AbhishekGhosh1998 commited on
Commit
7ff5ea2
·
verified ·
1 Parent(s): a644cf3

granularity results granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb @ 2026-09-21T07:47:10Z 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/KIMI_K2_L2/echo/tierb/manifest.txt +19 -0
  2. granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/pip_freeze.txt +317 -0
  3. granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/provenance.txt +90 -0
  4. granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep1/mem_samples.csv +240 -0
  5. granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep1/run.env +102 -0
  6. granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep1/train.log +1046 -0
  7. granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep2/mem_samples.csv +232 -0
  8. granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep2/run.env +102 -0
  9. granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep2/train.log +1046 -0
  10. granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep3/mem_samples.csv +240 -0
  11. granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep3/run.env +102 -0
  12. granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep3/train.log +1047 -0
  13. granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/tierb_gbs8/results.tsv +4 -0
granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/manifest.txt ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ timestamp: 2026-09-21T07:43:23+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-echo
6
+ study_scripts_commit: b8bd100
7
+ campaign: echo
8
+ backend: echo
9
+ phase: tierb
10
+ gbs: 8
11
+ seq: 4096
12
+ configs: G1 G2 G3 G4
13
+ capacity: 1.25
14
+ repeats: 3
15
+ nsys: 1
16
+ megatron_repo: /workspace/Megatron-MoE-Echo
17
+ megatron_branch_required: bracing-regions-through-kernels
18
+ megatron_branch_actual: bracing-regions-through-kernels
19
+ megatron_commit: 9a236793
granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/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/KIMI_K2_L2/echo/tierb/provenance.txt ADDED
@@ -0,0 +1,90 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ==============================================================
2
+ PROVENANCE: echo__tierb__gbs8
3
+ ==============================================================
4
+ captured_at : 2026-09-21T07:43:09+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/KIMI_K2_L2/echo/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, 35
23
+ 3, 150 MHz, 1965 MHz, 1000.00 W, 30
24
+ 4, 1965 MHz, 1965 MHz, 1000.00 W, 30
25
+ 5, 120 MHz, 1965 MHz, 1000.00 W, 34
26
+ 6, 120 MHz, 1965 MHz, 1000.00 W, 34
27
+ 7, 157 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-token-drop-analysis eaf829e7ad21
64
+ UltraEP main 94cab099b44f
65
+ DeepEP HEAD 34152ae28f80 (+3 untracked: build output)
66
+ MoonEP megatron-ddp-symmetric-params d8bc6667fad7
67
+ blackwell_run_scripts granularity-echo b8bd10097bb5
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/KIMI_K2_L2
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/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep1/mem_samples.csv ADDED
@@ -0,0 +1,240 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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, 1
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, 165
50
+ 1, 725
51
+ 2, 725
52
+ 3, 725
53
+ 4, 725
54
+ 5, 725
55
+ 6, 725
56
+ 7, 725
57
+ 0, 915
58
+ 1, 921
59
+ 2, 925
60
+ 3, 927
61
+ 4, 925
62
+ 5, 933
63
+ 6, 915
64
+ 7, 937
65
+ 0, 1927
66
+ 1, 1927
67
+ 2, 1927
68
+ 3, 1927
69
+ 4, 1927
70
+ 5, 1927
71
+ 6, 1927
72
+ 7, 1927
73
+ 0, 36271
74
+ 1, 36251
75
+ 2, 36289
76
+ 3, 36265
77
+ 4, 36315
78
+ 5, 36265
79
+ 6, 36279
80
+ 7, 36267
81
+ 0, 45201
82
+ 1, 45207
83
+ 2, 45203
84
+ 3, 45187
85
+ 4, 45205
86
+ 5, 45209
87
+ 6, 45203
88
+ 7, 45203
89
+ 0, 73339
90
+ 1, 73339
91
+ 2, 73339
92
+ 3, 73339
93
+ 4, 73339
94
+ 5, 73339
95
+ 6, 73339
96
+ 7, 73339
97
+ 0, 76315
98
+ 1, 76315
99
+ 2, 76315
100
+ 3, 76315
101
+ 4, 76315
102
+ 5, 76315
103
+ 6, 76315
104
+ 7, 76315
105
+ 0, 76315
106
+ 1, 76315
107
+ 2, 76315
108
+ 3, 76315
109
+ 4, 76315
110
+ 5, 77659
111
+ 6, 76315
112
+ 7, 77659
113
+ 0, 78881
114
+ 1, 78881
115
+ 2, 78881
116
+ 3, 78881
117
+ 4, 78881
118
+ 5, 78881
119
+ 6, 78881
120
+ 7, 78881
121
+ 0, 78881
122
+ 1, 78881
123
+ 2, 78881
124
+ 3, 78881
125
+ 4, 78881
126
+ 5, 78881
127
+ 6, 78881
128
+ 7, 78881
129
+ 0, 78881
130
+ 1, 78881
131
+ 2, 78881
132
+ 3, 78881
133
+ 4, 78881
134
+ 5, 78881
135
+ 6, 78881
136
+ 7, 78881
137
+ 0, 81891
138
+ 1, 81893
139
+ 2, 81891
140
+ 3, 81891
141
+ 4, 81891
142
+ 5, 81891
143
+ 6, 81905
144
+ 7, 81891
145
+ 0, 81897
146
+ 1, 81899
147
+ 2, 81897
148
+ 3, 81897
149
+ 4, 81897
150
+ 5, 82747
151
+ 6, 81911
152
+ 7, 82747
153
+ 0, 83395
154
+ 1, 83397
155
+ 2, 83395
156
+ 3, 83395
157
+ 4, 83395
158
+ 5, 83395
159
+ 6, 83409
160
+ 7, 83395
161
+ 0, 83397
162
+ 1, 83397
163
+ 2, 83397
164
+ 3, 83397
165
+ 4, 83397
166
+ 5, 83397
167
+ 6, 83411
168
+ 7, 83397
169
+ 0, 91415
170
+ 1, 91415
171
+ 2, 91415
172
+ 3, 91415
173
+ 4, 91415
174
+ 5, 91303
175
+ 6, 91317
176
+ 7, 91415
177
+ 0, 92759
178
+ 1, 92759
179
+ 2, 92759
180
+ 3, 92759
181
+ 4, 92759
182
+ 5, 92647
183
+ 6, 92661
184
+ 7, 92759
185
+ 0, 92759
186
+ 1, 92759
187
+ 2, 92759
188
+ 3, 92759
189
+ 4, 92759
190
+ 5, 92647
191
+ 6, 92661
192
+ 7, 92759
193
+ 0, 104805
194
+ 1, 104815
195
+ 2, 104837
196
+ 3, 104863
197
+ 4, 104807
198
+ 5, 104705
199
+ 6, 104753
200
+ 7, 104825
201
+ 0, 105447
202
+ 1, 105441
203
+ 2, 105469
204
+ 3, 105467
205
+ 4, 105451
206
+ 5, 105369
207
+ 6, 105353
208
+ 7, 105465
209
+ 0, 106177
210
+ 1, 106235
211
+ 2, 106187
212
+ 3, 106217
213
+ 4, 106229
214
+ 5, 106095
215
+ 6, 106111
216
+ 7, 106283
217
+ 0, 122659
218
+ 1, 123571
219
+ 2, 122281
220
+ 3, 122771
221
+ 4, 122477
222
+ 5, 122631
223
+ 6, 122603
224
+ 7, 122617
225
+ 0, 132177
226
+ 1, 132179
227
+ 2, 132163
228
+ 3, 132219
229
+ 4, 132219
230
+ 5, 132203
231
+ 6, 131991
232
+ 7, 132219
233
+ 0, 25302
234
+ 1, 25686
235
+ 2, 132163
236
+ 3, 27222
237
+ 4, 25302
238
+ 5, 25302
239
+ 6, 25302
240
+ 7, 25302
granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep1/run.env ADDED
@@ -0,0 +1,102 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ backend=echo
2
+ config=KIMI_K2_L2
3
+ family=tierB
4
+ source_model=Kimi-K2
5
+ standard_num_layers=61
6
+ base_rung=
7
+ repo=/workspace/Megatron-MoE-Echo
8
+ branch=bracing-regions-through-kernels
9
+ commit=9a236793
10
+ num_layers=2
11
+ hidden=7168
12
+ ffn_hidden=18432
13
+ moe_ffn_hidden=2048
14
+ num_experts=384
15
+ topk=8
16
+ experts_per_rank=48
17
+ expert_params=44040192
18
+ heads=64
19
+ kv_groups=64
20
+ kv_channels=64
21
+ qk_layernorm=1
22
+ shared_ffn=2048
23
+ shared_expert_overlap=0
24
+ moe_token_dispatcher=flex
25
+ mla=1
26
+ q_lora_rank=1536
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=1
32
+ moe_layers=1
33
+ rotary_base=50000
34
+ rotary_percent=1.0
35
+ rotary_scaling=32
36
+ mscale=1.0
37
+ mscale_all_dim=1.0
38
+ qkv_bias=0
39
+ router_balancing=seq_aux_loss
40
+ aux_loss_coeff=0
41
+ router_pre_softmax=0
42
+ router_fusion=0
43
+ router_dtype=fp32
44
+ router_groups=1
45
+ router_group_topk=1
46
+ router_score_function=sigmoid
47
+ router_expert_bias=1
48
+ router_bias_update_rate=1e-3
49
+ router_topk_scaling_factor=2.827
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-6
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=682
73
+ slot_policy=matched
74
+ slots_per_rank=48
75
+ capacity_mode=static
76
+ capacity=1.17578125
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=
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/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep1/train.log ADDED
@@ -0,0 +1,1046 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [07:43:23] WARN: bracing-regions-through-kernels lacks --moe-router-fusion; KIMI_K2_L2 runs unfused (timing only, routing identical)
2
+ ==============================================================
3
+ KIMI_K2_L2 / echo (slots/rank=48 policy=matched)
4
+ E=384 K=8 H'=2048 m=48
5
+ L=2 seq=4096 gbs=8 (1 microbatches)
6
+ cap_mode=static cap=1.17578125 cg=1 recompute=none
7
+ memory_dump=none snapshots=/workspace/granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep1/memory_snapshots/snapshot.pickle
8
+ repo=/workspace/Megatron-MoE-Echo @ 9a236793
9
+ ==============================================================
10
+ W0921 07:43:25.195000 501453 torch/distributed/run.py:766]
11
+ W0921 07:43:25.195000 501453 torch/distributed/run.py:766] *****************************************
12
+ W0921 07:43:25.195000 501453 torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.
13
+ W0921 07:43:25.195000 501453 torch/distributed/run.py:766] *****************************************
14
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
15
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
16
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
17
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
18
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
19
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
20
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
21
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
22
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
23
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
24
+ 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
25
+ using world size: 8, data-parallel size: 8, context-parallel size: 1, hierarchical context-parallel sizes: None, tensor-model-parallel size: 1, pipeline-model-parallel size: 1
26
+ WARNING: overriding default arguments for tokenizer_type:GPT2BPETokenizer with tokenizer_type:GPT2BPETokenizer
27
+ Number of virtual stages per pipeline stage: None
28
+ accumulate and all-reduce gradients in fp32 for bfloat16 data type.
29
+ using torch.bfloat16 for parameters ...
