AbhishekGhosh1998 commited on
Commit
d1f1ecf
·
verified ·
1 Parent(s): 9d350fe

granularity results granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb @ 2026-09-21T08:40:19Z from sp-aiinfra-interactive-aghosh-8-mb99k-8xb200-moe-cg-0

Browse files
Files changed (13) hide show
  1. granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/manifest.txt +19 -0
  2. granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/pip_freeze.txt +317 -0
  3. granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/provenance.txt +90 -0
  4. granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep1/mem_samples.csv +200 -0
  5. granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep1/run.env +102 -0
  6. granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep1/train.log +1258 -0
  7. granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep2/mem_samples.csv +208 -0
  8. granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep2/run.env +102 -0
  9. granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep2/train.log +1258 -0
  10. granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep3/mem_samples.csv +200 -0
  11. granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep3/run.env +102 -0
  12. granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep3/train.log +1258 -0
  13. granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/results.tsv +4 -0
granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/manifest.txt ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ timestamp: 2026-09-21T08:37:05+00:00
2
+ host: sp-aiinfra-interactive-aghosh-8-mb99k-8xb200-moe-cg-0
3
+ gpu: NVIDIA B200 x8
4
+ driver: 580.95.05
5
+ study_scripts_branch: granularity-ultraep
6
+ study_scripts_commit: 05dd2cb
7
+ campaign: ultraep
8
+ backend: ultraep
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-UltraEP
17
+ megatron_branch_required: ultraep-bracing-regions-through-kernels
18
+ megatron_branch_actual: ultraep-bracing-regions-through-kernels
19
+ megatron_commit: 03d57ef7
granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/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==1.32.0
79
+ hypothesis==6.130.8
80
+ idna==3.10
81
+ importlib_metadata==8.7.0
82
+ iniconfig==2.1.0
83
+ intel-openmp==2021.4.0
84
+ ipykernel==6.29.5
85
+ ipython==9.3.0
86
+ ipython_pygments_lexers==1.1.1
87
+ isoduration==20.11.0
88
+ isort==6.0.1
89
+ jedi==0.19.2
90
+ Jinja2==3.1.6
91
+ joblib==1.5.1
92
+ json5==0.12.0
93
+ jsonpointer==3.0.0
94
+ jsonschema==4.24.0
95
+ jsonschema-specifications==2025.4.1
96
+ jupyter-events==0.12.0
97
+ jupyter-lsp==2.2.5
98
+ jupyter_client==8.6.3
99
+ jupyter_core==5.8.1
100
+ jupyter_server==2.16.0
101
+ jupyter_server_terminals==0.5.3
102
+ jupyterlab==4.4.3
103
+ jupyterlab_code_formatter==3.0.2
104
+ jupyterlab_pygments==0.3.0
105
+ jupyterlab_server==2.27.3
106
+ jupyterlab_tensorboard_pro @ file:///builder/jupyterlab_tensorboard_pro-4.0.0-py2.py3-none-any.whl#sha256=6bfcb5ced3358eb72150f4b2ab91312faef2f0ca14d845e8379d8fa52df6902e
107
+ jupytext==1.17.2
108
+ kiwisolver==1.4.8
109
+ kvikio @ file:///rapids/kvikio-25.4.0-cp312-cp312-linux_x86_64.whl#sha256=8b40fd8ea9e4a2974c04b50b7fc26b06094ac807f9a233281d0bafbaa71b1d85
110
+ langcodes==3.5.0
111
+ language_data==1.3.0
112
+ lazy_loader==0.4
113
+ libcudf @ file:///rapids/libcudf-25.4.0-py3-none-linux_x86_64.whl#sha256=238320b4958c1737bba2a609673e2539a78216b811b5498a68e00395d9b021a4
114
+ libcugraph @ file:///rapids/libcugraph-25.4.0-py3-none-linux_x86_64.whl#sha256=29085e901b2f7569ff2653bb1355fe6fad20f2fcaa585f5649da9867d8df766d
115
+ libcuml @ file:///rapids/libcuml-25.4.0-py3-none-linux_x86_64.whl#sha256=8e2e444c72b0e995e15a48da37ff12fb410a47a2baa69ba86ccc9b1a3db91e8c
116
+ libcuvs @ file:///rapids/libcuvs-25.4.0-py3-none-linux_x86_64.whl#sha256=e2f8db61c92b769c590dd752f43439ddd62b2f6592a3c337362b6c1ef220c3b9
117
+ libkvikio @ file:///rapids/libkvikio-25.4.0-py3-none-linux_x86_64.whl#sha256=da8a5713b05fc478bf40529a3e0fb61a2d605bb5a70ca4d461f96a6005dbe8b7
118
+ libraft @ file:///rapids/libraft-25.4.0-py3-none-linux_x86_64.whl#sha256=b921dec005aa4a321dd156776a18376feb47c96a5b67eec125c8a8e19edd1e24
119
+ librmm @ file:///rapids/librmm-25.4.0-py3-none-linux_x86_64.whl#sha256=ff951811bda729d4fa1109ab906225ae1f489a6ec2eeab5d1b674cc34bd4422f
120
+ librmm-cu12 @ file:///rapids/librmm_cu12-25.4.0-py3-none-any.whl#sha256=29296ef2b0e90679a0db280e86015be12d6ab9f4e62a181aa085f62dbadd06b3
121
+ librosa==0.11.0
122
+ libucxx @ file:///rapids/libucxx-0.43.0-py3-none-linux_x86_64.whl#sha256=f15849d7ac2fb68f0fd8998838a13321ddcfb0c594f7e6cbc401a3bcc4750566
123
+ lightning-thunder @ file:///opt/pytorch/lightning-thunder
124
+ lightning-utilities==0.14.3
125
+ lintrunner==0.12.7
126
+ llvmlite==0.42.0
127
+ locket @ file:///rapids/locket-1.0.0-py2.py3-none-any.whl#sha256=b6c819a722f7b6bd955b80781788e4a66a55628b858d347536b7e81325a3a5e3
128
+ looseversion==1.3.0
129
+ marisa-trie==1.2.1
130
+ Markdown==3.8
131
+ markdown-it-py==3.0.0
132
+ MarkupSafe==3.0.2
133
+ matplotlib==3.10.3
134
+ matplotlib-inline==0.1.7
135
+ mdit-py-plugins==0.4.2
136
+ mdurl==0.1.2
137
+ -e git+ssh://git@github.com/sriram-17-17/my-Megatron-MoE-Echo.git@9a23679390d7ab1fe462bc61fa31d90d393db49c#egg=megatron_core
138
+ mistune==3.1.3
139
+ mkl==2021.1.1
140
+ mkl-devel==2021.1.1
141
+ mkl-include==2021.1.1
142
+ ml_dtypes==0.6.0
143
+ mock==5.2.0
144
+ -e git+ssh://git@github.com/sriram-17-17/MoonEP.git@d8bc6667fad7177e6e09b637d773fa67f08bae93#egg=moonep
145
+ mpmath==1.3.0
146
+ msgpack==1.1.0
147
+ multidict @ file:///rapids/multidict-6.4.4-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl#sha256=a920f9cf2abdf6e493c519492d892c362007f113c94da4c239ae88429835bad1
148
+ multiprocess==0.70.19
149
+ murmurhash==1.0.13
150
+ mypy_extensions==1.1.0
151
+ nbclient==0.10.2
152
+ nbconvert==7.16.6
153
+ nbformat==5.10.4
154
+ nest-asyncio==1.6.0
155
+ networkx==3.5
156
+ ninja==1.11.1.4
157
+ notebook==7.4.3
158
+ notebook_shim==0.2.4
159
+ numba==0.59.1
160
+ numba-cuda @ file:///rapids/numba_cuda-0.4.0-py3-none-any.whl#sha256=5b95a5f42068771e37142944d44fe844fa87f1b2f89051461a4cb0af91bcc28a
161
+ numcodecs @ file:///rapids/numcodecs-0.13.1-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl#sha256=da2230484e6102e5fa3cc1a5dd37ca1f92dfbd183d91662074d6f7574e3e8f53
162
+ numpy==1.26.4
163
+ nvdlfw_inspect @ git+https://github.com/NVIDIA/nvidia-dlfw-inspect.git@6b60eb0e675606fb2fbbfcb12667ecbec75eaf87
164
+ nvfuser==0.2.27a0+9bf5aca
165
+ nvidia-cuda-cccl-cu12==12.9.27
166
+ nvidia-cudnn-frontend @ file:///opt/pytorch/lightning-thunder/tmp/cudnn_frontend
167
+ nvidia-cutlass-dsl==4.4.2
168
+ nvidia-cutlass-dsl-libs-base==4.4.2
169
+ nvidia-dali-cuda120==1.50.0
170
+ nvidia-ml-py @ file:///rapids/nvidia_ml_py-12.575.51-py3-none-any.whl#sha256=eb8641800d98ce40a22f479873f34b482e214a7e80349c63be51c3919845446e
171
+ nvidia-modelopt==0.29.0
172
+ nvidia-modelopt-core==0.29.0
173
+ nvidia-nvcomp-cu12==4.2.0.14
174
+ nvidia-nvimgcodec-cu12==0.5.0.13
175
+ nvidia-nvjpeg-cu12==12.4.0.16
176
+ nvidia-nvjpeg2k-cu12==0.8.1.40
177
+ nvidia-nvshmem-cu12==3.7.2
178
+ nvidia-nvtiff-cu12==0.5.0.67
179
+ nvidia-resiliency-ext==0.4.0
180
+ nvtx @ file:///rapids/nvtx-0.2.11-cp312-cp312-linux_x86_64.whl#sha256=73521e37cba42b4db0b3cd68fc5eb814e55c500a61eeed6edcdf6746315ac53e
181
+ nx-cugraph @ file:///rapids/nx_cugraph-25.4.0-py3-none-any.whl#sha256=e625444486cc899a77bd48e7da76a2edd95c0006dd95fbf323745feae7f33111
182
+ onnx @ file:///opt/pytorch/pytorch/third_party/onnx
183
+ onnx-ir==1.0.0
184
+ onnxscript==0.7.2
185
+ opt_einsum==3.4.0
186
+ optree==0.16.0
187
+ overrides==7.7.0
188
+ packaging==23.2
189
+ pandas @ file:///rapids/pandas-2.2.3-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl#sha256=fffb8ae78d8af97f849404f21411c95062db1496aeb3e56f146f0355c9989319
190
+ pandocfilters==1.5.1
191
+ parso==0.8.4
192
+ partd @ file:///rapids/partd-1.4.2-py3-none-any.whl#sha256=978e4ac767ec4ba5b86c6eaa52e5a2a3bc748a2ca839e8cc798f1cc6ce6efb0f
193
+ pathspec==0.12.1
194
+ pexpect==4.9.0
195
+ pillow==11.2.1
196
+ platformdirs==4.3.8
197
+ pluggy==1.6.0
198
+ ply @ file:///rapids/ply-3.11-py2.py3-none-any.whl#sha256=096f9b8350b65ebd2fd1346b12452efe5b9607f7482813ffca50c22722a807ce
199
+ polars @ file:///rapids/polars-1.25.2-cp39-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl#sha256=f7fcbb4f476784384ccda48757fca4e8c2e2c5a0a3aef3717aaf56aee4e30e09
200
+ polygraphy==0.49.20
201
+ pooch==1.8.2
202
+ preshed==3.0.10
203
+ prometheus_client==0.22.1
204
+ prompt_toolkit==3.0.51
205
+ propcache @ file:///rapids/propcache-0.3.1-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl#sha256=563f9d8c03ad645597b8d010ef4e9eab359faeb11a0a2ac9f7b4bc8c28ebef25
206
+ protobuf==4.24.4
207
+ psutil==7.0.0
208
+ ptyprocess==0.7.0
209
+ PuLP==3.2.1
210
+ pure_eval==0.2.3
211
+ pyarrow==25.0.1
212
+ pybind11==2.13.6
213
+ pybind11_global==2.13.6
214
+ pycocotools @ git+https://github.com/nvidia/cocoapi.git@08350a157b41f0755a536441b5abec89ab934c5b#subdirectory=PythonAPI
215
+ pycparser==2.22
216
+ pydantic==2.11.5
217
+ pydantic_core==2.33.2
218
+ Pygments==2.19.1
219
+ pylibcudf @ file:///rapids/pylibcudf-25.4.0-cp312-cp312-linux_x86_64.whl#sha256=401b8bfd022d06f84e120779ca7c3d01a28b68762aaa8f7873b10d6b1975cf6a
220
+ pylibcugraph @ file:///rapids/pylibcugraph-25.4.0-cp312-cp312-linux_x86_64.whl#sha256=061c7389a88c815b54fd5d3a3584fb87bf693e5602f5be335260d0ba575ba557
221
+ pylibcugraphops @ file:///rapids/pylibcugraphops-25.4.0-cp312-cp312-linux_x86_64.whl#sha256=a25ef4b22166c80718d5c7e5bddb755e81a2fe340fa5becd0ac36ec3edfb208a
222
+ pylibraft @ file:///rapids/pylibraft-25.4.0-cp312-cp312-linux_x86_64.whl#sha256=f2ecd23fe51f9c2caa87bbbe85cb32fcaff4ecfb52ceb9a9991bc50f82782c2e
223
+ pylibwholegraph @ file:///rapids/pylibwholegraph-25.4.0-cp312-cp312-linux_x86_64.whl#sha256=b82913d1e38bda76d8c272fbd8cc9072d987209d3d47cc93a464a37c2925aca2
224
+ pynvjitlink @ file:///rapids/pynvjitlink-0.3.0-cp312-cp312-linux_x86_64.whl#sha256=066caf1314730643d3b48c9847bd65789defdd6a7c78993176eb3d8bbe3f4919
225
+ pynvml @ file:///rapids/pynvml-12.0.0-py3-none-any.whl#sha256=fdff84b62a27dbe98e08e1a647eb77342bef1aebe0878bcd15e99a83fcbecb9e
226
+ pyparsing==3.2.3
227
+ pytest==8.1.1
228
+ pytest-flakefinder==1.1.0
229
+ pytest-rerunfailures==15.1
230
+ pytest-shard==0.1.2
231
+ pytest-xdist==3.7.0
232
+ python-dateutil==2.9.0.post0
233
+ python-hostlist==2.2.1
234
+ python-json-logger==3.3.0
235
+ pytorch-triton @ file:///tmp/dist/pytorch_triton-3.3.0%2Bgit96316ce52.nvinternal-cp312-cp312-linux_x86_64.whl#sha256=6d6921772b9dfd0243c055e0d054de42a113e92cabcfdd1d3e80a5ad9c08cf7d
236
+ pytz @ file:///rapids/pytz-2023.4-py2.py3-none-any.whl#sha256=f90ef520d95e7c46951105338d918664ebfd6f1d995bd7d153127ce90efafa6a
237
+ PyYAML==6.0.2
238
+ pyzmq==26.4.0
239
+ raft-dask @ file:///rapids/raft_dask-25.4.0-cp312-cp312-linux_x86_64.whl#sha256=36cf37ba6ce35454c4729c2cb597a360f22780eefb11d514a16a07137da5eb6c
240
+ rapids-dask-dependency @ file:///rapids/rapids_dask_dependency-25.4.0a0-py3-none-any.whl#sha256=484faca626e7760112a7d8ad3426f5bdd720b609006d8dca87c2607076a348bd
241
+ rapids-logger @ file:///rapids/rapids_logger-0.1.18-py3-none-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl#sha256=3617b2d86291ac6cb957542895f20fc9737ace29773ca90337e6fb04c73bd120
242
+ referencing==0.36.2
243
+ regex==2024.11.6
244
+ requests==2.32.3
245
+ rfc3339-validator==0.1.4
246
+ rfc3986-validator==0.1.1
247
+ rich==14.0.0
248
+ rmm @ file:///rapids/rmm-25.4.0-cp312-cp312-linux_x86_64.whl#sha256=1824bcab3836296666980a997ddcb9c4bca24dfb873012cfa0a7ecea583ca712
249
+ rpds-py==0.25.1
250
+ safetensors==0.5.3
251
+ scikit-learn==1.6.1
252
+ scipy==1.15.3
253
+ Send2Trash==1.8.3
254
+ setuptools==78.1.1
255
+ shellingham==1.5.4
256
+ six==1.16.0
257
+ smart-open==7.1.0
258
+ sniffio==1.3.1
259
+ sortedcontainers==2.4.0
260
+ soundfile==0.13.1
261
+ soupsieve==2.7
262
+ soxr==0.5.0.post1
263
+ spacy==3.7.5
264
+ spacy-legacy==3.0.12
265
+ spacy-loggers==1.0.5
266
+ srsly==2.5.1
267
+ stack-data==0.6.3
268
+ sympy==1.14.0
269
+ tabulate==0.9.0
270
+ tbb==2021.13.1
271
+ tblib @ file:///rapids/tblib-3.1.0-py3-none-any.whl#sha256=670bb4582578134b3d81a84afa1b016128b429f3d48e6cbbaecc9d15675e984e
272
+ tensorboard==2.16.2
273
+ tensorboard-data-server==0.7.2
274
+ tensorrt @ file:///workspace/TensorRT-10.11.0.33/python/tensorrt-10.11.0.33-cp312-none-linux_x86_64.whl#sha256=c2f18f72758a3d54a71901495c8fda9e7094a8bcdd2a5b6b58967fbf05733b3a
275
+ terminado==0.18.1
276
+ thinc==8.2.5
277
+ threadpoolctl==3.6.0
278
+ thriftpy2 @ file:///rapids/thriftpy2-0.5.2-cp312-cp312-linux_x86_64.whl#sha256=c0c7d939bdc5ccf7739154959938445d5fdf72f1d6e12bd563947a36f67e2147
279
+ tinycss2==1.4.0
280
+ toolz @ file:///rapids/toolz-1.0.0-py3-none-any.whl#sha256=292c8f1c4e7516bf9086f8850935c799a874039c8bcf959d47b600e4c44a6236
281
+ torch @ file:///opt/transfer/torch-2.8.0a0%2B5228986c39.nv25.6-29946474-cp312-cp312-linux_x86_64.whl#sha256=7a3c9ff5288d8102a31296b3d62491f3f47c8820a5c9aaefc7dc0fe20e1e8ad5
282
+ torch_tensorrt @ file:///opt/pytorch/torch_tensorrt/dist/torch_tensorrt-2.8.0a0-cp312-cp312-linux_x86_64.whl#sha256=26f4d7a04e7e6e017a5b917b8b98b83720bb2bb0195f48d25f52d567d286bfe1
283
+ torchao @ file:///opt/pytorch/ao
284
+ torchprofile==0.0.4
285
+ torchvision @ file:///opt/transfer/torchvision-0.22.0a0%2B95f10a4e-29946474-cp312-cp312-manylinux_2_28_x86_64.whl#sha256=746a3c674cc61680a86f4a14b374b21af1b5c21f629c42765b40c9494f6b234d
286
+ tornado==6.5.1
287
+ tqdm==4.67.1
288
+ traitlets==5.14.3
289
+ transformer_engine @ file:///pvcdata/aiinfra/b-aghosh/wheels/transformer_engine-2.10.0.dev0%2B10f6128d-cp312-cp312-linux_x86_64.whl#sha256=35cf5f4b83b2baad864e12e512bb422e0929409ac05610812fa521d4942d0189
290
+ treelite @ file:///rapids/treelite-4.4.1-py3-none-linux_x86_64.whl#sha256=174741724205540a1a5b37da81483aeff55d89548f42b3461cf07acac7ae96eb
291
+ typer==0.16.0
292
+ types-dataclasses==0.6.6
293
+ types-python-dateutil==2.9.0.20250516
294
+ typing-inspection==0.4.1
295
+ typing_extensions==4.14.0
296
+ tzdata @ file:///rapids/tzdata-2025.2-py2.py3-none-any.whl#sha256=1a403fada01ff9221ca8044d701868fa132215d84beb92242d9acd2147f667a8
297
+ ucx-py @ file:///rapids/ucx_py-0.43.0-cp312-cp312-linux_x86_64.whl#sha256=4a452fd8d8c983283ada3d49086e3657dc2ee94fbad08efdd708d15ab6e0b73a
298
+ ucxx @ file:///rapids/ucxx-0.43.0-cp312-cp312-linux_x86_64.whl#sha256=f1b4e55cb45c66fe88a2d8f9c2b87751165f3cfec21fbdc9c47c4ddff75e8a5f
299
+ -e git+https://github.com/Dots-Infra/UltraEP.git@94cab099b44fffa99a82fea99e7c12d89cf65e4f#egg=ultra_ep
300
+ uri-template==1.3.0
301
+ urllib3==2.0.7
302
+ wasabi==1.1.3
303
+ wcwidth==0.2.13
304
+ weasel==0.4.1
305
+ webcolors==24.11.1
306
+ webencodings==0.5.1
307
+ websocket-client==1.8.0
308
+ Werkzeug==3.1.3
309
+ wheel==0.45.1
310
+ wrapt==1.17.2
311
+ xdoctest==1.0.2
312
+ xgboost @ file:///rapids/xgboost-2.1.4-py3-none-manylinux_2_39_x86_64.whl#sha256=25eeca27f4048fc491940b3a0b4b5782ee48c33cb3aec059cef71f8da542930b
313
+ xxhash==4.0.1
314
+ yarl @ file:///rapids/yarl-1.20.0-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl#sha256=bf099e2432131093cc611623e0b0bcc399b8cddd9a91eded8bfb50402ec35018
315
+ zarr @ file:///rapids/zarr-2.18.7-py3-none-any.whl#sha256=ac3dc4033e9ae4e9d7b5e27c97ea3eaf1003cc0a07f010bd83d5134bf8c4b223
316
+ zict @ file:///rapids/zict-3.0.0-py2.py3-none-any.whl#sha256=5796e36bd0e0cc8cf0fbc1ace6a68912611c1dbd74750a3f3026b9b9d6a327ae
317
+ zipp==3.22.0
granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/provenance.txt ADDED
@@ -0,0 +1,90 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ==============================================================
2
+ PROVENANCE: ultraep__tierb__gbs8
3
+ ==============================================================
4
+ captured_at : 2026-09-21T08:36:51+00:00
5
+ hostname : sp-aiinfra-interactive-aghosh-8-mb99k-8xb200-moe-cg-0
6
+ kernel : 6.8.0-1043-nvidia
7
+ cwd : /workspace/blackwell_run_scripts
8
+ command : run_ladder.sh --phase tierb --gbs 8 --out /workspace/granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb --nsys
9
+
10
+ ---- GPU / driver --------------------------------------------
11
+ 0, NVIDIA B200, 10.0, 183359 MiB, 580.95.05
12
+ 1, NVIDIA B200, 10.0, 183359 MiB, 580.95.05
13
+ 2, NVIDIA B200, 10.0, 183359 MiB, 580.95.05
14
+ 3, NVIDIA B200, 10.0, 183359 MiB, 580.95.05
15
+ 4, NVIDIA B200, 10.0, 183359 MiB, 580.95.05
16
+ 5, NVIDIA B200, 10.0, 183359 MiB, 580.95.05
17
+ 6, NVIDIA B200, 10.0, 183359 MiB, 580.95.05
18
+ 7, NVIDIA B200, 10.0, 183359 MiB, 580.95.05
19
+ persistence/clocks:
20
+ 0, 120 MHz, 1965 MHz, 1000.00 W, 29
21
+ 1, 1965 MHz, 1965 MHz, 1000.00 W, 35
22
+ 2, 120 MHz, 1965 MHz, 1000.00 W, 34
23
+ 3, 120 MHz, 1965 MHz, 1000.00 W, 30
24
+ 4, 1965 MHz, 1965 MHz, 1000.00 W, 30
25
+ 5, 135 MHz, 1965 MHz, 1000.00 W, 34
26
+ 6, 120 MHz, 1965 MHz, 1000.00 W, 34
27
+ 7, 120 MHz, 1965 MHz, 1000.00 W, 30
28
+
29
+ NVLink topology:
30
+ GPU0 GPU1 GPU2 GPU3 GPU4 GPU5 GPU6 GPU7 NIC0 NIC1 NIC2 NIC3 NIC4 NIC5 NIC6 NIC7 NIC8 CPU Affinity NUMA Affinity GPU NUMA ID
31
+ GPU0 X NV18 NV18 NV18 NV18 NV18 NV18 NV18 PIX NODE SYS SYS SYS SYS SYS SYS NODE 0-27 0 N/A
32
+ GPU1 NV18 X NV18 NV18 NV18 NV18 NV18 NV18 NODE PIX SYS SYS SYS SYS SYS SYS NODE 0-27 0 N/A
33
+ GPU2 NV18 NV18 X NV18 NV18 NV18 NV18 NV18 SYS SYS PIX NODE SYS SYS SYS SYS SYS 28-55 1 N/A
34
+ GPU3 NV18 NV18 NV18 X NV18 NV18 NV18 NV18 SYS SYS NODE PIX SYS SYS SYS SYS SYS 28-55 1 N/A
35
+ GPU4 NV18 NV18 NV18 NV18 X NV18 NV18 NV18 SYS SYS SYS SYS PIX NODE SYS SYS SYS 56-83 2 N/A
36
+ GPU5 NV18 NV18 NV18 NV18 NV18 X NV18 NV18 SYS SYS SYS SYS NODE PIX SYS SYS SYS 56-83 2 N/A
37
+ GPU6 NV18 NV18 NV18 NV18 NV18 NV18 X NV18 SYS SYS SYS SYS SYS SYS PIX NODE SYS 84-111 3 N/A
38
+ GPU7 NV18 NV18 NV18 NV18 NV18 NV18 NV18 X SYS SYS SYS SYS SYS SYS NODE PIX SYS 84-111 3 N/A
39
+ NIC0 PIX NODE SYS SYS SYS SYS SYS SYS X NODE SYS SYS SYS SYS SYS SYS NODE
40
+ NIC1 NODE PIX SYS SYS SYS SYS SYS SYS NODE X SYS SYS SYS SYS SYS SYS NODE
41
+ NIC2 SYS SYS PIX NODE SYS SYS SYS SYS SYS SYS X NODE SYS SYS SYS SYS SYS
42
+ NIC3 SYS SYS NODE PIX SYS SYS SYS SYS SYS SYS NODE X SYS SYS SYS SYS SYS
43
+ NIC4 SYS SYS SYS SYS PIX NODE SYS SYS SYS SYS SYS SYS X NODE SYS SYS SYS
44
+ unavailable
45
+
46
+ ---- toolchain -----------------------------------------------
47
+ nvcc : release 12.9, V12.9.86
48
+ gcc : gcc (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0
49
+ python : 3.12.3
50
+ torch : 2.8.0a0+5228986c39.nv25.06 cuda=12.9 nccl=2.27.3
51
+ torch arch : ['sm_75', 'sm_80', 'sm_86', 'sm_90', 'sm_100', 'sm_120', 'compute_120']
52
+ transformer_engine: 2.10.0.dev0+10f6128d [/usr/local/lib/python3.12/dist-packages/transformer_engine/__init__.py]
53
+ deep_ep : 1.2.1+34152ae [/workspace/DeepEP/deep_ep/__init__.py]
54
+ ultra_ep : 1.0.0 [/workspace/UltraEP/ultra_ep/__init__.py]
55
+ moonep : 0.0.1 [/workspace/MoonEP/moonep/__init__.py]
56
+ megatron : 0.14.0rc5 [None]
57
+ flash_attn : 2.7.4.post1 [/usr/local/lib/python3.12/dist-packages/flash_attn/__init__.py]
58
+ numpy : 1.26.4 [/usr/local/lib/python3.12/dist-packages/numpy/__init__.py]
59
+ huggingface_hub: 1.32.0 [/usr/local/lib/python3.12/dist-packages/huggingface_hub/__init__.py]
60
+
61
+ ---- repositories --------------------------------------------
62
+ Megatron-MoE-Echo bracing-regions-through-kernels 9a23679390d7
63
+ Megatron-UltraEP ultraep-bracing-regions-through-kernels 03d57ef7d476
64
+ UltraEP main 94cab099b44f
65
+ DeepEP HEAD 34152ae28f80 (+3 untracked: build output)
66
+ MoonEP megatron-ddp-symmetric-params d8bc6667fad7
67
+ blackwell_run_scripts granularity-ultraep 05dd2cbfb1eb
68
+
69
+ ---- environment ---------------------------------------------
70
+ CUDA_ARCH_LIST=7.5 8.0 8.6 9.0 10.0 12.0
71
+ CUDA_DRIVER_VERSION=575.57.08
72
+ CUDA_HOME=/usr/local/cuda
73
+ CUDA_MODULE_LOADING=LAZY
74
+ CUDA_VERSION=12.9.1.010
75
+ HF_REMOTE_ROOT=granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16
76
+ NCCL_VERSION=2.27.3
77
+ NVSHMEM_VERSION=3.2.5
78
+ PYTORCH_BUILD_NUMBER=0
79
+ PYTORCH_BUILD_VERSION=2.8.0a0+5228986
80
+ PYTORCH_HOME=/opt/pytorch/pytorch
81
+ PYTORCH_VERSION=2.8.0a0+5228986
82
+ TORCH_ALLOW_TF32_CUBLAS_OVERRIDE=1
83
+ TORCH_CUDA_ARCH_LIST=7.5 8.0 8.6 9.0 10.0 12.0+PTX
84
+ TORCH_NCCL_USE_COMM_NONBLOCKING=0
85
+
86
+ ---- shared library resolution -------------------------------
87
+ libnvshmem_host.so: not in ldconfig
88
+ libnccl.so: /lib/x86_64-linux-gnu/libnccl.so.2
89
+ lrwxrwxrwx 1 root root 20 Sep 21 04:57 /usr/local/lib/python3.12/dist-packages/nvidia/nvshmem/lib/libnvshmem_host.so -> libnvshmem_host.so.3
90
+ -rw-r--r-- 1 root root 179045744 Sep 21 04:54 /usr/local/lib/python3.12/dist-packages/nvidia/nvshmem/lib/libnvshmem_host.so.3
granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep1/mem_samples.csv ADDED
@@ -0,0 +1,200 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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, 0
10
+ 1, 0
11
+ 2, 0
12
+ 3, 0
13
+ 4, 0
14
+ 5, 0
15
+ 6, 0
16
+ 7, 0
17
+ 0, 4
18
+ 1, 1
19
+ 2, 1
20
+ 3, 1
21
+ 4, 1
22
+ 5, 1
23
+ 6, 1
24
+ 7, 1
25
+ 0, 4
26
+ 1, 4
27
+ 2, 4
28
+ 3, 4
29
+ 4, 2
30
+ 5, 1
31
+ 6, 1
32
+ 7, 1
33
+ 0, 4
34
+ 1, 4
35
+ 2, 4
36
+ 3, 4
37
+ 4, 4
38
+ 5, 4
39
+ 6, 4
40
+ 7, 4
41
+ 0, 821
42
+ 1, 819
43
+ 2, 821
44
+ 3, 819
45
+ 4, 821
46
+ 5, 821
47
+ 6, 821
48
+ 7, 819
49
+ 0, 1537
50
+ 1, 1537
51
+ 2, 1537
52
+ 3, 1537
53
+ 4, 1537
54
+ 5, 1537
55
+ 6, 1537
56
+ 7, 1537
57
+ 0, 1823
58
+ 1, 1823
59
+ 2, 1823
60
+ 3, 1823
61
+ 4, 1823
62
+ 5, 1823
63
+ 6, 1823
64
+ 7, 1823
65
+ 0, 2671
66
+ 1, 2671
67
+ 2, 2671
68
+ 3, 2671
69
+ 4, 2671
70
+ 5, 2671
71
+ 6, 2671
72
+ 7, 2671
73
+ 0, 15337
74
+ 1, 15285
75
+ 2, 15305
76
+ 3, 15331
77
+ 4, 15357
78
+ 5, 15297
79
+ 6, 15317
80
+ 7, 15297
81
+ 0, 16317
82
+ 1, 16335
83
+ 2, 16323
84
+ 3, 16333
85
+ 4, 16335
86
+ 5, 16319
87
+ 6, 16327
88
+ 7, 16337
89
+ 0, 16925
90
+ 1, 16927
91
+ 2, 16927
92
+ 3, 16929
93
+ 4, 16923
94
+ 5, 16927
95
+ 6, 16929
96
+ 7, 16923
97
+ 0, 17065
98
+ 1, 17067
99
+ 2, 17067
100
+ 3, 17069
101
+ 4, 17063
102
+ 5, 17067
103
+ 6, 17069
104
+ 7, 17063
105
+ 0, 17561
106
+ 1, 17563
107
+ 2, 17563
108
+ 3, 17565
109
+ 4, 17559
110
+ 5, 17563
111
+ 6, 17565
112
+ 7, 17559
113
+ 0, 24883
114
+ 1, 24863
115
+ 2, 24843
116
+ 3, 24877
117
+ 4, 24859
118
+ 5, 25035
119
+ 6, 25025
120
+ 7, 24879
121
+ 0, 28501
122
+ 1, 28503
123
+ 2, 28503
124
+ 3, 28485
125
+ 4, 28509
126
+ 5, 28483
127
+ 6, 28503
128
+ 7, 28509
129
+ 0, 28501
130
+ 1, 28503
131
+ 2, 28503
132
+ 3, 28485
133
+ 4, 28509
134
+ 5, 28483
135
+ 6, 28503
136
+ 7, 28509
137
+ 0, 28501
138
+ 1, 28503
139
+ 2, 28503
140
+ 3, 28485
141
+ 4, 28509
142
+ 5, 28483
143
+ 6, 28503
144
+ 7, 28509
145
+ 0, 28767
146
+ 1, 28803
147
+ 2, 28801
148
+ 3, 28753
149
+ 4, 28791
150
+ 5, 28817
151
+ 6, 28787
152
+ 7, 28785
153
+ 0, 29777
154
+ 1, 29789
155
+ 2, 29797
156
+ 3, 29779
157
+ 4, 29797
158
+ 5, 29827
159
+ 6, 29809
160
+ 7, 29831
161
+ 0, 37195
162
+ 1, 37201
163
+ 2, 37213
164
+ 3, 37197
165
+ 4, 37197
166
+ 5, 37193
167
+ 6, 37185
168
+ 7, 37197
169
+ 0, 39097
170
+ 1, 39087
171
+ 2, 38951
172
+ 3, 38931
173
+ 4, 39095
174
+ 5, 38941
175
+ 6, 39091
176
+ 7, 39095
177
+ 0, 36117
178
+ 1, 36107
179
+ 2, 35971
180
+ 3, 35951
181
+ 4, 36115
182
+ 5, 36025
183
+ 6, 36757
184
+ 7, 36435
185
+ 0, 34695
186
+ 1, 34685
187
+ 2, 34287
188
+ 3, 34529
189
+ 4, 34431
190
+ 5, 34539
191
+ 6, 35077
192
+ 7, 34693
193
+ 0, 129
194
+ 1, 34682
195
+ 2, 641
196
+ 3, 1
197
+ 4, 1
198
+ 5, 4
199
+ 6, 4
200
+ 7, 4
granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep1/run.env ADDED
@@ -0,0 +1,102 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ backend=ultraep
2
+ config=OLMOE_1B_7B_L16
3
+ family=tierB
4
+ source_model=OLMoE-1B-7B-0924
5
+ standard_num_layers=16
6
+ base_rung=
7
+ repo=/workspace/Megatron-UltraEP
8
+ branch=ultraep-bracing-regions-through-kernels
9
+ commit=03d57ef7
10
+ num_layers=16
11
+ hidden=2048
12
+ ffn_hidden=1024
13
+ moe_ffn_hidden=1024
14
+ num_experts=64
15
+ topk=8
16
+ experts_per_rank=8
17
+ expert_params=6291456
18
+ heads=16
19
+ kv_groups=16
20
+ kv_channels=128
21
+ qk_layernorm=0
22
+ shared_ffn=0
23
+ shared_expert_overlap=0
24
+ moe_token_dispatcher=flex
25
+ mla=0
26
+ q_lora_rank=
27
+ kv_lora_rank=512
28
+ qk_head_dim=128
29
+ qk_pos_emb_head_dim=64
30
+ v_head_dim=128
31
+ dense_first_layers=0
32
+ moe_layers=16
33
+ rotary_base=10000
34
+ rotary_percent=1.0
35
+ rotary_scaling=
36
+ mscale=
37
+ mscale_all_dim=
38
+ qkv_bias=0
39
+ router_balancing=aux_loss
40
+ aux_loss_coeff=1e-2
41
+ router_pre_softmax=1
42
+ router_fusion=0
43
+ router_dtype=fp32
44
+ router_groups=
45
+ router_group_topk=
46
+ router_score_function=softmax
47
+ router_expert_bias=0
48
+ router_bias_update_rate=
49
+ router_topk_scaling_factor=
50
+ shared_expert_gate=0
51
+ apply_layernorm_1p=0
52
+ attention_output_gate=0
53
+ linear_attention_type=
54
+ linear_attention_freq=
55
+ linear_conv_kernel_dim=
56
+ linear_key_head_dim=
57
+ linear_value_head_dim=
58
+ linear_num_key_heads=
59
+ linear_num_value_heads=
60
+ mtp_num_layers=0
61
+ mtp_loss_scaling_factor=0.1
62
+ no_weight_decay_cond_type=
63
+ max_position_embeddings=4096
64
+ init_method_std=0.02
65
+ device_initiated_grouped_gemm=1
66
+ norm_epsilon=1e-5
67
+ seq_length=4096
68
+ micro_batch_size=1
69
+ global_batch_size=8
70
+ num_microbatches=1
71
+ assignments_per_rank=32768
72
+ mean_tokens_per_expert=4096
73
+ slot_policy=matched
74
+ slots_per_rank=8
75
+ capacity_mode=static
76
+ capacity=1.08984375
77
+ cuda_graph=1
78
+ probe_mode=0
79
+ async_grad_reduce=0
80
+ branch_role=timing
81
+ memory_dump_mode=none
82
+ memory_snapshot_path=
83
+ cuda_graph_warmup_steps=3
84
+ recompute=none
85
+ force_balance=0
86
+ train_iters=12
87
+ ep_size=8
88
+ echo_dispatch_overlap=1
89
+ echo_dispatch_recompute=0
90
+ ultraep_quota_min_tokens=512
91
+ ultraep_weight_sync_join=
92
+ profile_enable=0
93
+ kernel_trace=0
94
+ fp8_mode=mxfp8
95
+ fp8_recipe=mxfp8
96
+ training_seed=1234
97
+ learning_rate=1e-4
98
+ min_learning_rate=1.0e-5
99
+ lr_warmup_iters=5
100
+ state_dump_enable=0
101
+ state_dump_dir=
102
+ moonep_assert_no_pending_reduce=0
granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep1/train.log ADDED
@@ -0,0 +1,1258 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ==============================================================
2
+ OLMOE_1B_7B_L16 / ultraep (slots/rank=8 policy=matched)
3
+ E=64 K=8 H'=1024 m=8
4
+ L=16 seq=4096 gbs=8 (1 microbatches)
5
+ cap_mode=static cap=1.08984375 cg=1 recompute=none
6
+ memory_dump=none snapshots=/workspace/granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep1/memory_snapshots/snapshot.pickle
7
+ repo=/workspace/Megatron-UltraEP @ 03d57ef7
8
+ ==============================================================
9
+ W0921 08:37:07.278000 694969 torch/distributed/run.py:766]
10
+ W0921 08:37:07.278000 694969 torch/distributed/run.py:766] *****************************************
11
