]:[W830 01:05:41.370937686 observer.cpp:116] Warning: Leaked callback handle: 1 (function operator())
1282
+ pure virtual method called
1283
+ terminate called without an active exception
1284
+ Collection stop is not allowed in this state.
1285
+ W0830 01:05:41.775000 363852 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 364073 closing signal SIGTERM
1286
+ W0830 01:05:41.776000 363852 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 364075 closing signal SIGTERM
1287
+ W0830 01:05:41.776000 363852 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 364076 closing signal SIGTERM
1288
+ W0830 01:05:41.778000 363852 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 364077 closing signal SIGTERM
1289
+ W0830 01:05:41.779000 363852 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 364078 closing signal SIGTERM
1290
+ W0830 01:05:41.779000 363852 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 364079 closing signal SIGTERM
1291
+ W0830 01:05:41.780000 363852 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 364080 closing signal SIGTERM
1292
+ E0830 01:05:44.562000 363852 torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: -6) local_rank: 1 (pid: 364074) of binary: /usr/bin/python
1293
+ Traceback (most recent call last):
1294
+ File "/usr/local/bin/torchrun", line 8, in <module>
1295
+ sys.exit(main())
1296
+ ^^^^^^
1297
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper
1298
+ return f(*args, **kwargs)
1299
+ ^^^^^^^^^^^^^^^^^^
1300
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 892, in main
1301
+ run(args)
1302
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 883, in run
1303
+ elastic_launch(
1304
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 139, in __call__
1305
+ return launch_agent(self._config, self._entrypoint, list(args))
1306
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1307
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 270, in launch_agent
1308
+ raise ChildFailedError(
1309
+ torch.distributed.elastic.multiprocessing.errors.ChildFailedError:
1310
+ ============================================================
1311
+ pretrain_gpt.py FAILED
1312
+ ------------------------------------------------------------
1313
+ Failures:
1314
+ <NO_OTHER_FAILURES>
1315
+ ------------------------------------------------------------
1316
+ Root Cause (first observed failure):
1317
+ [0]:
1318
+ time : 2026-08-30_01:05:41
1319
+ host : sp-aiinfra-interactive-aghosh-7jtjl-8xb200-moe-cg-0
1320
+ rank : 1 (local_rank: 1)
1321
+ exitcode : -6 (pid: 364074)
1322
+ error_file: <N/A>
1323
+ traceback : Signal 6 (SIGABRT) received by PID 364074
1324
+ ============================================================
1325
+ Generated:
1326
+ /workspace/tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/nsys/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async.nsys-rep
tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/nsys/verify_trace_ranks.log ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ report : /workspace/tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/nsys/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async.nsys-rep
2
+ kernels : 76400
3
+ devices : 8 of 8 -> [0, 1, 2, 3, 4, 5, 6, 7]
4
+ processes: 1
5
+ markers : 6528 moonep_trace__ kernels on devices [0, 1, 2, 3, 4, 5, 6, 7]
6
+ PASS: all ranks present and backend markers emitted
tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/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
tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/trace_extension_build.log ADDED
@@ -0,0 +1,58 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ Emitting ninja build file /workspace/Megatron-MoE-Echo/build/temp.linux-x86_64-cpython-312/build.ninja...
45
+ Compiling objects...
46
+ Allowing ninja to set a default number of workers... (overridable by setting the environment variable MAX_JOBS=N)
47
+ ninja: no work to do.
48
+ 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
49
+ building 'megatron.core.transformer.moe._moonep_trace_C' extension
50
+ Emitting ninja build file /workspace/Megatron-MoE-Echo/build/temp.linux-x86_64-cpython-312/build.ninja...
51
+ Compiling objects...
52
+ Allowing ninja to set a default number of workers... (overridable by setting the environment variable MAX_JOBS=N)
53
+ [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
54
+ [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
55
+ 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
56
+ copying build/lib.linux-x86_64-cpython-312/megatron/core/datasets/helpers_cpp.cpython-312-x86_64-linux-gnu.so -> megatron/core/datasets
57
+ 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
58
+ 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
tierb-fixed-moonep-full/moonep/tierbtrace/tierbtrace_gbs8/trace_results.tsv ADDED
@@ -0,0 +1 @@
 
 
1
+ model backend capacity status report async_overlap_json