30
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
31
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
32
+ ------------------------ arguments ------------------------
33
+ account_for_embedding_in_pipeline_split ......... False
34
+ account_for_loss_in_pipeline_split .............. False
35
+ accumulate_allreduce_grads_in_fp32 .............. True
36
+ adam_beta1 ...................................... 0.9
37
+ adam_beta2 ...................................... 0.999
38
+ adam_eps ........................................ 1e-08
39
+ add_bias_linear ................................. False
40
+ add_position_embedding .......................... False
41
+ add_qkv_bias .................................... False
42
+ adlr_autoresume ................................. False
43
+ adlr_autoresume_interval ........................ 1000
44
+ align_grad_reduce ............................... True
45
+ align_param_gather .............................. False
46
+ app_tag_run_name ................................ None
47
+ app_tag_run_version ............................. 0.0.0
48
+ apply_layernorm_1p .............................. False
49
+ apply_query_key_layer_scaling ................... False
50
+ apply_residual_connection_post_layernorm ........ False
51
+ apply_rope_fusion ............................... False
52
+ async_save ...................................... None
53
+ async_tensor_model_parallel_allreduce ........... True
54
+ attention_backend ............................... AttnBackend.auto
55
+ attention_dropout ............................... 0.0
56
+ attention_softmax_in_fp32 ....................... False
57
+ auto_detect_ckpt_format ......................... False
58
+ barrier_with_L1_time ............................ True
59
+ bert_binary_head ................................ True
60
+ bert_embedder_type .............................. megatron
61
+ bert_load ....................................... None
62
+ bf16 ............................................ True
63
+ bias_dropout_fusion ............................. True
64
+ bias_gelu_fusion ................................ False
65
+ bias_swiglu_fusion .............................. True
66
+ biencoder_projection_dim ........................ 0
67
+ biencoder_shared_query_context_model ............ False
68
+ block_data_path ................................. None
69
+ calc_ft_timeouts ................................ False
70
+ calculate_per_token_loss ........................ False
71
+ check_for_large_grads ........................... False
72
+ check_for_nan_in_loss_and_grad .................. False
73
+ check_for_spiky_loss ............................ False
74
+ check_weight_hash_across_dp_replicas_interval ... None
75
+ ckpt_assume_constant_structure .................. False
76
+ ckpt_convert_format ............................. None
77
+ ckpt_convert_save ............................... None
78
+ ckpt_convert_update_legacy_dist_opt_format ...... False
79
+ ckpt_format ..................................... torch_dist
80
+ ckpt_fully_parallel_load ........................ False
81
+ ckpt_fully_parallel_save ........................ True
82
+ ckpt_fully_parallel_save_deprecated ............. False
83
+ ckpt_step ....................................... None
84
+ classes_fraction ................................ 1.0
85
+ clip_grad ....................................... 1.0
86
+ clone_scatter_output_in_embedding ............... True
87
+ config_logger_dir ...............................
88
+ consumed_train_samples .......................... 0
89
+ consumed_valid_samples .......................... 0
90
+ context_parallel_size ........................... 1
91
+ cp_comm_type .................................... ['p2p']
92
+ create_attention_mask_in_dataloader ............. True
93
+ cross_entropy_fusion_impl ....................... native
94
+ cross_entropy_loss_fusion ....................... False
95
+ cuda_graph_scope ................................ full_iteration
96
+ cuda_graph_warmup_steps ......................... 3
97
+ data_args_path .................................. None
98
+ data_cache_path ................................. None
99
+ data_parallel_random_init ....................... False
100
+ data_parallel_sharding_strategy ................. no_shard
101
+ data_parallel_size .............................. 8
102
+ data_path ....................................... ['/workspace/data/wikitext/wikitext103_gpt2_text_document']
103
+ data_per_class_fraction ......................... 1.0
104
+ data_sharding ................................... True
105
+ dataloader_type ................................. single
106
+ ddp_average_in_collective ....................... False
107
+ ddp_bucket_size ................................. None
108
+ ddp_num_buckets ................................. None
109
+ ddp_pad_buckets_for_high_nccl_busbw ............. False
110
+ decoder_first_pipeline_num_layers ............... None
111
+ decoder_last_pipeline_num_layers ................ None
112
+ decoder_num_layers .............................. None
113
+ decoder_seq_length .............................. None
114
+ decoupled_lr .................................... None
115
+ decoupled_min_lr ................................ None
116
+ decrease_batch_size_if_needed ................... False
117
+ defer_embedding_wgrad_compute ................... False
118
+ delay_wgrad_compute ............................. False
119
+ deprecated_use_mcore_models ..................... True
120
+ deterministic_mode .............................. False
121
+ dino_bottleneck_size ............................ 256
122
+ dino_freeze_last_layer .......................... 1
123
+ dino_head_hidden_size ........................... 2048
124
+ dino_local_crops_number ......................... 10
125
+ dino_local_img_size ............................. 96
126
+ dino_norm_last_layer ............................ False
127
+ dino_teacher_temp ............................... 0.07
128
+ dino_warmup_teacher_temp ........................ 0.04
129
+ dino_warmup_teacher_temp_epochs ................. 30
130
+ disable_bf16_reduced_precision_matmul ........... False
131
+ disable_mamba_mem_eff_path ...................... False
132
+ disable_straggler_on_startup .................... False
133
+ dist_ckpt_format_deprecated ..................... None
134
+ dist_ckpt_strictness ............................ assume_ok_unexpected
135
+ distribute_saved_activations .................... False
136
+ distributed_backend ............................. nccl
137
+ distributed_timeout_minutes ..................... 10
138
+ embedding_init_method_std ....................... None
139
+ embedding_path .................................. None
140
+ empty_unused_memory_level ....................... 0
141
+ enable_cuda_graph ............................... True
142
+ enable_experimental ............................. False
143
+ enable_ft_package ............................... False
144
+ enable_gloo_process_groups ...................... True
145
+ enable_msc ...................................... True
146
+ enable_one_logger ............................... True
147
+ encoder_num_layers .............................. 2
148
+ encoder_seq_length .............................. 4096
149
+ end_weight_decay ................................ 0.1
150
+ eod_mask_loss ................................... False
151
+ error_injection_rate ............................ 0
152
+ error_injection_type ............................ transient_error
153
+ eval_interval ................................... 1000000
154
+ eval_iters ...................................... 0
155
+ evidence_data_path .............................. None
156
+ exit_duration_in_mins ........................... None
157
+ exit_interval ................................... None
158
+ exit_on_missing_checkpoint ...................... False
159
+ exit_signal_handler ............................. False
160
+ exp_avg_dtype ................................... torch.float32
161
+ exp_avg_sq_dtype ................................ torch.float32
162
+ expert_model_parallel_size ...................... 8
163
+ expert_tensor_parallel_size ..................... 1
164
+ export_force_local_attention .................... False
165
+ export_kd_cfg ................................... None
166
+ export_kd_teacher_ckpt_format ................... None
167
+ export_kd_teacher_load .......................... None
168
+ export_kv_cache_quant ........................... False
169
+ export_legacy_megatron .......................... False
170
+ export_model_type ............................... GPTModel
171
+ export_moe_apply_probs_on_input ................. False
172
+ export_qk_l2_norm ............................... False
173
+ export_quant_cfg ................................ None
174
+ export_real_quant_cfg ........................... None
175
+ export_te_mcore_model ........................... False
176
+ external_cuda_graph ............................. False
177
+ ffn_hidden_size ................................. 18432
178
+ finetune ........................................ False
179
+ finetune_data_split ............................. train
180
+ finetune_hf_dataset ............................. None
181
+ first_last_layers_bf16 .......................... False
182
+ flash_decode .................................... False
183
+ fp16 ............................................ False
184
+ fp16_lm_cross_entropy ........................... False
185
+ fp32_residual_connection ........................ False
186
+ fp8 ............................................. e4m3
187
+ fp8_amax_compute_algo ........................... most_recent
188
+ fp8_amax_history_len ............................ 1
189
+ fp8_interval .................................... 1
190
+ fp8_margin ...................................... 0
191
+ fp8_param_gather ................................ False
192
+ fp8_recipe ...................................... mxfp8
193
+ fp8_wgrad ....................................... True
194
+ fsdp_double_buffer .............................. False
195
+ global_batch_size ............................... 8
196
+ grad_reduce_in_bf16 ............................. False
197
+ gradient_accumulation_fusion .................... True
198
+ gradient_reduce_div_fusion ...................... True
199
+ group_query_attention ........................... False
200
+ head_lr_mult .................................... 1.0
201
+ heterogeneous_layers_config_encoded_json ........ None
202
+ heterogeneous_layers_config_path ................ None
203
+ hidden_dropout .................................. 0.0
204
+ hidden_size ..................................... 7168
205
+ hierarchical_context_parallel_sizes ............. None
206
+ high_priority_stream_groups ..................... []
207
+ hybrid_attention_ratio .......................... 0.0
208
+ hybrid_mlp_ratio ................................ 0.0
209
+ hybrid_override_pattern ......................... None
210
+ hysteresis ...................................... 2
211
+ ict_head_size ................................... None
212
+ ict_load ........................................ None
213
+ img_h ........................................... 224
214
+ img_w ........................................... 224
215
+ indexer_batch_size .............................. 128
216
+ indexer_log_interval ............................ 1000
217
+ inference_batch_times_seqlen_threshold .......... -1
218
+ inference_dynamic_batching ...................... False
219
+ inference_dynamic_batching_buffer_guaranteed_fraction 0.2
220
+ inference_dynamic_batching_buffer_overflow_factor None
221
+ inference_dynamic_batching_buffer_size_gb ....... 40.0
222
+ inference_dynamic_batching_chunk_size ........... 256
223
+ inference_dynamic_batching_max_requests_override None
224
+ inference_dynamic_batching_max_tokens_override .. None
225
+ inference_dynamic_batching_num_cuda_graphs ...... 16
226
+ inference_max_batch_size ........................ 8
227
+ inference_max_seq_length ........................ 2560
228
+ inference_rng_tracker ........................... False
229
+ init_method_std ................................. 0.02
230
+ init_method_xavier_uniform ...................... False
231
+ init_model_with_meta_device ..................... False
232
+ initial_loss_scale .............................. 4294967296
233
+ inprocess_active_world_size ..................... 8
234
+ inprocess_barrier_timeout ....................... 120
235
+ inprocess_completion_timeout .................... 120
236
+ inprocess_empty_cuda_cache ...................... False
237
+ inprocess_granularity ........................... node
238
+ inprocess_hard_timeout .......................... 90
239
+ inprocess_heartbeat_interval .................... 30
240
+ inprocess_heartbeat_timeout ..................... 60
241
+ inprocess_last_call_wait ........................ 1
242
+ inprocess_max_iterations ........................ None
243
+ inprocess_monitor_process_interval .............. 1.0
244
+ inprocess_monitor_thread_interval ............... 1.0
245
+ inprocess_progress_watchdog_interval ............ 1.0
246
+ inprocess_restart ............................... False
247
+ inprocess_soft_timeout .......................... 60
248
+ inprocess_termination_grace_time ................ 1
249
+ is_hybrid_model ................................. False
250
+ iter_per_epoch .................................. 1250
251
+ iterations_to_skip .............................. []
252
+ keep_fp8_transpose_cache_when_using_custom_fsdp . False
253
+ kitchen_config_file ............................. None
254
+ kitchen_recipe_number ........................... None
255
+ kv_channels ..................................... 64
256
+ kv_lora_rank .................................... 512
257
+ lazy_mpu_init ................................... None
258
+ load ............................................ None
259
+ load_main_params_from_ckpt ...................... None
260
+ load_model_opt_format ........................... False
261
+ local_rank ...................................... 0
262
+ log_energy ...................................... False
263
+ log_interval .................................... 1
264
+ log_loss_scale_to_tensorboard ................... True
265
+ log_memory_to_tensorboard ....................... False
266
+ log_num_zeros_in_grad ........................... False
267
+ log_params_norm ................................. False
268
+ log_progress .................................... False
269
+ log_straggler ................................... False
270
+ log_throughput .................................. True
271
+ log_timers_to_tensorboard ....................... False
272
+ log_validation_ppl_to_tensorboard ............... False
273
+ log_world_size_to_tensorboard ................... False
274
+ logging_level ................................... None
275
+ loss_scale ...................................... None
276
+ loss_scale_window ............................... 1000
277
+ lr .............................................. 0.0001
278
+ lr_decay_iters .................................. 8
279
+ lr_decay_samples ................................ None
280
+ lr_decay_style .................................. cosine
281
+ lr_warmup_fraction .............................. None
282
+ lr_warmup_init .................................. 0.0
283
+ lr_warmup_iters ................................. 5
284
+ lr_warmup_samples ............................... 0
285
+ lr_wsd_decay_iters .............................. None
286
+ lr_wsd_decay_samples ............................ None
287
+ lr_wsd_decay_style .............................. exponential
288
+ main_grads_dtype ................................ torch.float32
289
+ main_params_dtype ............................... torch.float32
290
+ make_vocab_size_divisible_by .................... 128
291
+ mamba_head_dim .................................. 64
292
+ mamba_num_groups ................................ 8
293
+ mamba_num_heads ................................. None
294
+ mamba_state_dim ................................. 128
295
+ manual_gc ....................................... False
296
+ manual_gc_eval .................................. True