+ W0921 08:37:07.278000 694969 torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.
12
+ W0921 08:37:07.278000 694969 torch/distributed/run.py:766] *****************************************
13
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
14
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
15
+ 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
16
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
17
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
18
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
19
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
20
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
21
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
22
+ 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
23
+ WARNING: overriding default arguments for tokenizer_type:GPT2BPETokenizer with tokenizer_type:GPT2BPETokenizer
24
+ Number of virtual stages per pipeline stage: None
25
+ accumulate and all-reduce gradients in fp32 for bfloat16 data type.
26
+ using torch.bfloat16 for parameters ...
27
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
28
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
29
+ ------------------------ arguments ------------------------
30
+ account_for_embedding_in_pipeline_split ......... False
31
+ account_for_loss_in_pipeline_split .............. False
32
+ accumulate_allreduce_grads_in_fp32 .............. True
33
+ adam_beta1 ...................................... 0.9
34
+ adam_beta2 ...................................... 0.999
35
+ adam_eps ........................................ 1e-08
36
+ add_bias_linear ................................. False
37
+ add_position_embedding .......................... False
38
+ add_qkv_bias .................................... False
39
+ adlr_autoresume ................................. False
40
+ adlr_autoresume_interval ........................ 1000
41
+ align_grad_reduce ............................... True
42
+ align_param_gather .............................. False
43
+ app_tag_run_name ................................ None
44
+ app_tag_run_version ............................. 0.0.0
45
+ apply_layernorm_1p .............................. False
46
+ apply_query_key_layer_scaling ................... False
47
+ apply_residual_connection_post_layernorm ........ False
48
+ apply_rope_fusion ............................... True
49
+ async_save ...................................... None
50
+ async_tensor_model_parallel_allreduce ........... True
51
+ attention_backend ............................... AttnBackend.auto
52
+ attention_dropout ............................... 0.0
53
+ attention_softmax_in_fp32 ....................... False
54
+ auto_detect_ckpt_format ......................... False
55
+ barrier_with_L1_time ............................ True
56
+ bert_binary_head ................................ True
57
+ bert_embedder_type .............................. megatron
58
+ bert_load ....................................... None
59
+ bf16 ............................................ True
60
+ bias_dropout_fusion ............................. True
61
+ bias_gelu_fusion ................................ False
62
+ bias_swiglu_fusion .............................. True
63
+ biencoder_projection_dim ........................ 0
64
+ biencoder_shared_query_context_model ............ False
65
+ block_data_path ................................. None
66
+ calc_ft_timeouts ................................ False
67
+ calculate_per_token_loss ........................ False
68
+ check_for_large_grads ........................... False
69
+ check_for_nan_in_loss_and_grad .................. False
70
+ check_for_spiky_loss ............................ False
71
+ check_weight_hash_across_dp_replicas_interval ... None
72
+ ckpt_assume_constant_structure .................. False
73
+ ckpt_convert_format ............................. None
74
+ ckpt_convert_save ............................... None
75
+ ckpt_convert_update_legacy_dist_opt_format ...... False
76
+ ckpt_format ..................................... torch_dist
77
+ ckpt_fully_parallel_load ........................ False
78
+ ckpt_fully_parallel_save ........................ True
79
+ ckpt_fully_parallel_save_deprecated ............. False
80
+ ckpt_step ....................................... None
81
+ classes_fraction ................................ 1.0
82
+ clip_grad ....................................... 1.0
83
+ clone_scatter_output_in_embedding ............... True
84
+ config_logger_dir ...............................
85
+ consumed_train_samples .......................... 0
86
+ consumed_valid_samples .......................... 0
87
+ context_parallel_size ........................... 1
88
+ cp_comm_type .................................... ['p2p']
89
+ create_attention_mask_in_dataloader ............. True
90
+ cross_entropy_fusion_impl ....................... native
91
+ cross_entropy_loss_fusion ....................... False
92
+ cuda_graph_scope ................................ full_iteration
93
+ cuda_graph_warmup_steps ......................... 3
94
+ data_args_path .................................. None
95
+ data_cache_path ................................. None
96
+ data_parallel_random_init ....................... False
97
+ data_parallel_sharding_strategy ................. no_shard
98
+ data_parallel_size .............................. 8
99
+ data_path ....................................... ['/workspace/data/wikitext/wikitext103_gpt2_text_document']
100
+ data_per_class_fraction ......................... 1.0
101
+ data_sharding ................................... True
102
+ dataloader_type ................................. single
103
+ ddp_average_in_collective ....................... False
104
+ ddp_bucket_size ................................. None
105
+ ddp_num_buckets ................................. None
106
+ ddp_pad_buckets_for_high_nccl_busbw ............. False
107
+ decoder_first_pipeline_num_layers ............... None
108
+ decoder_last_pipeline_num_layers ................ None
109
+ decoder_num_layers .............................. None
110
+ decoder_seq_length .............................. None
111
+ decoupled_lr .................................... None
112
+ decoupled_min_lr ................................ None
113
+ decrease_batch_size_if_needed ................... False
114
+ defer_embedding_wgrad_compute ................... False
115
+ delay_wgrad_compute ............................. False
116
+ deprecated_use_mcore_models ..................... True
117
+ deterministic_mode .............................. False
118
+ dino_bottleneck_size ............................ 256
119
+ dino_freeze_last_layer .......................... 1
120
+ dino_head_hidden_size ........................... 2048
121
+ dino_local_crops_number ......................... 10
122
+ dino_local_img_size ............................. 96
123
+ dino_norm_last_layer ............................ False
124
+ dino_teacher_temp ............................... 0.07
125
+ dino_warmup_teacher_temp ........................ 0.04
126
+ dino_warmup_teacher_temp_epochs ................. 30
127
+ disable_bf16_reduced_precision_matmul ........... False
128
+ disable_mamba_mem_eff_path ...................... False
129
+ disable_straggler_on_startup .................... False
130
+ dist_ckpt_format_deprecated ..................... None
131
+ dist_ckpt_strictness ............................ assume_ok_unexpected
132
+ distribute_saved_activations .................... False
133
+ distributed_backend ............................. nccl
134
+ distributed_timeout_minutes ..................... 10
135
+ embedding_init_method_std ....................... None
136
+ embedding_path .................................. None
137
+ empty_unused_memory_level ....................... 0
138
+ enable_cuda_graph ............................... True
139
+ enable_experimental ............................. False
140
+ enable_ft_package ............................... False
141
+ enable_gloo_process_groups ...................... True
142
+ enable_msc ...................................... True
143
+ enable_one_logger ............................... True
144
+ encoder_num_layers .............................. 16
145
+ encoder_seq_length .............................. 4096
146
+ end_weight_decay ................................ 0.1
147
+ eod_mask_loss ................................... False
148
+ error_injection_rate ............................ 0
149
+ error_injection_type ............................ transient_error
150
+ eval_interval ................................... 1000000
151
+ eval_iters ...................................... 0
152
+ evidence_data_path .............................. None
153
+ exit_duration_in_mins ........................... None
154
+ exit_interval ................................... None
155
+ exit_on_missing_checkpoint ...................... False
156
+ exit_signal_handler ............................. False
157
+ exp_avg_dtype ................................... torch.float32
158
+ exp_avg_sq_dtype ................................ torch.float32
159
+ expert_model_parallel_size ...................... 8
160
+ expert_tensor_parallel_size ..................... 1
161
+ export_force_local_attention .................... False
162
+ export_kd_cfg ................................... None
163
+ export_kd_teacher_ckpt_format ................... None
164
+ export_kd_teacher_load .......................... None
165
+ export_kv_cache_quant ........................... False
166
+ export_legacy_megatron .......................... False
167
+ export_model_type ............................... GPTModel
168
+ export_moe_apply_probs_on_input ................. False
169
+ export_qk_l2_norm ............................... False
170
+ export_quant_cfg ................................ None
171
+ export_real_quant_cfg ........................... None
172
+ export_te_mcore_model ........................... False
173
+ external_cuda_graph ............................. False
174
+ ffn_hidden_size ................................. 1024
175
+ finetune ........................................ False
176
+ finetune_data_split ............................. train
177
+ finetune_hf_dataset ............................. None
178
+ first_last_layers_bf16 .......................... False
179
+ flash_decode .................................... False
180
+ fp16 ............................................ False
181
+ fp16_lm_cross_entropy ........................... False
182
+ fp32_residual_connection ........................ False
183
+ fp8 ............................................. e4m3
184
+ fp8_amax_compute_algo ........................... most_recent
185
+ fp8_amax_history_len ............................ 1
186
+ fp8_interval .................................... 1
187
+ fp8_margin ...................................... 0
188
+ fp8_param_gather ................................ False
189
+ fp8_recipe ...................................... mxfp8
190
+ fp8_wgrad ....................................... True
191
+ fsdp_double_buffer .............................. False
192
+ global_batch_size ............................... 8
193
+ grad_reduce_in_bf16 ............................. False
194
+ gradient_accumulation_fusion .................... True
195
+ gradient_reduce_div_fusion ...................... True
196
+ group_query_attention ........................... True
197
+ head_lr_mult .................................... 1.0
198
+ heterogeneous_layers_config_encoded_json ........ None
199
+ heterogeneous_layers_config_path ................ None
200
+ hidden_dropout .................................. 0.0
201
+ hidden_size ..................................... 2048
202
+ hierarchical_context_parallel_sizes ............. None
203
+ high_priority_stream_groups ..................... []
204
+ hybrid_attention_ratio .......................... 0.0
205
+ hybrid_mlp_ratio ................................ 0.0
206
+ hybrid_override_pattern ......................... None
207
+ hysteresis ...................................... 2
208
+ ict_head_size ................................... None
209
+ ict_load ........................................ None
210
+ img_h ........................................... 224
211
+ img_w ........................................... 224
212
+ indexer_batch_size .............................. 128
213
+ indexer_log_interval ............................ 1000
214
+ inference_batch_times_seqlen_threshold .......... -1
215
+ inference_dynamic_batching ...................... False
216
+ inference_dynamic_batching_buffer_guaranteed_fraction 0.2
217
+ inference_dynamic_batching_buffer_overflow_factor None
218
+ inference_dynamic_batching_buffer_size_gb ....... 40.0
219
+ inference_dynamic_batching_chunk_size ........... 256
220
+ inference_dynamic_batching_max_requests_override None
221
+ inference_dynamic_batching_max_tokens_override .. None
222
+ inference_dynamic_batching_num_cuda_graphs ...... 16
223
+ inference_max_batch_size ........................ 8
224
+ inference_max_seq_length ........................ 2560
225
+ inference_rng_tracker ........................... False
226
+ init_method_std ................................. 0.02
227
+ init_method_xavier_uniform ...................... False
228
+ init_model_with_meta_device ..................... False
229
+ initial_loss_scale .............................. 4294967296
230
+ inprocess_active_world_size ..................... 8
231
+ inprocess_barrier_timeout ....................... 120
232
+ inprocess_completion_timeout .................... 120
233
+ inprocess_empty_cuda_cache ...................... False
234
+ inprocess_granularity ........................... node
235
+ inprocess_hard_timeout .......................... 90
236
+ inprocess_heartbeat_interval .................... 30
237
+ inprocess_heartbeat_timeout ..................... 60
238
+ inprocess_last_call_wait ........................ 1
239
+ inprocess_max_iterations ........................ None
240
+ inprocess_monitor_process_interval .............. 1.0
241
+ inprocess_monitor_thread_interval ............... 1.0
242
+ inprocess_progress_watchdog_interval ............ 1.0
243
+ inprocess_restart ............................... False
244
+ inprocess_soft_timeout .......................... 60
245
+ inprocess_termination_grace_time ................ 1
246
+ is_hybrid_model ................................. False
247
+ iter_per_epoch .................................. 1250
248
+ iterations_to_skip .............................. []
249
+ keep_fp8_transpose_cache_when_using_custom_fsdp . False
250
+ kitchen_config_file ............................. None
251
+ kitchen_recipe_number ........................... None
252
+ kv_channels ..................................... 128
253
+ kv_lora_rank .................................... 32
254
+ lazy_mpu_init ................................... None
255
+ load ............................................ None
256
+ load_main_params_from_ckpt ...................... None
257
+ load_model_opt_format ........................... False
258
+ local_rank ...................................... 0
259
+ log_energy ...................................... False
260
+ log_interval .................................... 1
261
+ log_loss_scale_to_tensorboard ................... True
262
+ log_memory_to_tensorboard ....................... False
263
+ log_num_zeros_in_grad ........................... False
264
+ log_params_norm ................................. False
265
+ log_progress .................................... False
266
+ log_straggler ................................... False
267
+ log_throughput .................................. True
268
+ log_timers_to_tensorboard ....................... False
269
+ log_validation_ppl_to_tensorboard ............... False
270
+ log_world_size_to_tensorboard ................... False
271
+ logging_level ................................... None
272
+ loss_scale ...................................... None
273
+ loss_scale_window ............................... 1000
274
+ lr .............................................. 0.0001
275
+ lr_decay_iters .................................. 8
276
+ lr_decay_samples ................................ None
277
+ lr_decay_style .................................. cosine
278
+ lr_warmup_fraction .............................. None
279
+ lr_warmup_init .................................. 0.0
280
+ lr_warmup_iters ................................. 5
281
+ lr_warmup_samples ............................... 0
282
+ lr_wsd_decay_iters .............................. None
283
+ lr_wsd_decay_samples ............................ None
284
+ lr_wsd_decay_style .............................. exponential
285
+ main_grads_dtype ................................ torch.float32
286
+ main_params_dtype ............................... torch.float32
287
+ make_vocab_size_divisible_by .................... 128
288
+ mamba_head_dim .................................. 64
289
+ mamba_num_groups ................................ 8
290
+ mamba_num_heads ................................. None
291
+ mamba_state_dim ................................. 128
292
+ manual_gc ....................................... False
293
+ manual_gc_eval .................................. True
294
+ manual_gc_interval .............................. 0
295
+ mask_factor ..................................... 1.0
296
+ mask_prob ....................................... 0.15
297
+ mask_type ....................................... random
298
+ masked_softmax_fusion ........................... False
299
+ max_position_embeddings ......................... 4096
300
+ max_tokens_to_oom ............................... 12000
301
+ memory_snapshot_path ............................ snapshot.pickle
302
+ merge_file ...................................... /workspace/data/wikitext/gpt2-merges.txt
303
+ micro_batch_size ................................ 1
304
+ microbatch_group_size_per_vp_stage .............. None
305
+ mid_level_dataset_surplus ....................... 0.005
306
+ min_loss_scale .................................. 1.0
307
+ min_lr .......................................... 1e-05
308
+ mlp_chunks_for_prefill .......................... 1
309
+ mmap_bin_files .................................. True
310
+ mock_data ....................................... False
311
+ moe_apply_probs_on_input ........................ False
312
+ moe_aux_loss_coeff .............................. 0.01
313
+ moe_capacity_monitor ............................ True
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 ............................. True
322
+ moe_enable_ultraep .............................. True
323
+ moe_expert_capacity_factor ...................... None
324
+ moe_extended_tp ................................. False
325
+ moe_ffn_hidden_size ............................. 1024
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_num_echo_experts ............................ None
332
+ moe_num_redundant_experts_per_rank .............. 8
333
+ moe_pad_expert_input_to_capacity ................ False
334
+ moe_per_layer_logging ........................... False
335
+ moe_permute_fusion .............................. False
336
+ moe_received_token_capacity ..................... 1.08984375
337
+ moe_router_bias_update_rate ..................... 0.001
338
+ moe_router_dtype ................................ fp32
339
+ moe_router_enable_expert_bias ................... False
340
+ moe_router_force_load_balancing ................. False
341
+ moe_router_group_topk ........................... None
342
+ moe_router_load_balancing_type .................. aux_loss
343
+ moe_router_num_groups ........................... None
344
+ moe_router_padding_for_fp8 ...................... False
345
+ moe_router_pre_softmax .......................... True
346
+ moe_router_score_function ....................... softmax
347
+ moe_router_topk ................................. 8
348
+ moe_router_topk_scaling_factor .................. None
349
+ moe_shared_expert_intermediate_size ............. None
350
+ moe_shared_expert_overlap ....................... False
351
+ moe_token_dispatcher_type ....................... flex
352
+ moe_token_drop_policy ........................... probs
353
+ moe_ultraep_max_microbatches .................... None
354
+ moe_ultraep_overlap_weight_sync ................. True
355
+ moe_ultraep_weight_sync_join .................... before_dispatch
356
+ moe_upcycling_granularity ....................... 1
357
+ moe_use_device_initiated_grouped_gemm ........... True
358
+ moe_use_legacy_grouped_gemm ..................... False
359
+ moe_use_upcycling ............................... False
360
+ moe_z_loss_coeff ................................ None
361
+ mrope_section ................................... None
362
+ mscale .......................................... 1.0
363
+ mscale_all_dim .................................. 0.0
364
+ mtp_loss_scaling_factor ......................... 0.1
365
+ mtp_num_layers .................................. None
366
+ multi_latent_attention .......................... False
367
+ nccl_all_reduce_for_prefill ..................... False
368
+ nccl_communicator_config_path ................... None
369
+ nccl_ub ......................................... False
370
+ no_load_optim ................................... None
371
+ no_load_rng ..................................... None
372
+ no_persist_layer_norm ........................... False
373
+ no_rope_freq .................................... None
374
+ no_save_optim ................................... None
375
+ no_save_rng ..................................... None
376
+ non_persistent_ckpt_type ........................ None
377
+ non_persistent_global_ckpt_dir .................. None
378
+ non_persistent_local_ckpt_algo .................. fully_parallel
379
+ non_persistent_local_ckpt_dir ................... None
380
+ non_persistent_save_interval .................... None
381
+ norm_epsilon .................................... 1e-05
382
+ normalization ................................... RMSNorm
383
+ num_attention_heads ............................. 16
384
+ num_channels .................................... 3
385
+ num_classes ..................................... 1000
386
+ num_dataset_builder_threads ..................... 1
387
+ num_distributed_optimizer_instances ............. 1
388
+ num_experts ..................................... 64
389
+ num_layers ...................................... 16
390
+ num_layers_at_end_in_bf16 ....................... 1
391
+ num_layers_at_start_in_bf16 ..................... 1
392
+ num_layers_per_virtual_pipeline_stage ........... None
393
+ num_query_groups ................................ 16
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 .................................... False
433
+ qk_pos_emb_head_dim ............................. 64
434
+ query_in_block_prob ............................. 0.1
435
+ rampup_batch_size ............................... None
436
+ rank ............................................ 0
437
+ recompute_granularity ........................... None
438
+ recompute_method ................................ None
439
+ recompute_modules ............................... None
440
+ recompute_num_layers ............................ None
441
+ record_memory_history ........................... False
442
+ relative_attention_max_distance ................. 128
443
+ relative_attention_num_buckets .................. 32
444
+ replication ..................................... False
445
+ replication_factor .............................. 2
446
+ replication_jump ................................ None
447
+ rerun_mode ...................................... validate_results
448
+ reset_attention_mask ............................ False
449
+ reset_position_ids .............................. False
450
+ result_rejected_tracker_filename ................ None
451
+ retriever_report_topk_accuracies ................ []
452
+ retriever_score_scaling ......................... False
453
+ retriever_seq_length ............................ 256
454
+ retro_add_retriever ............................. False
455
+ retro_attention_gate ............................ 1
456
+ retro_cyclic_train_iters ........................ None
457
+ retro_encoder_attention_dropout ................. 0.1
458
+ retro_encoder_hidden_dropout .................... 0.1
459
+ retro_encoder_layers ............................ 2
460
+ retro_num_neighbors ............................. 2
461
+ retro_num_retrieved_chunks ...................... 2
462
+ retro_project_dir ............................... None
463
+ retro_verify_neighbor_count ..................... True
464
+ reuse_grad_buf_for_mxfp8_param_ag ............... False
465
+ rope_scaling_factor ............................. 8.0
466
+ rope_type ....................................... None
467
+ rotary_base ..................................... 10000
468
+ rotary_interleaved .............................. False
469
+ rotary_percent .................................. 1.0
470
+ rotary_scaling_factor ........................... 1.0
471
+ rotary_seq_len_interpolation_factor ............. None
472
+ run_workload_inspector_server ................... False
473
+ sample_rate ..................................... 1.0
474
+ save ............................................ None
475
+ save_interval ................................... None
476
+ save_retain_interval ............................ None
477
+ scatter_gather_tensors_in_pipeline .............. True
478
+ seed ............................................ 1234
479
+ seq_length ...................................... 4096
480
+ sequence_parallel ............................... False
481
+ sft ............................................. False
482
+ sft_tokenizer_prompt_format ..................... nemotron-h-aligned
483
+ sgd_momentum .................................... 0.9
484
+ short_seq_prob .................................. 0.1
485
+ skip_train ...................................... False
486
+ skipped_train_samples ........................... 0
487
+ spec ............................................ None
488
+ split ........................................... 100,0,0
489
+ squared_relu .................................... False
490
+ start_weight_decay .............................. 0.1
491
+ straggler_ctrlr_port ............................ 65535
492
+ straggler_minmax_count .......................... 1
493
+ suggested_communication_unit_size ............... None
494
+ swiglu .......................................... True
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
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
574
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
575
+ > padded vocab (size: 50257) with 47 dummy tokens (new size: 50304)
576
+ WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.VALIDATE_RESULTS
577
+ > initializing torch distributed ...
