AbhishekGhosh1998 commited on
Commit
6f02126
·
verified ·
1 Parent(s): 950d778

granularity results tierb-fixed-moonep-full/moonep/tierb @ 2026-08-29T23:46:57Z from sp-aiinfra-interactive-aghosh-7jtjl-8xb200-moe-cg-0

Browse files
tierb-fixed-moonep-full/moonep/tierb/manifest.txt ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ timestamp: 2026-08-29T23:44:54+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: tierb
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/tierb/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/tierb/provenance.txt CHANGED
@@ -27,3 +27,66 @@ command : run_ladder.sh --phase tierb --gbs 8 --out /workspace/tierb-fixed
27
  7, 120 MHz, 1965 MHz, 1000.00 W, 29
28
 
29
  NVLink topology:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
27
  7, 120 MHz, 1965 MHz, 1000.00 W, 29
28
 
29
  NVLink topology:
30
+ GPU0 GPU1 GPU2 GPU3 GPU4 GPU5 GPU6 GPU7 NIC0 NIC1 NIC2 NIC3 NIC4 NIC5 NIC6 NIC7 NIC8 CPU Affinity NUMA Affinity GPU NUMA ID
31
+ GPU0 X NV18 NV18 NV18 NV18 NV18 NV18 NV18 PIX NODE SYS SYS SYS SYS SYS SYS NODE 0-27 0 N/A
32
+ GPU1 NV18 X NV18 NV18 NV18 NV18 NV18 NV18 NODE PIX SYS SYS SYS SYS SYS SYS NODE 0-27 0 N/A
33
+ GPU2 NV18 NV18 X NV18 NV18 NV18 NV18 NV18 SYS SYS PIX NODE SYS SYS SYS SYS SYS 28-55 1 N/A
34
+ GPU3 NV18 NV18 NV18 X NV18 NV18 NV18 NV18 SYS SYS NODE PIX SYS SYS SYS SYS SYS 28-55 1 N/A
35
+ GPU4 NV18 NV18 NV18 NV18 X NV18 NV18 NV18 SYS SYS SYS SYS PIX NODE SYS SYS SYS 56-83 2 N/A
36
+ GPU5 NV18 NV18 NV18 NV18 NV18 X NV18 NV18 SYS SYS SYS SYS NODE PIX SYS SYS SYS 56-83 2 N/A
37
+ GPU6 NV18 NV18 NV18 NV18 NV18 NV18 X NV18 SYS SYS SYS SYS SYS SYS PIX NODE SYS 84-111 3 N/A
38
+ GPU7 NV18 NV18 NV18 NV18 NV18 NV18 NV18 X SYS SYS SYS SYS SYS SYS NODE PIX SYS 84-111 3 N/A
39
+ NIC0 PIX NODE SYS SYS SYS SYS SYS SYS X NODE SYS SYS SYS SYS SYS SYS NODE
40
+ NIC1 NODE PIX SYS SYS SYS SYS SYS SYS NODE X SYS SYS SYS SYS SYS SYS NODE
41
+ NIC2 SYS SYS PIX NODE SYS SYS SYS SYS SYS SYS X NODE SYS SYS SYS SYS SYS
42
+ NIC3 SYS SYS NODE PIX SYS SYS SYS SYS SYS SYS NODE X SYS SYS SYS SYS SYS
43
+ NIC4 SYS SYS SYS SYS PIX NODE SYS SYS SYS SYS SYS SYS X NODE SYS SYS SYS
44
+ unavailable
45
+
46
+ ---- toolchain -----------------------------------------------
47
+ nvcc : release 12.9, V12.9.86
48
+ gcc : gcc (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0
49
+ python : 3.12.3
50
+ torch : 2.8.0a0+5228986c39.nv25.06 cuda=12.9 nccl=2.27.3
51
+ torch arch : ['sm_75', 'sm_80', 'sm_86', 'sm_90', 'sm_100', 'sm_120', 'compute_120']
52
+ transformer_engine: 2.10.0.dev0+10f6128d [/usr/local/lib/python3.12/dist-packages/transformer_engine/__init__.py]
53
+ deep_ep : 1.2.1+34152ae [/workspace/DeepEP/deep_ep/__init__.py]
54
+ ultra_ep : 1.0.0 [/workspace/UltraEP/ultra_ep/__init__.py]
55
+ moonep : 0.0.1 [/workspace/MoonEP/moonep/__init__.py]
56
+ megatron : 0.14.0rc5 [None]
57
+ flash_attn : 2.7.4.post1 [/usr/local/lib/python3.12/dist-packages/flash_attn/__init__.py]
58
+ numpy : 1.26.4 [/usr/local/lib/python3.12/dist-packages/numpy/__init__.py]
59
+ huggingface_hub: 1.29.0 [/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-token-drop-analysis eaf829e7ad21
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/tierb/tierb_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/rep1/mem_samples.csv ADDED
@@ -0,0 +1,224 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ 0, 0
2
+ 1, 0
3
+ 2, 0
4
+ 3, 0
5
+ 4, 0
6
+ 5, 0
7
+ 6, 0
8
+ 7, 0
9
+ 0, 4
10
+ 1, 4
11
+ 2, 1
12
+ 3, 1
13
+ 4, 1
14
+ 5, 1
15
+ 6, 1
16
+ 7, 1
17
+ 0, 4
18
+ 1, 4
19
+ 2, 4
20
+ 3, 4
21
+ 4, 4
22
+ 5, 4
23
+ 6, 4
24
+ 7, 1
25
+ 0, 1
26
+ 1, 1
27
+ 2, 1
28
+ 3, 1
29
+ 4, 3
30
+ 5, 4
31
+ 6, 4
32
+ 7, 4
33
+ 0, 0
34
+ 1, 0
35
+ 2, 0
36
+ 3, 0
37
+ 4, 0
38
+ 5, 0
39
+ 6, 0
40
+ 7, 0
41
+ 0, 1
42
+ 1, 1
43
+ 2, 1
44
+ 3, 1
45
+ 4, 1
46
+ 5, 1
47
+ 6, 1
48
+ 7, 1
49
+ 0, 4
50
+ 1, 4
51
+ 2, 4
52
+ 3, 4
53
+ 4, 1
54
+ 5, 1
55
+ 6, 1
56
+ 7, 1
57
+ 0, 4
58
+ 1, 4
59
+ 2, 4
60
+ 3, 4
61
+ 4, 4
62
+ 5, 4
63
+ 6, 4
64
+ 7, 4
65
+ 0, 4
66
+ 1, 4
67
+ 2, 4
68
+ 3, 4
69
+ 4, 4
70
+ 5, 4
71
+ 6, 4
72
+ 7, 4
73
+ 0, 821
74
+ 1, 821
75
+ 2, 821
76
+ 3, 819
77
+ 4, 821
78
+ 5, 821
79
+ 6, 821
80
+ 7, 821
81
+ 0, 1537
82
+ 1, 1537
83
+ 2, 1537
84
+ 3, 1537
85
+ 4, 1537
86
+ 5, 1537
87
+ 6, 1537
88
+ 7, 1537
89
+ 0, 2551
90
+ 1, 2551
91
+ 2, 2551
92
+ 3, 2551
93
+ 4, 2551
94
+ 5, 2551
95
+ 6, 2551
96
+ 7, 2551
97
+ 0, 46603
98
+ 1, 46619
99
+ 2, 46599
100
+ 3, 46619
101
+ 4, 46635
102
+ 5, 46639
103
+ 6, 46631
104
+ 7, 46637
105
+ 0, 56813
106
+ 1, 56831
107
+ 2, 56771
108
+ 3, 56791
109
+ 4, 56519
110
+ 5, 56847
111
+ 6, 56779
112
+ 7, 56849
113
+ 0, 59169
114
+ 1, 59167
115
+ 2, 59167
116
+ 3, 59161
117
+ 4, 58875
118
+ 5, 59167
119
+ 6, 59159
120
+ 7, 59181
121
+ 0, 59733
122
+ 1, 59731
123
+ 2, 59731
124
+ 3, 59733
125
+ 4, 59443
126
+ 5, 59731
127
+ 6, 59733
128
+ 7, 59731
129
+ 0, 59733
130
+ 1, 59731
131
+ 2, 59733
132
+ 3, 59733
133
+ 4, 59445
134
+ 5, 59733
135
+ 6, 59733
136
+ 7, 59733
137
+ 0, 86921
138
+ 1, 86961
139
+ 2, 86961
140
+ 3, 86961
141
+ 4, 86921
142
+ 5, 86961
143
+ 6, 86961
144
+ 7, 86961
145
+ 0, 87633
146
+ 1, 87673
147
+ 2, 87673
148
+ 3, 87673
149
+ 4, 87633
150
+ 5, 87673
151
+ 6, 87673
152
+ 7, 87673
153
+ 0, 87833
154
+ 1, 87869
155
+ 2, 87873
156
+ 3, 87873
157
+ 4, 87869
158
+ 5, 87939
159
+ 6, 87899
160
+ 7, 87937
161
+ 0, 94181
162
+ 1, 95013
163
+ 2, 95133
164
+ 3, 95229
165
+ 4, 94733
166
+ 5, 95325
167
+ 6, 94797
168
+ 7, 94821
169
+ 0, 97481
170
+ 1, 97527
171
+ 2, 97499
172
+ 3, 97465
173
+ 4, 97433
174
+ 5, 97525
175
+ 6, 97527
176
+ 7, 97523
177
+ 0, 120267
178
+ 1, 120303
179
+ 2, 120303
180
+ 3, 120303
181
+ 4, 120303
182
+ 5, 120303
183
+ 6, 120303
184
+ 7, 120303
185
+ 0, 120267
186
+ 1, 120303
187
+ 2, 120303
188
+ 3, 120303
189
+ 4, 120303
190
+ 5, 120303
191
+ 6, 120303
192
+ 7, 120303
193
+ 0, 119621
194
+ 1, 119657
195
+ 2, 119657
196
+ 3, 119657
197
+ 4, 119657
198
+ 5, 119657
199
+ 6, 119657
200
+ 7, 119657
201
+ 0, 117395
202
+ 1, 117431
203
+ 2, 117431
204
+ 3, 117431
205
+ 4, 117431
206
+ 5, 117431
207
+ 6, 117431
208
+ 7, 117431
209
+ 0, 92863
210
+ 1, 92899
211
+ 2, 92897
212
+ 3, 115751
213
+ 4, 92899
214
+ 5, 115751
215
+ 6, 115751
216
+ 7, 92899
217
+ 0, 1
218
+ 1, 4
219
+ 2, 4
220
+ 3, 114021
221
+ 4, 4
222
+ 5, 4
223
+ 6, 4
224
+ 7, 4
tierb-fixed-moonep-full/moonep/tierb/tierb_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/rep1/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=12
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=0
64
+ kernel_trace=0
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/tierb/tierb_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/rep1/train.log ADDED
@@ -0,0 +1,1314 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [23:45:03] 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/tierb/tierb_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/rep1/memory_snapshots/snapshot.pickle
8
+ repo=/workspace/Megatron-MoE-Echo @ 7bc235bf
9
+ ==============================================================
10
+ W0829 23:45:04.937000 237461 torch/distributed/run.py:766]
11
+ W0829 23:45:04.937000 237461 torch/distributed/run.py:766] *****************************************
12
+ W0829 23:45:04.937000 237461 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
+ W0829 23:45:04.937000 237461 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
+ 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
17
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
18
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
19
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
20
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
21
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
22
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
23
+ 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
24
+ WARNING: overriding default arguments for tokenizer_type:GPT2BPETokenizer with tokenizer_type:GPT2BPETokenizer
25
+ Number of virtual stages per pipeline stage: None
26
+ accumulate and all-reduce gradients in fp32 for bfloat16 data type.
27
+ using torch.bfloat16 for parameters ...