297
+ manual_gc_interval .............................. 0
298
+ mask_factor ..................................... 1.0
299
+ mask_prob ....................................... 0.15
300
+ mask_type ....................................... random
301
+ masked_softmax_fusion ........................... False
302
+ max_position_embeddings ......................... 4096
303
+ max_tokens_to_oom ............................... 12000
304
+ memory_snapshot_path ............................ snapshot.pickle
305
+ merge_file ...................................... /workspace/data/wikitext/gpt2-merges.txt
306
+ micro_batch_size ................................ 1
307
+ microbatch_group_size_per_vp_stage .............. None
308
+ mid_level_dataset_surplus ....................... 0.005
309
+ min_loss_scale .................................. 1.0
310
+ min_lr .......................................... 1e-05
311
+ mlp_chunks_for_prefill .......................... 1
312
+ mmap_bin_files .................................. True
313
+ mock_data ....................................... False
314
+ moe_apply_probs_on_input ........................ False
315
+ moe_aux_loss_coeff .............................. 0.0
316
+ moe_deepep_num_sms .............................. 20
317
+ moe_echo_enable_random_offloading ............... False
318
+ moe_echo_expert_dispatch_overlap ................ True
319
+ moe_echo_expert_dispatcher_type ................. hybridep
320
+ moe_echo_recompute_expert_dispatch .............. False
321
+ moe_enable_deepep ............................... False
322
+ moe_enable_echo ................................. True
323
+ moe_enable_hybridep ............................. True
324
+ moe_expert_capacity_factor ...................... None
325
+ moe_extended_tp ................................. False
326
+ moe_ffn_hidden_size ............................. 2048
327
+ moe_grouped_gemm ................................ True
328
+ moe_hybridep_num_sms ............................ 16
329
+ moe_input_jitter_eps ............................ None
330
+ moe_layer_freq .................................. [0, 1]
331
+ moe_layer_recompute ............................. False
332
+ moe_num_echo_experts ............................ 384
333
+ moe_pad_expert_input_to_capacity ................ False
334
+ moe_per_layer_logging ........................... False
335
+ moe_permute_fusion .............................. False
336
+ moe_received_token_capacity ..................... 1.17578125
337
+ moe_router_bias_update_rate ..................... 0.001
338
+ moe_router_dtype ................................ fp32/workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
339
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
340
+
341
+ moe_router_enable_expert_bias ................... True
342
+ moe_router_force_load_balancing ................. False
343
+ moe_router_group_topk ........................... 1
344
+ moe_router_load_balancing_type .................. seq_aux_loss
345
+ moe_router_num_groups ........................... 1
346
+ moe_router_padding_for_fp8 ...................... False
347
+ moe_router_pre_softmax .......................... False
348
+ moe_router_score_function ....................... sigmoid
349
+ moe_router_topk ................................. 8
350
+ moe_router_topk_scaling_factor .................. 2.827
351
+ moe_shared_expert_intermediate_size ............. 2048
352
+ moe_shared_expert_overlap ....................... False
353
+ moe_token_dispatcher_type ....................... flex
354
+ moe_token_drop_policy ........................... probs
355
+ moe_upcycling_granularity ....................... 1
356
+ moe_use_device_initiated_grouped_gemm ........... True
357
+ moe_use_legacy_grouped_gemm ..................... False
358
+ moe_use_upcycling ............................... False
359
+ moe_z_loss_coeff ................................ None
360
+ mrope_section ................................... None
361
+ mscale .......................................... 1.0
362
+ mscale_all_dim .................................. 1.0
363
+ mtp_loss_scaling_factor ......................... 0.1
364
+ mtp_num_layers .................................. None
365
+ multi_latent_attention .......................... True
366
+ nccl_all_reduce_for_prefill ..................... False
367
+ nccl_communicator_config_path ................... None
368
+ nccl_ub ......................................... False
369
+ no_load_optim ................................... None
370
+ no_load_rng ..................................... None
371
+ no_persist_layer_norm ........................... False
372
+ no_rope_freq .................................... None
373
+ no_save_optim ................................... None
374
+ no_save_rng ..................................... None
375
+ non_persistent_ckpt_type ........................ None
376
+ non_persistent_global_ckpt_dir .................. None
377
+ non_persistent_local_ckpt_algo .................. fully_parallel
378
+ non_persistent_local_ckpt_dir ................... None
379
+ non_persistent_save_interval .................... None
380
+ norm_epsilon .................................... 1e-06
381
+ normalization ................................... RMSNorm
382
+ num_attention_heads ............................. 64
383
+ num_channels .................................... 3
384
+ num_classes ..................................... 1000
385
+ num_dataset_builder_threads ..................... 1
386
+ num_distributed_optimizer_instances ............. 1
387
+ num_experts ..................................... 384
388
+ num_layers ...................................... 2
389
+ num_layers_at_end_in_bf16 ....................... 1
390
+ num_layers_at_start_in_bf16 ..................... 1
391
+ num_layers_per_virtual_pipeline_stage ........... None
392
+ num_query_groups ................................ 1
393
+ num_virtual_stages_per_pipeline_rank ............ None
394
+ num_workers ..................................... 2
395
+ object_storage_cache_path ....................... None
396
+ one_logger_async ................................ False
397
+ one_logger_project .............................. megatron-lm
398
+ one_logger_run_name ............................. None
399
+ onnx_safe ....................................... None
400
+ openai_gelu ..................................... False
401
+ optimizer ....................................... adam
402
+ optimizer_cpu_offload ........................... False
403
+ optimizer_offload_fraction ...................... 1.0
404
+ output_bert_embeddings .......................... False
405
+ overlap_cpu_optimizer_d2h_h2d ................... False
406
+ overlap_grad_reduce ............................. True
407
+ overlap_moe_expert_parallel_comm ................ False
408
+ overlap_p2p_comm ................................ False
409
+ overlap_p2p_comm_warmup_flush ................... False
410
+ overlap_param_gather ............................ True
411
+ overlap_param_gather_with_optimizer_step ........ False
412
+ override_opt_param_scheduler .................... False
413
+ params_dtype .................................... torch.bfloat16
414
+ patch_dim ....................................... 16
415
+ per_split_data_args_path ........................ None
416
+ perform_initialization .......................... True
417
+ pin_cpu_grads ................................... True
418
+ pin_cpu_params .................................. True
419
+ pipeline_model_parallel_comm_backend ............ None
420
+ pipeline_model_parallel_layout .................. None
421
+ pipeline_model_parallel_size .................... 1
422
+ position_embedding_type ......................... rope
423
+ pretrained_checkpoint ........................... None
424
+ profile ......................................... False
425
+ profile_ranks ................................... [0]
426
+ profile_step_end ................................ 12
427
+ profile_step_start .............................. 10
428
+ q_lora_rank ..................................... 1536
429
+ qk_head_dim ..................................... 128
430
+ qk_l2_norm ...................................... False
431
+ qk_layernorm .................................... True
432
+ qk_pos_emb_head_dim ............................. 64
433
+ query_in_block_prob ............................. 0.1
434
+ rampup_batch_size ............................... None
435
+ rank ............................................ 0
436
+ recompute_granularity ........................... None
437
+ recompute_method ................................ None
438
+ recompute_modules ............................... None
439
+ recompute_num_layers ............................ None
440
+ record_memory_history ........................... False
441
+ relative_attention_max_distance ................. 128
442
+ relative_attention_num_buckets .................. 32
443
+ replication ..................................... False
444
+ replication_factor .............................. 2
445
+ replication_jump ................................ None
446
+ rerun_mode ...................................... validate_results
447
+ reset_attention_mask ............................ False
448
+ reset_position_ids .............................. False
449
+ result_rejected_tracker_filename ................ None
450
+ retriever_report_topk_accuracies ................ []
451
+ retriever_score_scaling ......................... False
452
+ retriever_seq_length ............................ 256
453
+ retro_add_retriever ............................. False
454
+ retro_attention_gate ............................ 1
455
+ retro_cyclic_train_iters ........................ None
456
+ retro_encoder_attention_dropout ................. 0.1
457
+ retro_encoder_hidden_dropout .................... 0.1
458
+ retro_encoder_layers ............................ 2
459
+ retro_num_neighbors ............................. 2
460
+ retro_num_retrieved_chunks ...................... 2
461
+ retro_project_dir ............................... None
462
+ retro_verify_neighbor_count ..................... True
463
+ reuse_grad_buf_for_mxfp8_param_ag ............... False
464
+ rope_scaling_factor ............................. 8.0
465
+ rope_type ....................................... None
466
+ rotary_base ..................................... 50000
467
+ rotary_interleaved .............................. False
468
+ rotary_percent .................................. 1.0
469
+ rotary_scaling_factor ........................... 32.0
470
+ rotary_seq_len_interpolation_factor ............. None
471
+ run_workload_inspector_server ................... False
472
+ sample_rate ..................................... 1.0
473
+ save ............................................ None
474
+ save_interval ................................... None
475
+ save_retain_interval ............................ None
476
+ scatter_gather_tensors_in_pipeline .............. True
477
+ seed ............................................ 1234
478
+ seq_length ...................................... 4096
479
+ sequence_parallel ............................... False
480
+ sft ............................................. False
481
+ sft_tokenizer_prompt_format ..................... nemotron-h-aligned
482
+ sgd_momentum .................................... 0.9
483
+ short_seq_prob .................................. 0.1
484
+ skip_train ...................................... False
485
+ skipped_train_samples ........................... 0
486
+ spec ............................................ None
487
+ split ........................................... 100,0,0
488
+ squared_relu .................................... False
489
+ start_weight_decay .............................. 0.1
490
+ straggler_ctrlr_port ............................ 65535
491
+ straggler_minmax_count .......................... 1
492
+ suggested_communication_unit_size ............... None
493
+ swiglu .......................................... True
494
+ swin_backbone_type .............................. tiny
495
+ symmetric_ar_type ............................... None
496
+ te_rng_tracker .................................. True
497
+ tensor_model_parallel_size ...................... 1
498
+ tensorboard_dir ................................. None
499
+ tensorboard_log_interval ........................ 1
500
+ tensorboard_queue_size .......................... 1000
501
+ test_data_path .................................. None
502
+ test_mode ....................................... False
503
+ tiktoken_num_special_tokens ..................... 1000
504
+ tiktoken_pattern ................................ None
505
+ tiktoken_special_tokens ......................... None
506
+ timing_log_level ................................ 0
507
+ timing_log_option ............................... minmax
508
+ titles_data_path ................................ None
509
+ tokenizer_model ................................. None
510
+ tokenizer_type .................................. GPT2BPETokenizer
511
+ torch_fsdp2_reshard_after_forward ............... True
512
+ tp_comm_bootstrap_backend ....................... nccl
513
+ tp_comm_bulk_dgrad .............................. True
514
+ tp_comm_bulk_wgrad .............................. True
515
+ tp_comm_overlap ................................. False
516
+ tp_comm_overlap_ag .............................. True
517
+ tp_comm_overlap_cfg ............................. None
518
+ tp_comm_overlap_rs .............................. True
519
+ tp_comm_overlap_rs_dgrad ........................ False
520
+ tp_comm_split_ag ................................ True
521
+ tp_comm_split_rs ................................ True
522
+ train_data_path ................................. None
523
+ train_iters ..................................... 12
524
+ train_samples ................................... None
525
+ train_sync_interval ............................. None
526
+ transformer_impl ................................ transformer_engine
527
+ transformer_pipeline_model_parallel_size ........ 1
528
+ untie_embeddings_and_output_weights ............. True
529
+ use_checkpoint_args ............................. False
530
+ use_checkpoint_opt_param_scheduler .............. False
531
+ use_cpu_initialization .......................... None
532
+ use_custom_fsdp ................................. False
533
+ use_dist_ckpt ................................... True
534
+ use_dist_ckpt_deprecated ........................ False
535
+ use_distributed_optimizer ....................... True
536
+ use_flash_attn .................................. False
537
+ use_legacy_models ............................... False
538
+ use_mp_args_from_checkpoint_args ................ False
539
+ use_one_sent_docs ............................... False
540
+ use_persistent_ckpt_worker ...................... False
541
+ use_precision_aware_optimizer ................... False
542
+ use_pytorch_profiler ............................ False
543
+ use_ring_exchange_p2p ........................... False
544
+ use_rope_scaling ................................ False
545
+ use_rotary_position_embeddings .................. False
546
+ use_sharp ....................................... False
547
+ use_tokenizer_model_from_checkpoint_args ........ True
548
+ use_torch_fsdp2 ................................. False
549
+ use_torch_optimizer_for_cpu_offload ............. False
550
+ use_tp_pp_dp_mapping ............................ False
551
+ v_head_dim ...................................... 128
552
+ valid_data_path ................................. None
553
+ variable_seq_lengths ............................ False
554
+ virtual_pipeline_model_parallel_size ............ None
555
+ vision_backbone_type ............................ vit
556
+ vision_pretraining .............................. False
557
+ vision_pretraining_type ......................... classify
558
+ vocab_extra_ids ................................. 0
559
+ vocab_file ...................................... /workspace/data/wikitext/gpt2-vocab.json
560
+ vocab_size ...................................... None
561
+ wandb_exp_name ..................................