578
+ /workspace/Megatron-UltraEP/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
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
581
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
582
+ [Gloo] Rank 2 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 5 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
585
+ [Gloo] Rank [Gloo] Rank [Gloo] Rank [Gloo] Rank 4637 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
586
+
587
+
588
+
589
+ [Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
590
+ [Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
591
+ [Gloo] Rank [Gloo] Rank [Gloo] Rank [Gloo] Rank [Gloo] Rank 164 is connected to 3 is connected to [Gloo] Rank is connected to 5[Gloo] Rank 7 is connected to 77 is connected to peer ranks. 77 peer ranks. peer ranks. 7Expected number of connected peer ranks is : is connected to peer ranks. Expected number of connected peer ranks is : Expected number of connected peer ranks is : peer ranks. 77
592
+ Expected number of connected peer ranks is : 77Expected number of connected peer ranks is : peer ranks. 7
593
+
594
+
595
+ 7Expected number of connected peer ranks is :
596
+ 7
597
+ 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
598
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
599
+ > initialized tensor model parallel with size 1
600
+ > initialized pipeline model parallel with size 1
601
+ > setting random seeds to 1234 ...
602
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
603
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
604
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
605
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
606
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
607
+ [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
608
+
609
+ [rank1]:[W921 08:37:23.431240675 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]:[W921 08:37:23.431249514 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
+ [rank5]:[W921 08:37:23.431328953 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.
612
+ [rank6]:[W921 08:37:23.431330166 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.
613
+ [rank7]:[W921 08:37:23.431410535 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
+ [rank2]:[W921 08:37:23.431430255 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.
615
+ [rank3]:[W921 08:37:23.433029110 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.
616
+ > compiling dataset index builder ...
617
+ make: Entering directory '/workspace/Megatron-UltraEP/megatron/core/datasets'
618
+ make: Nothing to be done for 'default'.
619
+ make: Leaving directory '/workspace/Megatron-UltraEP/megatron/core/datasets'
620
+ >>> done with dataset index builder. Compilation time: 0.034 seconds
621
+ WARNING: constraints for invoking optimized fused softmax kernel are not met. We default back to unfused kernel invocations.
622
+ > compiling and loading fused kernels ...
623
+ [rank0]:[W921 08:37:23.586409329 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.
624
+ >>> done with compiling and loading fused kernels. Compilation time: 3.850 seconds
625
+ time to initialize megatron (seconds): 7.097
626
+ [after megatron is initialized] datetime: 2026-09-21 08:37:29
627
+ building GPT model ...
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
+ /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.
643
+ warnings.warn(
644
+ [INFO] Use MNNVL fabric: False
645
+ [WARN] MAX_NUM_NVL_PEERS is not set. Using detected value 8.
646
+ NVSHMEM v3.7.2
647
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
648
+
649
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
650
+
651
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
652
+
653
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
654
+
655
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
656
+
657
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
658
+
659
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
660
+
661
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
662
+
663
+ UltraEP: NVL domain size <= 8; forcing weight_sync plan mode to direct (was adaptive).
664
+ > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 2087258112
665
+ 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)
666
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 7
667
+ Params for bucket 1 (103022592 elements, 103022592 padded size):
668
+ module.output_layer.weight
669
+ Params for bucket 2 (46542848 elements, 46542848 padded size):
670
+ module.decoder.final_layernorm.weight
671
+ module.decoder.layers.15.mlp.router.weight
672
+ module.decoder.layers.15.self_attention.linear_qkv.layer_norm_weight
673
+ module.decoder.layers.15.self_attention.linear_proj.weight
674
+ module.decoder.layers.14.mlp.router.weight
675
+ module.decoder.layers.14.self_attention.linear_qkv.layer_norm_weight
676
+ module.decoder.layers.14.self_attention.linear_proj.weight
677
+ module.decoder.layers.13.pre_mlp_layernorm.weight
678
+ module.decoder.layers.13.self_attention.linear_qkv.weight
679
+ module.decoder.layers.15.self_attention.linear_qkv.weight
680
+ module.decoder.layers.15.pre_mlp_layernorm.weight
681
+ module.decoder.layers.14.pre_mlp_layernorm.weight
682
+ module.decoder.layers.14.self_attention.linear_qkv.weight
683
+ module.decoder.layers.13.mlp.router.weight
684
+ Params for bucket 3 (50737152 elements, 50737152 padded size):
685
+ module.decoder.layers.12.pre_mlp_layernorm.weight
686
+ module.decoder.layers.12.self_attention.linear_qkv.weight
687
+ module.decoder.layers.11.pre_mlp_layernorm.weight
688
+ module.decoder.layers.11.self_attention.linear_qkv.weight
689
+ module.decoder.layers.10.pre_mlp_layernorm.weight
690
+ module.decoder.layers.10.self_attention.linear_qkv.weight
691
+ module.decoder.layers.13.self_attention.linear_qkv.layer_norm_weight
692
+ module.decoder.layers.12.mlp.router.weight
693
+ module.decoder.layers.13.self_attention.linear_proj.weight
694
+ module.decoder.layers.12.self_attention.linear_qkv.layer_norm_weight
695
+ module.decoder.layers.12.self_attention.linear_proj.weight
696
+ module.decoder.layers.11.mlp.router.weight
697
+ module.decoder.layers.11.self_attention.linear_qkv.layer_norm_weight
698
+ module.decoder.layers.11.self_attention.linear_proj.weight
699
+ module.decoder.layers.10.mlp.router.weight
700
+ Params for bucket 4 (50737152 elements, 50737152 padded size):
701
+ module.decoder.layers.9.self_attention.linear_qkv.weight
702
+ module.decoder.layers.9.pre_mlp_layernorm.weight
703
+ module.decoder.layers.8.pre_mlp_layernorm.weight
704
+ module.decoder.layers.8.self_attention.linear_qkv.weight
705
+ module.decoder.layers.7.pre_mlp_layernorm.weight
706
+ module.decoder.layers.7.self_attention.linear_qkv.weight
707
+ module.decoder.layers.10.self_attention.linear_qkv.layer_norm_weight
708
+ module.decoder.layers.10.self_attention.linear_proj.weight
709
+ module.decoder.layers.9.mlp.router.weight
710
+ module.decoder.layers.9.self_attention.linear_qkv.layer_norm_weight
711
+ module.decoder.layers.9.self_attention.linear_proj.weight
712
+ module.decoder.layers.8.mlp.router.weight
713
+ module.decoder.layers.8.self_attention.linear_qkv.layer_norm_weight
714
+ module.decoder.layers.8.self_attention.linear_proj.weight
715
+ module.decoder.layers.7.mlp.router.weight
716
+ Params for bucket 5 (50737152 elements, 50737152 padded size):
717
+ module.decoder.layers.6.mlp.router.weight
718
+ module.decoder.layers.6.self_attention.linear_qkv.layer_norm_weight
719
+ module.decoder.layers.6.self_attention.linear_proj.weight
720
+ module.decoder.layers.5.pre_mlp_layernorm.weight
721
+ module.decoder.layers.5.self_attention.linear_qkv.weight
722
+ module.decoder.layers.5.self_attention.linear_proj.weight
723
+ module.decoder.layers.4.mlp.router.weight
724
+ module.decoder.layers.7.self_attention.linear_qkv.layer_norm_weight
725
+ module.decoder.layers.7.self_attention.linear_proj.weight
726
+ module.decoder.layers.6.pre_mlp_layernorm.weight
727
+ module.decoder.layers.6.self_attention.linear_qkv.weight
728
+ module.decoder.layers.5.mlp.router.weight
729
+ module.decoder.layers.5.self_attention.linear_qkv.layer_norm_weight
730
+ module.decoder.layers.4.pre_mlp_layernorm.weight
731
+ module.decoder.layers.4.self_attention.linear_qkv.weight
732
+ Params for bucket 6 (50737152 elements, 50737152 padded size):
733
+ module.decoder.layers.4.self_attention.linear_qkv.layer_norm_weight
734
+ module.decoder.layers.4.self_attention.linear_proj.weight
735
+ module.decoder.layers.3.mlp.router.weight
736
+ module.decoder.layers.3.self_attention.linear_qkv.layer_norm_weight
737
+ module.decoder.layers.3.self_attention.linear_proj.weight
738
+ module.decoder.layers.2.mlp.router.weight
739
+ module.decoder.layers.2.self_attention.linear_qkv.layer_norm_weight
740
+ module.decoder.layers.2.self_attention.linear_proj.weight
741
+ module.decoder.layers.1.pre_mlp_layernorm.weight
742
+ module.decoder.layers.1.self_attention.linear_qkv.weight
743
+ module.decoder.layers.3.pre_mlp_layernorm.weight
744
+ module.decoder.layers.3.self_attention.linear_qkv.weight
745
+ module.decoder.layers.2.pre_mlp_layernorm.weight
746
+ module.decoder.layers.2.self_attention.linear_qkv.weight
747
+ module.decoder.layers.1.mlp.router.weight
748
+ Params for bucket 7 (124131328 elements, 124131328 padded size):
749
+ module.decoder.layers.0.self_attention.linear_qkv.weight
750
+ module.decoder.layers.0.pre_mlp_layernorm.weight
751
+ module.decoder.layers.0.mlp.router.weight
752
+ module.embedding.word_embeddings.weight
753
+ module.decoder.layers.1.self_attention.linear_qkv.layer_norm_weight
754
+ module.decoder.layers.1.self_attention.linear_proj.weight
755
+ module.decoder.layers.0.self_attention.linear_qkv.layer_norm_weight
756
+ module.decoder.layers.0.self_attention.linear_proj.weight
757
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 20
758
+ Params for bucket 1 (41943040 elements, 41943040 padded size):
759
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight7
760
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight5
761
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight3
762
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight1
763
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight6
764
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight4
765
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight2
766
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight6
767
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight4
768
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight2
769
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight0
770
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight7
771
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight5
772
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight3
773
+ Params for bucket 2 (41943040 elements, 41943040 padded size):
774
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight6
775
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight0
776
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight4
777
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight2
778
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight0
779
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight6
780
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight4
781
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight1
782
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight5
783
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight7
784
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight3
785
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight1
786
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight7
787
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight5
788
+ Params for bucket 3 (41943040 elements, 41943040 padded size):
789
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight2
790
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight0
791
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight6
792
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight4
793
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight2
794
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight0
795
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight7
796
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight3
797
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight7
798
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight1
799
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight5
800
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight3
801
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight1
802
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight6
803
+ Params for bucket 4 (41943040 elements, 41943040 padded size):
804
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight5
805
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight1
806
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight3
807
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight6
808
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight4
809
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight2
810
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight0
811
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight2
812
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight4
813
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight0
814
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight7
815
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight5
816
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight3
817
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight1
818
+ Params for bucket 5 (41943040 elements, 41943040 padded size):
819
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight7
820
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight3
821
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight5
822
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight1
823
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight6
824
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight4
825
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight6
826
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight4
827
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight2
828
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight0
829
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight7
830
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight5
831
+ Params for bucket 6 (41943040 elements, 41943040 padded size):
832
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight7
833
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight2
834
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight0
835
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight5
836
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight3
837
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight1
838
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight3
839
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight1
840
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight6
841
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight4
842
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight2
843
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight0
844
+ Params for bucket 7 (41943040 elements, 41943040 padded size):
845
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight6
846
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight4
847
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight2
848
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight0
849
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight7
850
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight5
851
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight3
852
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight7
853
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight3
854
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight5
855
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight1
856
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight6
857
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight4
858
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight2
859
+ Params for bucket 8 (41943040 elements, 41943040 padded size):
860
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight1
861
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight6
862
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight4
863
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight2
864
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight0
865
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight7
866
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight5
867
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight7
868
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight0
869
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight5
870
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight3
871
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight1
872
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight6
873
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight4
874
+ Params for bucket 9 (41943040 elements, 41943040 padded size):
875
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight3
876
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight1
877
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight6
878
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight4
879
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight2
880
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight0
881
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight7
882
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight0
883
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight7
884
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight2
885
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight5
886
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight3
887
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight1
888
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight6
889
+ Params for bucket 10 (41943040 elements, 41943040 padded size):
890
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight5
891
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight3
892
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight1
893
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight6
894
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight4
895
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight2
896
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight0
897
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight4
898
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight2
899
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight0
900
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight7
901
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight5
902
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight3
903
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight1
904
+ Params for bucket 11 (41943040 elements, 41943040 padded size):
905
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight7
906
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight3
907
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight5
908
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight1
909
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight6
910
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight4
911
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight6
912
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight4
913
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight2
914
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight0
915
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight7
916
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight5
917
+ Params for bucket 12 (41943040 elements, 41943040 padded size):
918
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight0
919
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight2
920
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight6
921
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight4
922
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight2
923
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight0
924
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight3
925
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight1
926
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight7
927
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight5
928
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight3
929
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight1
930
+ Params for bucket 13 (41943040 elements, 41943040 padded size):
931
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight7
932
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight3
933
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight5
934
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight1
935
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight6
936
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight4
937
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight2
938
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight4
939
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight6
940
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight2
941
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight0
942
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight7
943
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight5
944
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight3
945
+ Params for bucket 14 (41943040 elements, 41943040 padded size):
946
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight7
947
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight5
948
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight0
949
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight3
950
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight1
951
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight6
952
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight4
953
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight1
954
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight6
955
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight4
956
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight2
957
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight0
958
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight7
959
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight5
960
+ Params for bucket 15 (41943040 elements, 41943040 padded size):
961
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight0
962
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight7
963
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight2
964
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight5
965
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight3
966
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight1
967
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight6
968
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight3
969
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight1
970
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight6
971
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight4
972
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight2
973
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight0
974
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight7
975
+ Params for bucket 16 (41943040 elements, 41943040 padded size):
976
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight2
977
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight4
978
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight0
979
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight6
980
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight4
981
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight2
982
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight0
983
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight5
984
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight1
985
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight3
986
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight7
987
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight5
988
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight3
989
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight1
990
+ Params for bucket 17 (41943040 elements, 41943040 padded size):
991
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight4
992
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight6
993
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight2
994
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight0
995
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight6
996
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight4
997
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight7
998
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight3
999
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight5
1000
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight1
1001
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight7
1002
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight5
1003
+ Params for bucket 18 (41943040 elements, 41943040 padded size):
1004
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight7
1005
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight2
1006
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight0
1007
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight5
1008
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight3
1009
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight1
1010
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight3
1011
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight1
1012
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight6
1013
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight4
1014
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight2
1015
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight0
1016
+ Params for bucket 19 (41943040 elements, 41943040 padded size):
1017
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight6
1018
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight4
1019
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight2
1020
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight0
1021
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight7
1022
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight5
1023
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight3
1024
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight7
1025
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight3
1026
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight5
1027
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight1
1028
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight6
1029
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight4
1030
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight2
1031
+ Params for bucket 20 (8388608 elements, 8388608 padded size):
1032
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight1
1033
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight0
1034
+ 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 0x7e9bdee7c620>, config_logger_dir='')
1035
+ INFO:megatron.core.optimizer_param_scheduler:> learning rate decay style: cosine
1036
+ [after model, optimizer, and learning rate scheduler are built] datetime: 2026-09-21 08:37:33
1037
+ > building train, validation, and test datasets ...
1038
+ > datasets target sizes (minimum size):
1039
+ train: 96
1040
+ validation: 0
1041
+ test: 0
1042
+ INFO:megatron.core.datasets.blended_megatron_dataset_config:Let split_matrix = [(0, 1.0), None, None]
1043
+ > building train, validation, and test datasets for GPT ...
1044
+ INFO:megatron.core.datasets.blended_megatron_dataset_builder:Building GPTDataset splits with sizes=(96, 0, 0) and config=GPTDatasetConfig(random_seed=1234, sequence_length=4096, blend=(['/workspace/data/wikitext/wikitext103_gpt2_text_document'], None), blend_per_split=None, split='100,0,0', split_matrix=[(0, 1.0), None, None], num_dataset_builder_threads=1, path_to_cache=None, mmap_bin_files=True, mock=False, tokenizer=<megatron.training.tokenizer.tokenizer._GPT2BPETokenizer object at 0x7e9be062e900>, 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)
1045
+ INFO:megatron.core.datasets.indexed_dataset:Load the _IndexReader from /workspace/data/wikitext/wikitext103_gpt2_text_document.idx
1046
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence lengths
1047
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence pointers
1048
+ INFO:megatron.core.datasets.indexed_dataset: Extract the document indices
1049
+ INFO:megatron.core.datasets.indexed_dataset:> total number of sequences: 1165029
1050
+ INFO:megatron.core.datasets.indexed_dataset:> total number of documents: 1801350
1051
+ INFO:megatron.core.datasets.gpt_dataset:Load the GPTDataset train indices
1052
+ INFO:megatron.core.datasets.gpt_dataset: Load the document index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-document_index.npy
1053
+ INFO:megatron.core.datasets.gpt_dataset: Load the sample index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-sample_index.npy
1054
+ INFO:megatron.core.datasets.gpt_dataset: Load the shuffle index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-shuffle_index.npy
1055
+ INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 29073
1056
+ > finished creating GPT datasets ...
1057
+ [after dataloaders are built] datetime: 2026-09-21 08:37:33
1058
+ done with setup ...
1059
+ (min, max) time across ranks (ms):
1060
+ model-and-optimizer-setup ......................: (3526.64, 3528.51)
1061
+ train/valid/test-data-iterators-setup ..........: (71.74, 92.66)
1062
+ training ...
1063
+ Overwriting rerun_state_machine.current_iteration from -1 to 0...
1064
+ [before the start of training step] datetime: 2026-09-21 08:37:33
1065
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1066
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1067
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1068
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1069
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1070
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1071
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1072
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1073
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1074
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1075
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1076
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1077
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1078
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1079
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1080
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1081
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1082
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1083
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1084
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1085
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1086
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1087
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1088
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1089
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1090
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1091
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1092
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1093
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1094
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1095
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1096
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1097
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1098
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1099
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1100
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1101
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1102
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1103
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1104
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1105
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1106
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1107
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1108
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1109
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1110
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1111
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1112
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1113
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1114
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1115
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1116
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1117
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1118
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1119
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1120
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1121
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1122
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1123
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1124
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1125
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1126
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1127
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1128
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1129
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1130
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1131
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1132
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1133
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1134
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1135
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1136
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1137
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1138
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1139
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1140
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1141
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1142
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1143
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1144
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1145
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1146
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1147
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1148
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1149
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1150
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1151
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1152
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1153
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1154
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1155
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1156
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1157
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1158
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1159
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1160
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1161
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1162
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1163
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1164
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1165
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1166
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1167
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1168
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1169
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1170
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1171
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1172
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1173
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1174
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1175
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1176
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1177
+ [2026-09-21 08:37:58] iteration 1/ 12 | consumed samples: 8 | elapsed time per iteration (ms): 25703.7 | throughput per GPU (TFLOP/s/GPU): 1.3 | learning rate: 2.000000E-05 | global batch size: 8 | lm loss: 1.118507E+01 | loss scale: 1.0 | grad norm: 35.484 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 33920 | min safe capacity: 1.0352 |Number of parameters in transformer block in billions: 6.71
1178
+
1179
+ Number of parameters in embedding layers in billions: 0.21
1180
+ Total number of parameters in billions: 6.92
1181
+ Number of parameters in most loaded shard in billions: 6.9171
1182
+ Theoretical memory footprints: weight and optimizer=49474.72 MB
1183
+ [Rank 0] (after 1 iterations) memory (MB) | allocated: 18080.58447265625 | max allocated: 21748.86669921875 | reserved: 22168.0 | max reserved: 22168.0
1184
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1185
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1186
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1187
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1188
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1189
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1190
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1191
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1192
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1193
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1194
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1195
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1196
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1197
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1198
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1199
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1200
+ [2026-09-21 08:37:59] iteration 2/ 12 | consumed samples: 16 | elapsed time per iteration (ms): 247.7 | throughput per GPU (TFLOP/s/GPU): 130.1 | learning rate: 4.000000E-05 | global batch size: 8 | lm loss: 1.120943E+01 | loss scale: 1.0 | grad norm: 36.301 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 33920 | min safe capacity: 1.0352 |
1201
+ [2026-09-21 08:37:59] iteration 3/ 12 | consumed samples: 24 | elapsed time per iteration (ms): 552.6 | throughput per GPU (TFLOP/s/GPU): 58.3 | learning rate: 6.000000E-05 | global batch size: 8 | lm loss: 1.006678E+01 | loss scale: 1.0 | grad norm: 101.512 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 33920 | min safe capacity: 1.0352 |
1202
+ Capture CUDA graph for training!!!