28
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
29
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
30
+ ------------------------ arguments ------------------------
31
+ account_for_embedding_in_pipeline_split ......... False
32
+ account_for_loss_in_pipeline_split .............. False
33
+ accumulate_allreduce_grads_in_fp32 .............. True
34
+ adam_beta1 ...................................... 0.9
35
+ adam_beta2 ...................................... 0.999
36
+ adam_eps ........................................ 1e-08
37
+ add_bias_linear ................................. False
38
+ add_position_embedding .......................... False
39
+ add_qkv_bias .................................... False
40
+ adlr_autoresume ................................. False
41
+ adlr_autoresume_interval ........................ 1000
42
+ align_grad_reduce ............................... True
43
+ align_param_gather .............................. False
44
+ app_tag_run_name ................................ None
45
+ app_tag_run_version ............................. 0.0.0
46
+ apply_layernorm_1p .............................. False
47
+ apply_query_key_layer_scaling ................... False
48
+ apply_residual_connection_post_layernorm ........ False
49
+ apply_rope_fusion ............................... True
50
+ async_save ...................................... None
51
+ async_tensor_model_parallel_allreduce ........... True
52
+ attention_backend ............................... AttnBackend.auto
53
+ attention_dropout ............................... 0.0
54
+ attention_softmax_in_fp32 ....................... False
55
+ auto_detect_ckpt_format ......................... False
56
+ barrier_with_L1_time ............................ True
57
+ bert_binary_head ................................ True
58
+ bert_embedder_type .............................. megatron
59
+ bert_load ....................................... None
60
+ bf16 ............................................ True
61
+ bias_dropout_fusion ............................. True
62
+ bias_gelu_fusion ................................ False
63
+ bias_swiglu_fusion .............................. True
64
+ biencoder_projection_dim ........................ 0
65
+ biencoder_shared_query_context_model ............ False
66
+ block_data_path ................................. None
67
+ calc_ft_timeouts ................................ False
68
+ calculate_per_token_loss ........................ False
69
+ check_for_large_grads ........................... False
70
+ check_for_nan_in_loss_and_grad .................. False
71
+ check_for_spiky_loss ............................ False
72
+ check_weight_hash_across_dp_replicas_interval ... None
73
+ ckpt_assume_constant_structure .................. False
74
+ ckpt_convert_format ............................. None
75
+ ckpt_convert_save ............................... None
76
+ ckpt_convert_update_legacy_dist_opt_format ...... False
77
+ ckpt_format ..................................... torch_dist
78
+ ckpt_fully_parallel_load ........................ False
79
+ ckpt_fully_parallel_save ........................ True
80
+ ckpt_fully_parallel_save_deprecated ............. False
81
+ ckpt_step ....................................... None
82
+ classes_fraction ................................ 1.0
83
+ clip_grad ....................................... 1.0
84
+ clone_scatter_output_in_embedding ............... True
85
+ config_logger_dir ...............................
86
+ consumed_train_samples .......................... 0
87
+ consumed_valid_samples .......................... 0
88
+ context_parallel_size ........................... 1
89
+ cp_comm_type .................................... ['p2p']
90
+ create_attention_mask_in_dataloader ............. True
91
+ cross_entropy_fusion_impl ....................... native
92
+ cross_entropy_loss_fusion ....................... False
93
+ cuda_graph_scope ................................ full_iteration
94
+ cuda_graph_warmup_steps ......................... 3
95
+ data_args_path .................................. None
96
+ data_cache_path ................................. None
97
+ data_parallel_random_init ....................... False
98
+ data_parallel_sharding_strategy ................. no_shard
99
+ data_parallel_size .............................. 8
100
+ data_path ....................................... ['/workspace/data/wikitext/wikitext103_gpt2_text_document']
101
+ data_per_class_fraction ......................... 1.0
102
+ data_sharding ................................... True
103
+ dataloader_type ................................. single
104
+ ddp_average_in_collective ....................... False
105
+ ddp_bucket_size ................................. None
106
+ ddp_num_buckets ................................. None
107
+ ddp_pad_buckets_for_high_nccl_busbw ............. False
108
+ decoder_first_pipeline_num_layers ............... None
109
+ decoder_last_pipeline_num_layers ................ None
110
+ decoder_num_layers .............................. None
111
+ decoder_seq_length .............................. None
112
+ decoupled_lr .................................... None
113
+ decoupled_min_lr ................................ None
114
+ decrease_batch_size_if_needed ................... False
115
+ defer_embedding_wgrad_compute ................... False
116
+ delay_wgrad_compute ............................. False
117
+ deprecated_use_mcore_models ..................... True
118
+ deterministic_mode .............................. False
119
+ dino_bottleneck_size ............................ 256
120
+ dino_freeze_last_layer .......................... 1
121
+ dino_head_hidden_size ........................... 2048
122
+ dino_local_crops_number ......................... 10
123
+ dino_local_img_size ............................. 96
124
+ dino_norm_last_layer ............................ False
125
+ dino_teacher_temp ............................... 0.07
126
+ dino_warmup_teacher_temp ........................ 0.04
127
+ dino_warmup_teacher_temp_epochs ................. 30
128
+ disable_bf16_reduced_precision_matmul ........... False
129
+ disable_mamba_mem_eff_path ...................... False
130
+ disable_straggler_on_startup .................... False
131
+ dist_ckpt_format_deprecated ..................... None
132
+ dist_ckpt_strictness ............................ assume_ok_unexpected
133
+ distribute_saved_activations .................... False
134
+ distributed_backend ............................. nccl
135
+ distributed_timeout_minutes ..................... 10
136
+ embedding_init_method_std ....................... None
137
+ embedding_path .................................. None
138
+ empty_unused_memory_level ....................... 0
139
+ enable_cuda_graph ............................... True
140
+ enable_experimental ............................. False
141
+ enable_ft_package ............................... False
142
+ enable_gloo_process_groups ...................... True
143
+ enable_msc ...................................... True
144
+ enable_one_logger ............................... True
145
+ encoder_num_layers .............................. 12
146
+ encoder_seq_length .............................. 8192
147
+ end_weight_decay ................................ 0.1
148
+ eod_mask_loss ................................... False
149
+ error_injection_rate ............................ 0
150
+ error_injection_type ............................ transient_error
151
+ eval_interval ................................... 1000000
152
+ eval_iters ...................................... 0
153
+ evidence_data_path .............................. None
154
+ exit_duration_in_mins ........................... None
155
+ exit_interval ................................... None
156
+ exit_on_missing_checkpoint ...................... False
157
+ exit_signal_handler ............................. False
158
+ exp_avg_dtype ................................... torch.float32
159
+ exp_avg_sq_dtype ................................ torch.float32
160
+ expert_model_parallel_size ...................... 8
161
+ expert_tensor_parallel_size ..................... 1
162
+ export_force_local_attention .................... False
163
+ export_kd_cfg ................................... None
164
+ export_kd_teacher_ckpt_format ................... None
165
+ export_kd_teacher_load .......................... None
166
+ export_kv_cache_quant ........................... False
167
+ export_legacy_megatron .......................... False
168
+ export_model_type ............................... GPTModel
169
+ export_moe_apply_probs_on_input ................. False
170
+ export_qk_l2_norm ............................... False
171
+ export_quant_cfg ................................ None
172
+ export_real_quant_cfg ........................... None
173
+ export_te_mcore_model ........................... False
174
+ external_cuda_graph ............................. False
175
+ ffn_hidden_size ................................. 12288
176
+ finetune ........................................ False
177
+ finetune_data_split ............................. train
178
+ finetune_hf_dataset ............................. None
179
+ first_last_layers_bf16 .......................... False
180
+ flash_decode .................................... False
181
+ fp16 ............................................ False
182
+ fp16_lm_cross_entropy ........................... False
183
+ fp32_residual_connection ........................ False
184
+ fp8 ............................................. e4m3
185
+ fp8_amax_compute_algo ........................... most_recent
186
+ fp8_amax_history_len ............................ 1
187
+ fp8_interval .................................... 1
188
+ fp8_margin ...................................... 0
189
+ fp8_param_gather ................................ False
190
+ fp8_recipe ...................................... mxfp8
191
+ fp8_wgrad ....................................... True
192
+ fsdp_double_buffer .............................. False
193
+ global_batch_size ............................... 8
194
+ grad_reduce_in_bf16 ............................. False
195
+ gradient_accumulation_fusion .................... True
196
+ gradient_reduce_div_fusion ...................... True
197
+ group_query_attention ........................... True
198
+ head_lr_mult .................................... 1.0
199
+ heterogeneous_layers_config_encoded_json ........ None
200
+ heterogeneous_layers_config_path ................ None
201
+ hidden_dropout .................................. 0.0
202
+ hidden_size ..................................... 4096
203
+ hierarchical_context_parallel_sizes ............. None
204
+ high_priority_stream_groups ..................... []
205
+ hybrid_attention_ratio .......................... 0.0
206
+ hybrid_mlp_ratio ................................ 0.0
207
+ hybrid_override_pattern ......................... None
208
+ hysteresis ...................................... 2
209
+ ict_head_size ................................... None
210
+ ict_load ........................................ None
211
+ img_h ........................................... 224
212
+ img_w ........................................... 224
213
+ indexer_batch_size .............................. 128
214
+ indexer_log_interval ............................ 1000
215
+ inference_batch_times_seqlen_threshold .......... -1
216
+ inference_dynamic_batching ...................... False
217
+ inference_dynamic_batching_buffer_guaranteed_fraction 0.2
218
+ inference_dynamic_batching_buffer_overflow_factor None
219
+ inference_dynamic_batching_buffer_size_gb ....... 40.0
220
+ inference_dynamic_batching_chunk_size ........... 256
221
+ inference_dynamic_batching_max_requests_override None
222
+ inference_dynamic_batching_max_tokens_override .. None
223
+ inference_dynamic_batching_num_cuda_graphs ...... 16
224
+ inference_max_batch_size ........................ 8
225
+ inference_max_seq_length ........................ 2560
226
+ inference_rng_tracker ........................... False
227
+ init_method_std ................................. 0.01
228
+ init_method_xavier_uniform ...................... False
229
+ init_model_with_meta_device ..................... False
230
+ initial_loss_scale .............................. 4294967296
231
+ inprocess_active_world_size ..................... 8
232
+ inprocess_barrier_timeout ....................... 120
233
+ inprocess_completion_timeout .................... 120
234
+ inprocess_empty_cuda_cache ...................... False
235
+ inprocess_granularity ........................... node
236
+ inprocess_hard_timeout .......................... 90
237
+ inprocess_heartbeat_interval .................... 30
238
+ inprocess_heartbeat_timeout ..................... 60
239
+ inprocess_last_call_wait ........................ 1
240
+ inprocess_max_iterations ........................ None
241
+ inprocess_monitor_process_interval .............. 1.0
242
+ inprocess_monitor_thread_interval ............... 1.0
243
+ inprocess_progress_watchdog_interval ............ 1.0
244
+ inprocess_restart ............................... False
245
+ inprocess_soft_timeout .......................... 60
246
+ inprocess_termination_grace_time ................ 1
247
+ is_hybrid_model ................................. False
248
+ iter_per_epoch .................................. 1250
249
+ iterations_to_skip .............................. []
250
+ keep_fp8_transpose_cache_when_using_custom_fsdp . False
251
+ kitchen_config_file ............................. None
252
+ kitchen_recipe_number ........................... None
253
+ kv_channels ..................................... 128
254
+ kv_lora_rank .................................... 32
255
+ lazy_mpu_init ................................... None
256
+ load ............................................ None
257
+ load_main_params_from_ckpt ...................... None
258
+ load_model_opt_format ........................... False
259
+ local_rank ...................................... 0
260
+ log_energy ...................................... False
261
+ log_interval .................................... 1
262
+ log_loss_scale_to_tensorboard ................... True
263
+ log_memory_to_tensorboard ....................... False
264
+ log_num_zeros_in_grad ........................... False
265
+ log_params_norm ................................. False
266
+ log_progress .................................... False
267
+ log_straggler ................................... False
268
+ log_throughput .................................. True
269
+ log_timers_to_tensorboard ....................... False
270
+ log_validation_ppl_to_tensorboard ............... False
271
+ log_world_size_to_tensorboard ................... False
272
+ logging_level ................................... None
273
+ loss_scale ...................................... None
274
+ loss_scale_window ............................... 1000
275
+ lr .............................................. 0.0001
276
+ lr_decay_iters .................................. 8
277
+ lr_decay_samples ................................ None
278
+ lr_decay_style .................................. cosine
279
+ lr_warmup_fraction .............................. None
280
+ lr_warmup_init .................................. 0.0
281
+ lr_warmup_iters ................................. 5
282
+ lr_warmup_samples ............................... 0
283
+ lr_wsd_decay_iters .............................. None
284
+ lr_wsd_decay_samples ............................ None
285
+ lr_wsd_decay_style .............................. exponential
286
+ main_grads_dtype ................................ torch.float32
287
+ main_params_dtype ............................... torch.float32
288
+ make_vocab_size_divisible_by .................... 128
289
+ mamba_head_dim .................................. 64
290
+ mamba_num_groups ................................ 8
291
+ mamba_num_heads ................................. None
292
+ mamba_state_dim ................................. 128
293
+ manual_gc ....................................... False
294
+ manual_gc_eval .................................. True
295
+ manual_gc_interval .............................. 0
296
+ mask_factor ..................................... 1.0
297
+ mask_prob ....................................... 0.15
298
+ mask_type ....................................... random
299
+ masked_softmax_fusion ........................... False
300
+ max_position_embeddings ......................... 32768
301
+ max_tokens_to_oom ............................... 12000