562
+ wandb_project ...................................
563
+ wandb_save_dir ..................................
564
+ weight_decay .................................... 0.1
565
+ weight_decay_incr_style ......................... constant
566
+ wgrad_deferral_limit ............................ 0
567
+ world_size ...................................... 8
568
+ yaml_cfg ........................................ None
569
+ -------------------- end of arguments ---------------------
570
+ INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 1
571
+ > building GPT2BPETokenizer tokenizer ...
572
+ > padded vocab (size: 50257) with 47 dummy tokens (new size: 50304)
573
+ WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.VALIDATE_RESULTS
574
+ > initializing torch distributed ...
575
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
576
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
577
+ [Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
578
+ [Gloo] Rank [Gloo] Rank 0 is connected to 71 peer ranks. is connected to Expected number of connected peer ranks is : 77 peer ranks.
579
+ Expected number of connected peer ranks is : 7
580
+ [Gloo] Rank 4 is connected to 7 peer ranks. [Gloo] Rank Expected number of connected peer ranks is : 73
581
+ is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
582
+ [Gloo] Rank [Gloo] Rank 6[Gloo] Rank is connected to 577 is connected to is connected to peer ranks. 77Expected number of connected peer ranks is : peer ranks. peer ranks. 7Expected number of connected peer ranks is : Expected number of connected peer ranks is :
583
+ 77
584
+
585
+ [Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
586
+ [Gloo] Rank 7 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
587
+ [Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
588
+ [Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
589
+ [Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
590
+ [Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
591
+ [Gloo] Rank [Gloo] Rank 5 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 :
592
+ 7
593
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
594
+ > initialized tensor model parallel with size 1
595
+ > initialized pipeline model parallel with size 1
596
+ > setting random seeds to 1234 ...
597
+ [Gloo] Rank 0 is connected to 0[Gloo] Rank peer ranks. Expected number of connected peer ranks is : 00
598
+ is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
599
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
600
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
601
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
602
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
603
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
604
+ [rank4]:[W921 07:43:42.288600181 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.
605
+ [rank7]:[W921 07:43:42.288637376 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.
606
+ [rank6]:[W921 07:43:42.288703549 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.
607
+ [rank2]:[W921 07:43:42.288746903 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
608
+ [rank3]:[W921 07:43:42.288782321 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.
609
+ [rank5]:[W921 07:43:42.290176909 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.
610
+ [rank1]:[W921 07:43:42.290448243 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
+ > compiling dataset index builder ...
612
+ make: Entering directory '/workspace/Megatron-MoE-Echo/megatron/core/datasets'
613
+ make: Nothing to be done for 'default'.
614
+ make: Leaving directory '/workspace/Megatron-MoE-Echo/megatron/core/datasets'
615
+ >>> done with dataset index builder. Compilation time: 0.036 seconds
616
+ WARNING: constraints for invoking optimized fused softmax kernel are not met. We default back to unfused kernel invocations.
617
+ > compiling and loading fused kernels ...
618
+ [rank0]:[W921 07:43:42.479186495 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.
619
+ >>> done with compiling and loading fused kernels. Compilation time: 4.042 seconds
620
+ time to initialize megatron (seconds): 7.124
621
+ [after megatron is initialized] datetime: 2026-09-21 07:43:48
622
+ building GPT model ...
623
+ /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.
624
+ warnings.warn(
625
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
626
+ warnings.warn(
627
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
628
+ warnings.warn(
629
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
630
+ warnings.warn(
631
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
632
+ warnings.warn(
633
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
634
+ warnings.warn(
635
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
636
+ warnings.warn(
637
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
638
+ warnings.warn(
639
+ > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 3480525824
640
+ 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)
641
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 9
642
+ Params for bucket 1 (360579072 elements, 360579072 padded size):
643
+ module.output_layer.weight
644
+ Params for bucket 2 (44047360 elements, 44047360 padded size):
645
+ module.decoder.final_layernorm.weight
646
+ module.decoder.layers.1.mlp.shared_experts.linear_fc2.weight
647
+ module.decoder.layers.1.mlp.shared_experts.linear_fc1.weight
648
+ Params for bucket 3 (45163520 elements, 45163520 padded size):
649
+ module.decoder.layers.1.mlp.router.weight
650
+ module.decoder.layers.1.self_attention.linear_kv_down_proj.weight
651
+ module.decoder.layers.1.self_attention.linear_kv_up_proj.layer_norm_weight
652
+ module.decoder.layers.1.self_attention.linear_q_up_proj.weight
653
+ module.decoder.layers.1.pre_mlp_layernorm.weight
654
+ module.decoder.layers.1.self_attention.linear_kv_up_proj.weight
655
+ module.decoder.layers.1.self_attention.linear_q_up_proj.layer_norm_weight
656
+ module.decoder.layers.1.self_attention.linear_q_down_proj.weight
657
+ Params for bucket 4 (58720256 elements, 58720256 padded size):
658
+ module.decoder.layers.1.self_attention.linear_proj.weight
659
+ Params for bucket 5 (132127744 elements, 132127744 padded size):
660
+ module.decoder.layers.1.input_layernorm.weight
661
+ module.decoder.layers.0.mlp.linear_fc2.weight
662
+ Params for bucket 6 (264241152 elements, 264241152 padded size):
663
+ module.decoder.layers.0.mlp.linear_fc1.weight
664
+ Params for bucket 7 (42411008 elements, 42411008 padded size):
665
+ module.decoder.layers.0.mlp.linear_fc1.layer_norm_weight
666
+ module.decoder.layers.0.self_attention.linear_kv_up_proj.weight
667
+ module.decoder.layers.0.self_attention.linear_q_up_proj.layer_norm_weight
668
+ module.decoder.layers.0.self_attention.linear_kv_up_proj.layer_norm_weight
669
+ module.decoder.layers.0.self_attention.linear_q_up_proj.weight
670
+ module.decoder.layers.0.self_attention.linear_kv_down_proj.weight
671
+ module.decoder.layers.0.self_attention.linear_q_down_proj.weight
672
+ Params for bucket 8 (58720256 elements, 58720256 padded size):
673
+ module.decoder.layers.0.self_attention.linear_proj.weight
674
+ Params for bucket 9 (360586240 elements, 360586240 padded size):
675
+ module.decoder.layers.0.input_layernorm.weight
676
+ module.embedding.word_embeddings.weight
677
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 40
678
+ Params for bucket 1 (44040192 elements, 44040192 padded size):
679
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight47
680
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight45
681
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight46
682
+ Params for bucket 2 (44040192 elements, 44040192 padded size):
683
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight44
684
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight42
685
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight43
686
+ Params for bucket 3 (44040192 elements, 44040192 padded size):
687
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight41
688
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight39
689
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight40
690
+ Params for bucket 4 (44040192 elements, 44040192 padded size):
691
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight38
692
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight37
693
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight36
694
+ Params for bucket 5 (44040192 elements, 44040192 padded size):
695
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight35
696
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight34
697
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight33
698
+ Params for bucket 6 (44040192 elements, 44040192 padded size):
699
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight32
700
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight30
701
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight31
702
+ Params for bucket 7 (44040192 elements, 44040192 padded size):
703
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight29
704
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight27
705
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight28
706
+ Params for bucket 8 (44040192 elements, 44040192 padded size):
707
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight26
708
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight24
709
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight25
710
+ Params for bucket 9 (44040192 elements, 44040192 padded size):
711
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight23
712
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight21
713
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight22
714
+ Params for bucket 10 (44040192 elements, 44040192 padded size):
715
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight20
716
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight18
717
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight19
718
+ Params for bucket 11 (44040192 elements, 44040192 padded size):
719
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight17
720
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight15
721
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight16
722
+ Params for bucket 12 (44040192 elements, 44040192 padded size):
723
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight14
724
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight12
725
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight13
726
+ Params for bucket 13 (44040192 elements, 44040192 padded size):
727
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight11
728
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight9
729
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight10
730
+ Params for bucket 14 (44040192 elements, 44040192 padded size):
731
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight8
732
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight6
733
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight7
734
+ Params for bucket 15 (44040192 elements, 44040192 padded size):
735
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight5
736
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight4
737
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight3
738
+ Params for bucket 16 (44040192 elements, 44040192 padded size):
739
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight2
740
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight1
741
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight0
742
+ Params for bucket 17 (58720256 elements, 58720256 padded size):
743
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight47
744
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight46
745
+ Params for bucket 18 (58720256 elements, 58720256 padded size):
746
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight45
747
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight44
748
+ Params for bucket 19 (58720256 elements, 58720256 padded size):
749
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight43
750
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight42
751
+ Params for bucket 20 (58720256 elements, 58720256 padded size):
752
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight41
753
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight40
754
+ Params for bucket 21 (58720256 elements, 58720256 padded size):
755
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight39
756
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight38
757
+ Params for bucket 22 (58720256 elements, 58720256 padded size):
758
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight37
759
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight36
760
+ Params for bucket 23 (58720256 elements, 58720256 padded size):
761
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight35
762
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight34
763
+ Params for bucket 24 (58720256 elements, 58720256 padded size):
764
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight33
765
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight32
766
+ Params for bucket 25 (58720256 elements, 58720256 padded size):
767
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight31
768
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight30
769
+ Params for bucket 26 (58720256 elements, 58720256 padded size):
770
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight29
771
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight28
772
+ Params for bucket 27 (58720256 elements, 58720256 padded size):
773
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight27
774
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight26
775
+ Params for bucket 28 (58720256 elements, 58720256 padded size):
776
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight25
777
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight24
778
+ Params for bucket 29 (58720256 elements, 58720256 padded size):
779
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight23
780
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight22
781
+ Params for bucket 30 (58720256 elements, 58720256 padded size):
782
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight21
783
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight20
784
+ Params for bucket 31 (58720256 elements, 58720256 padded size):
785
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight19
786
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight18
787
+ Params for bucket 32 (58720256 elements, 58720256 padded size):
788
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight17
789
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight16
790
+ Params for bucket 33 (58720256 elements, 58720256 padded size):
791
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight15
792
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight14
793
+ Params for bucket 34 (58720256 elements, 58720256 padded size):
794
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight13
795
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight12
796
+ Params for bucket 35 (58720256 elements, 58720256 padded size):
797
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight11
798
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight10
799
+ Params for bucket 36 (58720256 elements, 58720256 padded size):
800
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight9
801
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight8
802
+ Params for bucket 37 (58720256 elements, 58720256 padded size):
803
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight7
804
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight6
805
+ Params for bucket 38 (58720256 elements, 58720256 padded size):
806
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight5
807
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight4
808
+ Params for bucket 39 (58720256 elements, 58720256 padded size):
809
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight3
810
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight2
811
+ Params for bucket 40 (58720256 elements, 58720256 padded size):
812
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight1
813
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight0
814
+ 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 0x7852aff7ddc0>, config_logger_dir='')
815
+ INFO:megatron.core.optimizer_param_scheduler:> learning rate decay style: cosine
816
+ [after model, optimizer, and learning rate scheduler are built] datetime: 2026-09-21 07:43:48
817
+ > building train, validation, and test datasets ...