1203
+ CUDA graph capture done!!!
1204
+ [2026-09-21 08:38:00] iteration 4/ 12 | consumed samples: 32 | elapsed time per iteration (ms): 798.2 | throughput per GPU (TFLOP/s/GPU): 40.4 | learning rate: 8.000000E-05 | global batch size: 8 | lm loss: 1.023777E+01 | loss scale: 1.0 | grad norm: 28.800 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 33920 | min safe capacity: 1.0352 |
1205
+ [2026-09-21 08:38:00] iteration 5/ 12 | consumed samples: 40 | elapsed time per iteration (ms): 104.1 | throughput per GPU (TFLOP/s/GPU): 309.4 | learning rate: 1.000000E-04 | global batch size: 8 | lm loss: 1.073676E+01 | loss scale: 1.0 | grad norm: 20.808 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 33920 | min safe capacity: 1.0352 |
1206
+ [2026-09-21 08:38:00] iteration 6/ 12 | consumed samples: 48 | elapsed time per iteration (ms): 104.2 | throughput per GPU (TFLOP/s/GPU): 309.3 | learning rate: 7.750000E-05 | global batch size: 8 | lm loss: 9.626024E+00 | loss scale: 1.0 | grad norm: 8.173 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 34048 | min safe capacity: 1.0391 |
1207
+ [2026-09-21 08:38:00] iteration 7/ 12 | consumed samples: 56 | elapsed time per iteration (ms): 103.2 | throughput per GPU (TFLOP/s/GPU): 312.3 | learning rate: 3.250000E-05 | global batch size: 8 | lm loss: 9.077704E+00 | loss scale: 1.0 | grad norm: 16.540 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 34048 | min safe capacity: 1.0391 |
1208
+ [2026-09-21 08:38:00] iteration 8/ 12 | consumed samples: 64 | elapsed time per iteration (ms): 103.4 | throughput per GPU (TFLOP/s/GPU): 311.7 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.816444E+00 | loss scale: 1.0 | grad norm: 8.716 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 34048 | min safe capacity: 1.0391 |
1209
+ [2026-09-21 08:38:01] iteration 9/ 12 | consumed samples: 72 | elapsed time per iteration (ms): 103.8 | throughput per GPU (TFLOP/s/GPU): 310.3 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.709606E+00 | loss scale: 1.0 | grad norm: 3.792 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 34048 | min safe capacity: 1.0391 |
1210
+ [2026-09-21 08:38:01] iteration 10/ 12 | consumed samples: 80 | elapsed time per iteration (ms): 103.7 | throughput per GPU (TFLOP/s/GPU): 310.6 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.652786E+00 | loss scale: 1.0 | grad norm: 3.495 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 34048 | min safe capacity: 1.0391 |
1211
+ [2026-09-21 08:38:01] iteration 11/ 12 | consumed samples: 88 | elapsed time per iteration (ms): 103.3 | throughput per GPU (TFLOP/s/GPU): 311.9 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.641448E+00 | loss scale: 1.0 | grad norm: 3.468 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 34048 | min safe capacity: 1.0391 |
1212
+ [2026-09-21 08:38:01] iteration 12/ 12 | consumed samples: 96 | elapsed time per iteration (ms): 102.0 | throughput per GPU (TFLOP/s/GPU): 316.0 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.591211E+00 | loss scale: 1.0 | grad norm: 3.152 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 34048 | min safe capacity: 1.0391 |
1213
+ [after training is done] datetime: 2026-09-21 08:38:01
1214
+ [rank0]:[W921 08:38:02.361969628 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())
1215
+ [rank4]:[W921 08:38:03.632670301 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())
1216
+ [rank3]:[W921 08:38:03.638645185 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())
1217
+ [rank2]:[W921 08:38:03.646140146 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())
1218
+ [rank7]:[W921 08:38:03.646567799 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())
1219
+ [rank5]:[W921 08:38:03.649296049 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())
1220
+ [rank1]:[W921 08:38:03.865345681 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())
1221
+ [rank6]:[W921 08:38:04.021920427 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())
1222
+ W0921 08:38:05.963000 694969 torch/distributed/elastic/agent/server/api.py:719] Received 15 death signal, shutting down workers
1223
+ W0921 08:38:05.964000 694969 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 695238 closing signal SIGTERM
1224
+ W0921 08:38:05.964000 694969 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 695239 closing signal SIGTERM
1225
+ W0921 08:38:05.965000 694969 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 695241 closing signal SIGTERM
1226
+ W0921 08:38:05.965000 694969 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 695242 closing signal SIGTERM
1227
+ W0921 08:38:05.966000 694969 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 695244 closing signal SIGTERM
1228
+ W0921 08:38:05.966000 694969 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 695245 closing signal SIGTERM
1229
+ W0921 08:38:05.967000 694969 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 695246 closing signal SIGTERM
1230
+ W0921 08:38:05.967000 694969 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 695247 closing signal SIGTERM
1231
+ Traceback (most recent call last):
1232
+ File "/usr/local/bin/torchrun", line 8, in <module>
1233
+ sys.exit(main())
1234
+ ^^^^^^
1235
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper
1236
+ return f(*args, **kwargs)
1237
+ ^^^^^^^^^^^^^^^^^^
1238
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 892, in main
1239
+ run(args)
1240
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 883, in run
1241
+ elastic_launch(
1242
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 139, in __call__
1243
+ return launch_agent(self._config, self._entrypoint, list(args))
1244
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1245
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 261, in launch_agent
1246
+ result = agent.run()
1247
+ ^^^^^^^^^^^
1248
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/metrics/api.py", line 138, in wrapper
1249
+ result = f(*args, **kwargs)
1250
+ ^^^^^^^^^^^^^^^^^^
1251
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 711, in run
1252
+ result = self._invoke_run(role)
1253
+ ^^^^^^^^^^^^^^^^^^^^^^
1254
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 870, in _invoke_run
1255
+ time.sleep(monitor_interval)
1256
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/api.py", line 84, in _terminate_process_handler
1257
+ raise SignalException(f"Process {os.getpid()} got signal: {sigval}", sigval=sigval)
1258
+ torch.distributed.elastic.multiprocessing.api.SignalException: Process 694969 got signal: 15
granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep2/mem_samples.csv ADDED
@@ -0,0 +1,208 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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, 0
10
+ 1, 0
11
+ 2, 0
12
+ 3, 0
13
+ 4, 0
14
+ 5, 0
15
+ 6, 0
16
+ 7, 0
17
+ 0, 4
18
+ 1, 2
19
+ 2, 1
20
+ 3, 1
21
+ 4, 1
22
+ 5, 1
23
+ 6, 1
24
+ 7, 1
25
+ 0, 4
26
+ 1, 4
27
+ 2, 4
28
+ 3, 4
29
+ 4, 4
30
+ 5, 1
31
+ 6, 1
32
+ 7, 1
33
+ 0, 4
34
+ 1, 4
35
+ 2, 4
36
+ 3, 4
37
+ 4, 4
38
+ 5, 4
39
+ 6, 4
40
+ 7, 4
41
+ 0, 4
42
+ 1, 4
43
+ 2, 4
44
+ 3, 4
45
+ 4, 4
46
+ 5, 4
47
+ 6, 4
48
+ 7, 4
49
+ 0, 821
50
+ 1, 821
51
+ 2, 819
52
+ 3, 819
53
+ 4, 819
54
+ 5, 821
55
+ 6, 819
56
+ 7, 819
57
+ 0, 1537
58
+ 1, 1537
59
+ 2, 1537
60
+ 3, 1537
61
+ 4, 1537
62
+ 5, 1537
63
+ 6, 1537
64
+ 7, 1537
65
+ 0, 1823
66
+ 1, 1823
67
+ 2, 1823
68
+ 3, 1823
69
+ 4, 1823
70
+ 5, 1823
71
+ 6, 1823
72
+ 7, 1823
73
+ 0, 2671
74
+ 1, 2671
75
+ 2, 2671
76
+ 3, 2671
77
+ 4, 2671
78
+ 5, 2671
79
+ 6, 2671
80
+ 7, 2671
81
+ 0, 15291
82
+ 1, 15315
83
+ 2, 15303
84
+ 3, 15295
85
+ 4, 15321
86
+ 5, 15295
87
+ 6, 15287
88
+ 7, 15335
89
+ 0, 16321
90
+ 1, 16315
91
+ 2, 16319
92
+ 3, 16309
93
+ 4, 16313
94
+ 5, 16323
95
+ 6, 16325
96
+ 7, 16309
97
+ 0, 16925
98
+ 1, 16927
99
+ 2, 16927
100
+ 3, 16929
101
+ 4, 16923
102
+ 5, 16927
103
+ 6, 16929
104
+ 7, 16923
105
+ 0, 17065
106
+ 1, 17067
107
+ 2, 17067
108
+ 3, 17069
109
+ 4, 17063
110
+ 5, 17067
111
+ 6, 17069
112
+ 7, 17063
113
+ 0, 17561
114
+ 1, 17563
115
+ 2, 17563
116
+ 3, 17565
117
+ 4, 17559
118
+ 5, 17563
119
+ 6, 17565
120
+ 7, 17559
121
+ 0, 24553
122
+ 1, 24695
123
+ 2, 24535
124
+ 3, 24687
125
+ 4, 24703
126
+ 5, 24687
127
+ 6, 24699
128
+ 7, 24703
129
+ 0, 28501
130
+ 1, 28503
131
+ 2, 28503
132
+ 3, 28485
133
+ 4, 28509
134
+ 5, 28483
135
+ 6, 28503
136
+ 7, 28509
137
+ 0, 28501
138
+ 1, 28503
139
+ 2, 28503
140
+ 3, 28485
141
+ 4, 28509
142
+ 5, 28483
143
+ 6, 28503
144
+ 7, 28509
145
+ 0, 28501
146
+ 1, 28503
147
+ 2, 28503
148
+ 3, 28485
149
+ 4, 28509
150
+ 5, 28483
151
+ 6, 28503
152
+ 7, 28509
153
+ 0, 28763
154
+ 1, 28795
155
+ 2, 28775
156
+ 3, 28801
157
+ 4, 28811
158
+ 5, 28805
159
+ 6, 28827
160
+ 7, 28811
161
+ 0, 29773
162
+ 1, 29789
163
+ 2, 29755
164
+ 3, 29765
165
+ 4, 29805
166
+ 5, 29781
167
+ 6, 29799
168
+ 7, 29801
169
+ 0, 37195
170
+ 1, 37201
171
+ 2, 37213
172
+ 3, 37197
173
+ 4, 37197
174
+ 5, 37193
175
+ 6, 37185
176
+ 7, 37197
177
+ 0, 39097
178
+ 1, 39087
179
+ 2, 38951
180
+ 3, 38931
181
+ 4, 39095
182
+ 5, 38941
183
+ 6, 39091
184
+ 7, 39095
185
+ 0, 36763
186
+ 1, 36753
187
+ 2, 36617
188
+ 3, 36597
189
+ 4, 36761
190
+ 5, 36607
191
+ 6, 36757
192
+ 7, 36761
193
+ 0, 34695
194
+ 1, 35073
195
+ 2, 34937
196
+ 3, 34917
197
+ 4, 35081
198
+ 5, 34927
199
+ 6, 35077
200
+ 7, 35081
201
+ 0, 1
202
+ 1, 3
203
+ 2, 4
204
+ 3, 4
205
+ 4, 4
206
+ 5, 132
207
+ 6, 35077
208
+ 7, 644
granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep2/run.env ADDED
@@ -0,0 +1,102 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ backend=ultraep
2
+ config=OLMOE_1B_7B_L16
3
+ family=tierB
4
+ source_model=OLMoE-1B-7B-0924
5
+ standard_num_layers=16
6
+ base_rung=
7
+ repo=/workspace/Megatron-UltraEP
8
+ branch=ultraep-bracing-regions-through-kernels
9
+ commit=03d57ef7
10
+ num_layers=16
11
+ hidden=2048
12
+ ffn_hidden=1024
13
+ moe_ffn_hidden=1024
14
+ num_experts=64
15
+ topk=8
16
+ experts_per_rank=8
17
+ expert_params=6291456
18
+ heads=16
19
+ kv_groups=16
20
+ kv_channels=128
21
+ qk_layernorm=0
22
+ shared_ffn=0
23
+ shared_expert_overlap=0
24
+ moe_token_dispatcher=flex
25
+ mla=0
26
+ q_lora_rank=
27
+ kv_lora_rank=512
28
+ qk_head_dim=128
29
+ qk_pos_emb_head_dim=64
30
+ v_head_dim=128
31
+ dense_first_layers=0
32
+ moe_layers=16
33
+ rotary_base=10000
34
+ rotary_percent=1.0
35
+ rotary_scaling=
36
+ mscale=
37
+ mscale_all_dim=
38
+ qkv_bias=0
39
+ router_balancing=aux_loss
40
+ aux_loss_coeff=1e-2
41
+ router_pre_softmax=1
42
+ router_fusion=0
43
+ router_dtype=fp32
44
+ router_groups=
45
+ router_group_topk=
46
+ router_score_function=softmax
47
+ router_expert_bias=0
48
+ router_bias_update_rate=
49
+ router_topk_scaling_factor=
50
+ shared_expert_gate=0
51
+ apply_layernorm_1p=0
52
+ attention_output_gate=0
53
+ linear_attention_type=
54
+ linear_attention_freq=
55
+ linear_conv_kernel_dim=
56
+ linear_key_head_dim=
57
+ linear_value_head_dim=
58
+ linear_num_key_heads=
59
+ linear_num_value_heads=
60
+ mtp_num_layers=0
61
+ mtp_loss_scaling_factor=0.1
62
+ no_weight_decay_cond_type=
63
+ max_position_embeddings=4096
64
+ init_method_std=0.02
65
+ device_initiated_grouped_gemm=1
66
+ norm_epsilon=1e-5
67
+ seq_length=4096
68
+ micro_batch_size=1
69
+ global_batch_size=8
70
+ num_microbatches=1
71
+ assignments_per_rank=32768
72
+ mean_tokens_per_expert=4096
73
+ slot_policy=matched
74
+ slots_per_rank=8
75
+ capacity_mode=static
76
+ capacity=1.08984375
77
+ cuda_graph=1
78
+ probe_mode=0
79
+ async_grad_reduce=0
80
+ branch_role=timing
81
+ memory_dump_mode=none
82
+ memory_snapshot_path=
83
+ cuda_graph_warmup_steps=3
84
+ recompute=none
85
+ force_balance=0
86
+ train_iters=12
87
+ ep_size=8
88
+ echo_dispatch_overlap=1
89
+ echo_dispatch_recompute=0
90
+ ultraep_quota_min_tokens=512
91
+ ultraep_weight_sync_join=
92
+ profile_enable=0
93
+ kernel_trace=0
94
+ fp8_mode=mxfp8
95
+ fp8_recipe=mxfp8
96
+ training_seed=1234
97
+ learning_rate=1e-4
98
+ min_learning_rate=1.0e-5
99
+ lr_warmup_iters=5
100
+ state_dump_enable=0
101
+ state_dump_dir=
102
+ moonep_assert_no_pending_reduce=0
granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep2/train.log ADDED
@@ -0,0 +1,1258 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ==============================================================
2
+ OLMOE_1B_7B_L16 / ultraep (slots/rank=8 policy=matched)
3
+ E=64 K=8 H'=1024 m=8
4
+ L=16 seq=4096 gbs=8 (1 microbatches)
5
+ cap_mode=static cap=1.08984375 cg=1 recompute=none
6
+ memory_dump=none snapshots=/workspace/granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep2/memory_snapshots/snapshot.pickle
7
+ repo=/workspace/Megatron-UltraEP @ 03d57ef7
8
+ ==============================================================
9
+ W0921 08:38:11.460000 703807 torch/distributed/run.py:766]
10
+ W0921 08:38:11.460000 703807 torch/distributed/run.py:766] *****************************************
11
+ W0921 08:38:11.460000 703807 torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.
12
+ W0921 08:38:11.460000 703807 torch/distributed/run.py:766] *****************************************
13
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
14
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
15
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
16
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
17
+ /workspace/Megatron-UltraEP/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
+ 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
20
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
21
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
22
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
23
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
24
+ 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
25
+ WARNING: overriding default arguments for tokenizer_type:GPT2BPETokenizer with tokenizer_type:GPT2BPETokenizer
26
+ Number of virtual stages per pipeline stage: None
27
+ accumulate and all-reduce gradients in fp32 for bfloat16 data type.
28
+ using torch.bfloat16 for parameters ...
29
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
30
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
31
+ ------------------------ arguments ------------------------
32
+ account_for_embedding_in_pipeline_split ......... False
33
+ account_for_loss_in_pipeline_split .............. False
34
+ accumulate_allreduce_grads_in_fp32 .............. True
35
+ adam_beta1 ...................................... 0.9
36
+ adam_beta2 ...................................... 0.999
37
+ adam_eps ........................................ 1e-08
38
+ add_bias_linear ................................. False
39
+ add_position_embedding .......................... False
40
+ add_qkv_bias .................................... False
41
+ adlr_autoresume ................................. False
42
+ adlr_autoresume_interval ........................ 1000
43
+ align_grad_reduce ............................... True
44
+ align_param_gather .............................. False
45
+ app_tag_run_name ................................ None
46
+ app_tag_run_version ............................. 0.0.0
47
+ apply_layernorm_1p .............................. False
48
+ apply_query_key_layer_scaling ................... False
49
+ apply_residual_connection_post_layernorm ........ False
50
+ apply_rope_fusion ............................... True
51
+ async_save ...................................... None
52
+ async_tensor_model_parallel_allreduce ........... True
53
+ attention_backend ............................... AttnBackend.auto
54
+ attention_dropout ............................... 0.0
55
+ attention_softmax_in_fp32 ....................... False
56
+ auto_detect_ckpt_format ......................... False
57
+ barrier_with_L1_time ............................ True
58
+ bert_binary_head ................................ True
59
+ bert_embedder_type .............................. megatron
60
+ bert_load ....................................... None
61
+ bf16 ............................................ True
62
+ bias_dropout_fusion ............................. True
63
+ bias_gelu_fusion ................................ False
64
+ bias_swiglu_fusion .............................. True
65
+ biencoder_projection_dim ........................ 0
66
+ biencoder_shared_query_context_model ............ False
67
+ block_data_path ................................. None
68
+ calc_ft_timeouts ................................ False
69
+ calculate_per_token_loss ........................ False
70
+ check_for_large_grads ........................... False
71
+ check_for_nan_in_loss_and_grad .................. False
72
+ check_for_spiky_loss ............................ False
73
+ check_weight_hash_across_dp_replicas_interval ... None
74
+ ckpt_assume_constant_structure .................. False
75
+ ckpt_convert_format ............................. None
76
+ ckpt_convert_save ............................... None
77
+ ckpt_convert_update_legacy_dist_opt_format ...... False
78
+ ckpt_format ..................................... torch_dist
79
+ ckpt_fully_parallel_load ........................ False
80
+ ckpt_fully_parallel_save ........................ True
81
+ ckpt_fully_parallel_save_deprecated ............. False
82
+ ckpt_step ....................................... None
83
+ classes_fraction ................................ 1.0
84
+ clip_grad ....................................... 1.0
85
+ clone_scatter_output_in_embedding ............... True
86
+ config_logger_dir ...............................