302
+ memory_snapshot_path ............................ snapshot.pickle
303
+ merge_file ...................................... /workspace/data/wikitext/gpt2-merges.txt
304
+ micro_batch_size ................................ 1
305
+ microbatch_group_size_per_vp_stage .............. None
306
+ mid_level_dataset_surplus ....................... 0.005
307
+ min_loss_scale .................................. 1.0
308
+ min_lr .......................................... 1e-05
309
+ mlp_chunks_for_prefill .......................... 1
310
+ mmap_bin_files .................................. True
311
+ mock_data ....................................... False
312
+ moe_apply_probs_on_input ........................ False
313
+ moe_aux_loss_coeff .............................. 0.001
314
+ moe_deepep_num_sms .............................. 20
315
+ moe_echo_enable_random_offloading ............... False
316
+ moe_echo_expert_dispatch_overlap ................ False
317
+ moe_echo_expert_dispatcher_type ................. hybridep
318
+ moe_echo_recompute_expert_dispatch .............. False
319
+ moe_enable_deepep ............................... False
320
+ moe_enable_echo ................................. False
321
+ moe_enable_hybridep ............................. False
322
+ moe_enable_moonep ............................... True
323
+ moe_expert_capacity_factor ...................... None
324
+ moe_extended_tp ................................. False
325
+ moe_ffn_hidden_size ............................. 1536
326
+ moe_grouped_gemm ................................ True
327
+ moe_hybridep_num_sms ............................ 16
328
+ moe_input_jitter_eps ............................ None
329
+ moe_layer_freq .................................. 1
330
+ moe_layer_recompute ............................. False
331
+ moe_moonep_async_grad_reduce .................... True
332
+ moe_moonep_num_sms .............................. None
333
+ moe_num_echo_experts ............................ None
334
+ moe_pad_expert_input_to_capacity ................ False
335
+ moe_per_layer_logging ........................... False
336
+ moe_permute_fusion .............................. False
337
+ moe_received_token_capacity ..................... None
338
+ moe_router_bias_update_rate ..................... 0.001
339
+ moe_router_dtype ................................ fp32
340
+ moe_router_enable_expert_bias ................... False
341
+ moe_router_force_load_balancing ................. False
342
+ moe_router_group_topk ........................... None
343
+ moe_router_load_balancing_type .................. aux_loss
344
+ moe_router_num_groups ........................... None
345
+ moe_router_padding_for_fp8 ...................... False
346
+ moe_router_pre_softmax .......................... False
347
+ moe_router_score_function ....................... softmax
348
+ moe_router_topk ................................. 8
349
+ moe_router_topk_scaling_factor .................. None
350
+ moe_shared_expert_intermediate_size ............. None
351
+ moe_shared_expert_overlap ....................... False
352
+ moe_token_dispatcher_type ....................... flex
353
+ moe_token_drop_policy ........................... probs
354
+ moe_upcycling_granularity ....................... 1
355
+ moe_use_device_initiated_grouped_gemm ........... True
356
+ moe_use_legacy_grouped_gemm ..................... False
357
+ moe_use_upcycling ............................... False
358
+ moe_z_loss_coeff ................................ None
359
+ mrope_section ................................... None
360
+ mscale .......................................... 1.0
361
+ mscale_all_dim .................................. 0.0
362
+ mtp_loss_scaling_factor ......................... 0.1
363
+ mtp_num_layers .................................. None
364
+ multi_latent_attention .......................... False
365
+ nccl_all_reduce_for_prefill ..................... False
366
+ nccl_communicator_config_path ................... None
367
+ nccl_ub ......................................... False
368
+ no_load_optim ................................... None
369
+ no_load_rng ..................................... None
370
+ no_persist_layer_norm ........................... False
371
+ no_rope_freq .................................... None
372
+ no_save_optim ................................... None
373
+ no_save_rng ..................................... None
374
+ non_persistent_ckpt_type ........................ None
375
+ non_persistent_global_ckpt_dir .................. None
376
+ non_persistent_local_ckpt_algo .................. fully_parallel
377
+ non_persistent_local_ckpt_dir ................... None
378
+ non_persistent_save_interval .................... None
379
+ norm_epsilon .................................... 1e-06
380
+ normalization ................................... RMSNorm
381
+ num_attention_heads ............................. 64
382
+ num_channels .................................... 3
383
+ num_classes ..................................... 1000
384
+ num_dataset_builder_threads ..................... 1
385
+ num_distributed_optimizer_instances ............. 1
386
+ num_experts ..................................... 128
387
+ num_layers ...................................... 12
388
+ num_layers_at_end_in_bf16 ....................... 1
389
+ num_layers_at_start_in_bf16 ..................... 1
390
+ num_layers_per_virtual_pipeline_stage ........... None
391
+ num_query_groups ................................ 4
392
+ num_virtual_stages_per_pipeline_rank ............ None
393
+ num_workers ..................................... 2
394
+ object_storage_cache_path ....................... None
395
+ one_logger_async ................................ False
396
+ one_logger_project .............................. megatron-lm
397
+ one_logger_run_name ............................. None
398
+ onnx_safe ....................................... None
399
+ openai_gelu ..................................... False
400
+ optimizer ....................................... adam
401
+ optimizer_cpu_offload ........................... False
402
+ optimizer_offload_fraction ...................... 1.0
403
+ output_bert_embeddings .......................... False
404
+ overlap_cpu_optimizer_d2h_h2d ................... False
405
+ overlap_grad_reduce ............................. True
406
+ overlap_moe_expert_parallel_comm ................ False
407
+ overlap_p2p_comm ................................ False
408
+ overlap_p2p_comm_warmup_flush ................... False
409
+ overlap_param_gather ............................ True
410
+ overlap_param_gather_with_optimizer_step ........ False
411
+ override_opt_param_scheduler .................... False
412
+ params_dtype .................................... torch.bfloat16
413
+ patch_dim ....................................... 16
414
+ per_split_data_args_path ........................ None
415
+ perform_initialization .......................... True
416
+ pin_cpu_grads ................................... True
417
+ pin_cpu_params .................................. True
418
+ pipeline_model_parallel_comm_backend ............ None
419
+ pipeline_model_parallel_layout .................. None
420
+ pipeline_model_parallel_size .................... 1
421
+ position_embedding_type ......................... rope
422
+ pretrained_checkpoint ........................... None
423
+ profile ......................................... False
424
+ profile_ranks ................................... [0]
425
+ profile_step_end ................................ 12
426
+ profile_step_start .............................. 10
427
+ q_lora_rank ..................................... None
428
+ qk_head_dim ..................................... 128
429
+ qk_l2_norm ...................................... False
430
+ qk_layernorm .................................... True
431
+ qk_pos_emb_head_dim ............................. 64
432
+ query_in_block_prob ............................. 0.1
433
+ rampup_batch_size ............................... None
434
+ rank ............................................ 0
435
+ recompute_granularity ........................... None
436
+ recompute_method ................................ None
437
+ recompute_modules ............................... None
438
+ recompute_num_layers ............................ None
439
+ record_memory_history ........................... False
440
+ relative_attention_max_distance ................. 128
441
+ relative_attention_num_buckets .................. 32
442
+ replication ..................................... False
443
+ replication_factor .............................. 2
444
+ replication_jump ................................ None
445
+ rerun_mode ...................................... validate_results
446
+ reset_attention_mask ............................ False
447
+ reset_position_ids .............................. False
448
+ result_rejected_tracker_filename ................ None
449
+ retriever_report_topk_accuracies ................ []
450
+ retriever_score_scaling ......................... False
451
+ retriever_seq_length ............................ 256
452
+ retro_add_retriever ............................. False
453
+ retro_attention_gate ............................ 1
454
+ retro_cyclic_train_iters ........................ None
455
+ retro_encoder_attention_dropout ................. 0.1
456
+ retro_encoder_hidden_dropout .................... 0.1
457
+ retro_encoder_layers ............................ 2
458
+ retro_num_neighbors ............................. 2
459
+ retro_num_retrieved_chunks ...................... 2
460
+ retro_project_dir ............................... None
461
+ retro_verify_neighbor_count ..................... True
462
+ reuse_grad_buf_for_mxfp8_param_ag ............... False
463
+ rope_scaling_factor ............................. 8.0
464
+ rope_type ....................................... None
465
+ rotary_base ..................................... 1000000
466
+ rotary_interleaved .............................. False
467
+ rotary_percent .................................. 1.0
468
+ rotary_scaling_factor ........................... 1.0
469
+ rotary_seq_len_interpolation_factor ............. None
470
+ run_workload_inspector_server ................... False
471
+ sample_rate ..................................... 1.0
472
+ save ............................................ None
473
+ save_interval ................................... None
474
+ save_retain_interval ............................ None
475
+ scatter_gather_tensors_in_pipeline .............. True
476
+ seed ............................................ 1234
477
+ seq_length ...................................... 8192
478
+ sequence_parallel ............................... False
479
+ sft ............................................. False
480
+ sft_tokenizer_prompt_format ..................... nemotron-h-aligned
481
+ sgd_momentum .................................... 0.9
482
+ short_seq_prob .................................. 0.1
483
+ skip_train ...................................... False
484
+ skipped_train_samples ........................... 0
485
+ spec ............................................ None
486
+ split ........................................... 100,0,0
487
+ squared_relu .................................... False
488
+ start_weight_decay .............................. 0.1
489
+ straggler_ctrlr_port ............................ 65535
490
+ straggler_minmax_count .......................... 1
491
+ suggested_communication_unit_size ............... None
492
+ swiglu .......................................... True
493
+ swin_backbone_type .............................. tiny
494
+ symmetric_ar_type ............................... None
495
+ te_rng_tracker .................................. True
496
+ tensor_model_parallel_size ...................... 1
497
+ tensorboard_dir ................................. None
498
+ tensorboard_log_interval ........................ 1
499
+ tensorboard_queue_size .......................... 1000
500
+ test_data_path .................................. None
501
+ test_mode ....................................... False
502
+ tiktoken_num_special_tokens ..................... 1000
503
+ tiktoken_pattern ................................ None
504
+ tiktoken_special_tokens ......................... None
505
+ timing_log_level ................................ 0
506
+ timing_log_option ............................... minmax
507
+ titles_data_path ................................ None
508
+ tokenizer_model ................................. None
509
+ tokenizer_type .................................. GPT2BPETokenizer
510
+ torch_fsdp2_reshard_after_forward ............... True
511
+ tp_comm_bootstrap_backend ....................... nccl
512
+ tp_comm_bulk_dgrad .............................. True
513
+ tp_comm_bulk_wgrad .............................. True
514
+ tp_comm_overlap ................................. False
515
+ tp_comm_overlap_ag .............................. True
516
+ tp_comm_overlap_cfg ............................. None
517
+ tp_comm_overlap_rs .............................. True
518
+ tp_comm_overlap_rs_dgrad ........................ False
519
+ tp_comm_split_ag ................................ True
520
+ tp_comm_split_rs ................................ True
521
+ train_data_path ................................. None
522
+ train_iters ..................................... 12
523
+ train_samples ................................... None
524
+ train_sync_interval ............................. None
525
+ transformer_impl ................................ transformer_engine
526
+ transformer_pipeline_model_parallel_size ........ 1
527
+ untie_embeddings_and_output_weights ............. True
528
+ use_checkpoint_args ............................. False
529
+ use_checkpoint_opt_param_scheduler .............. False
530
+ use_cpu_initialization .......................... None
531
+ use_custom_fsdp ................................. False
532
+ use_dist_ckpt ................................... True
533
+ use_dist_ckpt_deprecated ........................ False
534
+ use_distributed_optimizer ....................... True
535
+ use_flash_attn .................................. False
536
+ use_legacy_models ............................... False
537
+ use_mp_args_from_checkpoint_args ................ False
538
+ use_one_sent_docs ............................... False
539
+ use_persistent_ckpt_worker ...................... False
540
+ use_precision_aware_optimizer ................... False
541
+ use_pytorch_profiler ............................ False
542
+ use_ring_exchange_p2p ........................... False
543
+ use_rope_scaling ................................ False
544
+ use_rotary_position_embeddings .................. False
545
+ use_sharp ....................................... False
546
+ use_tokenizer_model_from_checkpoint_args ........ True
547
+ use_torch_fsdp2 ................................. False
548
+ use_torch_optimizer_for_cpu_offload ............. False
549
+ use_tp_pp_dp_mapping ............................ False
550
+ v_head_dim ...................................... 128
551
+ valid_data_path ................................. None
552
+ variable_seq_lengths ............................ False
553
+ virtual_pipeline_model_parallel_size ............ None
554
+ vision_backbone_type ............................ vit
555
+ vision_pretraining .............................. False
556
+ vision_pretraining_type ......................... classify
557
+ vocab_extra_ids ................................. 0
558
+ vocab_file ...................................... /workspace/data/wikitext/gpt2-vocab.json
559
+ vocab_size ...................................... None
560
+ wandb_exp_name ..................................
561
+ wandb_project ...................................
562
+ wandb_save_dir ..................................
563
+ weight_decay .................................... 0.1
564
+ weight_decay_incr_style ......................... constant
565
+ wgrad_deferral_limit ............................ 0
566
+ world_size ...................................... 8
567
+ yaml_cfg ........................................ None
568
+ -------------------- end of arguments ---------------------
569
+ INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 1
570
+ > building GPT2BPETokenizer tokenizer ...
571
+ > padded vocab (size: 50257) with 47 dummy tokens (new size: 50304)
572
+ WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.VALIDATE_RESULTS
573
+ > initializing torch distributed ...