818
+ > datasets target sizes (minimum size):
819
+ train: 96
820
+ validation: 0
821
+ test: 0
822
+ INFO:megatron.core.datasets.blended_megatron_dataset_config:Let split_matrix = [(0, 1.0), None, None]
823
+ > building train, validation, and test datasets for GPT ...
824
+ 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 0x7852b2228c20>, 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)
825
+ INFO:megatron.core.datasets.indexed_dataset:Load the _IndexReader from /workspace/data/wikitext/wikitext103_gpt2_text_document.idx
826
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence lengths
827
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence pointers
828
+ INFO:megatron.core.datasets.indexed_dataset: Extract the document indices
829
+ INFO:megatron.core.datasets.indexed_dataset:> total number of sequences: 1165029
830
+ INFO:megatron.core.datasets.indexed_dataset:> total number of documents: 1801350
831
+ INFO:megatron.core.datasets.gpt_dataset:Load the GPTDataset train indices
832
+ INFO:megatron.core.datasets.gpt_dataset: Load the document index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-document_index.npy
833
+ INFO:megatron.core.datasets.gpt_dataset: Load the sample index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-sample_index.npy
834
+ INFO:megatron.core.datasets.gpt_dataset: Load the shuffle index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-shuffle_index.npy
835
+ INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 29073
836
+ > finished creating GPT datasets ...
837
+ [after dataloaders are built] datetime: 2026-09-21 07:43:49
838
+ done with setup ...
839
+ (min, max) time across ranks (ms):
840
+ model-and-optimizer-setup ......................: (371.23, 376.32)
841
+ train/valid/test-data-iterators-setup ..........: (64.20, 67.48)
842
+ training ...
843
+ Overwriting rerun_state_machine.current_iteration from -1 to 0...
844
+ [before the start of training step] datetime: 2026-09-21 07:43:49
845
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
846
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
847
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
848
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
849
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
850
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
851
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
852
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
853
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
854
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
855
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
856
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
857
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
858
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
859
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
860
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
861
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
862
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
863
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
864
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
865
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
866
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
867
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
868
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
869
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
870
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
871
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
872
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
873
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
874
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
875
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
876
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
877
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
878
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
879
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
880
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
881
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
882
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
883
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
884
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
885
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
886
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
887
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
888
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
889
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
890
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
891
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
892
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
893
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
894
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
895
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
896
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
897
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
898
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
899
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
900
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
901
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
902
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
903
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
904
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
905
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
906
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
907
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
908
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
909
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
910
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
911
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
912
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
913
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
914
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
915
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
916
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
917
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
918
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
919
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
920
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
921
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
922
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
923
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
924
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
925
+ [2026-09-21 07:44:29] iteration 1/ 12 | consumed samples: 8 | elapsed time per iteration (ms): 40794.5 | throughput per GPU (TFLOP/s/GPU): 0.9 | learning rate: 2.000000E-05 | global batch size: 8 | lm loss: 1.227922E+01 | loss scale: 1.0 | grad norm: 12.616 | number of skipped iterations: 0 | number of nan iterations: 0 |
926
+ Number of parameters in transformer block in billions: 17.55
927
+ Number of parameters in embedding layers in billions: 0.72
928
+ Total number of parameters in billions: 18.28
929
+ Number of parameters in most loaded shard in billions: 18.2753
930
+ Theoretical memory footprints: weight and optimizer=130715.23 MB
931
+ [Rank 0] (after 1 iterations) memory (MB) | allocated: 53862.45263671875 | max allocated: 57609.2568359375 | reserved: 74666.0 | max reserved: 74666.0
932
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
933
+ warnings.warn(
934
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
935
+ warnings.warn(
936
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
937
+ warnings.warn(
938
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
939
+ warnings.warn(
940
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
941
+ warnings.warn(
942
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
943
+ warnings.warn(
944
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
945
+ warnings.warn(
946
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
947
+ warnings.warn(
948
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
949
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
950
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
951
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
952
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
953
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
954
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
955
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
956
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
957
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
958
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
959
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
960
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
961
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
962
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
963
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
964
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
965
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
966
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
967
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
968
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
969
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
970
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
971
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
972
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
973
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
974
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
975
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
976
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
977
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
978
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
979
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
980
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
981
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
982
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
983
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
984
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
985
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
986
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
987
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
988
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
989
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
990
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
991
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
992
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
993
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
994
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
995
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
996
+ [2026-09-21 07:44:30] iteration 2/ 12 | consumed samples: 16 | elapsed time per iteration (ms): 279.3 | throughput per GPU (TFLOP/s/GPU): 126.6 | learning rate: 4.000000E-05 | global batch size: 8 | lm loss: 1.227086E+01 | loss scale: 1.0 | grad norm: 12.533 | number of skipped iterations: 0 | number of nan iterations: 0 |
997
+ [2026-09-21 07:44:30] iteration 3/ 12 | consumed samples: 24 | elapsed time per iteration (ms): 127.9 | throughput per GPU (TFLOP/s/GPU): 276.5 | learning rate: 6.000000E-05 | global batch size: 8 | lm loss: 1.088333E+01 | loss scale: 1.0 | grad norm: 10.978 | number of skipped iterations: 0 | number of nan iterations: 0 |
998
+ Capture CUDA graph for training!!!
999
+ CUDA graph capture done!!!
1000
+ [2026-09-21 07:44:32] iteration 4/ 12 | consumed samples: 32 | elapsed time per iteration (ms): 1729.0 | throughput per GPU (TFLOP/s/GPU): 20.5 | learning rate: 8.000000E-05 | global batch size: 8 | lm loss: 1.448197E+01 | loss scale: 1.0 | grad norm: 72.852 | number of skipped iterations: 0 | number of nan iterations: 0 |
1001
+ [2026-09-21 07:44:32] iteration 5/ 12 | consumed samples: 40 | elapsed time per iteration (ms): 120.6 | throughput per GPU (TFLOP/s/GPU): 293.3 | learning rate: 1.000000E-04 | global batch size: 8 | lm loss: 1.808492E+01 | loss scale: 1.0 | grad norm: 107.033 | number of skipped iterations: 0 | number of nan iterations: 0 |
1002
+ [2026-09-21 07:44:32] iteration 6/ 12 | consumed samples: 48 | elapsed time per iteration (ms): 119.8 | throughput per GPU (TFLOP/s/GPU): 295.3 | learning rate: 7.750000E-05 | global batch size: 8 | lm loss: 1.488903E+01 | loss scale: 1.0 | grad norm: 83.729 | number of skipped iterations: 0 | number of nan iterations: 0 |
1003
+ [2026-09-21 07:44:32] iteration 7/ 12 | consumed samples: 56 | elapsed time per iteration (ms): 119.0 | throughput per GPU (TFLOP/s/GPU): 297.3 | learning rate: 3.250000E-05 | global batch size: 8 | lm loss: 1.142221E+01 | loss scale: 1.0 | grad norm: 26.935 | number of skipped iterations: 0 | number of nan iterations: 0 |
1004
+ [2026-09-21 07:44:32] iteration 8/ 12 | consumed samples: 64 | elapsed time per iteration (ms): 119.3 | throughput per GPU (TFLOP/s/GPU): 296.5 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 1.163188E+01 | loss scale: 1.0 | grad norm: 12.747 | number of skipped iterations: 0 | number of nan iterations: 0 |
1005
+ [2026-09-21 07:44:32] iteration 9/ 12 | consumed samples: 72 | elapsed time per iteration (ms): 121.1 | throughput per GPU (TFLOP/s/GPU): 292.1 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 1.140652E+01 | loss scale: 1.0 | grad norm: 14.164 | number of skipped iterations: 0 | number of nan iterations: 0 |
1006