87
+ consumed_train_samples .......................... 0
88
+ consumed_valid_samples .......................... 0
89
+ context_parallel_size ........................... 1
90
+ cp_comm_type .................................... ['p2p']
91
+ create_attention_mask_in_dataloader ............. True
92
+ cross_entropy_fusion_impl ....................... native
93
+ cross_entropy_loss_fusion ....................... False
94
+ cuda_graph_scope ................................ full_iteration
95
+ cuda_graph_warmup_steps ......................... 3
96
+ data_args_path .................................. None
97
+ data_cache_path ................................. None
98
+ data_parallel_random_init ....................... False
99
+ data_parallel_sharding_strategy ................. no_shard
100
+ data_parallel_size .............................. 8
101
+ data_path ....................................... ['/workspace/data/wikitext/wikitext103_gpt2_text_document']
102
+ data_per_class_fraction ......................... 1.0
103
+ data_sharding ................................... True
104
+ dataloader_type ................................. single
105
+ ddp_average_in_collective ....................... False
106
+ ddp_bucket_size ................................. None
107
+ ddp_num_buckets ................................. None
108
+ ddp_pad_buckets_for_high_nccl_busbw ............. False
109
+ decoder_first_pipeline_num_layers ............... None
110
+ decoder_last_pipeline_num_layers ................ None
111
+ decoder_num_layers .............................. None
112
+ decoder_seq_length .............................. None
113
+ decoupled_lr .................................... None
114
+ decoupled_min_lr ................................ None
115
+ decrease_batch_size_if_needed ................... False
116
+ defer_embedding_wgrad_compute ................... False
117
+ delay_wgrad_compute ............................. False
118
+ deprecated_use_mcore_models ..................... True
119
+ deterministic_mode .............................. False
120
+ dino_bottleneck_size ............................ 256
121
+ dino_freeze_last_layer .......................... 1
122
+ dino_head_hidden_size ........................... 2048
123
+ dino_local_crops_number ......................... 10
124
+ dino_local_img_size ............................. 96
125
+ dino_norm_last_layer ............................ False
126
+ dino_teacher_temp ............................... 0.07
127
+ dino_warmup_teacher_temp ........................ 0.04
128
+ dino_warmup_teacher_temp_epochs ................. 30
129
+ disable_bf16_reduced_precision_matmul ........... False
130
+ disable_mamba_mem_eff_path ...................... False
131
+ disable_straggler_on_startup .................... False
132
+ dist_ckpt_format_deprecated ..................... None
133
+ dist_ckpt_strictness ............................ assume_ok_unexpected
134
+ distribute_saved_activations .................... False
135
+ distributed_backend ............................. nccl
136
+ distributed_timeout_minutes ..................... 10
137
+ embedding_init_method_std ....................... None
138
+ embedding_path .................................. None
139
+ empty_unused_memory_level ....................... 0
140
+ enable_cuda_graph ............................... True
141
+ enable_experimental ............................. False
142
+ enable_ft_package ............................... False
143
+ enable_gloo_process_groups ...................... True
144
+ enable_msc ...................................... True
145
+ enable_one_logger ............................... True
146
+ encoder_num_layers .............................. 16
147
+ encoder_seq_length .............................. 4096
148
+ end_weight_decay ................................ 0.1
149
+ eod_mask_loss ................................... False
150
+ error_injection_rate ............................ 0
151
+ error_injection_type ............................ transient_error
152
+ eval_interval ................................... 1000000
153
+ eval_iters ...................................... 0
154
+ evidence_data_path .............................. None
155
+ exit_duration_in_mins ........................... None
156
+ exit_interval ................................... None
157
+ exit_on_missing_checkpoint ...................... False
158
+ exit_signal_handler ............................. False
159
+ exp_avg_dtype ................................... torch.float32
160
+ exp_avg_sq_dtype ................................ torch.float32
161
+ expert_model_parallel_size ...................... 8
162
+ expert_tensor_parallel_size ..................... 1
163
+ export_force_local_attention .................... False
164
+ export_kd_cfg ................................... None
165
+ export_kd_teacher_ckpt_format ................... None
166
+ export_kd_teacher_load .......................... None
167
+ export_kv_cache_quant ........................... False
168
+ export_legacy_megatron .......................... False
169
+ export_model_type ............................... GPTModel
170
+ export_moe_apply_probs_on_input ................. False
171
+ export_qk_l2_norm ............................... False
172
+ export_quant_cfg ................................ None
173
+ export_real_quant_cfg ........................... None
174
+ export_te_mcore_model ........................... False
175
+ external_cuda_graph ............................. False
176
+ ffn_hidden_size ................................. 1024
177
+ finetune ........................................ False
178
+ finetune_data_split ............................. train
179
+ finetune_hf_dataset ............................. None
180
+ first_last_layers_bf16 .......................... False
181
+ flash_decode .................................... False
182
+ fp16 ............................................ False
183
+ fp16_lm_cross_entropy ........................... False
184
+ fp32_residual_connection ........................ False
185
+ fp8 ............................................. e4m3
186
+ fp8_amax_compute_algo ........................... most_recent
187
+ fp8_amax_history_len ............................ 1
188
+ fp8_interval .................................... 1
189
+ fp8_margin ...................................... 0
190
+ fp8_param_gather ................................ False
191
+ fp8_recipe ...................................... mxfp8
192
+ fp8_wgrad ....................................... True
193
+ fsdp_double_buffer .............................. False
194
+ global_batch_size ............................... 8
195
+ grad_reduce_in_bf16 ............................. False
196
+ gradient_accumulation_fusion .................... True
197
+ gradient_reduce_div_fusion ...................... True
198
+ group_query_attention ........................... True
199
+ head_lr_mult .................................... 1.0
200
+ heterogeneous_layers_config_encoded_json ........ None
201
+ heterogeneous_layers_config_path ................ None
202
+ hidden_dropout .................................. 0.0
203
+ hidden_size ..................................... 2048
204
+ hierarchical_context_parallel_sizes ............. None
205
+ high_priority_stream_groups ..................... []
206
+ hybrid_attention_ratio .......................... 0.0
207
+ hybrid_mlp_ratio ................................ 0.0
208
+ hybrid_override_pattern ......................... None
209
+ hysteresis ...................................... 2
210
+ ict_head_size ................................... None
211
+ ict_load ........................................ None
212
+ img_h ........................................... 224
213
+ img_w ........................................... 224
214
+ indexer_batch_size .............................. 128
215
+ indexer_log_interval ............................ 1000
216
+ inference_batch_times_seqlen_threshold .......... -1
217
+ inference_dynamic_batching ...................... False
218
+ inference_dynamic_batching_buffer_guaranteed_fraction 0.2
219
+ inference_dynamic_batching_buffer_overflow_factor None
220
+ inference_dynamic_batching_buffer_size_gb ....... 40.0
221
+ inference_dynamic_batching_chunk_size ........... 256
222
+ inference_dynamic_batching_max_requests_override None
223
+ inference_dynamic_batching_max_tokens_override .. None
224
+ inference_dynamic_batching_num_cuda_graphs ...... 16
225
+ inference_max_batch_size ........................ 8
226
+ inference_max_seq_length ........................ 2560
227
+ inference_rng_tracker ........................... False
228
+ init_method_std ................................. 0.02
229
+ init_method_xavier_uniform ...................... False
230
+ init_model_with_meta_device ..................... False
231
+ initial_loss_scale .............................. 4294967296
232
+ inprocess_active_world_size ..................... 8
233
+ inprocess_barrier_timeout ....................... 120
234
+ inprocess_completion_timeout .................... 120
235
+ inprocess_empty_cuda_cache ...................... False
236
+ inprocess_granularity ........................... node
237
+ inprocess_hard_timeout .......................... 90
238
+ inprocess_heartbeat_interval .................... 30
239
+ inprocess_heartbeat_timeout ..................... 60
240
+ inprocess_last_call_wait ........................ 1
241
+ inprocess_max_iterations ........................ None
242
+ inprocess_monitor_process_interval .............. 1.0
243
+ inprocess_monitor_thread_interval ............... 1.0
244
+ inprocess_progress_watchdog_interval ............ 1.0
245
+ inprocess_restart ............................... False
246
+ inprocess_soft_timeout .......................... 60
247
+ inprocess_termination_grace_time ................ 1
248
+ is_hybrid_model ................................. False
249
+ iter_per_epoch .................................. 1250
250
+ iterations_to_skip .............................. []
251
+ keep_fp8_transpose_cache_when_using_custom_fsdp . False
252
+ kitchen_config_file ............................. None
253
+ kitchen_recipe_number ........................... None
254
+ kv_channels ..................................... 128
255
+ kv_lora_rank .................................... 32
256
+ lazy_mpu_init ................................... None
257
+ load ............................................ None
258
+ load_main_params_from_ckpt ...................... None
259
+ load_model_opt_format ........................... False
260
+ local_rank ...................................... 0
261
+ log_energy ...................................... False
262
+ log_interval .................................... 1
263
+ log_loss_scale_to_tensorboard ................... True
264
+ log_memory_to_tensorboard ....................... False
265
+ log_num_zeros_in_grad ........................... False
266
+ log_params_norm ................................. False
267
+ log_progress .................................... False
268
+ log_straggler ................................... False
269
+ log_throughput .................................. True
270
+ log_timers_to_tensorboard ....................... False
271
+ log_validation_ppl_to_tensorboard ............... False
272
+ log_world_size_to_tensorboard ................... False
273
+ logging_level ................................... None
274
+ loss_scale ...................................... None
275
+ loss_scale_window ............................... 1000
276
+ lr .............................................. 0.0001
277
+ lr_decay_iters .................................. 8
278
+ lr_decay_samples ................................ None
279
+ lr_decay_style .................................. cosine
280
+ lr_warmup_fraction .............................. None
281
+ lr_warmup_init .................................. 0.0
282
+ lr_warmup_iters ................................. 5
283
+ lr_warmup_samples ............................... 0
284
+ lr_wsd_decay_iters .............................. None
285
+ lr_wsd_decay_samples ............................ None
286
+ lr_wsd_decay_style .............................. exponential
287
+ main_grads_dtype ................................ torch.float32
288
+ main_params_dtype ............................... torch.float32
289
+ make_vocab_size_divisible_by .................... 128
290
+ mamba_head_dim .................................. 64
291
+ mamba_num_groups ................................ 8
292
+ mamba_num_heads ................................. None
293
+ mamba_state_dim ................................. 128
294
+ manual_gc ....................................... False
295
+ manual_gc_eval .................................. True
296
+ manual_gc_interval .............................. 0
297
+ mask_factor ..................................... 1.0
298
+ mask_prob ....................................... 0.15
299
+ mask_type ....................................... random
300
+ masked_softmax_fusion ........................... False
301
+ max_position_embeddings ......................... 4096
302
+ max_tokens_to_oom ............................... 12000
303
+ memory_snapshot_path ............................ snapshot.pickle
304
+ merge_file ...................................... /workspace/data/wikitext/gpt2-merges.txt
305
+ micro_batch_size ................................ 1
306
+ microbatch_group_size_per_vp_stage .............. None
307
+ mid_level_dataset_surplus ....................... 0.005
308
+ min_loss_scale .................................. 1.0
309
+ min_lr .......................................... 1e-05
310
+ mlp_chunks_for_prefill .......................... 1
311
+ mmap_bin_files .................................. True
312
+ mock_data ....................................... False
313
+ moe_apply_probs_on_input ........................ False
314
+ moe_aux_loss_coeff .............................. 0.01
315
+ moe_capacity_monitor ............................ True
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 ............................. True
324
+ moe_enable_ultraep .............................. True
325
+ moe_expert_capacity_factor ...................... None
326
+ moe_extended_tp ................................. False
327
+ moe_ffn_hidden_size ............................. 1024
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_num_echo_experts ............................ None
334
+ moe_num_redundant_experts_per_rank .............. 8
335
+ moe_pad_expert_input_to_capacity ................ False
336
+ moe_per_layer_logging ........................... False
337
+ moe_permute_fusion .............................. False
338
+ moe_received_token_capacity ..................... 1.08984375
339
+ moe_router_bias_update_rate ..................... 0.001
340
+ moe_router_dtype ................................ fp32
341
+ moe_router_enable_expert_bias ................... False
342
+ moe_router_force_load_balancing ................. False
343
+ moe_router_group_topk ........................... None
344
+ moe_router_load_balancing_type .................. aux_loss
345
+ moe_router_num_groups ........................... None
346
+ moe_router_padding_for_fp8 ...................... False
347
+ moe_router_pre_softmax .......................... True
348
+ moe_router_score_function ....................... softmax
349
+ moe_router_topk ................................. 8
350
+ moe_router_topk_scaling_factor .................. None
351
+ moe_shared_expert_intermediate_size ............. None
352
+ moe_shared_expert_overlap ....................... False
353
+ moe_token_dispatcher_type ....................... flex
354
+ moe_token_drop_policy ........................... probs
355
+ moe_ultraep_max_microbatches .................... None
356
+ moe_ultraep_overlap_weight_sync ................. True
357
+ moe_ultraep_weight_sync_join .................... before_dispatch
358
+ moe_upcycling_granularity ....................... 1
359
+ moe_use_device_initiated_grouped_gemm ........... True
360
+ moe_use_legacy_grouped_gemm ..................... False
361
+ moe_use_upcycling ............................... False
362
+ moe_z_loss_coeff ................................ None
363
+ mrope_section ................................... None
364
+ mscale .......................................... 1.0
365
+ mscale_all_dim .................................. 0.0
366
+ mtp_loss_scaling_factor ......................... 0.1
367
+ mtp_num_layers .................................. None
368
+ multi_latent_attention .......................... False
369
+ nccl_all_reduce_for_prefill ..................... False
370
+ nccl_communicator_config_path ................... None
371
+ nccl_ub ......................................... False
372
+ no_load_optim ................................... None
373
+ no_load_rng ..................................... None
374
+ no_persist_layer_norm ........................... False
375
+ no_rope_freq .................................... None
376
+ no_save_optim ................................... None
377
+ no_save_rng ..................................... None
378
+ non_persistent_ckpt_type ........................ None
379
+ non_persistent_global_ckpt_dir .................. None
380
+ non_persistent_local_ckpt_algo .................. fully_parallel
381
+ non_persistent_local_ckpt_dir ................... None
382
+ non_persistent_save_interval .................... None
383
+ norm_epsilon .................................... 1e-05
384
+ normalization ................................... RMSNorm
385
+ num_attention_heads ............................. 16
386
+ num_channels .................................... 3
387
+ num_classes ..................................... 1000
388
+ num_dataset_builder_threads ..................... 1
389
+ num_distributed_optimizer_instances ............. 1
390
+ num_experts ..................................... 64
391
+ num_layers ...................................... 16
392
+ num_layers_at_end_in_bf16 ....................... 1
393
+ num_layers_at_start_in_bf16 ..................... 1
394
+ num_layers_per_virtual_pipeline_stage ........... None
395
+ num_query_groups ................................ 16
396
+ num_virtual_stages_per_pipeline_rank ............ None
397
+ num_workers ..................................... 2
398
+ object_storage_cache_path ....................... None
399
+ one_logger_async ................................ False
400
+ one_logger_project .............................. megatron-lm
401
+ one_logger_run_name ............................. None
402
+ onnx_safe ....................................... None
403
+ openai_gelu ..................................... False
404
+ optimizer ....................................... adam
405
+ optimizer_cpu_offload ........................... False
406
+ optimizer_offload_fraction ...................... 1.0
407
+ output_bert_embeddings .......................... False
408
+ overlap_cpu_optimizer_d2h_h2d ................... False
409
+ overlap_grad_reduce ............................. True
410
+ overlap_moe_expert_parallel_comm ................ False
411
+ overlap_p2p_comm ................................ False
412
+ overlap_p2p_comm_warmup_flush ................... False
413
+ overlap_param_gather ............................ True
414
+ overlap_param_gather_with_optimizer_step ........ False
415
+ override_opt_param_scheduler .................... False
416
+ params_dtype .................................... torch.bfloat16
417
+ patch_dim ....................................... 16
418
+ per_split_data_args_path ........................ None
419
+ perform_initialization .......................... True
420
+ pin_cpu_grads ................................... True
421
+ pin_cpu_params .................................. True
422
+ pipeline_model_parallel_comm_backend ............ None
423
+ pipeline_model_parallel_layout .................. None
424
+ pipeline_model_parallel_size .................... 1
425
+ position_embedding_type ......................... rope
426
+ pretrained_checkpoint ........................... None
427
+ profile ......................................... False
428
+ profile_ranks ................................... [0]
429
+ profile_step_end ................................ 12
430
+ profile_step_start .............................. 10
431
+ q_lora_rank ..................................... None
432
+ qk_head_dim ..................................... 128
433
+ qk_l2_norm ...................................... False
434
+ qk_layernorm .................................... False
435
+ qk_pos_emb_head_dim ............................. 64
436
+ query_in_block_prob ............................. 0.1
437
+ rampup_batch_size ............................... None
438
+ rank ............................................ 0
439
+ recompute_granularity ........................... None
440
+ recompute_method ................................ None
441
+ recompute_modules ............................... None
442
+ recompute_num_layers ............................ None
443
+ record_memory_history ........................... False
444
+ relative_attention_max_distance ................. 128
445
+ relative_attention_num_buckets .................. 32
446
+ replication ..................................... False
447
+ replication_factor .............................. 2
448
+ replication_jump ................................ None
449
+ rerun_mode ...................................... validate_results
450
+ reset_attention_mask ............................ False
451
+ reset_position_ids .............................. False
452
+ result_rejected_tracker_filename ................ None
453
+ retriever_report_topk_accuracies ................ []
454
+ retriever_score_scaling ......................... False
455
+ retriever_seq_length ............................ 256
456
+ retro_add_retriever ............................. False
457
+ retro_attention_gate ............................ 1
458
+ retro_cyclic_train_iters ........................ None
459
+ retro_encoder_attention_dropout ................. 0.1
460
+ retro_encoder_hidden_dropout .................... 0.1
461
+ retro_encoder_layers ............................ 2
462
+ retro_num_neighbors ............................. 2
463
+ retro_num_retrieved_chunks ...................... 2
464
+ retro_project_dir ............................... None
465
+ retro_verify_neighbor_count ..................... True
466
+ reuse_grad_buf_for_mxfp8_param_ag ............... False
467
+ rope_scaling_factor ............................. 8.0
468
+ rope_type ....................................... None
469
+ rotary_base ..................................... 10000
470
+ rotary_interleaved .............................. False
471
+ rotary_percent .................................. 1.0
472
+ rotary_scaling_factor ........................... 1.0
473
+ rotary_seq_len_interpolation_factor ............. None
474
+ run_workload_inspector_server ................... False
475
+ sample_rate ..................................... 1.0
476
+ save ............................................ None
477
+ save_interval ................................... None
478
+ save_retain_interval ............................ None
479
+ scatter_gather_tensors_in_pipeline .............. True
480
+ seed ............................................ 1234
481
+ seq_length ...................................... 4096
482
+ sequence_parallel ............................... False
483
+ sft ............................................. False
484
+ sft_tokenizer_prompt_format ..................... nemotron-h-aligned
485
+ sgd_momentum .................................... 0.9
486
+ short_seq_prob .................................. 0.1
487
+ skip_train ...................................... False
488
+ skipped_train_samples ........................... 0
489
+ spec ............................................ None
490
+ split ........................................... 100,0,0
491
+ squared_relu .................................... False
492
+ start_weight_decay .............................. 0.1
493
+ straggler_ctrlr_port ............................ 65535
494
+ straggler_minmax_count .......................... 1
495
+ suggested_communication_unit_size ............... None
496
+ swiglu .......................................... True
497
+ swin_backbone_type .............................. tiny
498
+ symmetric_ar_type ............................... None
499
+ te_rng_tracker .................................. True
500
+ tensor_model_parallel_size ...................... 1
501
+ tensorboard_dir ................................. None
502
+ tensorboard_log_interval ........................ 1
503
+ tensorboard_queue_size .......................... 1000
504
+ test_data_path .................................. None
505
+ test_mode ....................................... False
506
+ tiktoken_num_special_tokens ..................... 1000
507
+ tiktoken_pattern ................................ None
508
+ tiktoken_special_tokens ......................... None
509
+ timing_log_level ................................ 0
510
+ timing_log_option ............................... minmax
511
+ titles_data_path ................................ None
512
+ tokenizer_model ................................. None
513
+ tokenizer_type .................................. GPT2BPETokenizer
514
+ torch_fsdp2_reshard_after_forward ............... True
515
+ tp_comm_bootstrap_backend ....................... nccl
516
+ tp_comm_bulk_dgrad .............................. True
517
+ tp_comm_bulk_wgrad .............................. True
518
+ tp_comm_overlap ................................. False
519
+ tp_comm_overlap_ag .............................. True
520
+ tp_comm_overlap_cfg ............................. None
521
+ tp_comm_overlap_rs .............................. True
522
+ tp_comm_overlap_rs_dgrad ........................ False
523
+ tp_comm_split_ag ................................ True
524
+ tp_comm_split_rs ................................ True
525
+ train_data_path ................................. None
526
+ train_iters ..................................... 12
527
+ train_samples ................................... None
528
+ train_sync_interval ............................. None
529
+ transformer_impl ................................ transformer_engine
530
+ transformer_pipeline_model_parallel_size ........ 1
531
+ untie_embeddings_and_output_weights ............. True
532
+ use_checkpoint_args ............................. False
533
+ use_checkpoint_opt_param_scheduler .............. False
534
+ use_cpu_initialization .......................... None
535
+ use_custom_fsdp ................................. False
536
+ use_dist_ckpt ................................... True
537
+ use_dist_ckpt_deprecated ........................ False
538
+ use_distributed_optimizer ....................... True
539
+ use_flash_attn .................................. False
540
+ use_legacy_models ............................... False
541
+ use_mp_args_from_checkpoint_args ................ False
542
+ use_one_sent_docs ............................... False
543
+ use_persistent_ckpt_worker ...................... False
544
+ use_precision_aware_optimizer ................... False
545
+ use_pytorch_profiler ............................ False
546
+ use_ring_exchange_p2p ........................... False
547
+ use_rope_scaling ................................ False
548
+ use_rotary_position_embeddings .................. False
549
+ use_sharp ....................................... False
550
+ use_tokenizer_model_from_checkpoint_args ........ True
551
+ use_torch_fsdp2 ................................. False
552
+ use_torch_optimizer_for_cpu_offload ............. False
553
+ use_tp_pp_dp_mapping ............................ False
554
+ v_head_dim ...................................... 128
555
+ valid_data_path ................................. None
556
+ variable_seq_lengths ............................ False
557
+ virtual_pipeline_model_parallel_size ............ None
558
+ vision_backbone_type ............................ vit
559
+ vision_pretraining .............................. False
560
+ vision_pretraining_type ......................... classify
561
+ vocab_extra_ids ................................. 0
562
+ vocab_file ...................................... /workspace/data/wikitext/gpt2-vocab.json
563
+ vocab_size ...................................... None
564
+ wandb_exp_name ..................................
565
+ wandb_project ...................................
566
+ wandb_save_dir ..................................
567
+ weight_decay .................................... 0.1
568
+ weight_decay_incr_style ......................... constant
569
+ wgrad_deferral_limit ............................ 0
570
+ world_size ...................................... 8
571
+ yaml_cfg ........................................ None
572
+ -------------------- end of arguments ---------------------
573
+ INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 1
574
+ > building GPT2BPETokenizer tokenizer ...
575
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
576
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
577
+ > padded vocab (size: 50257) with 47 dummy tokens (new size: 50304)
578
+ WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.VALIDATE_RESULTS
579
+ > initializing torch distributed ...
580
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
581
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
582
+ [Gloo] Rank [Gloo] Rank 0 is connected to 75 peer ranks. is connected to Expected number of connected peer ranks is : 77 peer ranks. Expected number of connected peer ranks is :
583
+ 7
584
+ [Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
585
+ [Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
586
+ [Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
587
+ [Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
588
+ [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 :
589
+ 7
590
+ [Gloo] Rank [Gloo] Rank 1 is connected to 07 is connected to peer ranks. 7Expected number of connected peer ranks is : peer ranks. 7Expected number of connected peer ranks is : [Gloo] Rank
591
+ 7[Gloo] Rank 2
592
+ 3 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
593
+
594
+ [Gloo] Rank [Gloo] Rank 4 is connected to 75 peer ranks. is connected to Expected number of connected peer ranks is : 77 peer ranks. Expected number of connected peer ranks is :
595
+ 7
596
+ [Gloo] Rank 6 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
597
+ [Gloo] Rank 7 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
598
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
599
+ > initialized tensor model parallel with size 1
600
+ > initialized pipeline model parallel with size 1
601
+ > setting random seeds to 1234 ...
602
+ [Gloo] Rank [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 [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
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
608
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
609
+ [rank3]:[W921 08:38:27.743710852 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
610
+ [rank4]:[W921 08:38:27.743711569 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
+ [rank1]:[W921 08:38:27.743719427 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
612
+ [rank5]:[W921 08:38:27.743771075 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.
613
+ [rank7]:[W921 08:38:27.743867003 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
+ [rank2]:[W921 08:38:27.744016435 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.
615
+ [rank6]:[W921 08:38:27.744115842 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.
616
+ > compiling dataset index builder ...
617
+ make: Entering directory '/workspace/Megatron-UltraEP/megatron/core/datasets'
618
+ make: Nothing to be done for 'default'.
619
+ make: Leaving directory '/workspace/Megatron-UltraEP/megatron/core/datasets'
620
+ >>> done with dataset index builder. Compilation time: 0.033 seconds
621
+ WARNING: constraints for invoking optimized fused softmax kernel are not met. We default back to unfused kernel invocations.
622
+ > compiling and loading fused kernels ...
623
+ [rank0]:[W921 08:38:27.905172503 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.
624
+ >>> done with compiling and loading fused kernels. Compilation time: 3.915 seconds
625
+ time to initialize megatron (seconds): 7.458
626
+ [after megatron is initialized] datetime: 2026-09-21 08:38:33
627
+ building GPT model ...
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
+ /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.
643
+ warnings.warn(
644
+ [INFO] Use MNNVL fabric: False
645
+ [WARN] MAX_NUM_NVL_PEERS is not set. Using detected value 8.
646
+ NVSHMEM v3.7.2
647
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
648
+
649
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
650
+
651
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
652
+
653
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
654
+
655
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
656
+
657
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
658
+
659
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
660
+
661
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
662
+
663
+ UltraEP: NVL domain size <= 8; forcing weight_sync plan mode to direct (was adaptive).
664
+ > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 2087258112
665
+ 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)
666
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 7
667
+ Params for bucket 1 (103022592 elements, 103022592 padded size):
668
+ module.output_layer.weight
669
+ Params for bucket 2 (46542848 elements, 46542848 padded size):
670
+ module.decoder.layers.15.self_attention.linear_qkv.weight
671
+ module.decoder.layers.15.pre_mlp_layernorm.weight
672
+ module.decoder.layers.14.pre_mlp_layernorm.weight
673
+ module.decoder.layers.14.self_attention.linear_qkv.weight
674
+ module.decoder.layers.13.pre_mlp_layernorm.weight
675
+ module.decoder.layers.13.self_attention.linear_qkv.weight
676
+ module.decoder.final_layernorm.weight
677
+ module.decoder.layers.15.self_attention.linear_qkv.layer_norm_weight
678
+ module.decoder.layers.15.mlp.router.weight
679
+ module.decoder.layers.15.self_attention.linear_proj.weight
680
+ module.decoder.layers.14.mlp.router.weight
681
+ module.decoder.layers.14.self_attention.linear_qkv.layer_norm_weight
682
+ module.decoder.layers.14.self_attention.linear_proj.weight
683
+ module.decoder.layers.13.mlp.router.weight
684
+ Params for bucket 3 (50737152 elements, 50737152 padded size):
685
+ module.decoder.layers.12.pre_mlp_layernorm.weight
686
+ module.decoder.layers.12.self_attention.linear_qkv.weight
687
+ module.decoder.layers.11.mlp.router.weight
688
+ module.decoder.layers.11.self_attention.linear_qkv.layer_norm_weight
689
+ module.decoder.layers.11.self_attention.linear_proj.weight
690
+ module.decoder.layers.10.mlp.router.weight
691
+ module.decoder.layers.13.self_attention.linear_qkv.layer_norm_weight
692
+ module.decoder.layers.12.mlp.router.weight
693
+ module.decoder.layers.13.self_attention.linear_proj.weight
694
+ module.decoder.layers.12.self_attention.linear_qkv.layer_norm_weight
695
+ module.decoder.layers.12.self_attention.linear_proj.weight
696
+ module.decoder.layers.11.pre_mlp_layernorm.weight
697
+ module.decoder.layers.11.self_attention.linear_qkv.weight
698
+ module.decoder.layers.10.pre_mlp_layernorm.weight
699
+ module.decoder.layers.10.self_attention.linear_qkv.weight
700
+ Params for bucket 4 (50737152 elements, 50737152 padded size):
701
+ module.decoder.layers.10.self_attention.linear_qkv.layer_norm_weight
702
+ module.decoder.layers.9.mlp.router.weight
703
+ module.decoder.layers.10.self_attention.linear_proj.weight
704
+ module.decoder.layers.9.self_attention.linear_qkv.layer_norm_weight
705
+ module.decoder.layers.9.self_attention.linear_proj.weight
706
+ module.decoder.layers.8.mlp.router.weight
707
+ module.decoder.layers.8.self_attention.linear_qkv.layer_norm_weight
708
+ module.decoder.layers.8.self_attention.linear_proj.weight
709
+ module.decoder.layers.7.mlp.router.weight
710
+ module.decoder.layers.9.pre_mlp_layernorm.weight
711
+ module.decoder.layers.9.self_attention.linear_qkv.weight
712
+ module.decoder.layers.8.pre_mlp_layernorm.weight
713
+ module.decoder.layers.8.self_attention.linear_qkv.weight
714
+ module.decoder.layers.7.pre_mlp_layernorm.weight
715
+ module.decoder.layers.7.self_attention.linear_qkv.weight
716
+ Params for bucket 5 (50737152 elements, 50737152 padded size):
717
+ module.decoder.layers.7.self_attention.linear_qkv.layer_norm_weight
718
+ module.decoder.layers.6.mlp.router.weight
719
+ module.decoder.layers.7.self_attention.linear_proj.weight
720
+ module.decoder.layers.6.self_attention.linear_qkv.layer_norm_weight
721
+ module.decoder.layers.6.self_attention.linear_proj.weight
722
+ module.decoder.layers.5.mlp.router.weight
723
+ module.decoder.layers.5.self_attention.linear_qkv.layer_norm_weight
724
+ module.decoder.layers.4.pre_mlp_layernorm.weight
725
+ module.decoder.layers.4.self_attention.linear_qkv.weight
726
+ module.decoder.layers.6.self_attention.linear_qkv.weight
727
+ module.decoder.layers.6.pre_mlp_layernorm.weight
728
+ module.decoder.layers.5.pre_mlp_layernorm.weight
729
+ module.decoder.layers.5.self_attention.linear_qkv.weight
730
+ module.decoder.layers.5.self_attention.linear_proj.weight
731
+ module.decoder.layers.4.mlp.router.weight
732
+ Params for bucket 6 (50737152 elements, 50737152 padded size):
733
+ module.decoder.layers.3.mlp.router.weight
734
+ module.decoder.layers.3.self_attention.linear_qkv.layer_norm_weight
735
+ module.decoder.layers.3.self_attention.linear_proj.weight
736
+ module.decoder.layers.2.mlp.router.weight
737
+ module.decoder.layers.2.self_attention.linear_qkv.layer_norm_weight
738
+ module.decoder.layers.2.self_attention.linear_proj.weight
739
+ module.decoder.layers.1.mlp.router.weight
740
+ module.decoder.layers.4.self_attention.linear_qkv.layer_norm_weight
741
+ module.decoder.layers.3.pre_mlp_layernorm.weight
742
+ module.decoder.layers.4.self_attention.linear_proj.weight
743
+ module.decoder.layers.3.self_attention.linear_qkv.weight
744
+ module.decoder.layers.2.pre_mlp_layernorm.weight
745
+ module.decoder.layers.2.self_attention.linear_qkv.weight
746
+ module.decoder.layers.1.pre_mlp_layernorm.weight
747
+ module.decoder.layers.1.self_attention.linear_qkv.weight
748
+ Params for bucket 7 (124131328 elements, 124131328 padded size):
749
+ module.decoder.layers.1.self_attention.linear_qkv.layer_norm_weight
750
+ module.decoder.layers.1.self_attention.linear_proj.weight
751
+ module.decoder.layers.0.pre_mlp_layernorm.weight
752
+ module.decoder.layers.0.self_attention.linear_qkv.weight
753
+ module.decoder.layers.0.self_attention.linear_proj.weight
754
+ module.embedding.word_embeddings.weight
755
+ module.decoder.layers.0.mlp.router.weight
756
+ module.decoder.layers.0.self_attention.linear_qkv.layer_norm_weight
757
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 20
758
+ Params for bucket 1 (41943040 elements, 41943040 padded size):
759
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight4
760
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight6
761
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight2
762
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight0
763
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight7
764
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight5
765
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight3
766
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight7
767
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight5
768
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight3
769
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight1
770
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight6
771
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight4
772
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight2
773
+ Params for bucket 2 (41943040 elements, 41943040 padded size):
774
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight1
775
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight7
776
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight5
777
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight3
778
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight1
779
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight6
780
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight4
781
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight0
782
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight6
783
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight4
784
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight2
785
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight0
786
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight7
787
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight5
788
+ Params for bucket 3 (41943040 elements, 41943040 padded size):
789
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight2
790
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight0
791
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight6
792
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight4
793
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight2
794
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight0
795
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight7
796
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight3
797
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight1
798
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight7
799
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight5
800
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight3
801
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight1
802
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight6
803
+ Params for bucket 4 (41943040 elements, 41943040 padded size):
804
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight5
805
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight1
806
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight3
807
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight6
808
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight4
809
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight2
810
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight0
811
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight4
812
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight2
813
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight0
814
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight7
815
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight5
816
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight3
817
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight1
818
+ Params for bucket 5 (41943040 elements, 41943040 padded size):
819
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight7
820
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight3
821
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight5
822
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight1
823
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight6
824
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight4
825
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight6
826
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight4
827
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight2
828
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight0
829
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight7
830
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight5
831
+ Params for bucket 6 (41943040 elements, 41943040 padded size):
832
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight0
833
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight7
834
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight2
835
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight5
836
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight3
837
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight1
838
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight3
839
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight1
840
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight6
841
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight4
842
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight2
843
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight0
844
+ Params for bucket 7 (41943040 elements, 41943040 padded size):
845
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight7
846
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight5
847
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight3
848
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight1
849
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight6
850
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight4
851
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight2
852
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight4
853
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight6
854
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight2
855
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight0
856
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight7
857
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight5
858
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight3
859
+ Params for bucket 8 (41943040 elements, 41943040 padded size):
860
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight5
861
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight0
862
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight7
863
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight3
864
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight1
865
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight6
866
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight4
867
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight1
868
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight6
869
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight4
870
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight2
871
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight0
872
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight7
873
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight5
874
+ Params for bucket 9 (41943040 elements, 41943040 padded size):
875
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight2
876
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight7
877
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight0
878
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight5
879
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight3
880
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight1
881
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight6
882
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight3
883
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight1
884
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight6
885
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight4
886
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight2
887
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight0
888
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight7
889
+ Params for bucket 10 (41943040 elements, 41943040 padded size):
890
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight4
891
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight2
892
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight0
893
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight7
894
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight5
895
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight3
896
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight1
897
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight5
898
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight3
899
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight1
900
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight6
901
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight4
902
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight2
903
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight0
904
+ Params for bucket 11 (41943040 elements, 41943040 padded size):
905
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight4
906
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight6
907
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight2
908
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight0
909
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight7
910
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight5
911
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight7
912
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight5
913
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight3
914
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight1
915
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight6
916
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight4
917
+ Params for bucket 12 (41943040 elements, 41943040 padded size):
918
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight3
919
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight1
920
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight6
921
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight4
922
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight2
923
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight0
924
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight2
925
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight0
926
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight7
927
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight5
928
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight3
929
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight1
930
+ Params for bucket 13 (41943040 elements, 41943040 padded size):
931
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight6
932
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight4
933
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight2
934
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight0
935
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight7
936
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight5
937
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight3
938
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight7
939
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight5
940
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight3
941
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight1
942
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight6
943
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight4
944
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight2
945
+ Params for bucket 14 (41943040 elements, 41943040 padded size):
946
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight1
947
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight6
948
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight4
949
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight2
950
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight0
951
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight7
952
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight5
953
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight0
954
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight5
955
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight7
956
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight3
957
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight1
958
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight6
959
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight4
960
+ Params for bucket 15 (41943040 elements, 41943040 padded size):
961
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight3
962
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight1
963
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight7
964
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight5
965
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight3
966
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight1
967
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight6
968
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight0
969
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight2
970
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight6
971
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight4
972
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight2
973
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight0
974
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight7
975
+ Params for bucket 16 (41943040 elements, 41943040 padded size):
976
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight4
977
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight2
978
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight0
979
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight6
980
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight4
981
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight2
982
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight0
983
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight5
984
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight3
985
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight1
986
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight7
987
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight5
988
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight3
989
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight1
990
+ Params for bucket 17 (41943040 elements, 41943040 padded size):
991
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight7
992
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight3
993
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight5
994
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight1
995
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight7
996
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight5
997
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight6
998
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight4
999
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight2
1000
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight0
1001
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight6
1002
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight4
1003
+ Params for bucket 18 (41943040 elements, 41943040 padded size):
1004
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight3
1005
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight1
1006
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight6
1007
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight4
1008
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight2
1009
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight0
1010
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight7
1011
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight0
1012
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight2
1013
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight5
1014
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight3
1015
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight1
1016
+ Params for bucket 19 (41943040 elements, 41943040 padded size):
1017
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight4
1018
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight6
1019
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight2
1020
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight0
1021
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight6
1022
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight4
1023
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight2
1024
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight7
1025
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight5
1026
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight3
1027
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight1
1028
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight7
1029
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight5
1030
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight3
1031
+ Params for bucket 20 (8388608 elements, 8388608 padded size):
1032
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight1
1033
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight0
1034
+ 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 0x7c1ad3c3e3c0>, config_logger_dir='')
1035
+ INFO:megatron.core.optimizer_param_scheduler:> learning rate decay style: cosine
1036
+ [after model, optimizer, and learning rate scheduler are built] datetime: 2026-09-21 08:38:37
1037
+ > building train, validation, and test datasets ...