574
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
575
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
576
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
577
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
578
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
579
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
580
+ [Gloo] Rank [Gloo] Rank 0 is connected to 77 is connected to peer ranks. 7Expected number of connected peer ranks is : peer ranks. 7Expected number of connected peer ranks is : 7
581
+
582
+ [Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
583
+ [Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
584
+ [Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
585
+ [Gloo] Rank [Gloo] Rank 5 is connected to 72 peer ranks. is connected to Expected number of connected peer ranks is : 77 peer ranks. Expected number of connected peer ranks is :
586
+ 7
587
+ [Gloo] Rank 6 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
588
+ [Gloo] Rank 3 is connected to [Gloo] Rank 7 peer ranks. 2Expected number of connected peer ranks is : is connected to 77
589
+ peer ranks. Expected number of connected peer ranks is : 7
590
+ [Gloo] Rank 5 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
591
+ [Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
592
+ [Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
593
+ [Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
594
+ [Gloo] Rank 6 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
595
+ [Gloo] Rank 7 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
596
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
597
+ > initialized tensor model parallel with size 1
598
+ > initialized pipeline model parallel with size 1
599
+ > setting random seeds to 1234 ...
600
+ [Gloo] Rank [Gloo] Rank 00 is connected to is connected to 00 peer ranks. peer ranks. Expected number of connected peer ranks is : Expected number of connected peer ranks is : 00
601
+
602
+ [Gloo] Rank [Gloo] Rank 00 is connected to is connected to 00 peer ranks. peer ranks. Expected number of connected peer ranks is : Expected number of connected peer ranks is : 00
603
+
604
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
605
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
606
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
607
+ [rank2]:[W829 23:45:23.527949468 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.
608
+ [rank5]:[W829 23:45:23.528057503 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
609
+ [rank1]:[W829 23:45:23.528113700 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.
610
+ [rank4]:[W829 23:45:23.528140272 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.
611
+ [rank6]:[W829 23:45:23.528167906 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.
612
+ [rank3]:[W829 23:45:23.529332491 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.
613
+ [rank7]:[W829 23:45:23.529419740 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.
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: 0.033 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]:[W829 23:45:23.696190153 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: 3.725 seconds
623
+ time to initialize megatron (seconds): 7.357
624
+ [after megatron is initialized] datetime: 2026-08-29 23:45:29
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.pre_mlp_layernorm.weight
649
+ module.decoder.layers.11.mlp.router.weight
650
+ module.decoder.layers.11.self_attention.q_layernorm.weight
651
+ module.decoder.layers.11.self_attention.linear_qkv.weight
652
+ module.decoder.layers.11.self_attention.linear_qkv.layer_norm_weight
653
+ module.decoder.final_layernorm.weight
654
+ module.decoder.layers.11.self_attention.k_layernorm.weight
655
+ module.decoder.layers.11.self_attention.linear_proj.weight
656
+ Params for bucket 3 (71835904 elements, 71835904 padded size):
657
+ module.decoder.layers.10.mlp.router.weight
658
+ module.decoder.layers.10.pre_mlp_layernorm.weight
659
+ module.decoder.layers.10.self_attention.k_layernorm.weight
660
+ module.decoder.layers.10.self_attention.q_layernorm.weight
661
+ module.decoder.layers.10.self_attention.linear_qkv.weight
662
+ module.decoder.layers.10.self_attention.linear_qkv.layer_norm_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.mlp.router.weight
666
+ module.decoder.layers.9.pre_mlp_layernorm.weight
667
+ module.decoder.layers.9.self_attention.linear_proj.weight
668
+ module.decoder.layers.9.self_attention.q_layernorm.weight
669
+ module.decoder.layers.9.self_attention.k_layernorm.weight
670
+ module.decoder.layers.9.self_attention.linear_qkv.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.mlp.router.weight
674
+ module.decoder.layers.8.pre_mlp_layernorm.weight
675
+ module.decoder.layers.8.self_attention.k_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.q_layernorm.weight
684
+ module.decoder.layers.7.self_attention.k_layernorm.weight
685
+ module.decoder.layers.7.self_attention.linear_qkv.layer_norm_weight
686
+ module.decoder.layers.7.self_attention.linear_qkv.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.pre_mlp_layernorm.weight
691
+ module.decoder.layers.6.self_attention.k_layernorm.weight
692
+ module.decoder.layers.6.self_attention.linear_qkv.weight
693
+ module.decoder.layers.6.self_attention.q_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.mlp.router.weight
698
+ module.decoder.layers.5.pre_mlp_layernorm.weight
699
+ module.decoder.layers.5.self_attention.linear_qkv.weight
700
+ module.decoder.layers.5.self_attention.q_layernorm.weight
701
+ module.decoder.layers.5.self_attention.linear_qkv.layer_norm_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.mlp.router.weight
706
+ module.decoder.layers.4.pre_mlp_layernorm.weight
707
+ module.decoder.layers.4.self_attention.k_layernorm.weight
708
+ module.decoder.layers.4.self_attention.q_layernorm.weight
709
+ module.decoder.layers.4.self_attention.linear_qkv.weight
710
+ module.decoder.layers.4.self_attention.linear_qkv.layer_norm_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.linear_qkv.weight
716
+ module.decoder.layers.3.self_attention.q_layernorm.weight
717
+ module.decoder.layers.3.self_attention.k_layernorm.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.k_layernorm.weight
723
+ module.decoder.layers.2.self_attention.q_layernorm.weight
724
+ module.decoder.layers.2.self_attention.linear_qkv.weight
725
+ module.decoder.layers.2.self_attention.linear_qkv.layer_norm_weight
726
+ module.decoder.layers.2.self_attention.linear_proj.weight
727
+ module.decoder.layers.2.pre_mlp_layernorm.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.q_layernorm.weight
732
+ module.decoder.layers.1.pre_mlp_layernorm.weight
733
+ module.decoder.layers.1.self_attention.k_layernorm.weight
734
+ module.decoder.layers.1.self_attention.linear_qkv.layer_norm_weight
735
+ module.decoder.layers.1.self_attention.linear_proj.weight
736
+ Params for bucket 13 (71835904 elements, 71835904 padded size):
737
+ module.decoder.layers.0.pre_mlp_layernorm.weight
738
+ module.decoder.layers.0.mlp.router.weight
739
+ module.decoder.layers.0.self_attention.q_layernorm.weight
740
+ module.decoder.layers.0.self_attention.linear_qkv.weight
741
+ module.decoder.layers.0.self_attention.k_layernorm.weight
742
+ module.decoder.layers.0.self_attention.linear_qkv.layer_norm_weight
743
+ module.decoder.layers.0.self_attention.linear_proj.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.weight11
761
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight9
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_fc1.weight0
767
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight15
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.weight6
772
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight4
773
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight5
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.weight10
777
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight9
778
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight8
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.weight13
782
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight14
783
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight12
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.weight3
787
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight1
788
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight0
789
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight2
790
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight4
791
+ Params for bucket 8 (44040192 elements, 44040192 padded size):
792
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight5
793
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight7
794
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight9
795
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight11
796
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight6
797
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight8
798
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight10
799
+ Params for bucket 9 (50331648 elements, 50331648 padded size):
800
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight13
801
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight15
802
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight0
803
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight12
804
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight14
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.weight7
833
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight5
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_fc1.weight0
840
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight13
841
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight11
842
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight15
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.weight4
846
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight2
847
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight3
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.weight7
851
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight8
852
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight6
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.weight10
856
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight11
857
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight12
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.weight14
861
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight0
862
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight15
863
+ Params for bucket 20 (44040192 elements, 44040192 padded size):
864
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight4
865
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight2
866
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight6
867
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight1
868
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight3
869
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight5
870
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight7
871
+ Params for bucket 21 (44040192 elements, 44040192 padded size):
872
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight8
873
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight10
874
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight12
875
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight14
876
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight9
877
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight11
878
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight13
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.weight1
882
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight0
883
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight2
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.weight4
887
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight5
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.weight10
892
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight8
893
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight9
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.weight13
897
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight14
898
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight12
899
+ Params for bucket 26 (44040192 elements, 44040192 padded size):
900
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight15
901
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight3
902
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight1
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_fc1.weight0
917
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight13
918
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight12
919
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight14
920
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight1
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.weight3
924
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight4
925
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight5
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.weight7
929
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight9
930
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight8
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.weight13
934
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight12
935
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight11
936
+ Params for bucket 32 (44040192 elements, 44040192 padded size):
937
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight14
938
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight2
939
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight0
940
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight1
941
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight15
942
+ Params for bucket 33 (44040192 elements, 44040192 padded size):
943
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight6
944
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight4
945
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight8
946
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight3
947
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight5
948
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight7
949
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight9
950
+ Params for bucket 34 (50331648 elements, 50331648 padded size):
951
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight10
952
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight14
953
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight12
954
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight11
955
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight13
956
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight15
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.weight2
961
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight3
962
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight4
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.weight8
966
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight6
967
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight7
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.weight11
971
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight12
972
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight10
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.weight1
980
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight5
981
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight3
982
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight7
983
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight2
984
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight4
985
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight6
986
+ Params for bucket 40 (44040192 elements, 44040192 padded size):
987
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight11
988
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight9
989
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight13
990
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight8
991
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight12
992
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight10
993
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight14
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.weight0
997
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight1
998
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight2
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.weight6
1002
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight4
1003
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight5
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.weight10
1007
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight8
1008
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight9
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.weight13
1012
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight14
1013
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight12
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.weight0
1017
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight2
1018
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight4
1019
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight3
1020
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight1
1021
+ Params for bucket 46 (44040192 elements, 44040192 padded size):
1022
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight8
1023
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight6
1024
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight10
1025
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight5
1026
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight7
1027
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight9
1028
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight11
1029
+ Params for bucket 47 (50331648 elements, 50331648 padded size):
1030
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight12
1031
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight14
1032
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight0
1033
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight15
1034
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight13
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_fc2.weight13
1150
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight0
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_fc1.weight0
1185
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight11
1186
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight13
1187
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight15
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.weight4
1191
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight3
1192
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight2
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.weight8
1196
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight7
1197
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight6
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.weight11
1201
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight10
1202
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight12
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.weight14
1206
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight15
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 0x751c290313d0>, 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: 2026-08-29 23:45:32
1210
+ > building train, validation, and test datasets ...
1211
+ > datasets target sizes (minimum size):
1212
+ train: 96
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=(96, 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 0x751982c299a0>, 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 eadc3756447e39d609e8aa728a91f451-GPTDataset-train-document_index.npy
1226
+ INFO:megatron.core.datasets.gpt_dataset: Load the sample index from eadc3756447e39d609e8aa728a91f451-GPTDataset-train-sample_index.npy
1227
+ INFO:megatron.core.datasets.gpt_dataset: Load the shuffle index from eadc3756447e39d609e8aa728a91f451-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: 2026-08-29 23:45:32
1231
+ done with setup ...
1232
+ (min, max) time across ranks (ms):
1233
+ model-and-optimizer-setup ......................: (3261.11, 3293.81)
1234
+ train/valid/test-data-iterators-setup ..........: (63.05, 68.76)
1235
+ training ...
1236
+ Overwriting rerun_state_machine.current_iteration from -1 to 0...
1237
+ [before the start of training step] datetime: 2026-08-29 23:45:32
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
+ Number of parameters in transformer block in billions: 29.85
1250
+ Number of parameters in embedding layers in billions: 0.41
1251
+ Total number of parameters in billions: 30.26
1252
+ Number of parameters in most loaded shard in billions: 30.2590
1253
+ Theoretical memory footprints: weight and optimizer=216428.96 MB
1254
+ [2026-08-29 23:45:53] iteration 1/ 12 | consumed samples: 8 | elapsed time per iteration (ms): 21086.4 | 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 |
1255
+ [Rank 0] (after 1 iterations) memory (MB) | allocated: 67239.548828125 | max allocated: 67239.54931640625 | reserved: 68548.0 | max reserved: 68548.0
1256
+ [2026-08-29 23:45:54] iteration 2/ 12 | consumed samples: 16 | elapsed time per iteration (ms): 434.3 | throughput per GPU (TFLOP/s/GPU): 416.4 | learning rate: 4.000000E-05 | global batch size: 8 | lm loss: 1.117600E+01 | loss scale: 1.0 | grad norm: 44.335 | number of skipped iterations: 0 | number of nan iterations: 0 |
1257
+ [2026-08-29 23:45:54] iteration 3/ 12 | consumed samples: 24 | elapsed time per iteration (ms): 331.5 | throughput per GPU (TFLOP/s/GPU): 545.5 | learning rate: 6.000000E-05 | global batch size: 8 | lm loss: 9.686145E+00 | loss scale: 1.0 | grad norm: 62.918 | number of skipped iterations: 0 | number of nan iterations: 0 |
1258
+ Capture CUDA graph for training!!!
1259
+ CUDA graph capture done!!!