+ [2026-09-21 07:44:32] iteration 10/ 12 | consumed samples: 80 | elapsed time per iteration (ms): 118.3 | throughput per GPU (TFLOP/s/GPU): 299.0 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 1.102892E+01 | loss scale: 1.0 | grad norm: 9.158 | number of skipped iterations: 0 | number of nan iterations: 0 |
1007
+ [2026-09-21 07:44:32] iteration 11/ 12 | consumed samples: 88 | elapsed time per iteration (ms): 119.0 | throughput per GPU (TFLOP/s/GPU): 297.2 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 1.082034E+01 | loss scale: 1.0 | grad norm: 9.517 | number of skipped iterations: 0 | number of nan iterations: 0 |
1008
+ [2026-09-21 07:44:32] iteration 12/ 12 | consumed samples: 96 | elapsed time per iteration (ms): 119.3 | throughput per GPU (TFLOP/s/GPU): 296.5 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 1.056247E+01 | loss scale: 1.0 | grad norm: 7.645 | number of skipped iterations: 0 | number of nan iterations: 0 |
1009
+ [after training is done] datetime: 2026-09-21 07:44:32
1010
+ W0921 07:44:33.866000 501453 torch/distributed/elastic/agent/server/api.py:719] Received 15 death signal, shutting down workers
1011
+ W0921 07:44:33.867000 501453 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 501560 closing signal SIGTERM
1012
+ W0921 07:44:33.867000 501453 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 501561 closing signal SIGTERM
1013
+ W0921 07:44:33.868000 501453 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 501562 closing signal SIGTERM
1014
+ W0921 07:44:33.869000 501453 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 501563 closing signal SIGTERM
1015
+ W0921 07:44:33.870000 501453 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 501564 closing signal SIGTERM
1016
+ W0921 07:44:33.871000 501453 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 501565 closing signal SIGTERM
1017
+ W0921 07:44:33.873000 501453 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 501566 closing signal SIGTERM
1018
+ W0921 07:44:33.874000 501453 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 501567 closing signal SIGTERM
1019
+ Traceback (most recent call last):
1020
+ File "/usr/local/bin/torchrun", line 8, in <module>
1021
+ sys.exit(main())
1022
+ ^^^^^^
1023
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper
1024
+ return f(*args, **kwargs)
1025
+ ^^^^^^^^^^^^^^^^^^
1026
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 892, in main
1027
+ run(args)
1028
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 883, in run
1029
+ elastic_launch(
1030
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 139, in __call__
1031
+ return launch_agent(self._config, self._entrypoint, list(args))
1032
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1033
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 261, in launch_agent
1034
+ result = agent.run()
1035
+ ^^^^^^^^^^^
1036
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/metrics/api.py", line 138, in wrapper
1037
+ result = f(*args, **kwargs)
1038
+ ^^^^^^^^^^^^^^^^^^
1039
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 711, in run
1040
+ result = self._invoke_run(role)
1041
+ ^^^^^^^^^^^^^^^^^^^^^^
1042
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 870, in _invoke_run
1043
+ time.sleep(monitor_interval)
1044
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/api.py", line 84, in _terminate_process_handler
1045
+ raise SignalException(f"Process {os.getpid()} got signal: {sigval}", sigval=sigval)
1046
+ torch.distributed.elastic.multiprocessing.api.SignalException: Process 501453 got signal: 15
granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep2/mem_samples.csv ADDED
@@ -0,0 +1,232 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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, 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, 629
44
+ 3, 4
45
+ 4, 27
46
+ 5, 4
47
+ 6, 4
48
+ 7, 41
49
+ 0, 821
50
+ 1, 821
51
+ 2, 821
52
+ 3, 821
53
+ 4, 821
54
+ 5, 821
55
+ 6, 821
56
+ 7, 821
57
+ 0, 1927
58
+ 1, 1927
59
+ 2, 1927
60
+ 3, 1927
61
+ 4, 1927
62
+ 5, 1927
63
+ 6, 1927
64
+ 7, 1927
65
+ 0, 36175
66
+ 1, 36207
67
+ 2, 36189
68
+ 3, 36197
69
+ 4, 36175
70
+ 5, 36197
71
+ 6, 36203
72
+ 7, 36209
73
+ 0, 45251
74
+ 1, 45257
75
+ 2, 45249
76
+ 3, 45245
77
+ 4, 45283
78
+ 5, 45265
79
+ 6, 45255
80
+ 7, 45271
81
+ 0, 73339
82
+ 1, 73339
83
+ 2, 73339
84
+ 3, 73339
85
+ 4, 73339
86
+ 5, 73339
87
+ 6, 73339
88
+ 7, 73339
89
+ 0, 76315
90
+ 1, 76315
91
+ 2, 76315
92
+ 3, 76315
93
+ 4, 76315
94
+ 5, 76315
95
+ 6, 76315
96
+ 7, 76315
97
+ 0, 77659
98
+ 1, 77659
99
+ 2, 77659
100
+ 3, 77659
101
+ 4, 77659
102
+ 5, 77659
103
+ 6, 77659
104
+ 7, 77659
105
+ 0, 78881
106
+ 1, 78881
107
+ 2, 78881
108
+ 3, 78881
109
+ 4, 78881
110
+ 5, 78881
111
+ 6, 78881
112
+ 7, 78881
113
+ 0, 78881
114
+ 1, 78881
115
+ 2, 78881
116
+ 3, 78881
117
+ 4, 78881
118
+ 5, 78881
119
+ 6, 78881
120
+ 7, 78881
121
+ 0, 78881
122
+ 1, 78881
123
+ 2, 78881
124
+ 3, 78881
125
+ 4, 78881
126
+ 5, 78881
127
+ 6, 78881
128
+ 7, 78881
129
+ 0, 81891
130
+ 1, 81893
131
+ 2, 81891
132
+ 3, 81891
133
+ 4, 81891
134
+ 5, 81891
135
+ 6, 81905
136
+ 7, 81891
137
+ 0, 82749
138
+ 1, 82751
139
+ 2, 82749
140
+ 3, 82749
141
+ 4, 82749
142
+ 5, 82749
143
+ 6, 82763
144
+ 7, 82749
145
+ 0, 83397
146
+ 1, 83397
147
+ 2, 83397
148
+ 3, 83397
149
+ 4, 83397
150
+ 5, 83397
151
+ 6, 83411
152
+ 7, 83397
153
+ 0, 86295
154
+ 1, 86295
155
+ 2, 86295
156
+ 3, 86295
157
+ 4, 86295
158
+ 5, 86295
159
+ 6, 86309
160
+ 7, 86295
161
+ 0, 91415
162
+ 1, 91415
163
+ 2, 91415
164
+ 3, 91415
165
+ 4, 91415
166
+ 5, 91303
167
+ 6, 91317
168
+ 7, 91415
169
+ 0, 92759
170
+ 1, 92759
171
+ 2, 92759
172
+ 3, 92759
173
+ 4, 92759
174
+ 5, 92647
175
+ 6, 92661
176
+ 7, 92759
177
+ 0, 94491
178
+ 1, 94383
179
+ 2, 94435
180
+ 3, 94399
181
+ 4, 94449
182
+ 5, 100875
183
+ 6, 94367
184
+ 7, 102555
185
+ 0, 104837
186
+ 1, 104835
187
+ 2, 104833
188
+ 3, 104863
189
+ 4, 104863
190
+ 5, 104709
191
+ 6, 104745
192
+ 7, 104863
193
+ 0, 105409
194
+ 1, 105421
195
+ 2, 105409
196
+ 3, 105397
197
+ 4, 105431
198
+ 5, 105305
199
+ 6, 105313
200
+ 7, 105429
201
+ 0, 106209
202
+ 1, 106289
203
+ 2, 106227
204
+ 3, 106231
205
+ 4, 106207
206
+ 5, 106177
207
+ 6, 106153
208
+ 7, 106273
209
+ 0, 122225
210
+ 1, 121779
211
+ 2, 119131
212
+ 3, 121707
213
+ 4, 122435
214
+ 5, 121875
215
+ 6, 122379
216
+ 7, 88955
217
+ 0, 132177
218
+ 1, 132179
219
+ 2, 132163
220
+ 3, 132219
221
+ 4, 132219
222
+ 5, 132203
223
+ 6, 131991
224
+ 7, 132219
225
+ 0, 25299
226
+ 1, 25302
227
+ 2, 25686
228
+ 3, 132219
229
+ 4, 27222
230
+ 5, 25302
231
+ 6, 25302
232
+ 7, 25302
granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep2/run.env ADDED
@@ -0,0 +1,102 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ backend=echo
2
+ config=KIMI_K2_L2
3
+ family=tierB
4
+ source_model=Kimi-K2
5
+ standard_num_layers=61
6
+ base_rung=
7
+ repo=/workspace/Megatron-MoE-Echo
8
+ branch=bracing-regions-through-kernels
9
+ commit=9a236793
10
+ num_layers=2
11
+ hidden=7168
12
+ ffn_hidden=18432
13
+ moe_ffn_hidden=2048
14
+ num_experts=384
15
+ topk=8
16
+ experts_per_rank=48
17
+ expert_params=44040192
18
+ heads=64
19
+ kv_groups=64
20
+ kv_channels=64
21
+ qk_layernorm=1
22
+ shared_ffn=2048
23
+ shared_expert_overlap=0
24
+ moe_token_dispatcher=flex
25
+ mla=1
26
+ q_lora_rank=1536
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=1
32
+ moe_layers=1
33
+ rotary_base=50000
34
+ rotary_percent=1.0
35
+ rotary_scaling=32
36
+ mscale=1.0
37
+ mscale_all_dim=1.0
38
+ qkv_bias=0
39
+ router_balancing=seq_aux_loss
40
+ aux_loss_coeff=0
41
+ router_pre_softmax=0
42
+ router_fusion=0
43
+ router_dtype=fp32
44
+ router_groups=1
45
+ router_group_topk=1
46
+ router_score_function=sigmoid
47
+ router_expert_bias=1
48
+ router_bias_update_rate=1e-3
49
+ router_topk_scaling_factor=2.827
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-6
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=682
73
+ slot_policy=matched
74
+ slots_per_rank=48
75
+ capacity_mode=static
76
+ capacity=1.17578125
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=
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/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep2/train.log ADDED
@@ -0,0 +1,1046 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [07:44:39] WARN: bracing-regions-through-kernels lacks --moe-router-fusion; KIMI_K2_L2 runs unfused (timing only, routing identical)
2
+ ==============================================================
3
+ KIMI_K2_L2 / echo (slots/rank=48 policy=matched)
4
+ E=384 K=8 H'=2048 m=48
5
+ L=2 seq=4096 gbs=8 (1 microbatches)
6
+ cap_mode=static cap=1.17578125 cg=1 recompute=none
7
+ memory_dump=none snapshots=/workspace/granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep2/memory_snapshots/snapshot.pickle
8
+ repo=/workspace/Megatron-MoE-Echo @ 9a236793
9
+ ==============================================================
10
+ W0921 07:44:40.388000 504994 torch/distributed/run.py:766]
11
+ W0921 07:44:40.388000 504994 torch/distributed/run.py:766] *****************************************
12
+ W0921 07:44:40.388000 504994 torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.
13
+ W0921 07:44:40.388000 504994 torch/distributed/run.py:766] *****************************************
14
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
15
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
16
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
17
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
18
+ WARNING: one_logger package is required to enable e2e metrics tracking. please go to https://confluence.nvidia.com/display/MLWFO/Package+Repositories for details to install it
19
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
20
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
21
+ 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
22
+ WARNING: overriding default arguments for tokenizer_type:GPT2BPETokenizer with tokenizer_type:GPT2BPETokenizer
23
+ Number of virtual stages per pipeline stage: None
24
+ accumulate and all-reduce gradients in fp32 for bfloat16 data type.
25
+ using torch.bfloat16 for parameters ...
26
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
27
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
28
+ ------------------------ arguments ------------------------
29
+ account_for_embedding_in_pipeline_split ......... False
30
+ account_for_loss_in_pipeline_split .............. False
31
+ accumulate_allreduce_grads_in_fp32 .............. True
32
+ adam_beta1 ...................................... 0.9
33
+ adam_beta2 ...................................... 0.999
34
+ adam_eps ........................................ 1e-08
35
+ add_bias_linear ................................. False
36
+ add_position_embedding .......................... False
37
+ add_qkv_bias .................................... False
38
+ adlr_autoresume ................................. False
39
+ adlr_autoresume_interval ........................ 1000
40
+ align_grad_reduce ............................... True
41
+ align_param_gather .............................. False
42
+ app_tag_run_name ................................ None
43
+ app_tag_run_version ............................. 0.0.0
44
+ apply_layernorm_1p .............................. False
45
+ apply_query_key_layer_scaling ................... False
46
+ apply_residual_connection_post_layernorm ........ False
47
+ apply_rope_fusion ............................... False
48
+ async_save ...................................... None
49
+ async_tensor_model_parallel_allreduce ........... True
50
+ attention_backend ............................... AttnBackend.auto
51
+ attention_dropout ............................... 0.0
52
+ attention_softmax_in_fp32 ....................... False
53
+ auto_detect_ckpt_format ......................... False
54
+ barrier_with_L1_time ............................ True
55
+ bert_binary_head ................................ True
56
+ bert_embedder_type .............................. megatron
57
+ bert_load ....................................... None
58
+ bf16 ............................................ True
59
+ bias_dropout_fusion ............................. True
60
+ bias_gelu_fusion ................................ False
61
+ bias_swiglu_fusion .............................. True
62
+ biencoder_projection_dim ........................ 0
63
+ biencoder_shared_query_context_model ............ False
64
+ block_data_path ................................. None
65
+ calc_ft_timeouts ................................ False
66
+ calculate_per_token_loss ........................ False
67
+ check_for_large_grads ........................... False
68
+ check_for_nan_in_loss_and_grad .................. False
69
+ check_for_spiky_loss ............................ False
70
+ check_weight_hash_across_dp_replicas_interval ... None
71
+ ckpt_assume_constant_structure .................. False
72
+ ckpt_convert_format ............................. None
73
+ ckpt_convert_save ............................... None
74
+ ckpt_convert_update_legacy_dist_opt_format ...... False
75
+ ckpt_format ..................................... torch_dist
76
+ ckpt_fully_parallel_load ........................ False
77
+ ckpt_fully_parallel_save ........................ True
78
+ ckpt_fully_parallel_save_deprecated ............. False
79
+ ckpt_step ....................................... None
80
+ classes_fraction ................................ 1.0
81
+ clip_grad ....................................... 1.0
82
+ clone_scatter_output_in_embedding ............... True
83
+ config_logger_dir ...............................