1038
+ > datasets target sizes (minimum size):
1039
+ train: 96
1040
+ validation: 0
1041
+ test: 0
1042
+ INFO:megatron.core.datasets.blended_megatron_dataset_config:Let split_matrix = [(0, 1.0), None, None]
1043
+ > building train, validation, and test datasets for GPT ...
1044
+ INFO:megatron.core.datasets.blended_megatron_dataset_builder:Building GPTDataset splits with sizes=(96, 0, 0) and config=GPTDatasetConfig(random_seed=1234, sequence_length=4096, blend=(['/workspace/data/wikitext/wikitext103_gpt2_text_document'], None), blend_per_split=None, split='100,0,0', split_matrix=[(0, 1.0), None, None], num_dataset_builder_threads=1, path_to_cache=None, mmap_bin_files=True, mock=False, tokenizer=<megatron.training.tokenizer.tokenizer._GPT2BPETokenizer object at 0x7c1ad410cc20>, 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)
1045
+ INFO:megatron.core.datasets.indexed_dataset:Load the _IndexReader from /workspace/data/wikitext/wikitext103_gpt2_text_document.idx
1046
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence lengths
1047
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence pointers
1048
+ INFO:megatron.core.datasets.indexed_dataset: Extract the document indices
1049
+ INFO:megatron.core.datasets.indexed_dataset:> total number of sequences: 1165029
1050
+ INFO:megatron.core.datasets.indexed_dataset:> total number of documents: 1801350
1051
+ INFO:megatron.core.datasets.gpt_dataset:Load the GPTDataset train indices
1052
+ INFO:megatron.core.datasets.gpt_dataset: Load the document index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-document_index.npy
1053
+ INFO:megatron.core.datasets.gpt_dataset: Load the sample index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-sample_index.npy
1054
+ INFO:megatron.core.datasets.gpt_dataset: Load the shuffle index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-shuffle_index.npy
1055
+ INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 29073
1056
+ > finished creating GPT datasets ...
1057
+ [after dataloaders are built] datetime: 2026-09-21 08:38:37
1058
+ done with setup ...
1059
+ training ...
1060
+ Overwriting rerun_state_machine.current_iteration from -1 to 0...
1061
+ (min, max) time across ranks (ms):
1062
+ model-and-optimizer-setup ......................: (3465.27, 3470.01)
1063
+ train/valid/test-data-iterators-setup ..........: (70.01, 91.77)
1064
+ [before the start of training step] datetime: 2026-09-21 08:38:37
1065
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1066
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1067
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1068
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1069
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1070
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1071
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1072
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1073
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1074
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1075
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1076
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1077
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1078
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1079
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1080
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1081
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1082
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1083
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1084
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1085
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1086
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1087
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1088
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1089
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1090
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1091
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1092
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1093
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1094
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1095
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1096
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1097
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1098
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1099
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1100
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1101
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1102
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1103
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1104
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1105
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1106
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1107
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1108
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1109
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1110
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1111
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1112
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1113
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1114
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1115
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1116
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1117
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1118
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1119
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1120
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1121
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1122
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1123
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1124
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1125
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1126
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1127
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1128
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1129
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1130
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1131
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1132
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1133
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1134
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1135
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1136
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1137
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1138
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1139
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1140
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1141
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1142
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1143
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1144
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1145
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1146
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1147
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1148
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1149
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1150
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1151
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1152
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1153
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1154
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1155
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1156
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1157
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1158
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1159
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1160
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1161
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1162
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1163
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1164
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1165
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1166
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1167
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1168
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1169
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1170
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1171
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1172
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1173
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1174
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1175
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1176
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1177
+ [2026-09-21 08:39:03] iteration 1/ 12 | consumed samples: 8 | elapsed time per iteration (ms): 25606.9 | throughput per GPU (TFLOP/s/GPU): 1.3 | learning rate: 2.000000E-05 | global batch size: 8 | lm loss: 1.118507E+01 | loss scale: 1.0 | grad norm: 35.484 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 33920 | min safe capacity: 1.0352 |
1178
+ Number of parameters in transformer block in billions: 6.71
1179
+ Number of parameters in embedding layers in billions: 0.21
1180
+ Total number of parameters in billions: 6.92
1181
+ Number of parameters in most loaded shard in billions: 6.9171
1182
+ Theoretical memory footprints: weight and optimizer=49474.72 MB
1183
+ [Rank 0] (after 1 iterations) memory (MB) | allocated: 18080.58447265625 | max allocated: 21748.86669921875 | reserved: 22168.0 | max reserved: 22168.0
1184
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1185
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1186
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1187
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1188
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1189
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1190
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1191
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1192
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1193
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1194
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1195
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1196
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1197
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1198
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1199
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1200
+ [2026-09-21 08:39:03] iteration 2/ 12 | consumed samples: 16 | elapsed time per iteration (ms): 278.1 | throughput per GPU (TFLOP/s/GPU): 115.9 | learning rate: 4.000000E-05 | global batch size: 8 | lm loss: 1.120943E+01 | loss scale: 1.0 | grad norm: 36.299 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 33920 | min safe capacity: 1.0352 |
1201
+ [2026-09-21 08:39:03] iteration 3/ 12 | consumed samples: 24 | elapsed time per iteration (ms): 586.4 | throughput per GPU (TFLOP/s/GPU): 54.9 | learning rate: 6.000000E-05 | global batch size: 8 | lm loss: 1.006625E+01 | loss scale: 1.0 | grad norm: 101.431 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 33920 | min safe capacity: 1.0352 |
1202
+ Capture CUDA graph for training!!!
1203
+ CUDA graph capture done!!!
1204
+ [2026-09-21 08:39:04] iteration 4/ 12 | consumed samples: 32 | elapsed time per iteration (ms): 833.9 | throughput per GPU (TFLOP/s/GPU): 38.6 | learning rate: 8.000000E-05 | global batch size: 8 | lm loss: 1.023696E+01 | loss scale: 1.0 | grad norm: 28.610 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 33920 | min safe capacity: 1.0352 |
1205
+ [2026-09-21 08:39:04] iteration 5/ 12 | consumed samples: 40 | elapsed time per iteration (ms): 104.3 | throughput per GPU (TFLOP/s/GPU): 308.9 | learning rate: 1.000000E-04 | global batch size: 8 | lm loss: 1.072934E+01 | loss scale: 1.0 | grad norm: 20.712 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 33920 | min safe capacity: 1.0352 |
1206
+ [2026-09-21 08:39:05] iteration 6/ 12 | consumed samples: 48 | elapsed time per iteration (ms): 104.2 | throughput per GPU (TFLOP/s/GPU): 309.1 | learning rate: 7.750000E-05 | global batch size: 8 | lm loss: 9.619561E+00 | loss scale: 1.0 | grad norm: 8.168 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 33920 | min safe capacity: 1.0352 |
1207
+ [2026-09-21 08:39:05] iteration 7/ 12 | consumed samples: 56 | elapsed time per iteration (ms): 103.4 | throughput per GPU (TFLOP/s/GPU): 311.6 | learning rate: 3.250000E-05 | global batch size: 8 | lm loss: 9.073619E+00 | loss scale: 1.0 | grad norm: 16.804 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 33920 | min safe capacity: 1.0352 |
1208
+ [2026-09-21 08:39:05] iteration 8/ 12 | consumed samples: 64 | elapsed time per iteration (ms): 103.8 | throughput per GPU (TFLOP/s/GPU): 310.3 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.812750E+00 | loss scale: 1.0 | grad norm: 8.850 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 33920 | min safe capacity: 1.0352 |
1209
+ [2026-09-21 08:39:05] iteration 9/ 12 | consumed samples: 72 | elapsed time per iteration (ms): 104.0 | throughput per GPU (TFLOP/s/GPU): 309.7 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.707838E+00 | loss scale: 1.0 | grad norm: 3.731 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 33920 | min safe capacity: 1.0352 |
1210
+ [2026-09-21 08:39:05] iteration 10/ 12 | consumed samples: 80 | elapsed time per iteration (ms): 104.3 | throughput per GPU (TFLOP/s/GPU): 308.8 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.651522E+00 | loss scale: 1.0 | grad norm: 3.491 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 33920 | min safe capacity: 1.0352 |
1211
+ [2026-09-21 08:39:05] iteration 11/ 12 | consumed samples: 88 | elapsed time per iteration (ms): 103.4 | throughput per GPU (TFLOP/s/GPU): 311.4 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.640787E+00 | loss scale: 1.0 | grad norm: 3.453 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 33920 | min safe capacity: 1.0352 |
1212
+ [2026-09-21 08:39:05] iteration 12/ 12 | consumed samples: 96 | elapsed time per iteration (ms): 103.0 | throughput per GPU (TFLOP/s/GPU): 312.7 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.590766E+00 | loss scale: 1.0 | grad norm: 3.121 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 33920 | min safe capacity: 1.0352 |
1213
+ [after training is done] datetime: 2026-09-21 08:39:05
1214
+ [rank0]:[W921 08:39:06.753250045 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())
1215
+ [rank3]:[W921 08:39:08.239723497 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())
1216
+ [rank1]:[W921 08:39:08.240978281 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())
1217
+ [rank5]:[W921 08:39:08.241841270 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())
1218
+ [rank6]:[W921 08:39:08.244155752 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())
1219
+ [rank4]:[W921 08:39:08.244828400 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())
1220
+ [rank7]:[W921 08:39:08.317756492 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())
1221
+ [rank2]:[W921 08:39:08.812782740 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())
1222
+ W0921 08:39:10.151000 703807 torch/distributed/elastic/agent/server/api.py:719] Received 15 death signal, shutting down workers
1223
+ W0921 08:39:10.152000 703807 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 703895 closing signal SIGTERM
1224
+ W0921 08:39:10.152000 703807 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 703896 closing signal SIGTERM
1225
+ W0921 08:39:10.153000 703807 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 703897 closing signal SIGTERM
1226
+ W0921 08:39:10.153000 703807 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 703898 closing signal SIGTERM
1227
+ W0921 08:39:10.154000 703807 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 703899 closing signal SIGTERM
1228
+ W0921 08:39:10.154000 703807 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 703900 closing signal SIGTERM
1229
+ W0921 08:39:10.155000 703807 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 703901 closing signal SIGTERM
1230
+ W0921 08:39:10.155000 703807 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 703902 closing signal SIGTERM
1231
+ Traceback (most recent call last):
1232
+ File "/usr/local/bin/torchrun", line 8, in <module>
1233
+ sys.exit(main())
1234
+ ^^^^^^
1235
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper
1236
+ return f(*args, **kwargs)
1237
+ ^^^^^^^^^^^^^^^^^^
1238
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 892, in main
1239
+ run(args)
1240
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 883, in run
1241
+ elastic_launch(
1242
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 139, in __call__
1243
+ return launch_agent(self._config, self._entrypoint, list(args))
1244
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1245
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 261, in launch_agent
1246
+ result = agent.run()
1247
+ ^^^^^^^^^^^
1248
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/metrics/api.py", line 138, in wrapper
1249
+ result = f(*args, **kwargs)
1250
+ ^^^^^^^^^^^^^^^^^^
1251
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 711, in run
1252
+ result = self._invoke_run(role)
1253
+ ^^^^^^^^^^^^^^^^^^^^^^
1254
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 870, in _invoke_run
1255
+ time.sleep(monitor_interval)
1256
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/api.py", line 84, in _terminate_process_handler
1257
+ raise SignalException(f"Process {os.getpid()} got signal: {sigval}", sigval=sigval)
1258
+ torch.distributed.elastic.multiprocessing.api.SignalException: Process 703807 got signal: 15
granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep3/mem_samples.csv ADDED
@@ -0,0 +1,200 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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, 0
10
+ 1, 0
11
+ 2, 0
12
+ 3, 0
13
+ 4, 0
14
+ 5, 0
15
+ 6, 0
16
+ 7, 0
17
+ 0, 1
18
+ 1, 1
19
+ 2, 1
20
+ 3, 1
21
+ 4, 1
22
+ 5, 1
23
+ 6, 1
24
+ 7, 1
25
+ 0, 4
26
+ 1, 4
27
+ 2, 4
28
+ 3, 4
29
+ 4, 1
30
+ 5, 1
31
+ 6, 1
32
+ 7, 1
33
+ 0, 4
34
+ 1, 4
35
+ 2, 4
36
+ 3, 4
37
+ 4, 4
38
+ 5, 4
39
+ 6, 4
40
+ 7, 4
41
+ 0, 821
42
+ 1, 821
43
+ 2, 821
44
+ 3, 819
45
+ 4, 819
46
+ 5, 819
47
+ 6, 819
48
+ 7, 819
49
+ 0, 1537
50
+ 1, 1537
51
+ 2, 1537
52
+ 3, 1537
53
+ 4, 1537
54
+ 5, 1537
55
+ 6, 1537
56
+ 7, 1537
57
+ 0, 1823
58
+ 1, 1823
59
+ 2, 1823
60
+ 3, 1823
61
+ 4, 1823
62
+ 5, 1823
63
+ 6, 1823
64
+ 7, 1823
65
+ 0, 2671
66
+ 1, 2671
67
+ 2, 2671
68
+ 3, 2671
69
+ 4, 2671
70
+ 5, 2671
71
+ 6, 2671
72
+ 7, 2671
73
+ 0, 15337
74
+ 1, 15283
75
+ 2, 15339
76
+ 3, 15301
77
+ 4, 15303
78
+ 5, 15313
79
+ 6, 15315
80
+ 7, 15295
81
+ 0, 16331
82
+ 1, 16301
83
+ 2, 16313
84
+ 3, 16315
85
+ 4, 16341
86
+ 5, 16365
87
+ 6, 16341
88
+ 7, 16337
89
+ 0, 16925
90
+ 1, 16927
91
+ 2, 16927
92
+ 3, 16929
93
+ 4, 16923
94
+ 5, 16927
95
+ 6, 16929
96
+ 7, 16923
97
+ 0, 17065
98
+ 1, 17067
99
+ 2, 17067
100
+ 3, 17069
101
+ 4, 17063
102
+ 5, 17067
103
+ 6, 17069
104
+ 7, 17063
105
+ 0, 17561
106
+ 1, 17563
107
+ 2, 17563
108
+ 3, 17565
109
+ 4, 17559
110
+ 5, 17563
111
+ 6, 17565
112
+ 7, 17559
113
+ 0, 26069
114
+ 1, 26051
115
+ 2, 26091
116
+ 3, 26253
117
+ 4, 26117
118
+ 5, 26113
119
+ 6, 26111
120
+ 7, 26313
121
+ 0, 28501
122
+ 1, 28503
123
+ 2, 28503
124
+ 3, 28485
125
+ 4, 28509
126
+ 5, 28483
127
+ 6, 28503
128
+ 7, 28509
129
+ 0, 28501
130
+ 1, 28503
131
+ 2, 28503
132
+ 3, 28485
133
+ 4, 28509
134
+ 5, 28483
135
+ 6, 28503
136
+ 7, 28509
137
+ 0, 28501
138
+ 1, 28503
139
+ 2, 28503
140
+ 3, 28485
141
+ 4, 28509
142
+ 5, 28483
143
+ 6, 28503
144
+ 7, 28509
145
+ 0, 28905
146
+ 1, 28907
147
+ 2, 28907
148
+ 3, 28889
149
+ 4, 28913
150
+ 5, 28887
151
+ 6, 28907
152
+ 7, 28913
153
+ 0, 29869
154
+ 1, 29879
155
+ 2, 29867
156
+ 3, 29863
157
+ 4, 29849
158
+ 5, 29833
159
+ 6, 29891
160
+ 7, 29891
161
+ 0, 37195
162
+ 1, 37201
163
+ 2, 37213
164
+ 3, 37197
165
+ 4, 37197
166
+ 5, 37193
167
+ 6, 37185
168
+ 7, 37197
169
+ 0, 39097
170
+ 1, 39087
171
+ 2, 38951
172
+ 3, 38931
173
+ 4, 39095
174
+ 5, 38941
175
+ 6, 39091
176
+ 7, 39095
177
+ 0, 36763
178
+ 1, 36753
179
+ 2, 36617
180
+ 3, 36597
181
+ 4, 36761
182
+ 5, 36607
183
+ 6, 36757
184
+ 7, 36761
185
+ 0, 35083
186
+ 1, 34685
187
+ 2, 34937
188
+ 3, 34917
189
+ 4, 35081
190
+ 5, 34277
191
+ 6, 34689
192
+ 7, 35081
193
+ 0, 1
194
+ 1, 1
195
+ 2, 3
196
+ 3, 132
197
+ 4, 35081
198
+ 5, 644
199
+ 6, 4
200
+ 7, 4
granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep3/run.env ADDED
@@ -0,0 +1,102 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ backend=ultraep
2
+ config=OLMOE_1B_7B_L16
3
+ family=tierB
4
+ source_model=OLMoE-1B-7B-0924
5
+ standard_num_layers=16
6
+ base_rung=
7
+ repo=/workspace/Megatron-UltraEP
8
+ branch=ultraep-bracing-regions-through-kernels
9
+ commit=03d57ef7
10
+ num_layers=16
11
+ hidden=2048
12
+ ffn_hidden=1024
13
+ moe_ffn_hidden=1024
14
+ num_experts=64
15
+ topk=8
16
+ experts_per_rank=8
17
+ expert_params=6291456
18
+ heads=16
19
+ kv_groups=16
20
+ kv_channels=128
21
+ qk_layernorm=0
22
+ shared_ffn=0
23
+ shared_expert_overlap=0
24
+ moe_token_dispatcher=flex
25
+ mla=0
26
+ q_lora_rank=
27
+ kv_lora_rank=512
28
+ qk_head_dim=128
29
+ qk_pos_emb_head_dim=64
30
+ v_head_dim=128
31
+ dense_first_layers=0
32
+ moe_layers=16
33
+ rotary_base=10000
34
+ rotary_percent=1.0
35
+ rotary_scaling=
36
+ mscale=
37
+ mscale_all_dim=
38
+ qkv_bias=0
39
+ router_balancing=aux_loss
40
+ aux_loss_coeff=1e-2
41
+ router_pre_softmax=1
42
+ router_fusion=0
43
+ router_dtype=fp32
44
+ router_groups=
45
+ router_group_topk=
46
+ router_score_function=softmax
47
+ router_expert_bias=0
48
+ router_bias_update_rate=
49
+ router_topk_scaling_factor=
50
+ shared_expert_gate=0
51
+ apply_layernorm_1p=0
52
+ attention_output_gate=0
53
+ linear_attention_type=
54
+ linear_attention_freq=
55
+ linear_conv_kernel_dim=
56
+ linear_key_head_dim=
57
+ linear_value_head_dim=
58
+ linear_num_key_heads=
59
+ linear_num_value_heads=
60
+ mtp_num_layers=0
61
+ mtp_loss_scaling_factor=0.1
62
+ no_weight_decay_cond_type=
63
+ max_position_embeddings=4096
64
+ init_method_std=0.02
65
+ device_initiated_grouped_gemm=1
66
+ norm_epsilon=1e-5
67
+ seq_length=4096
68
+ micro_batch_size=1
69
+ global_batch_size=8
70
+ num_microbatches=1
71
+ assignments_per_rank=32768
72
+ mean_tokens_per_expert=4096
73
+ slot_policy=matched
74
+ slots_per_rank=8
75
+ capacity_mode=static
76
+ capacity=1.08984375
77
+ cuda_graph=1
78
+ probe_mode=0
79
+ async_grad_reduce=0
80
+ branch_role=timing
81
+ memory_dump_mode=none
82
+ memory_snapshot_path=
83
+ cuda_graph_warmup_steps=3
84
+ recompute=none
85
+ force_balance=0
86
+ train_iters=12
87
+ ep_size=8
88
+ echo_dispatch_overlap=1
89
+ echo_dispatch_recompute=0
90
+ ultraep_quota_min_tokens=512
91
+ ultraep_weight_sync_join=
92
+ profile_enable=0
93
+ kernel_trace=0
94
+ fp8_mode=mxfp8
95
+ fp8_recipe=mxfp8
96
+ training_seed=1234
97
+ learning_rate=1e-4
98
+ min_learning_rate=1.0e-5
99
+ lr_warmup_iters=5
100
+ state_dump_enable=0
101
+ state_dump_dir=
102
+ moonep_assert_no_pending_reduce=0
granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep3/train.log ADDED
@@ -0,0 +1,1258 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ==============================================================
2
+ OLMOE_1B_7B_L16 / ultraep (slots/rank=8 policy=matched)
3
+ E=64 K=8 H'=1024 m=8
4
+ L=16 seq=4096 gbs=8 (1 microbatches)
5
+ cap_mode=static cap=1.08984375 cg=1 recompute=none
6
+ memory_dump=none snapshots=/workspace/granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep3/memory_snapshots/snapshot.pickle
7
+ repo=/workspace/Megatron-UltraEP @ 03d57ef7
8
+ ==============================================================
9
+ W0921 08:39:15.651000 706795 torch/distributed/run.py:766]
10
+ W0921 08:39:15.651000 706795 torch/distributed/run.py:766] *****************************************
11
+ W0921 08:39:15.651000 706795 torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.
12
+ W0921 08:39:15.651000 706795 torch/distributed/run.py:766] *****************************************
13
+ 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
14
+ WARNING: overriding default arguments for tokenizer_type:GPT2BPETokenizer with tokenizer_type:GPT2BPETokenizer
15
+ Number of virtual stages per pipeline stage: None
16
+ accumulate and all-reduce gradients in fp32 for bfloat16 data type.
17
+ using torch.bfloat16 for parameters ...