1260
+ [2026-08-29 23:45:55] iteration 4/ 12 | consumed samples: 32 | elapsed time per iteration (ms): 982.5 | throughput per GPU (TFLOP/s/GPU): 184.1 | learning rate: 8.000000E-05 | global batch size: 8 | lm loss: 1.338087E+01 | loss scale: 1.0 | grad norm: 20.295 | number of skipped iterations: 0 | number of nan iterations: 0 |
1261
+ [2026-08-29 23:45:56] iteration 5/ 12 | consumed samples: 40 | elapsed time per iteration (ms): 324.9 | throughput per GPU (TFLOP/s/GPU): 556.6 | learning rate: 1.000000E-04 | global batch size: 8 | lm loss: 1.049608E+01 | loss scale: 1.0 | grad norm: 95.117 | number of skipped iterations: 0 | number of nan iterations: 0 |
1262
+ [2026-08-29 23:45:56] iteration 6/ 12 | consumed samples: 48 | elapsed time per iteration (ms): 321.3 | throughput per GPU (TFLOP/s/GPU): 562.7 | learning rate: 7.750000E-05 | global batch size: 8 | lm loss: 1.044566E+01 | loss scale: 1.0 | grad norm: 52.048 | number of skipped iterations: 0 | number of nan iterations: 0 |
1263
+ [2026-08-29 23:45:56] iteration 7/ 12 | consumed samples: 56 | elapsed time per iteration (ms): 321.2 | throughput per GPU (TFLOP/s/GPU): 562.9 | learning rate: 3.250000E-05 | global batch size: 8 | lm loss: 1.027068E+01 | loss scale: 1.0 | grad norm: 6.324 | number of skipped iterations: 0 | number of nan iterations: 0 |
1264
+ [2026-08-29 23:45:56] iteration 8/ 12 | consumed samples: 64 | elapsed time per iteration (ms): 321.6 | throughput per GPU (TFLOP/s/GPU): 562.2 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 9.808611E+00 | loss scale: 1.0 | grad norm: 5.543 | number of skipped iterations: 0 | number of nan iterations: 0 |
1265
+ [2026-08-29 23:45:57] iteration 9/ 12 | consumed samples: 72 | elapsed time per iteration (ms): 320.4 | throughput per GPU (TFLOP/s/GPU): 564.5 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 9.545443E+00 | loss scale: 1.0 | grad norm: 5.290 | number of skipped iterations: 0 | number of nan iterations: 0 |
1266
+ [2026-08-29 23:45:57] iteration 10/ 12 | consumed samples: 80 | elapsed time per iteration (ms): 323.8 | throughput per GPU (TFLOP/s/GPU): 558.4 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 9.446625E+00 | loss scale: 1.0 | grad norm: 5.018 | number of skipped iterations: 0 | number of nan iterations: 0 |
1267
+ [2026-08-29 23:45:57] iteration 11/ 12 | consumed samples: 88 | elapsed time per iteration (ms): 321.9 | throughput per GPU (TFLOP/s/GPU): 561.8 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 9.324770E+00 | loss scale: 1.0 | grad norm: 4.700 | number of skipped iterations: 0 | number of nan iterations: 0 |
1268
+ [2026-08-29 23:45:58] iteration 12/ 12 | consumed samples: 96 | elapsed time per iteration (ms): 321.0 | throughput per GPU (TFLOP/s/GPU): 563.4 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 9.210297E+00 | loss scale: 1.0 | grad norm: 4.466 | number of skipped iterations: 0 | number of nan iterations: 0 |
1269
+ [after training is done] datetime: 2026-08-29 23:45:58
1270
+ [rank0]:[W829 23:45:59.838424178 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())
1271
+ [rank4]:[W829 23:46:00.089107186 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())
1272
+ [rank6]:[W829 23:46:00.110120316 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())
1273
+ [rank2]:[W829 23:46:00.110384142 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
+ [rank7]:[W829 23:46:00.117783475 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
+ [rank5]:[W829 23:46:00.195426955 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
+ [rank3]:[W829 23:46:01.738688058 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
+ [rank1]:[W829 23:46:01.749461216 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
+ W0829 23:46:04.688000 237461 torch/distributed/elastic/agent/server/api.py:719] Received 15 death signal, shutting down workers
1279
+ W0829 23:46:04.689000 237461 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 237651 closing signal SIGTERM
1280
+ W0829 23:46:04.689000 237461 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 237652 closing signal SIGTERM
1281
+ W0829 23:46:04.690000 237461 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 237653 closing signal SIGTERM
1282
+ W0829 23:46:04.690000 237461 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 237654 closing signal SIGTERM
1283
+ W0829 23:46:04.690000 237461 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 237655 closing signal SIGTERM
1284
+ W0829 23:46:04.691000 237461 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 237656 closing signal SIGTERM
1285
+ W0829 23:46:04.691000 237461 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 237657 closing signal SIGTERM
1286
+ W0829 23:46:04.692000 237461 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 237658 closing signal SIGTERM
1287
+ Traceback (most recent call last):
1288
+ File "/usr/local/bin/torchrun", line 8, in <module>
1289
+ sys.exit(main())
1290
+ ^^^^^^
1291
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper
1292
+ return f(*args, **kwargs)
1293
+ ^^^^^^^^^^^^^^^^^^
1294
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 892, in main
1295
+ run(args)
1296
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 883, in run
1297
+ elastic_launch(
1298
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 139, in __call__
1299
+ return launch_agent(self._config, self._entrypoint, list(args))
1300
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1301
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 261, in launch_agent
1302
+ result = agent.run()
1303
+ ^^^^^^^^^^^
1304
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/metrics/api.py", line 138, in wrapper
1305
+ result = f(*args, **kwargs)
1306
+ ^^^^^^^^^^^^^^^^^^
1307
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 711, in run
1308
+ result = self._invoke_run(role)
1309
+ ^^^^^^^^^^^^^^^^^^^^^^
1310
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 870, in _invoke_run
1311
+ time.sleep(monitor_interval)
1312
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/api.py", line 84, in _terminate_process_handler
1313
+ raise SignalException(f"Process {os.getpid()} got signal: {sigval}", sigval=sigval)
1314
+ torch.distributed.elastic.multiprocessing.api.SignalException: Process 237461 got signal: 15
tierb-fixed-moonep-full/moonep/tierb/tierb_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/rep2/mem_samples.csv ADDED
@@ -0,0 +1,136 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ 0, 0
2
+ 1, 0
3
+ 2, 0
4
+ 3, 0
5
+ 4, 0
6
+ 5, 0
7
+ 6, 0
8
+ 7, 0
9
+ 0, 4
10
+ 1, 4
11
+ 2, 1
12
+ 3, 1
13
+ 4, 1
14
+ 5, 1
15
+ 6, 1
16
+ 7, 1
17
+ 0, 4
18
+ 1, 4
19
+ 2, 4
20
+ 3, 4
21
+ 4, 4
22
+ 5, 4
23
+ 6, 4
24
+ 7, 1
25
+ 0, 1
26
+ 1, 1
27
+ 2, 1
28
+ 3, 1
29
+ 4, 1
30
+ 5, 4
31
+ 6, 4
32
+ 7, 4
33
+ 0, 0
34
+ 1, 0
35
+ 2, 0
36
+ 3, 0
37
+ 4, 0
38
+ 5, 0
39
+ 6, 0
40
+ 7, 0
41
+ 0, 1
42
+ 1, 1
43
+ 2, 1
44
+ 3, 1
45
+ 4, 1
46
+ 5, 1
47
+ 6, 1
48
+ 7, 1
49
+ 0, 4
50
+ 1, 4
51
+ 2, 4
52
+ 3, 4
53
+ 4, 1
54
+ 5, 1
55
+ 6, 1
56
+ 7, 1
57
+ 0, 4
58
+ 1, 4
59
+ 2, 4
60
+ 3, 4
61
+ 4, 4
62
+ 5, 4
63
+ 6, 4
64
+ 7, 1
65
+ 0, 4
66
+ 1, 4
67
+ 2, 4
68
+ 3, 4
69
+ 4, 4
70
+ 5, 4
71
+ 6, 4
72
+ 7, 4
73
+ 0, 821
74
+ 1, 821
75
+ 2, 821
76
+ 3, 821
77
+ 4, 821
78
+ 5, 821
79
+ 6, 821
80
+ 7, 821
81
+ 0, 1975
82
+ 1, 1975
83
+ 2, 1975
84
+ 3, 1975
85
+ 4, 1975
86
+ 5, 1975
87
+ 6, 1975
88
+ 7, 1975
89
+ 0, 5201
90
+ 1, 4985
91
+ 2, 4913
92
+ 3, 5249
93
+ 4, 5297
94
+ 5, 5417
95
+ 6, 4913
96
+ 7, 5153
97
+ 0, 47099
98
+ 1, 47099
99
+ 2, 47099
100
+ 3, 47099
101
+ 4, 47099
102
+ 5, 47099
103
+ 6, 47099
104
+ 7, 47099
105
+ 0, 56813
106
+ 1, 56811
107
+ 2, 56807
108
+ 3, 56801
109
+ 4, 56569
110
+ 5, 56851
111
+ 6, 56843
112
+ 7, 56875
113
+ 0, 58799
114
+ 1, 58797
115
+ 2, 58799
116
+ 3, 58797
117
+ 4, 58511
118
+ 5, 58797
119
+ 6, 58797
120
+ 7, 58799
121
+ 0, 59733
122
+ 1, 59731
123
+ 2, 59733
124
+ 3, 59731
125
+ 4, 59445
126
+ 5, 59731
127
+ 6, 59731
128
+ 7, 59733
129
+ 0, 59733
130
+ 1, 59733
131
+ 2, 59733
132
+ 3, 59733
133
+ 4, 59445
134
+ 5, 59733
135
+ 6, 59731
136
+ 7, 59733
tierb-fixed-moonep-full/moonep/tierb/tierb_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/rep2/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=12
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=0
64
+ kernel_trace=0
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/tierb/tierb_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/rep2/train.log ADDED
@@ -0,0 +1,1248 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [23:46:17] 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/tierb/tierb_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/rep2/memory_snapshots/snapshot.pickle
8
+ repo=/workspace/Megatron-MoE-Echo @ 7bc235bf
9
+ ==============================================================
10
+ W0829 23:46:19.086000 240032 torch/distributed/run.py:766]
11
+ W0829 23:46:19.086000 240032 torch/distributed/run.py:766] *****************************************
12
+ W0829 23:46:19.086000 240032 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
+ W0829 23:46:19.086000 240032 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
+ 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
19
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
20
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
21
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
22
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
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
+ 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
26
+ WARNING: overriding default arguments for tokenizer_type:GPT2BPETokenizer with tokenizer_type:GPT2BPETokenizer
27
+ Number of virtual stages per pipeline stage: None
28
+ accumulate and all-reduce gradients in fp32 for bfloat16 data type.
29
+ using torch.bfloat16 for parameters ...
30
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
31
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
32
+ ------------------------ arguments ------------------------
33
+ account_for_embedding_in_pipeline_split ......... False
34
+ account_for_loss_in_pipeline_split .............. False
35
+ accumulate_allreduce_grads_in_fp32 .............. True
36
+ adam_beta1 ...................................... 0.9
37
+ adam_beta2 ...................................... 0.999
38
+ adam_eps ........................................ 1e-08
39
+ add_bias_linear ................................. False
40
+ add_position_embedding .......................... False
41
+ add_qkv_bias .................................... False
42
+ adlr_autoresume ................................. False
43
+ adlr_autoresume_interval ........................ 1000
44
+ align_grad_reduce ............................... True
45
+ align_param_gather .............................. False
46
+ app_tag_run_name ................................ None
47
+ app_tag_run_version ............................. 0.0.0
48
+ apply_layernorm_1p .............................. False
49
+ apply_query_key_layer_scaling ................... False
50
+ apply_residual_connection_post_layernorm ........ False
51
+ apply_rope_fusion ............................... True
52
+ async_save ...................................... None
53
+ async_tensor_model_parallel_allreduce ........... True
54
+ attention_backend ............................... AttnBackend.auto
55
+ attention_dropout ............................... 0.0
56
+ attention_softmax_in_fp32 ....................... False
57
+ auto_detect_ckpt_format ......................... False
58
+ barrier_with_L1_time ............................ True
59
+ bert_binary_head ................................ True
60
+ bert_embedder_type .............................. megatron
61
+ bert_load ....................................... None
62
+ bf16 ............................................ True
63
+ bias_dropout_fusion ............................. True
64
+ bias_gelu_fusion ................................ False
65
+ bias_swiglu_fusion .............................. True
66
+ biencoder_projection_dim ........................ 0
67
+ biencoder_shared_query_context_model ............ False
68
+ block_data_path ................................. None
69
+ calc_ft_timeouts ................................ False
70
+ calculate_per_token_loss ........................ False
71
+ check_for_large_grads ........................... False
72
+ check_for_nan_in_loss_and_grad .................. False
73
+ check_for_spiky_loss ............................ False
74
+ check_weight_hash_across_dp_replicas_interval ... None
75
+ ckpt_assume_constant_structure .................. False
76
+ ckpt_convert_format ............................. None
77
+ ckpt_convert_save ............................... None
78
+ ckpt_convert_update_legacy_dist_opt_format ...... False
79
+ ckpt_format ..................................... torch_dist
80
+ ckpt_fully_parallel_load ........................ False
81
+ ckpt_fully_parallel_save ........................ True
82
+ ckpt_fully_parallel_save_deprecated ............. False
83
+ ckpt_step ....................................... None
84
+ classes_fraction ................................ 1.0
85
+ clip_grad ....................................... 1.0
86
+ clone_scatter_output_in_embedding ............... True
87
+ config_logger_dir ...............................