84
+ consumed_train_samples .......................... 0
85
+ consumed_valid_samples .......................... 0
86
+ context_parallel_size ........................... 1
87
+ cp_comm_type .................................... ['p2p']
88
+ create_attention_mask_in_dataloader ............. True
89
+ cross_entropy_fusion_impl ....................... native
90
+ cross_entropy_loss_fusion ....................... False
91
+ cuda_graph_scope ................................ full_iteration
92
+ cuda_graph_warmup_steps ......................... 3
93
+ data_args_path .................................. None
94
+ data_cache_path ................................. None
95
+ data_parallel_random_init ....................... False
96
+ data_parallel_sharding_strategy ................. no_shard
97
+ data_parallel_size .............................. 8
98
+ data_path ....................................... ['/workspace/data/wikitext/wikitext103_gpt2_text_document']
99
+ data_per_class_fraction ......................... 1.0
100
+ data_sharding ................................... True
101
+ dataloader_type ................................. single
102
+ ddp_average_in_collective ....................... False
103
+ ddp_bucket_size ................................. None
104
+ ddp_num_buckets ................................. None
105
+ ddp_pad_buckets_for_high_nccl_busbw ............. False
106
+ decoder_first_pipeline_num_layers ............... None
107
+ decoder_last_pipeline_num_layers ................ None
108
+ decoder_num_layers .............................. None
109
+ decoder_seq_length .............................. None
110
+ decoupled_lr .................................... None
111
+ decoupled_min_lr ................................ None
112
+ decrease_batch_size_if_needed ................... False
113
+ defer_embedding_wgrad_compute ................... False
114
+ delay_wgrad_compute ............................. False
115
+ deprecated_use_mcore_models ..................... True
116
+ deterministic_mode .............................. False
117
+ dino_bottleneck_size ............................ 256
118
+ dino_freeze_last_layer .......................... 1
119
+ dino_head_hidden_size ........................... 2048
120
+ dino_local_crops_number ......................... 10
121
+ dino_local_img_size ............................. 96
122
+ dino_norm_last_layer ............................ False
123
+ dino_teacher_temp ............................... 0.07
124
+ dino_warmup_teacher_temp ........................ 0.04
125
+ dino_warmup_teacher_temp_epochs ................. 30
126
+ disable_bf16_reduced_precision_matmul ........... False
127
+ disable_mamba_mem_eff_path ...................... False
128
+ disable_straggler_on_startup .................... False
129
+ dist_ckpt_format_deprecated ..................... None
130
+ dist_ckpt_strictness ............................ assume_ok_unexpected
131
+ distribute_saved_activations .................... False
132
+ distributed_backend ............................. nccl
133
+ distributed_timeout_minutes ..................... 10
134
+ embedding_init_method_std ....................... None
135
+ embedding_path .................................. None
136
+ empty_unused_memory_level ....................... 0
137
+ enable_cuda_graph ............................... True
138
+ enable_experimental ............................. False
139
+ enable_ft_package ............................... False
140
+ enable_gloo_process_groups ...................... True
141
+ enable_msc ...................................... True
142
+ enable_one_logger ............................... True
143
+ encoder_num_layers .............................. 2
144
+ encoder_seq_length .............................. 4096
145
+ end_weight_decay ................................ 0.1
146
+ eod_mask_loss ................................... False
147
+ error_injection_rate ............................ 0
148
+ error_injection_type ............................ transient_error
149
+ eval_interval ................................... 1000000
150
+ eval_iters ...................................... 0
151
+ evidence_data_path .............................. None
152
+ exit_duration_in_mins ........................... None
153
+ exit_interval ................................... None
154
+ exit_on_missing_checkpoint ...................... False
155
+ exit_signal_handler ............................. False
156
+ exp_avg_dtype ................................... torch.float32
157
+ exp_avg_sq_dtype ................................ torch.float32
158
+ expert_model_parallel_size ...................... 8
159
+ expert_tensor_parallel_size ..................... 1
160
+ export_force_local_attention .................... False
161
+ export_kd_cfg ................................... None
162
+ export_kd_teacher_ckpt_format ................... None
163
+ export_kd_teacher_load .......................... None
164
+ export_kv_cache_quant ........................... False
165
+ export_legacy_megatron .......................... False
166
+ export_model_type ............................... GPTModel
167
+ export_moe_apply_probs_on_input ................. False
168
+ export_qk_l2_norm ............................... False
169
+ export_quant_cfg ................................ None
170
+ export_real_quant_cfg ........................... None
171
+ export_te_mcore_model ........................... False
172
+ external_cuda_graph ............................. False
173
+ ffn_hidden_size ................................. 18432
174
+ finetune ........................................ False
175
+ finetune_data_split ............................. train
176
+ finetune_hf_dataset ............................. None
177
+ first_last_layers_bf16 .......................... False
178
+ flash_decode .................................... False
179
+ fp16 ............................................ False
180
+ fp16_lm_cross_entropy ........................... False
181
+ fp32_residual_connection ........................ False
182
+ fp8 ............................................. e4m3
183
+ fp8_amax_compute_algo ........................... most_recent
184
+ fp8_amax_history_len ............................ 1
185
+ fp8_interval .................................... 1
186
+ fp8_margin ...................................... 0
187
+ fp8_param_gather ................................ False
188
+ fp8_recipe ...................................... mxfp8
189
+ fp8_wgrad ....................................... True
190
+ fsdp_double_buffer .............................. False
191
+ global_batch_size ............................... 8
192
+ grad_reduce_in_bf16 ............................. False
193
+ gradient_accumulation_fusion .................... True
194
+ gradient_reduce_div_fusion ...................... True
195
+ group_query_attention ........................... False
196
+ head_lr_mult .................................... 1.0
197
+ heterogeneous_layers_config_encoded_json ........ None
198
+ heterogeneous_layers_config_path ................ None
199
+ hidden_dropout .................................. 0.0
200
+ hidden_size ..................................... 7168
201
+ hierarchical_context_parallel_sizes ............. None
202
+ high_priority_stream_groups ..................... []
203
+ hybrid_attention_ratio .......................... 0.0
204
+ hybrid_mlp_ratio ................................ 0.0
205
+ hybrid_override_pattern ......................... None
206
+ hysteresis ...................................... 2
207
+ ict_head_size ................................... None
208
+ ict_load ........................................ None
209
+ img_h ........................................... 224
210
+ img_w ........................................... 224
211
+ indexer_batch_size .............................. 128
212
+ indexer_log_interval ............................ 1000
213
+ inference_batch_times_seqlen_threshold .......... -1
214
+ inference_dynamic_batching ...................... False
215
+ inference_dynamic_batching_buffer_guaranteed_fraction 0.2
216
+ inference_dynamic_batching_buffer_overflow_factor None
217
+ inference_dynamic_batching_buffer_size_gb ....... 40.0
218
+ inference_dynamic_batching_chunk_size ........... 256
219
+ inference_dynamic_batching_max_requests_override None
220
+ inference_dynamic_batching_max_tokens_override .. None
221
+ inference_dynamic_batching_num_cuda_graphs ...... 16
222
+ inference_max_batch_size ........................ 8
223
+ inference_max_seq_length ........................ 2560
224
+ inference_rng_tracker ........................... False
225
+ init_method_std ................................. 0.02
226
+ init_method_xavier_uniform ...................... False
227
+ init_model_with_meta_device ..................... False
228
+ initial_loss_scale .............................. 4294967296
229
+ inprocess_active_world_size ..................... 8
230
+ inprocess_barrier_timeout ....................... 120
231
+ inprocess_completion_timeout .................... 120
232
+ inprocess_empty_cuda_cache ...................... False
233
+ inprocess_granularity ........................... node
234
+ inprocess_hard_timeout .......................... 90
235
+ inprocess_heartbeat_interval .................... 30
236
+ inprocess_heartbeat_timeout ..................... 60
237
+ inprocess_last_call_wait ........................ 1
238
+ inprocess_max_iterations ........................ None
239
+ inprocess_monitor_process_interval .............. 1.0
240
+ inprocess_monitor_thread_interval ............... 1.0
241
+ inprocess_progress_watchdog_interval ............ 1.0
242
+ inprocess_restart ............................... False
243
+ inprocess_soft_timeout .......................... 60
244
+ inprocess_termination_grace_time ................ 1
245
+ is_hybrid_model ................................. False
246
+ iter_per_epoch .................................. 1250
247
+ iterations_to_skip .............................. []
248
+ keep_fp8_transpose_cache_when_using_custom_fsdp . False
249
+ kitchen_config_file ............................. None
250
+ kitchen_recipe_number ........................... None
251
+ kv_channels ..................................... 64
252
+ kv_lora_rank .................................... 512
253
+ lazy_mpu_init ................................... None
254
+ load ............................................ None
255
+ load_main_params_from_ckpt ...................... None
256
+ load_model_opt_format ........................... False
257
+ local_rank ...................................... 0
258
+ log_energy ...................................... False
259
+ log_interval .................................... 1
260
+ log_loss_scale_to_tensorboard ................... True
261
+ log_memory_to_tensorboard ....................... False
262
+ log_num_zeros_in_grad ........................... False
263
+ log_params_norm ................................. False
264
+ log_progress .................................... False
265
+ log_straggler ................................... False
266
+ log_throughput .................................. True
267
+ log_timers_to_tensorboard ....................... False
268
+ log_validation_ppl_to_tensorboard ............... False
269
+ log_world_size_to_tensorboard ................... False
270
+ logging_level ................................... None
271
+ loss_scale ...................................... None
272
+ loss_scale_window ............................... 1000
273
+ lr .............................................. 0.0001
274
+ lr_decay_iters .................................. 8
275
+ lr_decay_samples ................................ None
276
+ lr_decay_style .................................. cosine
277
+ lr_warmup_fraction .............................. None
278
+ lr_warmup_init .................................. 0.0
279
+ lr_warmup_iters ................................. 5
280
+ lr_warmup_samples ............................... 0
281
+ lr_wsd_decay_iters .............................. None
282
+ lr_wsd_decay_samples ............................ None
283
+ lr_wsd_decay_style .............................. exponential
284
+ main_grads_dtype ................................ torch.float32
285
+ main_params_dtype ............................... torch.float32
286
+ make_vocab_size_divisible_by .................... 128
287
+ mamba_head_dim .................................. 64
288
+ mamba_num_groups ................................ 8
289
+ mamba_num_heads ................................. None
290
+ mamba_state_dim ................................. 128
291
+ manual_gc ....................................... False
292
+ manual_gc_eval .................................. True
293
+ manual_gc_interval .............................. 0
294
+ mask_factor ..................................... 1.0
295
+ mask_prob ....................................... 0.15
296
+ mask_type ....................................... random
297
+ masked_softmax_fusion ........................... False
298
+ max_position_embeddings ......................... 4096
299
+ max_tokens_to_oom ............................... 12000
300
+ memory_snapshot_path ............................ snapshot.pickle
301
+ merge_file ...................................... /workspace/data/wikitext/gpt2-merges.txt
302
+ micro_batch_size ................................ 1
303
+ microbatch_group_size_per_vp_stage .............. None
304
+ mid_level_dataset_surplus ....................... 0.005
305
+ min_loss_scale .................................. 1.0
306
+ min_lr .......................................... 1e-05
307
+ mlp_chunks_for_prefill .......................... 1
308
+ mmap_bin_files .................................. True
309
+ mock_data ....................................... False
310
+ moe_apply_probs_on_input ........................ False
311
+ moe_aux_loss_coeff .............................. 0.0
312
+ moe_deepep_num_sms .............................. 20
313
+ moe_echo_enable_random_offloading ............... False
314
+ moe_echo_expert_dispatch_overlap ................ True
315
+ moe_echo_expert_dispatcher_type ................. hybridep
316
+ moe_echo_recompute_expert_dispatch .............. False
317
+ moe_enable_deepep ............................... False
318
+ moe_enable_echo ................................. True
319
+ moe_enable_hybridep ............................. True
320
+ moe_expert_capacity_factor ...................... None
321
+ moe_extended_tp ................................. False
322
+ moe_ffn_hidden_size ............................. 2048
323
+ moe_grouped_gemm ................................ True
324
+ moe_hybridep_num_sms ............................ 16
325
+ moe_input_jitter_eps ............................ None
326
+ moe_layer_freq .................................. [0, 1]
327
+ moe_layer_recompute ............................. False
328
+ moe_num_echo_experts ............................ 384
329
+ moe_pad_expert_input_to_capacity ................ False
330
+ moe_per_layer_logging ........................... False
331
+ moe_permute_fusion .............................. False
332
+ moe_received_token_capacity ..................... 1.17578125
333
+ moe_router_bias_update_rate ..................... 0.001
334
+ moe_router_dtype ................................ fp32
335
+ moe_router_enable_expert_bias ................... True
336
+ moe_router_force_load_balancing ................. False
337
+ moe_router_group_topk ........................... 1
338
+ moe_router_load_balancing_type .................. seq_aux_loss
339
+ moe_router_num_groups ........................... 1
340
+ moe_router_padding_for_fp8 ...................... False
341
+ moe_router_pre_softmax .......................... False
342
+ moe_router_score_function ....................... sigmoid
343
+ moe_router_topk ................................. 8
344
+ moe_router_topk_scaling_factor .................. 2.827
345
+ moe_shared_expert_intermediate_size ............. 2048
346
+ moe_shared_expert_overlap ....................... False
347
+ moe_token_dispatcher_type ....................... flex
348
+ moe_token_drop_policy ........................... probs
349
+ moe_upcycling_granularity ....................... 1
350
+ moe_use_device_initiated_grouped_gemm ........... True