18
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
19
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
20
+ ------------------------ arguments ------------------------
21
+ account_for_embedding_in_pipeline_split ......... False
22
+ account_for_loss_in_pipeline_split .............. False
23
+ accumulate_allreduce_grads_in_fp32 .............. True
24
+ adam_beta1 ...................................... 0.9
25
+ adam_beta2 ...................................... 0.999
26
+ adam_eps ........................................ 1e-08
27
+ add_bias_linear ................................. False
28
+ add_position_embedding .......................... False
29
+ add_qkv_bias .................................... False
30
+ adlr_autoresume ................................. False
31
+ adlr_autoresume_interval ........................ 1000
32
+ align_grad_reduce ............................... True
33
+ align_param_gather .............................. False
34
+ app_tag_run_name ................................ None
35
+ app_tag_run_version ............................. 0.0.0
36
+ apply_layernorm_1p .............................. False
37
+ apply_query_key_layer_scaling ................... False
38
+ apply_residual_connection_post_layernorm ........ False
39
+ apply_rope_fusion ............................... True
40
+ async_save ...................................... None
41
+ async_tensor_model_parallel_allreduce ........... True
42
+ attention_backend ............................... AttnBackend.auto
43
+ attention_dropout ............................... 0.0
44
+ attention_softmax_in_fp32 ....................... False
45
+ auto_detect_ckpt_format ......................... False
46
+ barrier_with_L1_time ............................ True
47
+ bert_binary_head ................................ True
48
+ bert_embedder_type .............................. megatron
49
+ bert_load ....................................... None
50
+ bf16 ............................................ True
51
+ bias_dropout_fusion ............................. True
52
+ bias_gelu_fusion ................................ False
53
+ bias_swiglu_fusion .............................. True
54
+ biencoder_projection_dim ........................ 0
55
+ biencoder_shared_query_context_model ............ False
56
+ block_data_path ................................. None
57
+ calc_ft_timeouts ................................ False
58
+ calculate_per_token_loss ........................ False
59
+ check_for_large_grads ........................... False
60
+ check_for_nan_in_loss_and_grad .................. False
61
+ check_for_spiky_loss ............................ False
62
+ check_weight_hash_across_dp_replicas_interval ... None
63
+ ckpt_assume_constant_structure .................. False
64
+ ckpt_convert_format ............................. None
65
+ ckpt_convert_save ............................... None
66
+ ckpt_convert_update_legacy_dist_opt_format ...... False
67
+ ckpt_format ..................................... torch_dist
68
+ ckpt_fully_parallel_load ........................ False
69
+ ckpt_fully_parallel_save ........................ True
70
+ ckpt_fully_parallel_save_deprecated ............. False
71
+ ckpt_step ....................................... None
72
+ classes_fraction ................................ 1.0
73
+ clip_grad ....................................... 1.0
74
+ clone_scatter_output_in_embedding ............... True
75
+ config_logger_dir ...............................
76
+ consumed_train_samples .......................... 0
77
+ consumed_valid_samples .......................... 0
78
+ context_parallel_size ........................... 1
79
+ cp_comm_type .................................... ['p2p']
80
+ create_attention_mask_in_dataloader ............. True
81
+ cross_entropy_fusion_impl ....................... native
82
+ cross_entropy_loss_fusion ....................... False
83
+ cuda_graph_scope ................................ full_iteration
84
+ cuda_graph_warmup_steps ......................... 3
85
+ data_args_path .................................. None
86
+ data_cache_path ................................. None
87
+ data_parallel_random_init ....................... False
88
+ data_parallel_sharding_strategy ................. no_shard
89
+ data_parallel_size .............................. 8
90
+ data_path ....................................... ['/workspace/data/wikitext/wikitext103_gpt2_text_document']
91
+ data_per_class_fraction ......................... 1.0
92
+ data_sharding ................................... True
93
+ dataloader_type ................................. single
94
+ ddp_average_in_collective ....................... False
95
+ ddp_bucket_size ................................. None
96
+ ddp_num_buckets ................................. None
97
+ ddp_pad_buckets_for_high_nccl_busbw ............. False
98
+ decoder_first_pipeline_num_layers ............... None
99
+ decoder_last_pipeline_num_layers ................ None
100
+ decoder_num_layers .............................. None
101
+ decoder_seq_length .............................. None
102
+ decoupled_lr .................................... None
103
+ decoupled_min_lr ................................ None
104
+ decrease_batch_size_if_needed ................... False
105
+ defer_embedding_wgrad_compute ................... False
106
+ delay_wgrad_compute ............................. False
107
+ deprecated_use_mcore_models ..................... True
108
+ deterministic_mode .............................. False
109
+ dino_bottleneck_size ............................ 256
110
+ dino_freeze_last_layer .......................... 1
111
+ dino_head_hidden_size ........................... 2048
112
+ dino_local_crops_number ......................... 10
113
+ dino_local_img_size ............................. 96
114
+ dino_norm_last_layer ............................ False
115
+ dino_teacher_temp ............................... 0.07
116
+ dino_warmup_teacher_temp ........................ 0.04
117
+ dino_warmup_teacher_temp_epochs ................. 30
118
+ disable_bf16_reduced_precision_matmul ........... False
119
+ disable_mamba_mem_eff_path ...................... False
120
+ disable_straggler_on_startup .................... False
121
+ dist_ckpt_format_deprecated ..................... None
122
+ dist_ckpt_strictness ............................ assume_ok_unexpected
123
+ distribute_saved_activations .................... False
124
+ distributed_backend ............................. nccl
125
+ distributed_timeout_minutes ..................... 10
126
+ embedding_init_method_std ....................... None
127
+ embedding_path .................................. None
128
+ empty_unused_memory_level ....................... 0
129
+ enable_cuda_graph ............................... True
130
+ enable_experimental ............................. False
131
+ enable_ft_package ............................... False
132
+ enable_gloo_process_groups ...................... True
133
+ enable_msc ...................................... True
134
+ enable_one_logger ............................... True
135
+ encoder_num_layers .............................. 16
136
+ encoder_seq_length .............................. 4096
137
+ end_weight_decay ................................ 0.1
138
+ eod_mask_loss ................................... False
139
+ error_injection_rate ............................ 0
140
+ error_injection_type ............................ transient_error
141
+ eval_interval ................................... 1000000
142
+ eval_iters ...................................... 0
143
+ evidence_data_path .............................. None
144
+ exit_duration_in_mins ........................... None
145
+ exit_interval ................................... None
146
+ exit_on_missing_checkpoint ...................... False
147
+ exit_signal_handler ............................. False
148
+ exp_avg_dtype ................................... torch.float32
149
+ exp_avg_sq_dtype ................................ torch.float32
150
+ expert_model_parallel_size ...................... 8
151
+ expert_tensor_parallel_size ..................... 1
152
+ export_force_local_attention .................... False
153
+ export_kd_cfg ................................... None
154
+ export_kd_teacher_ckpt_format ................... None
155
+ export_kd_teacher_load .......................... None
156
+ export_kv_cache_quant ........................... False
157
+ export_legacy_megatron .......................... False
158
+ export_model_type ............................... GPTModel
159
+ export_moe_apply_probs_on_input ................. False
160
+ export_qk_l2_norm ............................... False
161
+ export_quant_cfg ................................ None
162
+ export_real_quant_cfg ........................... None
163
+ export_te_mcore_model ........................... False
164
+ external_cuda_graph ............................. False
165
+ ffn_hidden_size ................................. 1024
166
+ finetune ........................................ False
167
+ finetune_data_split ............................. train
168
+ finetune_hf_dataset ............................. None
169
+ first_last_layers_bf16 .......................... False
170
+ flash_decode .................................... False
171
+ fp16 ............................................ False
172
+ fp16_lm_cross_entropy ........................... False
173
+ fp32_residual_connection ........................ False
174
+ fp8 ............................................. e4m3
175
+ fp8_amax_compute_algo ........................... most_recent
176
+ fp8_amax_history_len ............................ 1
177
+ fp8_interval .................................... 1
178
+ fp8_margin ...................................... 0
179
+ fp8_param_gather ................................ False
180
+ fp8_recipe ...................................... mxfp8
181
+ fp8_wgrad ....................................... True
182
+ fsdp_double_buffer .............................. False
183
+ global_batch_size ............................... 8
184
+ grad_reduce_in_bf16 ............................. False
185
+ gradient_accumulation_fusion .................... True
186
+ gradient_reduce_div_fusion ...................... True
187
+ group_query_attention ........................... True
188
+ head_lr_mult .................................... 1.0
189
+ heterogeneous_layers_config_encoded_json ........ None
190
+ heterogeneous_layers_config_path ................ None
191
+ hidden_dropout .................................. 0.0
192
+ hidden_size ..................................... 2048
193
+ hierarchical_context_parallel_sizes ............. None
194
+ high_priority_stream_groups ..................... []
195
+ hybrid_attention_ratio .......................... 0.0
196
+ hybrid_mlp_ratio ................................ 0.0
197
+ hybrid_override_pattern ......................... None
198
+ hysteresis ...................................... 2
199
+ ict_head_size ................................... None
200
+ ict_load ........................................ None
201
+ img_h ........................................... 224
202
+ img_w ........................................... 224
203
+ indexer_batch_size .............................. 128
204
+ indexer_log_interval ............................ 1000
205
+ inference_batch_times_seqlen_threshold .......... -1
206
+ inference_dynamic_batching ...................... False
207
+ inference_dynamic_batching_buffer_guaranteed_fraction 0.2
208
+ inference_dynamic_batching_buffer_overflow_factor None
209
+ inference_dynamic_batching_buffer_size_gb ....... 40.0
210
+ inference_dynamic_batching_chunk_size ........... 256
211
+ inference_dynamic_batching_max_requests_override None
212
+ inference_dynamic_batching_max_tokens_override .. None
213
+ inference_dynamic_batching_num_cuda_graphs ...... 16
214
+ inference_max_batch_size ........................ 8
215
+ inference_max_seq_length ........................ 2560
216
+ inference_rng_tracker ........................... False
217
+ init_method_std ................................. 0.02
218
+ init_method_xavier_uniform ...................... False
219
+ init_model_with_meta_device ..................... False
220
+ initial_loss_scale .............................. 4294967296
221
+ inprocess_active_world_size ..................... 8
222
+ inprocess_barrier_timeout ....................... 120
223
+ inprocess_completion_timeout .................... 120
224
+ inprocess_empty_cuda_cache ...................... False
225
+ inprocess_granularity ........................... node
226
+ inprocess_hard_timeout .......................... 90
227
+ inprocess_heartbeat_interval .................... 30
228
+ inprocess_heartbeat_timeout ..................... 60
229
+ inprocess_last_call_wait ........................ 1
230
+ inprocess_max_iterations ........................ None
231
+ inprocess_monitor_process_interval .............. 1.0
232
+ inprocess_monitor_thread_interval ............... 1.0
233
+ inprocess_progress_watchdog_interval ............ 1.0
234
+ inprocess_restart ............................... False
235
+ inprocess_soft_timeout .......................... 60
236
+ inprocess_termination_grace_time ................ 1
237
+ is_hybrid_model ................................. False
238
+ iter_per_epoch .................................. 1250
239
+ iterations_to_skip .............................. []
240
+ keep_fp8_transpose_cache_when_using_custom_fsdp . False
241
+ kitchen_config_file ............................. None
242
+ kitchen_recipe_number ........................... None
243
+ kv_channels ..................................... 128
244
+ kv_lora_rank .................................... 32
245
+ lazy_mpu_init ................................... None
246
+ load ............................................ None
247
+ load_main_params_from_ckpt ...................... None
248
+ load_model_opt_format ........................... False
249
+ local_rank ...................................... 0
250
+ log_energy ...................................... False
251
+ log_interval .................................... 1
252
+ log_loss_scale_to_tensorboard ................... True
253
+ log_memory_to_tensorboard ....................... False
254
+ log_num_zeros_in_grad ........................... False
255
+ log_params_norm ................................. False
256
+ log_progress .................................... False
257
+ log_straggler ................................... False
258
+ log_throughput .................................. True
259
+ log_timers_to_tensorboard ....................... False
260
+ log_validation_ppl_to_tensorboard ............... False
261
+ log_world_size_to_tensorboard ................... False
262
+ logging_level ................................... None
263
+ loss_scale ...................................... None
264
+ loss_scale_window ............................... 1000
265
+ lr .............................................. 0.0001
266
+ lr_decay_iters .................................. 8
267
+ lr_decay_samples ................................ None
268
+ lr_decay_style .................................. cosine
269
+ lr_warmup_fraction .............................. None
270
+ lr_warmup_init .................................. 0.0
271
+ lr_warmup_iters ................................. 5
272
+ lr_warmup_samples ............................... 0
273
+ lr_wsd_decay_iters .............................. None
274
+ lr_wsd_decay_samples ............................ None
275
+ lr_wsd_decay_style .............................. exponential
276
+ main_grads_dtype ................................ torch.float32
277
+ main_params_dtype ............................... torch.float32
278
+ make_vocab_size_divisible_by .................... 128
279
+ mamba_head_dim .................................. 64
280
+ mamba_num_groups ................................ 8
281
+ mamba_num_heads ................................. None
282
+ mamba_state_dim ................................. 128
283
+ manual_gc ....................................... False
284
+ manual_gc_eval .................................. True
285
+ manual_gc_interval .............................. 0
286
+ mask_factor ..................................... 1.0
287
+ mask_prob ....................................... 0.15
288
+ mask_type ....................................... random
289
+ masked_softmax_fusion ........................... False
290
+ max_position_embeddings ......................... 4096
291
+ max_tokens_to_oom ............................... 12000
292
+ memory_snapshot_path ............................ snapshot.pickle
293
+ merge_file ...................................... /workspace/data/wikitext/gpt2-merges.txt
294
+ micro_batch_size ................................ 1
295
+ microbatch_group_size_per_vp_stage .............. None
296
+ mid_level_dataset_surplus ....................... 0.005
297
+ min_loss_scale .................................. 1.0
298
+ min_lr .......................................... 1e-05
299
+ mlp_chunks_for_prefill .......................... 1
300
+ mmap_bin_files .................................. True
301
+ mock_data ....................................... False
302
+ moe_apply_probs_on_input ........................ False
303
+ moe_aux_loss_coeff .............................. 0.01
304
+ moe_capacity_monitor ............................ True
305
+ moe_deepep_num_sms .............................. 20
306
+ moe_echo_enable_random_offloading ............... False
307
+ moe_echo_expert_dispatch_overlap ................ False
308
+ moe_echo_expert_dispatcher_type ................. hybridep
309
+ moe_echo_recompute_expert_dispatch .............. False
310
+ moe_enable_deepep ............................... False
311
+ moe_enable_echo ................................. False
312
+ moe_enable_hybridep ............................. True
313
+ moe_enable_ultraep .............................. True
314
+ moe_expert_capacity_factor ...................... None
315
+ moe_extended_tp ................................. False
316
+ moe_ffn_hidden_size ............................. 1024
317
+ moe_grouped_gemm ................................ True
318
+ moe_hybridep_num_sms ............................ 16
319
+ moe_input_jitter_eps ............................ None
320
+ moe_layer_freq .................................. 1
321
+ moe_layer_recompute ............................. False
322
+ moe_num_echo_experts ............................ None
323
+ moe_num_redundant_experts_per_rank .............. 8
324
+ moe_pad_expert_input_to_capacity ................ False
325
+ moe_per_layer_logging ........................... False
326
+ moe_permute_fusion .............................. False
327
+ moe_received_token_capacity ..................... 1.08984375
328
+ moe_router_bias_update_rate ..................... 0.001
329
+ moe_router_dtype ................................ fp32
330
+ moe_router_enable_expert_bias ................... False
331
+ moe_router_force_load_balancing ................. False
332
+ moe_router_group_topk ........................... None
333
+ moe_router_load_balancing_type .................. aux_loss
334
+ moe_router_num_groups ........................... None
335
+ moe_router_padding_for_fp8 ...................... False
336
+ moe_router_pre_softmax .......................... True
337
+ moe_router_score_function ....................... softmax
338
+ moe_router_topk ................................. 8
339
+ moe_router_topk_scaling_factor .................. None
340
+ moe_shared_expert_intermediate_size ............. None
341
+ moe_shared_expert_overlap ....................... False
342
+ moe_token_dispatcher_type ....................... flex
343
+ moe_token_drop_policy ........................... probs
344
+ moe_ultraep_max_microbatches .................... None
345
+ moe_ultraep_overlap_weight_sync ................. True
346
+ moe_ultraep_weight_sync_join .................... before_dispatch
347
+ moe_upcycling_granularity ....................... 1
348
+ moe_use_device_initiated_grouped_gemm ........... True
349
+ moe_use_legacy_grouped_gemm ..................... False
350
+ moe_use_upcycling ............................... False
351
+ moe_z_loss_coeff ................................ None
352
+ mrope_section ................................... None
353
+ mscale .......................................... 1.0
354
+ mscale_all_dim .................................. 0.0
355
+ mtp_loss_scaling_factor ......................... 0.1
356
+ mtp_num_layers .................................. None
357
+ multi_latent_attention .......................... False
358
+ nccl_all_reduce_for_prefill ..................... False
359
+ nccl_communicator_config_path ................... None
360
+ nccl_ub ......................................... False
361
+ no_load_optim ................................... None
362
+ no_load_rng ..................................... None
363
+ no_persist_layer_norm ........................... False
364
+ no_rope_freq .................................... None
365
+ no_save_optim ................................... None
366
+ no_save_rng ..................................... None
367
+ non_persistent_ckpt_type ........................ None
368
+ non_persistent_global_ckpt_dir .................. None
369
+ non_persistent_local_ckpt_algo .................. fully_parallel
370
+ non_persistent_local_ckpt_dir ................... None
371
+ non_persistent_save_interval .................... None
372
+ norm_epsilon .................................... 1e-05
373
+ normalization ................................... RMSNorm
374
+ num_attention_heads ............................. 16
375
+ num_channels .................................... 3
376
+ num_classes ..................................... 1000
377
+ num_dataset_builder_threads ..................... 1
378
+ num_distributed_optimizer_instances ............. 1
379
+ num_experts ..................................... 64
380
+ num_layers ...................................... 16
381
+ num_layers_at_end_in_bf16 ....................... 1
382
+ num_layers_at_start_in_bf16 ..................... 1
383
+ num_layers_per_virtual_pipeline_stage ........... None
384
+ num_query_groups ................................ 16
385
+ num_virtual_stages_per_pipeline_rank ............ None
386
+ num_workers ..................................... 2
387
+ object_storage_cache_path ....................... None
388
+ one_logger_async ................................ False
389
+ one_logger_project .............................. megatron-lm
390
+ one_logger_run_name ............................. None
391
+ onnx_safe ....................................... None
392
+ openai_gelu ..................................... False
393
+ optimizer ....................................... adam
394
+ optimizer_cpu_offload ........................... False
395
+ optimizer_offload_fraction ...................... 1.0
396
+ output_bert_embeddings .......................... False
397
+ overlap_cpu_optimizer_d2h_h2d ................... False
398
+ overlap_grad_reduce ............................. True
399
+ overlap_moe_expert_parallel_comm ................ False
400
+ overlap_p2p_comm ................................ False
401
+ overlap_p2p_comm_warmup_flush ................... False
402
+ overlap_param_gather ............................ True
403
+ overlap_param_gather_with_optimizer_step ........ False
404
+ override_opt_param_scheduler .................... False
405
+ params_dtype .................................... torch.bfloat16
406
+ patch_dim ....................................... 16
407
+ per_split_data_args_path ........................ None
408
+ perform_initialization .......................... True
409
+ pin_cpu_grads ................................... True
410
+ pin_cpu_params .................................. True
411
+ pipeline_model_parallel_comm_backend ............ None
412
+ pipeline_model_parallel_layout .................. None
413
+ pipeline_model_parallel_size .................... 1
414
+ position_embedding_type ......................... rope
415
+ pretrained_checkpoint ........................... None
416
+ profile ......................................... False
417
+ profile_ranks ................................... [0]
418
+ profile_step_end ................................ 12
419
+ profile_step_start .............................. 10
420
+ q_lora_rank ..................................... None
421
+ qk_head_dim ..................................... 128
422
+ qk_l2_norm ...................................... False
423
+ qk_layernorm .................................... False
424
+ qk_pos_emb_head_dim ............................. 64
425
+ query_in_block_prob ............................. 0.1
426
+ rampup_batch_size ............................... None
427
+ rank ............................................ 0
428
+ recompute_granularity ........................... None
429
+ recompute_method ................................ None
430
+ recompute_modules ............................... None
431
+ recompute_num_layers ............................ None
432
+ record_memory_history ........................... False
433
+ relative_attention_max_distance ................. 128
434
+ relative_attention_num_buckets .................. 32
435
+ replication ..................................... False
436
+ replication_factor .............................. 2
437
+ replication_jump ................................ None
438
+ rerun_mode ...................................... validate_results
439
+ reset_attention_mask ............................ False
440
+ reset_position_ids .............................. False
441
+ result_rejected_tracker_filename ................ None
442
+ retriever_report_topk_accuracies ................ []
443
+ retriever_score_scaling ......................... False
444
+ retriever_seq_length ............................ 256
445
+ retro_add_retriever ............................. False
446
+ retro_attention_gate ............................ 1
447
+ retro_cyclic_train_iters ........................ None
448
+ retro_encoder_attention_dropout ................. 0.1
449
+ retro_encoder_hidden_dropout .................... 0.1
450
+ retro_encoder_layers ............................ 2
451
+ retro_num_neighbors ............................. 2
452
+ retro_num_retrieved_chunks ...................... 2
453
+ retro_project_dir ............................... None
454
+ retro_verify_neighbor_count ..................... True
455
+ reuse_grad_buf_for_mxfp8_param_ag ............... False
456
+ rope_scaling_factor ............................. 8.0
457
+ rope_type ....................................... None
458
+ rotary_base ..................................... 10000
459
+ rotary_interleaved .............................. False
460
+ rotary_percent .................................. 1.0
461
+ rotary_scaling_factor ........................... 1.0
462
+ rotary_seq_len_interpolation_factor ............. None
463
+ run_workload_inspector_server ................... False
464
+ sample_rate ..................................... 1.0
465
+ save ............................................ None
466
+ save_interval ................................... None
467
+ save_retain_interval ............................ None
468
+ scatter_gather_tensors_in_pipeline .............. True
469
+ seed ............................................ 1234
470
+ seq_length ...................................... 4096
471
+ sequence_parallel ............................... False
472
+ sft ............................................. False
473
+ sft_tokenizer_prompt_format ..................... nemotron-h-aligned
474
+ sgd_momentum .................................... 0.9
475
+ short_seq_prob .................................. 0.1
476
+ skip_train ...................................... False
477
+ skipped_train_samples ........................... 0
478
+ spec ............................................ None
479
+ split ........................................... 100,0,0
480
+ squared_relu .................................... False
481
+ start_weight_decay .............................. 0.1
482
+ straggler_ctrlr_port ............................ 65535
483
+ straggler_minmax_count .......................... 1
484
+ suggested_communication_unit_size ............... None
485
+ swiglu .......................................... True
486
+ swin_backbone_type .............................. tiny
487
+ symmetric_ar_type ............................... None
488
+ te_rng_tracker .................................. True
489
+ tensor_model_parallel_size ...................... 1
490
+ tensorboard_dir ................................. None
491
+ tensorboard_log_interval ........................ 1
492
+ tensorboard_queue_size .......................... 1000
493
+ test_data_path .................................. None
494
+ test_mode ....................................... False
495
+ tiktoken_num_special_tokens ..................... 1000
496
+ tiktoken_pattern ................................ None
497
+ tiktoken_special_tokens ......................... None
498
+ timing_log_level ................................ 0
499
+ timing_log_option ............................... minmax
500
+ titles_data_path ................................ None
501
+ tokenizer_model ................................. None
502
+ tokenizer_type .................................. GPT2BPETokenizer
503
+ torch_fsdp2_reshard_after_forward ............... True
504
+ tp_comm_bootstrap_backend ....................... nccl
505
+ tp_comm_bulk_dgrad .............................. True
506
+ tp_comm_bulk_wgrad .............................. True
507
+ tp_comm_overlap ................................. False
508
+ tp_comm_overlap_ag .............................. True
509
+ tp_comm_overlap_cfg ............................. None
510
+ tp_comm_overlap_rs .............................. True
511
+ tp_comm_overlap_rs_dgrad ........................ False
512
+ tp_comm_split_ag ................................ True
513
+ tp_comm_split_rs ................................ True
514
+ train_data_path ................................. None
515
+ train_iters ..................................... 12
516
+ train_samples ................................... None
517
+ train_sync_interval ............................. None
518
+ transformer_impl ................................ transformer_engine
519
+ transformer_pipeline_model_parallel_size ........ 1
520
+ untie_embeddings_and_output_weights ............. True
521
+ use_checkpoint_args ............................. False
522
+ use_checkpoint_opt_param_scheduler .............. False
523
+ use_cpu_initialization .......................... None
524
+ use_custom_fsdp ................................. False
525
+ use_dist_ckpt ................................... True
526
+ use_dist_ckpt_deprecated ........................ False
527
+ use_distributed_optimizer ....................... True
528
+ use_flash_attn .................................. False
529
+ use_legacy_models ............................... False
530
+ use_mp_args_from_checkpoint_args ................ False
531
+ use_one_sent_docs ............................... False
532
+ use_persistent_ckpt_worker ...................... False
533
+ use_precision_aware_optimizer ................... False
534
+ use_pytorch_profiler ............................ False
535
+ use_ring_exchange_p2p ........................... False
536
+ use_rope_scaling ................................ False
537
+ use_rotary_position_embeddings .................. False
538
+ use_sharp ....................................... False
539
+ use_tokenizer_model_from_checkpoint_args ........ True
540
+ use_torch_fsdp2 ................................. False
541
+ use_torch_optimizer_for_cpu_offload ............. False
542
+ use_tp_pp_dp_mapping ............................ False
543
+ v_head_dim ...................................... 128
544
+ valid_data_path ................................. None
545
+ variable_seq_lengths ............................ False
546
+ virtual_pipeline_model_parallel_size ............ None
547
+ vision_backbone_type ............................ vit
548
+ vision_pretraining .............................. False
549
+ vision_pretraining_type ......................... classify
550
+ vocab_extra_ids ................................. 0
551
+ vocab_file ...................................... /workspace/data/wikitext/gpt2-vocab.json
552
+ vocab_size ...................................... None
553
+ wandb_exp_name ..................................
554
+ wandb_project ...................................
555
+ wandb_save_dir ..................................
556
+ weight_decay .................................... 0.1
557
+ weight_decay_incr_style ......................... constant
558
+ wgrad_deferral_limit ............................ 0
559
+ world_size ...................................... 8
560
+ yaml_cfg ........................................ None
561
+ -------------------- end of arguments ---------------------
562
+ INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 1
563
+ > building GPT2BPETokenizer tokenizer ...
564
+ > padded vocab (size: 50257) with 47 dummy tokens (new size: 50304)
565
+ WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.VALIDATE_RESULTS
566
+ > initializing torch distributed ...
567
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
568
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
569
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
570
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
571
+ 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
572
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
573
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
574
+ /workspace/Megatron-UltraEP/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-UltraEP/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-UltraEP/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
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
581
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
582
+ [Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
583
+ [Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
584
+ [Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
585
+ [Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
586
+ [Gloo] Rank [Gloo] Rank 5 is connected to 76 peer ranks. is connected to Expected number of connected peer ranks is : 77 peer ranks.
587
+ Expected number of connected peer ranks is : 7
588
+ [Gloo] Rank [Gloo] Rank 4 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 :
589
+ 7
590
+ [Gloo] Rank [Gloo] Rank 7 is connected to 07 is connected to peer ranks. 7Expected number of connected peer ranks is : peer ranks. 7Expected number of connected peer ranks is :
591
+ 7
592
+ [Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
593
+ [Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
594
+ [Gloo] Rank [Gloo] Rank 4 is connected to 37 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 5 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
597
+ [Gloo] Rank 6 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
598
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
599
+ > initialized tensor model parallel with size 1
600
+ > initialized pipeline model parallel with size 1
601
+ > setting random seeds to 1234 ...