88
+ consumed_train_samples .......................... 0
89
+ consumed_valid_samples .......................... 0
90
+ context_parallel_size ........................... 1
91
+ cp_comm_type .................................... ['p2p']
92
+ create_attention_mask_in_dataloader ............. True
93
+ cross_entropy_fusion_impl ....................... native
94
+ cross_entropy_loss_fusion ....................... False
95
+ cuda_graph_scope ................................ full_iteration
96
+ cuda_graph_warmup_steps ......................... 3
97
+ data_args_path .................................. None
98
+ data_cache_path ................................. None
99
+ data_parallel_random_init ....................... False
100
+ data_parallel_sharding_strategy ................. no_shard
101
+ data_parallel_size .............................. 8
102
+ data_path ....................................... ['/workspace/data/wikitext/wikitext103_gpt2_text_document']
103
+ data_per_class_fraction ......................... 1.0
104
+ data_sharding ................................... True
105
+ dataloader_type ................................. single
106
+ ddp_average_in_collective ....................... False
107
+ ddp_bucket_size ................................. None
108
+ ddp_num_buckets ................................. None
109
+ ddp_pad_buckets_for_high_nccl_busbw ............. False
110
+ decoder_first_pipeline_num_layers ............... None
111
+ decoder_last_pipeline_num_layers ................ None
112
+ decoder_num_layers .............................. None
113
+ decoder_seq_length .............................. None
114
+ decoupled_lr .................................... None
115
+ decoupled_min_lr ................................ None
116
+ decrease_batch_size_if_needed ................... False
117
+ defer_embedding_wgrad_compute ................... False
118
+ delay_wgrad_compute ............................. False
119
+ deprecated_use_mcore_models ..................... True
120
+ deterministic_mode .............................. False
121
+ dino_bottleneck_size ............................ 256
122
+ dino_freeze_last_layer .......................... 1
123
+ dino_head_hidden_size ........................... 2048
124
+ dino_local_crops_number ......................... 10
125
+ dino_local_img_size ............................. 96
126
+ dino_norm_last_layer ............................ False
127
+ dino_teacher_temp ............................... 0.07
128
+ dino_warmup_teacher_temp ........................ 0.04
129
+ dino_warmup_teacher_temp_epochs ................. 30
130
+ disable_bf16_reduced_precision_matmul ........... False
131
+ disable_mamba_mem_eff_path ...................... False
132
+ disable_straggler_on_startup .................... False
133
+ dist_ckpt_format_deprecated ..................... None
134
+ dist_ckpt_strictness ............................ assume_ok_unexpected
135
+ distribute_saved_activations .................... False
136
+ distributed_backend ............................. nccl
137
+ distributed_timeout_minutes ..................... 10
138
+ embedding_init_method_std ....................... None
139
+ embedding_path .................................. None
140
+ empty_unused_memory_level ....................... 0
141
+ enable_cuda_graph ............................... True
142
+ enable_experimental ............................. False
143
+ enable_ft_package ............................... False
144
+ enable_gloo_process_groups ...................... True
145
+ enable_msc ...................................... True
146
+ enable_one_logger ............................... True
147
+ encoder_num_layers .............................. 12
148
+ encoder_seq_length .............................. 8192
149
+ end_weight_decay ................................ 0.1
150
+ eod_mask_loss ................................... False
151
+ error_injection_rate ............................ 0
152
+ error_injection_type ............................ transient_error
153
+ eval_interval ................................... 1000000
154
+ eval_iters ...................................... 0
155
+ evidence_data_path .............................. None
156
+ exit_duration_in_mins ........................... None
157
+ exit_interval ................................... None
158
+ exit_on_missing_checkpoint ...................... False
159
+ exit_signal_handler ............................. False
160
+ exp_avg_dtype ................................... torch.float32
161
+ exp_avg_sq_dtype ................................ torch.float32
162
+ expert_model_parallel_size ...................... 8
163
+ expert_tensor_parallel_size ..................... 1
164
+ export_force_local_attention .................... False
165
+ export_kd_cfg ................................... None
166
+ export_kd_teacher_ckpt_format ................... None
167
+ export_kd_teacher_load .......................... None
168
+ export_kv_cache_quant ........................... False
169
+ export_legacy_megatron .......................... False
170
+ export_model_type ............................... GPTModel
171
+ export_moe_apply_probs_on_input ................. False
172
+ export_qk_l2_norm ............................... False
173
+ export_quant_cfg ................................ None
174
+ export_real_quant_cfg ........................... None
175
+ export_te_mcore_model ........................... False
176
+ external_cuda_graph ............................. False
177
+ ffn_hidden_size ................................. 12288
178
+ finetune ........................................ False
179
+ finetune_data_split ............................. train
180
+ finetune_hf_dataset ............................. None
181
+ first_last_layers_bf16 .......................... False
182
+ flash_decode .................................... False
183
+ fp16 ............................................ False
184
+ fp16_lm_cross_entropy ........................... False
185
+ fp32_residual_connection ........................ False
186
+ fp8 ............................................. e4m3
187
+ fp8_amax_compute_algo ........................... most_recent
188
+ fp8_amax_history_len ............................ 1
189
+ fp8_interval .................................... 1
190
+ fp8_margin ...................................... 0
191
+ fp8_param_gather ................................ False
192
+ fp8_recipe ...................................... mxfp8
193
+ fp8_wgrad ....................................... True
194
+ fsdp_double_buffer .............................. False
195
+ global_batch_size ............................... 8
196
+ grad_reduce_in_bf16 ............................. False
197
+ gradient_accumulation_fusion .................... True
198
+ gradient_reduce_div_fusion ...................... True
199
+ group_query_attention ........................... True
200
+ head_lr_mult .................................... 1.0
201
+ heterogeneous_layers_config_encoded_json ........ None
202
+ heterogeneous_layers_config_path ................ None
203
+ hidden_dropout .................................. 0.0
204
+ hidden_size ..................................... 4096
205
+ hierarchical_context_parallel_sizes ............. None
206
+ high_priority_stream_groups ..................... []
207
+ hybrid_attention_ratio .......................... 0.0
208
+ hybrid_mlp_ratio ................................ 0.0
209
+ hybrid_override_pattern ......................... None
210
+ hysteresis ...................................... 2
211
+ ict_head_size ................................... None
212
+ ict_load ........................................ None
213
+ img_h ........................................... 224
214
+ img_w ........................................... 224
215
+ indexer_batch_size .............................. 128
216
+ indexer_log_interval ............................ 1000
217
+ inference_batch_times_seqlen_threshold .......... -1
218
+ inference_dynamic_batching ...................... False
219
+ inference_dynamic_batching_buffer_guaranteed_fraction 0.2
220
+ inference_dynamic_batching_buffer_overflow_factor None
221
+ inference_dynamic_batching_buffer_size_gb ....... 40.0
222
+ inference_dynamic_batching_chunk_size ........... 256
223
+ inference_dynamic_batching_max_requests_override None
224
+ inference_dynamic_batching_max_tokens_override .. None
225
+ inference_dynamic_batching_num_cuda_graphs ...... 16
226
+ inference_max_batch_size ........................ 8
227
+ inference_max_seq_length ........................ 2560
228
+ inference_rng_tracker ........................... False
229
+ init_method_std ................................. 0.01
230
+ init_method_xavier_uniform ...................... False
231
+ init_model_with_meta_device ..................... False
232
+ initial_loss_scale .............................. 4294967296
233
+ inprocess_active_world_size ..................... 8
234
+ inprocess_barrier_timeout ....................... 120
235
+ inprocess_completion_timeout .................... 120
236
+ inprocess_empty_cuda_cache ...................... False
237
+ inprocess_granularity ........................... node
238
+ inprocess_hard_timeout .......................... 90
239
+ inprocess_heartbeat_interval .................... 30
240
+ inprocess_heartbeat_timeout ..................... 60
241
+ inprocess_last_call_wait ........................ 1
242
+ inprocess_max_iterations ........................ None
243
+ inprocess_monitor_process_interval .............. 1.0
244
+ inprocess_monitor_thread_interval ............... 1.0
245
+ inprocess_progress_watchdog_interval ............ 1.0
246
+ inprocess_restart ............................... False
247
+ inprocess_soft_timeout .......................... 60
248
+ inprocess_termination_grace_time ................ 1
249
+ is_hybrid_model ................................. False
250
+ iter_per_epoch .................................. 1250
251
+ iterations_to_skip .............................. []
252
+ keep_fp8_transpose_cache_when_using_custom_fsdp . False
253
+ kitchen_config_file ............................. None
254
+ kitchen_recipe_number ........................... None
255
+ kv_channels ..................................... 128
256
+ kv_lora_rank .................................... 32
257
+ lazy_mpu_init ................................... None
258
+ load ............................................ None
259
+ load_main_params_from_ckpt ...................... None
260
+ load_model_opt_format ........................... False
261
+ local_rank ...................................... 0
262
+ log_energy ...................................... False
263
+ log_interval .................................... 1
264
+ log_loss_scale_to_tensorboard ................... True
265
+ log_memory_to_tensorboard ....................... False
266
+ log_num_zeros_in_grad ........................... False
267
+ log_params_norm ................................. False
268
+ log_progress .................................... False
269
+ log_straggler ................................... False
270
+ log_throughput .................................. True
271
+ log_timers_to_tensorboard ....................... False
272
+ log_validation_ppl_to_tensorboard ............... False
273
+ log_world_size_to_tensorboard ................... False
274
+ logging_level ................................... None
275
+ loss_scale ...................................... None
276
+ loss_scale_window ............................... 1000
277
+ lr .............................................. 0.0001
278
+ lr_decay_iters .................................. 8
279
+ lr_decay_samples ................................ None
280
+ lr_decay_style .................................. cosine
281
+ lr_warmup_fraction .............................. None
282
+ lr_warmup_init .................................. 0.0
283
+ lr_warmup_iters ................................. 5
284
+ lr_warmup_samples ............................... 0
285
+ lr_wsd_decay_iters .............................. None
286
+ lr_wsd_decay_samples ............................ None
287
+ lr_wsd_decay_style .............................. exponential
288
+ main_grads_dtype ................................ torch.float32
289
+ main_params_dtype ............................... torch.float32
290
+ make_vocab_size_divisible_by .................... 128
291
+ mamba_head_dim .................................. 64
292
+ mamba_num_groups ................................ 8
293
+ mamba_num_heads ................................. None
294
+ mamba_state_dim ................................. 128
295
+ manual_gc ....................................... False
296
+ manual_gc_eval .................................. True
297
+ manual_gc_interval .............................. 0
298
+ mask_factor ..................................... 1.0
299
+ mask_prob ....................................... 0.15
300
+ mask_type ....................................... random
301
+ masked_softmax_fusion ........................... False
302
+ max_position_embeddings ......................... 32768
303
+ max_tokens_to_oom ............................... 12000
304
+ memory_snapshot_path ............................ snapshot.pickle
305
+ merge_file ...................................... /workspace/data/wikitext/gpt2-merges.txt
306
+ micro_batch_size ................................ 1
307
+ microbatch_group_size_per_vp_stage .............. None
308
+ mid_level_dataset_surplus ....................... 0.005
309
+ min_loss_scale .................................. 1.0
310
+ min_lr .......................................... 1e-05
311
+ mlp_chunks_for_prefill .......................... 1
312
+ mmap_bin_files .................................. True
313
+ mock_data ....................................... False
314
+ moe_apply_probs_on_input ........................ False
315
+ moe_aux_loss_coeff .............................. 0.001
316
+ moe_deepep_num_sms .............................. 20
317
+ moe_echo_enable_random_offloading ............... False
318
+ moe_echo_expert_dispatch_overlap ................ False
319
+ moe_echo_expert_dispatcher_type ................. hybridep
320
+ moe_echo_recompute_expert_dispatch .............. False
321
+ moe_enable_deepep ............................... False
322
+ moe_enable_echo ................................. False
323
+ moe_enable_hybridep ............................. False
324
+ moe_enable_moonep ............................... True
325
+ moe_expert_capacity_factor ...................... None
326
+ moe_extended_tp ................................. False
327
+ moe_ffn_hidden_size ............................. 1536
328
+ moe_grouped_gemm ................................ True
329
+ moe_hybridep_num_sms ............................ 16
330
+ moe_input_jitter_eps ............................ None
331
+ moe_layer_freq .................................. 1
332
+ moe_layer_recompute ............................. False
333
+ moe_moonep_async_grad_reduce .................... True
334
+ moe_moonep_num_sms .............................. None
335
+ moe_num_echo_experts ............................ None
336
+ moe_pad_expert_input_to_capacity ................ False
337
+ moe_per_layer_logging ........................... False
338
+ moe_permute_fusion .............................. False
339
+ moe_received_token_capacity ..................... None
340
+ moe_router_bias_update_rate ..................... 0.001
341
+ moe_router_dtype ................................ fp32
342
+ moe_router_enable_expert_bias ................... False
343
+ moe_router_force_load_balancing ................. False
344
+ moe_router_group_topk ........................... None
345
+ moe_router_load_balancing_type .................. aux_loss
346
+ moe_router_num_groups ........................... None
347
+ moe_router_padding_for_fp8 ...................... False
348
+ moe_router_pre_softmax .......................... False
349
+ moe_router_score_function ....................... softmax
350
+ moe_router_topk ................................. 8
351
+ moe_router_topk_scaling_factor .................. None
352
+ moe_shared_expert_intermediate_size ............. None
353
+ moe_shared_expert_overlap ....................... False
354
+ moe_token_dispatcher_type ....................... flex
355
+ moe_token_drop_policy ........................... probs
356
+ moe_upcycling_granularity ....................... 1
357
+ moe_use_device_initiated_grouped_gemm ........... True
358
+ moe_use_legacy_grouped_gemm ..................... False
359
+ moe_use_upcycling ............................... False
360
+ moe_z_loss_coeff ................................ None
361
+ mrope_section ................................... None
362
+ mscale .......................................... 1.0
363
+ mscale_all_dim .................................. 0.0
364
+ mtp_loss_scaling_factor ......................... 0.1
365
+ mtp_num_layers .................................. None
366
+ multi_latent_attention .......................... False
367
+ nccl_all_reduce_for_prefill ..................... False
368
+ nccl_communicator_config_path ................... None
369
+ nccl_ub ......................................... False
370
+ no_load_optim ................................... None
371
+ no_load_rng ..................................... None
372
+ no_persist_layer_norm ........................... False
373
+ no_rope_freq .................................... None
374
+ no_save_optim ................................... None
375
+ no_save_rng ..................................... None
376
+ non_persistent_ckpt_type ........................ None
377
+ non_persistent_global_ckpt_dir .................. None
378
+ non_persistent_local_ckpt_algo .................. fully_parallel
379
+ non_persistent_local_ckpt_dir ................... None
380
+ non_persistent_save_interval .................... None
381
+ norm_epsilon .................................... 1e-06
382
+ normalization ................................... RMSNorm
383
+ num_attention_heads ............................. 64
384
+ num_channels .................................... 3
385
+ num_classes ..................................... 1000
386
+ num_dataset_builder_threads ..................... 1
387
+ num_distributed_optimizer_instances ............. 1
388
+ num_experts ..................................... 128
389
+ num_layers ...................................... 12
390
+ num_layers_at_end_in_bf16 ....................... 1
391
+ num_layers_at_start_in_bf16 ..................... 1
392
+ num_layers_per_virtual_pipeline_stage ........... None
393
+ num_query_groups ................................ 4
394
+ num_virtual_stages_per_pipeline_rank ............ None
395
+ num_workers ..................................... 2
396
+ object_storage_cache_path ....................... None
397
+ one_logger_async ................................ False
398
+ one_logger_project .............................. megatron-lm
399
+ one_logger_run_name ............................. None
400
+ onnx_safe ....................................... None
401
+ openai_gelu ..................................... False
402
+ optimizer ....................................... adam
403
+ optimizer_cpu_offload ........................... False
404
+ optimizer_offload_fraction ...................... 1.0
405
+ output_bert_embeddings .......................... False
406
+ overlap_cpu_optimizer_d2h_h2d ................... False
407
+ overlap_grad_reduce ............................. True
408
+ overlap_moe_expert_parallel_comm ................ False
409
+ overlap_p2p_comm ................................ False
410
+ overlap_p2p_comm_warmup_flush ................... False
411
+ overlap_param_gather ............................ True
412
+ overlap_param_gather_with_optimizer_step ........ False
413
+ override_opt_param_scheduler .................... False
414
+ params_dtype .................................... torch.bfloat16
415
+ patch_dim ....................................... 16
416
+ per_split_data_args_path ........................ None
417
+ perform_initialization .......................... True
418
+ pin_cpu_grads ................................... True
419
+ pin_cpu_params .................................. True
420
+ pipeline_model_parallel_comm_backend ............ None
421
+ pipeline_model_parallel_layout .................. None
422
+ pipeline_model_parallel_size .................... 1
423
+ position_embedding_type ......................... rope
424
+ pretrained_checkpoint ........................... None
425
+ profile ......................................... False
426
+ profile_ranks ................................... [0]
427
+ profile_step_end ................................ 12
428
+ profile_step_start .............................. 10
429
+ q_lora_rank ..................................... None
430
+ qk_head_dim ..................................... 128
431
+ qk_l2_norm ...................................... False
432
+ qk_layernorm .................................... True
433
+ qk_pos_emb_head_dim ............................. 64
434
+ query_in_block_prob ............................. 0.1
435
+ rampup_batch_size ............................... None
436
+ rank ............................................ 0
437
+ recompute_granularity ........................... None
438
+ recompute_method ................................ None
439
+ recompute_modules ............................... None
440
+ recompute_num_layers ............................ None
441
+ record_memory_history ........................... False
442
+ relative_attention_max_distance ................. 128
443
+ relative_attention_num_buckets .................. 32
444
+ replication ..................................... False
445
+ replication_factor .............................. 2
446
+ replication_jump ................................ None
447
+ rerun_mode ...................................... validate_results
448
+ reset_attention_mask ............................ False
449
+ reset_position_ids .............................. False
450
+ result_rejected_tracker_filename ................ None
451
+ retriever_report_topk_accuracies ................ []
452
+ retriever_score_scaling ......................... False
453
+ retriever_seq_length ............................ 256
454
+ retro_add_retriever ............................. False
455
+ retro_attention_gate ............................ 1
456
+ retro_cyclic_train_iters ........................ None
457
+ retro_encoder_attention_dropout ................. 0.1
458
+ retro_encoder_hidden_dropout .................... 0.1
459
+ retro_encoder_layers ............................ 2
460
+ retro_num_neighbors ............................. 2
461
+ retro_num_retrieved_chunks ...................... 2
462
+ retro_project_dir ............................... None
463
+ retro_verify_neighbor_count ..................... True
464
+ reuse_grad_buf_for_mxfp8_param_ag ............... False
465
+ rope_scaling_factor ............................. 8.0
466
+ rope_type ....................................... None
467
+ rotary_base ..................................... 1000000
468
+ rotary_interleaved .............................. False
469
+ rotary_percent .................................. 1.0
470
+ rotary_scaling_factor ........................... 1.0
471
+ rotary_seq_len_interpolation_factor ............. None
472
+ run_workload_inspector_server ................... False
473
+ sample_rate ..................................... 1.0
474
+ save ............................................ None
475
+ save_interval ................................... None
476
+ save_retain_interval ............................ None
477
+ scatter_gather_tensors_in_pipeline .............. True
478
+ seed ............................................ 1234
479
+ seq_length ...................................... 8192
480
+ sequence_parallel ............................... False
481
+ sft ............................................. False
482
+ sft_tokenizer_prompt_format ..................... nemotron-h-aligned
483
+ sgd_momentum .................................... 0.9
484
+ short_seq_prob .................................. 0.1
485
+ skip_train ...................................... False
486
+ skipped_train_samples ........................... 0
487
+ spec ............................................ None
488
+ split ........................................... 100,0,0
489
+ squared_relu .................................... False
490
+ start_weight_decay .............................. 0.1
491
+ straggler_ctrlr_port ............................ 65535
492
+ straggler_minmax_count .......................... 1
493
+ suggested_communication_unit_size ............... None
494
+ swiglu .......................................... True
495
+ swin_backbone_type .............................. tiny
496
+ symmetric_ar_type ............................... None
497
+ te_rng_tracker .................................. True
498
+ tensor_model_parallel_size ...................... 1
499
+ tensorboard_dir ................................. None
500
+ tensorboard_log_interval ........................ 1
501
+ tensorboard_queue_size .......................... 1000
502
+ test_data_path .................................. None
503
+ test_mode ....................................... False
504
+ tiktoken_num_special_tokens ..................... 1000
505
+ tiktoken_pattern ................................ None
506
+ tiktoken_special_tokens ......................... None
507
+ timing_log_level ................................ 0
508
+ timing_log_option ............................... minmax
509
+ titles_data_path ................................ None
510
+ tokenizer_model ................................. None
511
+ tokenizer_type .................................. GPT2BPETokenizer
512
+ torch_fsdp2_reshard_after_forward ............... True
513
+ tp_comm_bootstrap_backend ....................... nccl
514
+ tp_comm_bulk_dgrad .............................. True
515
+ tp_comm_bulk_wgrad .............................. True
516
+ tp_comm_overlap ................................. False
517
+ tp_comm_overlap_ag .............................. True
518
+ tp_comm_overlap_cfg ............................. None
519
+ tp_comm_overlap_rs .............................. True
520
+ tp_comm_overlap_rs_dgrad ........................ False
521
+ tp_comm_split_ag ................................ True
522
+ tp_comm_split_rs ................................ True
523
+ train_data_path ................................. None
524
+ train_iters ..................................... 12
525
+ train_samples ................................... None
526
+ train_sync_interval ............................. None
527
+ transformer_impl ................................ transformer_engine
528
+ transformer_pipeline_model_parallel_size ........ 1
529
+ untie_embeddings_and_output_weights ............. True
530
+ use_checkpoint_args ............................. False
531
+ use_checkpoint_opt_param_scheduler .............. False
532
+ use_cpu_initialization .......................... None
533
+ use_custom_fsdp ................................. False
534
+ use_dist_ckpt ................................... True
535
+ use_dist_ckpt_deprecated ........................ False
536
+ use_distributed_optimizer ....................... True
537
+ use_flash_attn .................................. False
538
+ use_legacy_models ............................... False
539
+ use_mp_args_from_checkpoint_args ................ False
540
+ use_one_sent_docs ............................... False
541
+ use_persistent_ckpt_worker ...................... False
542
+ use_precision_aware_optimizer ................... False
543
+ use_pytorch_profiler ............................ False
544
+ use_ring_exchange_p2p ........................... False
545
+ use_rope_scaling ................................ False
546
+ use_rotary_position_embeddings .................. False
547
+ use_sharp ....................................... False
548
+ use_tokenizer_model_from_checkpoint_args ........ True
549
+ use_torch_fsdp2 ................................. False
550
+ use_torch_optimizer_for_cpu_offload ............. False
551
+ use_tp_pp_dp_mapping ............................ False
552
+ v_head_dim ...................................... 128
553
+ valid_data_path ................................. None
554
+ variable_seq_lengths ............................ False
555
+ virtual_pipeline_model_parallel_size ............ None
556
+ vision_backbone_type ............................ vit
557
+ vision_pretraining .............................. False
558
+ vision_pretraining_type ......................... classify
559
+ vocab_extra_ids ................................. 0
560
+ vocab_file ...................................... /workspace/data/wikitext/gpt2-vocab.json
561
+ vocab_size ...................................... None
562
+ wandb_exp_name ..................................
563
+ wandb_project ...................................
564
+ wandb_save_dir ..................................
565
+ weight_decay .................................... 0.1
566
+ weight_decay_incr_style ......................... constant
567
+ wgrad_deferral_limit ............................ 0
568
+ world_size ...................................... 8
569
+ yaml_cfg ........................................ None
570
+ -------------------- end of arguments ---------------------
571
+ INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 1
572
+ > building GPT2BPETokenizer tokenizer ...
573
+ > padded vocab (size: 50257) with 47 dummy tokens (new size: 50304)
574
+ WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.VALIDATE_RESULTS
575
+ > initializing torch distributed ...
576
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
577
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
578
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
579
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
580
+ [Gloo] Rank [Gloo] Rank 20 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
581
+
582
+ [Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
583
+ [Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
584
+ [Gloo] Rank [Gloo] Rank [Gloo] Rank [Gloo] Rank 3657 is connected to is connected to is connected to is connected to 7777 peer ranks. peer ranks. peer ranks. peer ranks. Expected number of connected peer ranks is : Expected number of connected peer ranks is : Expected number of connected peer ranks is : Expected number of connected peer ranks is : 7777
585
+
586
+
587
+
588
+ [Gloo] Rank 0 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 5 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 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
593
+ [Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
594
+ [Gloo] Rank [Gloo] Rank 7 is connected to 67 is connected to peer ranks. 7Expected number of connected peer ranks is : peer ranks. 7Expected 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
+ > initialized tensor model parallel with size 1
598
+ > initialized pipeline model parallel with size 1
599
+ > setting random seeds to 1234 ...
600
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
601
+ [Gloo] Rank 0 is connected to 0 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
+ [Gloo] Rank [Gloo] Rank 00 is connected to is connected to 00 peer ranks. peer ranks. Expected number of connected peer ranks is : Expected number of connected peer ranks is : 00
605
+
606
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
607
+ [rank4]:[W829 23:46:37.521085384 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
608
+ [rank5]:[W829 23:46:37.521095894 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
609
+ [rank2]:[W829 23:46:37.521112608 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.
610
+ [rank1]:[W829 23:46:37.521143928 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.