602
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
603
+ [Gloo] Rank [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
604
+
605
+ [Gloo] Rank 0 is connected to 0[Gloo] Rank peer ranks. Expected number of connected peer ranks is : 00
606
+ is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
607
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
608
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
609
+ [rank4]:[W921 08:39:32.049181155 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.
610
+ [rank1]:[W921 08:39:32.049208181 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
+ [rank5]:[W921 08:39:32.049241144 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.
612
+ [rank2]:[W921 08:39:32.051174001 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
613
+ [rank7]:[W921 08:39:32.051228818 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
+ [rank6]:[W921 08:39:32.051296766 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.
615
+ [rank3]:[W921 08:39:32.051297460 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.
616
+ > compiling dataset index builder ...
617
+ make: Entering directory '/workspace/Megatron-UltraEP/megatron/core/datasets'
618
+ make: Nothing to be done for 'default'.
619
+ make: Leaving directory '/workspace/Megatron-UltraEP/megatron/core/datasets'
620
+ >>> done with dataset index builder. Compilation time: 0.036 seconds
621
+ WARNING: constraints for invoking optimized fused softmax kernel are not met. We default back to unfused kernel invocations.
622
+ > compiling and loading fused kernels ...
623
+ [rank0]:[W921 08:39:32.207693340 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.
624
+ >>> done with compiling and loading fused kernels. Compilation time: 3.854 seconds
625
+ time to initialize megatron (seconds): 7.221
626
+ [after megatron is initialized] datetime: 2026-09-21 08:39:38
627
+ building GPT model ...
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
+ /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.
643
+ warnings.warn(
644
+ [INFO] Use MNNVL fabric: False
645
+ [WARN] MAX_NUM_NVL_PEERS is not set. Using detected value 8.
646
+ NVSHMEM v3.7.2
647
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
648
+
649
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
650
+
651
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
652
+
653
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
654
+
655
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
656
+
657
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
658
+
659
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
660
+
661
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
662
+
663
+ UltraEP: NVL domain size <= 8; forcing weight_sync plan mode to direct (was adaptive).
664
+ > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 2087258112
665
+ 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)
666
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 7
667
+ Params for bucket 1 (103022592 elements, 103022592 padded size):
668
+ module.output_layer.weight
669
+ Params for bucket 2 (46542848 elements, 46542848 padded size):
670
+ module.decoder.final_layernorm.weight
671
+ module.decoder.layers.15.self_attention.linear_qkv.layer_norm_weight
672
+ module.decoder.layers.15.mlp.router.weight
673
+ module.decoder.layers.15.self_attention.linear_proj.weight
674
+ module.decoder.layers.14.mlp.router.weight
675
+ module.decoder.layers.14.self_attention.linear_qkv.layer_norm_weight
676
+ module.decoder.layers.14.self_attention.linear_proj.weight
677
+ module.decoder.layers.13.mlp.router.weight
678
+ module.decoder.layers.15.self_attention.linear_qkv.weight
679
+ module.decoder.layers.15.pre_mlp_layernorm.weight
680
+ module.decoder.layers.14.pre_mlp_layernorm.weight
681
+ module.decoder.layers.14.self_attention.linear_qkv.weight
682
+ module.decoder.layers.13.pre_mlp_layernorm.weight
683
+ module.decoder.layers.13.self_attention.linear_qkv.weight
684
+ Params for bucket 3 (50737152 elements, 50737152 padded size):
685
+ module.decoder.layers.13.self_attention.linear_qkv.layer_norm_weight
686
+ module.decoder.layers.13.self_attention.linear_proj.weight
687
+ module.decoder.layers.12.mlp.router.weight
688
+ module.decoder.layers.12.self_attention.linear_qkv.layer_norm_weight
689
+ module.decoder.layers.12.self_attention.linear_proj.weight
690
+ module.decoder.layers.11.mlp.router.weight
691
+ module.decoder.layers.11.self_attention.linear_qkv.layer_norm_weight
692
+ module.decoder.layers.11.self_attention.linear_proj.weight
693
+ module.decoder.layers.10.pre_mlp_layernorm.weight
694
+ module.decoder.layers.10.self_attention.linear_qkv.weight
695
+ module.decoder.layers.12.self_attention.linear_qkv.weight
696
+ module.decoder.layers.12.pre_mlp_layernorm.weight
697
+ module.decoder.layers.11.pre_mlp_layernorm.weight
698
+ module.decoder.layers.11.self_attention.linear_qkv.weight
699
+ module.decoder.layers.10.mlp.router.weight
700
+ Params for bucket 4 (50737152 elements, 50737152 padded size):
701
+ module.decoder.layers.9.self_attention.linear_qkv.weight
702
+ module.decoder.layers.9.pre_mlp_layernorm.weight
703
+ module.decoder.layers.8.pre_mlp_layernorm.weight
704
+ module.decoder.layers.8.self_attention.linear_qkv.weight
705
+ module.decoder.layers.7.pre_mlp_layernorm.weight
706
+ module.decoder.layers.7.self_attention.linear_qkv.weight
707
+ module.decoder.layers.10.self_attention.linear_qkv.layer_norm_weight
708
+ module.decoder.layers.10.self_attention.linear_proj.weight
709
+ module.decoder.layers.9.mlp.router.weight
710
+ module.decoder.layers.9.self_attention.linear_qkv.layer_norm_weight
711
+ module.decoder.layers.9.self_attention.linear_proj.weight
712
+ module.decoder.layers.8.mlp.router.weight
713
+ module.decoder.layers.8.self_attention.linear_qkv.layer_norm_weight
714
+ module.decoder.layers.8.self_attention.linear_proj.weight
715
+ module.decoder.layers.7.mlp.router.weight
716
+ Params for bucket 5 (50737152 elements, 50737152 padded size):
717
+ module.decoder.layers.6.mlp.router.weight
718
+ module.decoder.layers.6.self_attention.linear_qkv.layer_norm_weight
719
+ module.decoder.layers.6.self_attention.linear_proj.weight
720
+ module.decoder.layers.5.mlp.router.weight
721
+ module.decoder.layers.5.self_attention.linear_qkv.layer_norm_weight
722
+ module.decoder.layers.4.pre_mlp_layernorm.weight
723
+ module.decoder.layers.4.self_attention.linear_qkv.weight
724
+ module.decoder.layers.7.self_attention.linear_qkv.layer_norm_weight
725
+ module.decoder.layers.7.self_attention.linear_proj.weight
726
+ module.decoder.layers.6.pre_mlp_layernorm.weight
727
+ module.decoder.layers.6.self_attention.linear_qkv.weight
728
+ module.decoder.layers.5.pre_mlp_layernorm.weight
729
+ module.decoder.layers.5.self_attention.linear_qkv.weight
730
+ module.decoder.layers.5.self_attention.linear_proj.weight
731
+ module.decoder.layers.4.mlp.router.weight
732
+ Params for bucket 6 (50737152 elements, 50737152 padded size):
733
+ module.decoder.layers.3.pre_mlp_layernorm.weight
734
+ module.decoder.layers.3.self_attention.linear_qkv.weight
735
+ module.decoder.layers.2.mlp.router.weight
736
+ module.decoder.layers.2.self_attention.linear_qkv.layer_norm_weight
737
+ module.decoder.layers.2.self_attention.linear_proj.weight
738
+ module.decoder.layers.1.mlp.router.weight
739
+ module.decoder.layers.4.self_attention.linear_qkv.layer_norm_weight
740
+ module.decoder.layers.3.mlp.router.weight
741
+ module.decoder.layers.4.self_attention.linear_proj.weight
742
+ module.decoder.layers.3.self_attention.linear_qkv.layer_norm_weight
743
+ module.decoder.layers.3.self_attention.linear_proj.weight
744
+ module.decoder.layers.2.pre_mlp_layernorm.weight
745
+ module.decoder.layers.2.self_attention.linear_qkv.weight
746
+ module.decoder.layers.1.pre_mlp_layernorm.weight
747
+ module.decoder.layers.1.self_attention.linear_qkv.weight
748
+ Params for bucket 7 (124131328 elements, 124131328 padded size):
749
+ module.decoder.layers.1.self_attention.linear_qkv.layer_norm_weight
750
+ module.decoder.layers.0.pre_mlp_layernorm.weight
751
+ module.decoder.layers.1.self_attention.linear_proj.weight
752
+ module.decoder.layers.0.self_attention.linear_proj.weight
753
+ module.decoder.layers.0.self_attention.linear_qkv.weight
754
+ module.decoder.layers.0.mlp.router.weight
755
+ module.decoder.layers.0.self_attention.linear_qkv.layer_norm_weight
756
+ module.embedding.word_embeddings.weight
757
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 20
758
+ Params for bucket 1 (41943040 elements, 41943040 padded size):
759
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight7
760
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight3
761
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight5
762
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight1
763
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight6
764
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight4
765
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight2
766
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight4
767
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight6
768
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight2
769
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight0
770
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight7
771
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight5
772
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight3
773
+ Params for bucket 2 (41943040 elements, 41943040 padded size):
774
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight7
775
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight0
776
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight5
777
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight3
778
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight1
779
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight6
780
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight4
781
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight1
782
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight6
783
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight4
784
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight2
785
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight0
786
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight7
787
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight5
788
+ Params for bucket 3 (41943040 elements, 41943040 padded size):
789
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight2
790
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight7
791
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight0
792
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight5
793
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight3
794
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight1
795
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight6
796
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight3
797
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight1
798
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight6
799
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight4
800
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight2
801
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight0
802
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight7
803
+ Params for bucket 4 (41943040 elements, 41943040 padded size):
804
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight2
805
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight4
806
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight0
807
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight7
808
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight5
809
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight3
810
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight1
811
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight5
812
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight3
813
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight1
814
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight6
815
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight4
816
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight2
817
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight0
818
+ Params for bucket 5 (41943040 elements, 41943040 padded size):
819
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight4
820
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight6
821
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight2
822
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight0
823
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight7
824
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight5
825
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight7
826
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight3
827
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight5
828
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight1
829
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight6
830
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight4
831
+ Params for bucket 6 (41943040 elements, 41943040 padded size):
832
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight3
833
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight1
834
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight6
835
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight4
836
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight2
837
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight0
838
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight2
839
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight7
840
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight0
841
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight5
842
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight3
843
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight1
844
+ Params for bucket 7 (41943040 elements, 41943040 padded size):
845
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight6
846
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight4
847
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight2
848
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight0
849
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight7
850
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight5
851
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight3
852
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight7
853
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight3
854
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight5
855
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight1
856
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight6
857
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight4
858
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight2
859
+ Params for bucket 8 (41943040 elements, 41943040 padded size):
860
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight1
861
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight6
862
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight4
863
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight2
864
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight0
865
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight7
866
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight5
867
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight7
868
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight0
869
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight5
870
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight3
871
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight1
872
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight6
873
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight4
874
+ Params for bucket 9 (41943040 elements, 41943040 padded size):
875
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight3
876
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight1
877
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight6
878
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight4
879
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight2
880
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight0
881
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight7
882
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight7
883
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight0
884
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight2
885
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight5
886
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight3
887
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight1
888
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight6
889
+ Params for bucket 10 (41943040 elements, 41943040 padded size):
890
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight5
891
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight1
892
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight3
893
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight6
894
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight4
895
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight2
896
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight0
897
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight4
898
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight2
899
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight0
900
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight7
901
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight5
902
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight3
903
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight1
904
+ Params for bucket 11 (41943040 elements, 41943040 padded size):
905
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight7
906
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight3
907
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight5
908
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight1
909
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight6
910
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight4
911
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight6
912
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight4
913
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight2
914
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight0
915
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight7
916
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight5
917
+ Params for bucket 12 (41943040 elements, 41943040 padded size):
918
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight0
919
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight2
920
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight6
921
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight4
922
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight2
923
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight0
924
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight3
925
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight1
926
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight7
927
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight5
928
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight3
929
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight1
930
+ Params for bucket 13 (41943040 elements, 41943040 padded size):
931
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight7
932
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight3
933
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight5
934
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight1
935
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight7
936
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight5
937
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight3
938
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight4
939
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight6
940
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight2
941
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight0
942
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight6
943
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight4
944
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight2
945
+ Params for bucket 14 (41943040 elements, 41943040 padded size):
946
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight1
947
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight6
948
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight4
949
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight2
950
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight0
951
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight7
952
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight5
953
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight0
954
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight7
955
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight5
956
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight3
957
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight1
958
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight6
959
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight4
960
+ Params for bucket 15 (41943040 elements, 41943040 padded size):
961
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight3
962
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight7
963
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight1
964
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight5
965
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight3
966
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight1
967
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight7
968
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight0
969
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight2
970
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight6
971
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight4
972
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight2
973
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight0
974
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight6
975
+ Params for bucket 16 (41943040 elements, 41943040 padded size):
976
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight5
977
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight3
978
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight1
979
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight6
980
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight4
981
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight2
982
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight0
983
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight2
984
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight4
985
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight0
986
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight7
987
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight5
988
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight3
989
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight1
990
+ Params for bucket 17 (41943040 elements, 41943040 padded size):
991
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight7
992
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight3
993
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight5
994
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight1
995
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight7
996
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight5
997
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight4
998
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight6
999
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight2
1000
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight0
1001
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight6
1002
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight4
1003
+ Params for bucket 18 (41943040 elements, 41943040 padded size):
1004
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight3
1005
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight1
1006
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight6
1007
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight4
1008
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight2
1009
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight0
1010
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight7
1011
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight2
1012
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight0
1013
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight5
1014
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight3
1015
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight1
1016
+ Params for bucket 19 (41943040 elements, 41943040 padded size):
1017
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight7
1018
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight5
1019
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight3
1020
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight1
1021
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight6
1022
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight4
1023
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight3
1024
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight6
1025
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight4
1026
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight2
1027
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight0
1028
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight7
1029
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight5
1030
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight2
1031
+ Params for bucket 20 (8388608 elements, 8388608 padded size):
1032
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight1
1033
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight0
1034
+ 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 0x745ed3e6e4b0>, config_logger_dir='')
1035
+ INFO:megatron.core.optimizer_param_scheduler:> learning rate decay style: cosine
1036
+ [after model, optimizer, and learning rate scheduler are built] datetime: 2026-09-21 08:39:41
1037
+ > building train, validation, and test datasets ...
1038
+ > datasets target sizes (minimum size):
1039
+ train: 96
1040
+ validation: 0
1041
+ test: 0
1042
+ INFO:megatron.core.datasets.blended_megatron_dataset_config:Let split_matrix = [(0, 1.0), None, None]
1043
+ > building train, validation, and test datasets for GPT ...
1044
+ INFO:megatron.core.datasets.blended_megatron_dataset_builder:Building GPTDataset splits with sizes=(96, 0, 0) and config=GPTDatasetConfig(random_seed=1234, sequence_length=4096, blend=(['/workspace/data/wikitext/wikitext103_gpt2_text_document'], None), blend_per_split=None, split='100,0,0', split_matrix=[(0, 1.0), None, None], num_dataset_builder_threads=1, path_to_cache=None, mmap_bin_files=True, mock=False, tokenizer=<megatron.training.tokenizer.tokenizer._GPT2BPETokenizer object at 0x745ed6041580>, 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)
1045
+ INFO:megatron.core.datasets.indexed_dataset:Load the _IndexReader from /workspace/data/wikitext/wikitext103_gpt2_text_document.idx
1046
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence lengths
1047
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence pointers
1048
+ INFO:megatron.core.datasets.indexed_dataset: Extract the document indices
1049
+ INFO:megatron.core.datasets.indexed_dataset:> total number of sequences: 1165029
1050
+ INFO:megatron.core.datasets.indexed_dataset:> total number of documents: 1801350
1051
+ INFO:megatron.core.datasets.gpt_dataset:Load the GPTDataset train indices
1052
+ INFO:megatron.core.datasets.gpt_dataset: Load the document index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-document_index.npy
1053
+ INFO:megatron.core.datasets.gpt_dataset: Load the sample index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-sample_index.npy
1054
+ INFO:megatron.core.datasets.gpt_dataset: Load the shuffle index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-shuffle_index.npy
1055
+ INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 29073
1056
+ > finished creating GPT datasets ...
1057
+ [after dataloaders are built] datetime: 2026-09-21 08:39:41
1058
+ done with setup ...
1059
+ training ...
1060
+ (min, max) time across ranks (ms):
1061
+ model-and-optimizer-setup ......................: (3626.64, 3629.12)
1062
+ train/valid/test-data-iterators-setup ..........: (73.22, 92.67)
1063
+ Overwriting rerun_state_machine.current_iteration from -1 to 0...
1064
+ [before the start of training step] datetime: 2026-09-21 08:39:41
1065
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1066
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1067
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1068
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1069
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1070
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1071
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1072
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1073
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1074
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1075
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1076
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1077
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1078
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1079
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1080
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1081
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1082
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1083
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1084
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1085
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1086
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1087
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1088
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1089
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1090
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1091
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1092
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1093
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1094
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1095
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1096
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1097
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1098
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1099
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1100
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1101
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1102
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1103
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1104
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1105
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1106
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1107
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1108
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1109
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1110
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1111
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1112
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1113
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1114
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1115
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1116
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1117
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1118
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1119
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1120
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1121
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1122
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1123
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1124
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1125
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1126
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1127
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1128
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1129
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1130
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1131
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1132
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1133
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1134
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1135
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1136
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1137
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1138
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1139
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1140
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1141
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1142
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1143
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1144
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1145
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1146
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1147
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1148
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1149
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1150
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1151
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1152
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1153
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1154
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1155
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1156
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1157
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1158
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1159
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1160
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1161
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1162
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1163
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1164
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1165
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1166
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1167
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1168
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1169
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1170
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1171
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1172
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1173
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1174
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1175
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1176
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1177
+ Number of parameters in transformer block in billions: 6.71
1178
+ [2026-09-21 08:40:07] iteration 1/ 12 | consumed samples: 8 | elapsed time per iteration (ms): 25307.7 | throughput per GPU (TFLOP/s/GPU): 1.3 | learning rate: 2.000000E-05 | global batch size: 8 | lm loss: 1.118507E+01 | loss scale: 1.0 | grad norm: 35.484 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 33920 | min safe capacity: 1.0352 |Number of parameters in embedding layers in billions: 0.21
1179
+
1180
+ Total number of parameters in billions: 6.92
1181
+ Number of parameters in most loaded shard in billions: 6.9171
1182
+ Theoretical memory footprints: weight and optimizer=49474.72 MB
1183
+ [Rank 0] (after 1 iterations) memory (MB) | allocated: 18080.58447265625 | max allocated: 21748.86669921875 | reserved: 22168.0 | max reserved: 22168.0
1184
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1185
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1186
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1187
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1188
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1189
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1190
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1191
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1192
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1193
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1194
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1195
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1196
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1197
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1198
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1199
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1200
+ [2026-09-21 08:40:07] iteration 2/ 12 | consumed samples: 16 | elapsed time per iteration (ms): 273.7 | throughput per GPU (TFLOP/s/GPU): 117.7 | learning rate: 4.000000E-05 | global batch size: 8 | lm loss: 1.120943E+01 | loss scale: 1.0 | grad norm: 36.299 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 33920 | min safe capacity: 1.0352 |
1201
+ [2026-09-21 08:40:07] iteration 3/ 12 | consumed samples: 24 | elapsed time per iteration (ms): 506.8 | throughput per GPU (TFLOP/s/GPU): 63.6 | learning rate: 6.000000E-05 | global batch size: 8 | lm loss: 1.006762E+01 | loss scale: 1.0 | grad norm: 101.683 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 33920 | min safe capacity: 1.0352 |
1202
+ Capture CUDA graph for training!!!
1203
+ CUDA graph capture done!!!
1204
+ [2026-09-21 08:40:08] iteration 4/ 12 | consumed samples: 32 | elapsed time per iteration (ms): 819.0 | throughput per GPU (TFLOP/s/GPU): 39.3 | learning rate: 8.000000E-05 | global batch size: 8 | lm loss: 1.023827E+01 | loss scale: 1.0 | grad norm: 28.873 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 33920 | min safe capacity: 1.0352 |
1205
+ [2026-09-21 08:40:08] iteration 5/ 12 | consumed samples: 40 | elapsed time per iteration (ms): 104.4 | throughput per GPU (TFLOP/s/GPU): 308.6 | learning rate: 1.000000E-04 | global batch size: 8 | lm loss: 1.073976E+01 | loss scale: 1.0 | grad norm: 20.839 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 33920 | min safe capacity: 1.0352 |
1206
+ [2026-09-21 08:40:09] iteration 6/ 12 | consumed samples: 48 | elapsed time per iteration (ms): 104.3 | throughput per GPU (TFLOP/s/GPU): 309.0 | learning rate: 7.750000E-05 | global batch size: 8 | lm loss: 9.629029E+00 | loss scale: 1.0 | grad norm: 8.181 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 34048 | min safe capacity: 1.0391 |
1207
+ [2026-09-21 08:40:09] iteration 7/ 12 | consumed samples: 56 | elapsed time per iteration (ms): 103.1 | throughput per GPU (TFLOP/s/GPU): 312.6 | learning rate: 3.250000E-05 | global batch size: 8 | lm loss: 9.080976E+00 | loss scale: 1.0 | grad norm: 16.476 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 34048 | min safe capacity: 1.0391 |
1208
+ [2026-09-21 08:40:09] iteration 8/ 12 | consumed samples: 64 | elapsed time per iteration (ms): 103.6 | throughput per GPU (TFLOP/s/GPU): 310.9 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.817616E+00 | loss scale: 1.0 | grad norm: 8.584 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 34048 | min safe capacity: 1.0391 |
1209
+ [2026-09-21 08:40:09] iteration 9/ 12 | consumed samples: 72 | elapsed time per iteration (ms): 104.3 | throughput per GPU (TFLOP/s/GPU): 308.9 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.710125E+00 | loss scale: 1.0 | grad norm: 3.788 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 34048 | min safe capacity: 1.0391 |
1210
+ [2026-09-21 08:40:09] iteration 10/ 12 | consumed samples: 80 | elapsed time per iteration (ms): 103.8 | throughput per GPU (TFLOP/s/GPU): 310.4 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.653005E+00 | loss scale: 1.0 | grad norm: 3.497 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 34048 | min safe capacity: 1.0391 |
1211
+ [2026-09-21 08:40:09] iteration 11/ 12 | consumed samples: 88 | elapsed time per iteration (ms): 103.5 | throughput per GPU (TFLOP/s/GPU): 311.4 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.641299E+00 | loss scale: 1.0 | grad norm: 3.470 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 34048 | min safe capacity: 1.0391 |
1212
+ [2026-09-21 08:40:09] iteration 12/ 12 | consumed samples: 96 | elapsed time per iteration (ms): 102.3 | throughput per GPU (TFLOP/s/GPU): 315.0 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.590896E+00 | loss scale: 1.0 | grad norm: 3.165 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 34048 | min safe capacity: 1.0391 |
1213
+ [after training is done] datetime: 2026-09-21 08:40:09
1214
+ [rank0]:[W921 08:40:10.686426019 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())
1215
+ [rank3]:[W921 08:40:12.321458949 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())
1216
+ [rank5]:[W921 08:40:12.323678517 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())
1217
+ [rank1]:[W921 08:40:12.328984691 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())
1218
+ [rank7]:[W921 08:40:12.362367926 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())
1219
+ [rank6]:[W921 08:40:12.425163521 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())
1220
+ [rank4]:[W921 08:40:12.873551101 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())
1221
+ [rank2]:[W921 08:40:12.890572825 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())
1222
+ W0921 08:40:14.338000 706795 torch/distributed/elastic/agent/server/api.py:719] Received 15 death signal, shutting down workers
1223
+ W0921 08:40:14.339000 706795 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 706883 closing signal SIGTERM
1224
+ W0921 08:40:14.340000 706795 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 706884 closing signal SIGTERM
1225
+ W0921 08:40:14.341000 706795 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 706885 closing signal SIGTERM
1226
+ W0921 08:40:14.341000 706795 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 706886 closing signal SIGTERM
1227
+ W0921 08:40:14.341000 706795 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 706887 closing signal SIGTERM
1228
+ W0921 08:40:14.342000 706795 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 706888 closing signal SIGTERM
1229
+ W0921 08:40:14.343000 706795 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 706889 closing signal SIGTERM
1230
+ W0921 08:40:14.343000 706795 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 706890 closing signal SIGTERM
1231
+ Traceback (most recent call last):
1232
+ File "/usr/local/bin/torchrun", line 8, in <module>
1233
+ sys.exit(main())
1234
+ ^^^^^^
1235
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper
1236
+ return f(*args, **kwargs)
1237
+ ^^^^^^^^^^^^^^^^^^
1238
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 892, in main
1239
+ run(args)
1240
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 883, in run
1241
+ elastic_launch(
1242
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 139, in __call__
1243
+ return launch_agent(self._config, self._entrypoint, list(args))
1244
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1245
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 261, in launch_agent
1246
+ result = agent.run()
1247
+ ^^^^^^^^^^^
1248
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/metrics/api.py", line 138, in wrapper
1249
+ result = f(*args, **kwargs)
1250
+ ^^^^^^^^^^^^^^^^^^
1251
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 711, in run
1252
+ result = self._invoke_run(role)
1253
+ ^^^^^^^^^^^^^^^^^^^^^^
1254
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 870, in _invoke_run
1255
+ time.sleep(monitor_interval)
1256
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/api.py", line 84, in _terminate_process_handler
1257
+ raise SignalException(f"Process {os.getpid()} got signal: {sigval}", sigval=sigval)
1258
+ torch.distributed.elastic.multiprocessing.api.SignalException: Process 706795 got signal: 15
granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/results.tsv ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ config backend policy slots seq gbs cap recompute rep status iter_ms_mean iter_ms_min peak_mib run_dir
2
+ OLMOE_1B_7B_L16 ultraep matched 8 4096 8 1.08984375 none 1 OK 103.37 102.00 39097 /workspace/granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep1
3
+ OLMOE_1B_7B_L16 ultraep matched 8 4096 8 1.08984375 none 2 OK 103.55 102.00 39097 /workspace/granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep2
4
+ OLMOE_1B_7B_L16 ultraep matched 8 4096 8 1.08984375 none 3 OK 103.55 102.00 39097 /workspace/granularity_next_b200/tierb_models/attempts/OLMOE_1B_7B_L16/ultraep/tierb/tierb_gbs8/OLMOE_1B_7B_L16__ultraep__matched__seq4096__gbs8__cap1p08984375/rep3