611
+ [rank6]:[W829 23:46:37.522800573 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.
612
+ [rank7]:[W829 23:46:37.522847520 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.
613
+ [rank3]:[W829 23:46:37.522872548 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.
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: 0.035 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]:[W829 23:46:37.680689063 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: 3.980 seconds
623
+ time to initialize megatron (seconds): 7.571
624
+ [after megatron is initialized] datetime: 2026-08-29 23:46:43
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.pre_mlp_layernorm.weight
649
+ module.decoder.layers.11.self_attention.q_layernorm.weight
650
+ module.decoder.layers.11.mlp.router.weight
651
+ module.decoder.layers.11.self_attention.linear_qkv.weight
652
+ module.decoder.layers.11.self_attention.linear_qkv.layer_norm_weight
653
+ module.decoder.final_layernorm.weight
654
+ module.decoder.layers.11.self_attention.k_layernorm.weight
655
+ module.decoder.layers.11.self_attention.linear_proj.weight
656
+ Params for bucket 3 (71835904 elements, 71835904 padded size):
657
+ module.decoder.layers.10.mlp.router.weight
658
+ module.decoder.layers.10.pre_mlp_layernorm.weight
659
+ module.decoder.layers.10.self_attention.q_layernorm.weight
660
+ module.decoder.layers.10.self_attention.linear_qkv.weight
661
+ module.decoder.layers.10.self_attention.k_layernorm.weight
662
+ module.decoder.layers.10.self_attention.linear_qkv.layer_norm_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.mlp.router.weight
666
+ module.decoder.layers.9.pre_mlp_layernorm.weight
667
+ module.decoder.layers.9.self_attention.q_layernorm.weight
668
+ module.decoder.layers.9.self_attention.linear_qkv.layer_norm_weight
669
+ module.decoder.layers.9.self_attention.linear_qkv.weight
670
+ module.decoder.layers.9.self_attention.k_layernorm.weight
671
+ module.decoder.layers.9.self_attention.linear_proj.weight
672
+ Params for bucket 5 (71835904 elements, 71835904 padded size):
673
+ module.decoder.layers.8.mlp.router.weight
674
+ module.decoder.layers.8.self_attention.k_layernorm.weight
675
+ module.decoder.layers.8.self_attention.q_layernorm.weight
676
+ module.decoder.layers.8.pre_mlp_layernorm.weight
677
+ module.decoder.layers.8.self_attention.linear_qkv.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.q_layernorm.weight
685
+ module.decoder.layers.7.self_attention.linear_qkv.layer_norm_weight
686
+ module.decoder.layers.7.self_attention.linear_qkv.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.linear_qkv.weight
691
+ module.decoder.layers.6.self_attention.k_layernorm.weight
692
+ module.decoder.layers.6.self_attention.linear_qkv.layer_norm_weight
693
+ module.decoder.layers.6.pre_mlp_layernorm.weight
694
+ module.decoder.layers.6.self_attention.q_layernorm.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.mlp.router.weight
698
+ module.decoder.layers.5.self_attention.linear_qkv.weight
699
+ module.decoder.layers.5.self_attention.q_layernorm.weight
700
+ module.decoder.layers.5.pre_mlp_layernorm.weight
701
+ module.decoder.layers.5.self_attention.linear_qkv.layer_norm_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.linear_qkv.weight
707
+ module.decoder.layers.4.self_attention.q_layernorm.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.q_layernorm.weight
716
+ module.decoder.layers.3.self_attention.linear_qkv.weight
717
+ module.decoder.layers.3.self_attention.k_layernorm.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.pre_mlp_layernorm.weight
723
+ module.decoder.layers.2.self_attention.linear_qkv.weight
724
+ module.decoder.layers.2.self_attention.q_layernorm.weight
725
+ module.decoder.layers.2.self_attention.k_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.pre_mlp_layernorm.weight
731
+ module.decoder.layers.1.self_attention.q_layernorm.weight
732
+ module.decoder.layers.1.self_attention.linear_qkv.weight
733
+ module.decoder.layers.1.self_attention.k_layernorm.weight
734
+ module.decoder.layers.1.self_attention.linear_qkv.layer_norm_weight
735
+ module.decoder.layers.1.self_attention.linear_proj.weight
736
+ Params for bucket 13 (71835904 elements, 71835904 padded size):
737
+ module.decoder.layers.0.self_attention.q_layernorm.weight
738
+ module.decoder.layers.0.pre_mlp_layernorm.weight
739
+ module.decoder.layers.0.self_attention.linear_qkv.layer_norm_weight
740
+ module.decoder.layers.0.mlp.router.weight
741
+ module.decoder.layers.0.self_attention.linear_qkv.weight
742
+ module.decoder.layers.0.self_attention.k_layernorm.weight
743
+ module.decoder.layers.0.self_attention.linear_proj.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.weight1
753
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight3
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.weight8
757
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight10
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.weight11
761
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight9
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_fc1.weight2
767
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight15
768
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight0
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.weight4
772
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight5
773
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight6
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.weight8
777
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight10
778
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight9
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.weight14
782
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight13
783
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight12
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.weight3
787
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight1
788
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight0
789
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight2
790
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight4
791
+ Params for bucket 8 (44040192 elements, 44040192 padded size):
792
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight5
793
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight7
794
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight9
795
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight11
796
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight8
797
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight6
798
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight10
799
+ Params for bucket 9 (50331648 elements, 50331648 padded size):
800
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight13
801
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight0
802
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight15
803
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight12
804
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight14
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.weight4
809
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight5
810
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight3
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.weight7
814
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight8
815
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight9
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.weight13
819
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight11
820
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight12
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.weight1
824
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight2
825
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight0
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.weight3
829
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight5
830
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight7
831
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight9
832
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight6
833
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight4
834
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight8
835
+ Params for bucket 15 (50331648 elements, 50331648 padded size):
836
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight13
837
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight11
838
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight15
839
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight0
840
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight10
841
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight14
842
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight12
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.8.mlp.experts.linear_fc2.weight0
861
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight14
862
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight15
863
+ Params for bucket 20 (44040192 elements, 44040192 padded size):
864
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight4
865
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight2
866
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight6
867
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight1
868
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight5
869
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight3
870
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight7
871
+ Params for bucket 21 (44040192 elements, 44040192 padded size):
872
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight8
873
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight10
874
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight12
875
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight14
876
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight11
877
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight9
878
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight13
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.weight1
882
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight2
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.8.mlp.experts.linear_fc1.weight15
901
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight2
902
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight0
903
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight4
904
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight3
905
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight1
906
+ Params for bucket 27 (44040192 elements, 44040192 padded size):
907
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight8
908
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight6
909
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight10
910
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight5
911
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight7
912
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight9
913
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight11
914
+ Params for bucket 28 (50331648 elements, 50331648 padded size):
915
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight12
916
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight0
917
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight14
918
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight15
919
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight13
920
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight1
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.weight8
929
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight9
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.weight13
934
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight12
935
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight11
936
+ Params for bucket 32 (44040192 elements, 44040192 padded size):
937
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight14
938
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight2
939
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight0
940
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight1
941
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight15
942
+ Params for bucket 33 (44040192 elements, 44040192 padded size):
943
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight4
944
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight6
945
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight8
946
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight3
947
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight7
948
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight5
949
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight9
950
+ Params for bucket 34 (50331648 elements, 50331648 padded size):
951
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight10
952
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight12
953
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight14
954
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight11
955
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight13
956
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight15
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.weight7
966
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight8
967
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight6
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.weight11
972
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight12
973
+ Params for bucket 38 (44040192 elements, 44040192 padded size):
974
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight13
975
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight0
976
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight14
977
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight15
978
+ Params for bucket 39 (44040192 elements, 44040192 padded size):
979
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight1
980
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight3
981
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight5
982
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight7
983
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight4
984
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight2
985
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight6
986
+ Params for bucket 40 (44040192 elements, 44040192 padded size):
987
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight9
988
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight11
989
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight13
990
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight8
991
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight10
992
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight12
993
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight14
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.weight0
997
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight1
998
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight2
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.weight6
1002
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight4
1003
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight5
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.weight10
1007
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight8
1008
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight9
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.weight13
1012
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight14
1013
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight12
1014
+ Params for bucket 45 (44040192 elements, 44040192 padded size):
1015
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight2
1016
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight0
1017
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight4
1018
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight15
1019
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight3
1020
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight1
1021
+ Params for bucket 46 (44040192 elements, 44040192 padded size):
1022
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight8
1023
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight6
1024
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight10
1025
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight5
1026
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight9
1027
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight7
1028
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight11
1029
+ Params for bucket 47 (50331648 elements, 50331648 padded size):
1030
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight12
1031
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight14
1032
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight1
1033
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight0
1034
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight15
1035
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight13
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.weight4
1039
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight5
1040
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight3
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.weight7
1044
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight8
1045
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight9
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.weight11
1049
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight12
1050
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight13
1051
+ Params for bucket 51 (44040192 elements, 44040192 padded size):
1052
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight1
1053
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight15
1054
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight14
1055
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight2
1056
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight0
1057
+ Params for bucket 52 (44040192 elements, 44040192 padded size):
1058
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight3
1059
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight7
1060
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight5
1061
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight9
1062
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight6
1063
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight4
1064
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight8
1065
+ Params for bucket 53 (50331648 elements, 50331648 padded size):
1066
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight13
1067
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight11
1068
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight15
1069
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight0
1070
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight10
1071
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight14
1072
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight12
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.weight4
1076
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight2
1077
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight3
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.weight7
1081
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight8
1082
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight6
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.weight10
1086
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight11
1087
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight12
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.weight14
1091
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight0
1092
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight15
1093
+ Params for bucket 58 (44040192 elements, 44040192 padded size):
1094
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight1
1095
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight3
1096
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight5
1097
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight7
1098
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight2
1099
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight4
1100
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight6
1101
+ Params for bucket 59 (44040192 elements, 44040192 padded size):
1102
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight11
1103
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight9
1104
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight13
1105
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight8
1106
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight12
1107
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight10
1108
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight14
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.weight2
1112
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight0
1113
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight1
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.weight5
1117
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight6
1118
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight4
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.weight9
1122
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight10
1123
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight8
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.weight12
1127
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight13
1128
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight14
1129
+ Params for bucket 64 (44040192 elements, 44040192 padded size):
1130
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight2
1131
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight0
1132
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight4
1133
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight15
1134
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight1
1135
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight3
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.weight0
1148
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight15
1149
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight13
1150
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight1
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.weight3
1154
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight4
1155
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight5
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.weight9
1159
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight7
1160
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight8
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.weight12
1164
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight13
1165
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight11
1166
+ Params for bucket 70 (44040192 elements, 44040192 padded size):
1167
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight14
1168
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight0
1169
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight2
1170
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight1
1171
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight15
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.weight7
1178
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight5
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_fc1.weight0
1185
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight11
1186
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight13
1187
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight15
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.weight4
1192
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight2
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.weight14
1206
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight15
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 0x70bf386b5010>, 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: 2026-08-29 23:46:46
1210
+ > building train, validation, and test datasets ...
1211
+ > datasets target sizes (minimum size):
1212
+ train: 96
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=(96, 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 0x70bf389001a0>, 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 eadc3756447e39d609e8aa728a91f451-GPTDataset-train-document_index.npy
1226
+ INFO:megatron.core.datasets.gpt_dataset: Load the sample index from eadc3756447e39d609e8aa728a91f451-GPTDataset-train-sample_index.npy
1227
+ INFO:megatron.core.datasets.gpt_dataset: Load the shuffle index from eadc3756447e39d609e8aa728a91f451-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: 2026-08-29 23:46:47
1231
+ done with setup ...
1232
+ (min, max) time across ranks (ms):
1233
+ model-and-optimizer-setup ......................: (3290.03, 3304.28)
1234
+ train/valid/test-data-iterators-setup ..........: (63.09, 81.05)
1235
+ training ...
1236
+ Overwriting rerun_state_machine.current_iteration from -1 to 0...
1237
+ [before the start of training step] datetime: 2026-08-29 23:46:47
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]]
tierb-fixed-moonep-full/moonep/tierb/tierb_gbs8/results.tsv ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ config backend policy slots seq gbs cap recompute rep status iter_ms_mean iter_ms_min peak_mib run_dir
2
+ QWEN3_235B moonep matched 16 8192 8 none none 1 OK 321.60 320.40 120303 /workspace/tierb-fixed-moonep-full/moonep/tierb/tierb_gbs8/QWEN3_235B__moonep__matched__seq8192__gbs8__capnone__async/rep1