351
+ moe_use_legacy_grouped_gemm ..................... False
352
+ moe_use_upcycling ............................... False
353
+ moe_z_loss_coeff ................................ None
354
+ mrope_section ................................... None
355
+ mscale .......................................... 1.0
356
+ mscale_all_dim .................................. 1.0
357
+ mtp_loss_scaling_factor ......................... 0.1
358
+ mtp_num_layers .................................. None
359
+ multi_latent_attention .......................... True
360
+ nccl_all_reduce_for_prefill ..................... False
361
+ nccl_communicator_config_path ................... None
362
+ nccl_ub ......................................... False
363
+ no_load_optim ................................... None
364
+ no_load_rng ..................................... None
365
+ no_persist_layer_norm ........................... False
366
+ no_rope_freq .................................... None
367
+ no_save_optim ................................... None
368
+ no_save_rng ..................................... None
369
+ non_persistent_ckpt_type ........................ None
370
+ non_persistent_global_ckpt_dir .................. None
371
+ non_persistent_local_ckpt_algo .................. fully_parallel
372
+ non_persistent_local_ckpt_dir ................... None
373
+ non_persistent_save_interval .................... None
374
+ norm_epsilon .................................... 1e-06
375
+ normalization ................................... RMSNorm
376
+ num_attention_heads ............................. 64
377
+ num_channels .................................... 3
378
+ num_classes ..................................... 1000
379
+ num_dataset_builder_threads ..................... 1
380
+ num_distributed_optimizer_instances ............. 1
381
+ num_experts ..................................... 384
382
+ num_layers ...................................... 2
383
+ num_layers_at_end_in_bf16 ....................... 1
384
+ num_layers_at_start_in_bf16 ..................... 1
385
+ num_layers_per_virtual_pipeline_stage ........... None
386
+ num_query_groups ................................ 1
387
+ num_virtual_stages_per_pipeline_rank ............ None
388
+ num_workers ..................................... 2
389
+ object_storage_cache_path ....................... None
390
+ one_logger_async ................................ False
391
+ one_logger_project .............................. megatron-lm
392
+ one_logger_run_name ............................. None
393
+ onnx_safe ....................................... None
394
+ openai_gelu ..................................... False
395
+ optimizer ....................................... adam
396
+ optimizer_cpu_offload ........................... False
397
+ optimizer_offload_fraction ...................... 1.0
398
+ output_bert_embeddings .......................... False
399
+ overlap_cpu_optimizer_d2h_h2d ................... False
400
+ overlap_grad_reduce ............................. True
401
+ overlap_moe_expert_parallel_comm ................ False
402
+ overlap_p2p_comm ................................ False
403
+ overlap_p2p_comm_warmup_flush ................... False
404
+ overlap_param_gather ............................ True
405
+ overlap_param_gather_with_optimizer_step ........ False
406
+ override_opt_param_scheduler .................... False
407
+ params_dtype .................................... torch.bfloat16
408
+ patch_dim ....................................... 16
409
+ per_split_data_args_path ........................ None
410
+ perform_initialization .......................... True
411
+ pin_cpu_grads ................................... True
412
+ pin_cpu_params .................................. True
413
+ pipeline_model_parallel_comm_backend ............ None
414
+ pipeline_model_parallel_layout .................. None
415
+ pipeline_model_parallel_size .................... 1
416
+ position_embedding_type ......................... rope
417
+ pretrained_checkpoint ........................... None
418
+ profile ......................................... False
419
+ profile_ranks ................................... [0]
420
+ profile_step_end ................................ 12
421
+ profile_step_start .............................. 10
422
+ q_lora_rank ..................................... 1536
423
+ qk_head_dim ..................................... 128
424
+ qk_l2_norm ...................................... False
425
+ qk_layernorm .................................... True
426
+ qk_pos_emb_head_dim ............................. 64
427
+ query_in_block_prob ............................. 0.1
428
+ rampup_batch_size ............................... None
429
+ rank ............................................ 0
430
+ recompute_granularity ........................... None
431
+ recompute_method ................................ None
432
+ recompute_modules ............................... None
433
+ recompute_num_layers ............................ None
434
+ record_memory_history ........................... False
435
+ relative_attention_max_distance ................. 128
436
+ relative_attention_num_buckets .................. 32
437
+ replication ..................................... False
438
+ replication_factor .............................. 2
439
+ replication_jump ................................ None
440
+ rerun_mode ...................................... validate_results
441
+ reset_attention_mask ............................ False
442
+ reset_position_ids .............................. False
443
+ result_rejected_tracker_filename ................ None
444
+ retriever_report_topk_accuracies ................ []
445
+ retriever_score_scaling ......................... False
446
+ retriever_seq_length ............................ 256
447
+ retro_add_retriever ............................. False
448
+ retro_attention_gate ............................ 1
449
+ retro_cyclic_train_iters ........................ None
450
+ retro_encoder_attention_dropout ................. 0.1
451
+ retro_encoder_hidden_dropout .................... 0.1
452
+ retro_encoder_layers ............................ 2
453
+ retro_num_neighbors ............................. 2
454
+ retro_num_retrieved_chunks ...................... 2
455
+ retro_project_dir ............................... None
456
+ retro_verify_neighbor_count ..................... True
457
+ reuse_grad_buf_for_mxfp8_param_ag ............... False
458
+ rope_scaling_factor ............................. 8.0
459
+ rope_type ....................................... None
460
+ rotary_base ..................................... 50000
461
+ rotary_interleaved .............................. False
462
+ rotary_percent .................................. 1.0
463
+ rotary_scaling_factor ........................... 32.0
464
+ rotary_seq_len_interpolation_factor ............. None
465
+ run_workload_inspector_server ................... False
466
+ sample_rate ..................................... 1.0
467
+ save ............................................ None
468
+ save_interval ................................... None
469
+ save_retain_interval ............................ None
470
+ scatter_gather_tensors_in_pipeline .............. True
471
+ seed ............................................ 1234
472
+ seq_length ...................................... 4096
473
+ sequence_parallel ............................... False
474
+ sft ............................................. False
475
+ sft_tokenizer_prompt_format ..................... nemotron-h-aligned
476
+ sgd_momentum .................................... 0.9
477
+ short_seq_prob .................................. 0.1
478
+ skip_train ...................................... False
479
+ skipped_train_samples ........................... 0
480
+ spec ............................................ None
481
+ split ........................................... 100,0,0
482
+ squared_relu .................................... False
483
+ start_weight_decay .............................. 0.1
484
+ straggler_ctrlr_port ............................ 65535
485
+ straggler_minmax_count .......................... 1
486
+ suggested_communication_unit_size ............... None
487
+ swiglu .......................................... True
488
+ swin_backbone_type .............................. tiny
489
+ symmetric_ar_type ............................... None
490
+ te_rng_tracker .................................. True
491
+ tensor_model_parallel_size ...................... 1
492
+ tensorboard_dir ................................. None
493
+ tensorboard_log_interval ........................ 1
494
+ tensorboard_queue_size .......................... 1000
495
+ test_data_path .................................. None
496
+ test_mode ....................................... False
497
+ tiktoken_num_special_tokens ..................... 1000
498
+ tiktoken_pattern ................................ None
499
+ tiktoken_special_tokens ......................... None
500
+ timing_log_level ................................ 0
501
+ timing_log_option ............................... minmax
502
+ titles_data_path ................................ None
503
+ tokenizer_model ................................. None
504
+ tokenizer_type .................................. GPT2BPETokenizer
505
+ torch_fsdp2_reshard_after_forward ............... True
506
+ tp_comm_bootstrap_backend ....................... nccl
507
+ tp_comm_bulk_dgrad .............................. True
508
+ tp_comm_bulk_wgrad .............................. True
509
+ tp_comm_overlap ................................. False
510
+ tp_comm_overlap_ag .............................. True
511
+ tp_comm_overlap_cfg ............................. None
512
+ tp_comm_overlap_rs .............................. True
513
+ tp_comm_overlap_rs_dgrad ........................ False
514
+ tp_comm_split_ag ................................ True
515
+ tp_comm_split_rs ................................ True
516
+ train_data_path ................................. None
517
+ train_iters ..................................... 12
518
+ train_samples ................................... None
519
+ train_sync_interval ............................. None
520
+ transformer_impl ................................ transformer_engine
521
+ transformer_pipeline_model_parallel_size ........ 1
522
+ untie_embeddings_and_output_weights ............. True
523
+ use_checkpoint_args ............................. False
524
+ use_checkpoint_opt_param_scheduler .............. False
525
+ use_cpu_initialization .......................... None
526
+ use_custom_fsdp ................................. False
527
+ use_dist_ckpt ................................... True
528
+ use_dist_ckpt_deprecated ........................ False
529
+ use_distributed_optimizer ....................... True
530
+ use_flash_attn .................................. False
531
+ use_legacy_models ............................... False
532
+ use_mp_args_from_checkpoint_args ................ False
533
+ use_one_sent_docs ............................... False
534
+ use_persistent_ckpt_worker ...................... False
535
+ use_precision_aware_optimizer ................... False
536
+ use_pytorch_profiler ............................ False
537
+ use_ring_exchange_p2p ........................... False
538
+ use_rope_scaling ................................ False
539
+ use_rotary_position_embeddings .................. False
540
+ use_sharp ....................................... False
541
+ use_tokenizer_model_from_checkpoint_args ........ True
542
+ use_torch_fsdp2 ................................. False
543
+ use_torch_optimizer_for_cpu_offload ............. False
544
+ use_tp_pp_dp_mapping ............................ False
545
+ v_head_dim ...................................... 128
546
+ valid_data_path ................................. None
547
+ variable_seq_lengths ............................ False
548
+ virtual_pipeline_model_parallel_size ............ None
549
+ vision_backbone_type ............................ vit
550
+ vision_pretraining .............................. False
551
+ vision_pretraining_type ......................... classify
552
+ vocab_extra_ids ................................. 0
553
+ vocab_file ...................................... /workspace/data/wikitext/gpt2-vocab.json
554
+ vocab_size ...................................... None
555
+ wandb_exp_name ..................................
556
+ wandb_project ...................................
557
+ wandb_save_dir ..................................
558
+ weight_decay .................................... 0.1
559
+ weight_decay_incr_style ......................... constant
560
+ wgrad_deferral_limit ............................ 0
561
+ world_size ...................................... 8
562
+ yaml_cfg ........................................ None
563
+ -------------------- end of arguments ---------------------
564
+ INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 1
565
+ > building GPT2BPETokenizer tokenizer ...
566
+ > padded vocab (size: 50257) with 47 dummy tokens (new size: 50304)
567
+ WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.VALIDATE_RESULTS
568
+ > initializing torch distributed ...
569
+ /workspace/Megatron-MoE-Echo/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
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
572
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
573
+ /workspace/Megatron-MoE-Echo/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
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
576
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
577
+ [Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
578
+ [Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
579
+ [Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
580
+ [Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
581
+ [Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
582
+ [Gloo] Rank 5 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
583
+ [Gloo] Rank [Gloo] Rank 6 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
584
+ 7 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 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
587
+ [Gloo] Rank 0 is connected to 7 peer ranks. [Gloo] Rank Expected number of connected peer ranks is : 73 is connected to 7
588
+ peer ranks. Expected number of connected peer ranks is : 7
589
+ [Gloo] Rank 4 is connected to 7 peer ranks. [Gloo] Rank Expected number of connected peer ranks is : 75
590
+ is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
591
+ [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 :
592
+ 7
593
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
594
+ > initialized tensor model parallel with size 1
595
+ > initialized pipeline model parallel with size 1
596
+ > setting random seeds to 1234 ...
597
+ [Gloo] Rank [Gloo] Rank 0 is connected to 00 is connected to peer ranks. 0Expected number of connected peer ranks is : peer ranks. 0Expected number of connected peer ranks is :
598
+ 0
599
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
600
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
601
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
602
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
603
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
604
+ [rank5]:[W921 07:44:57.321169406 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.
605
+ [rank2]:[W921 07:44:57.321175144 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.
606
+ [rank1]:[W921 07:44:57.321190438 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.
607
+ [rank7]:[W921 07:44:57.321212308 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.
608
+ [rank3]:[W921 07:44:57.321290464 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.
609
+ [rank6]:[W921 07:44:57.321334066 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
610
+ [rank4]:[W921 07:44:57.321337557 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
+ > compiling dataset index builder ...
612
+ make: Entering directory '/workspace/Megatron-MoE-Echo/megatron/core/datasets'
613
+ make: Nothing to be done for 'default'.
614
+ make: Leaving directory '/workspace/Megatron-MoE-Echo/megatron/core/datasets'
615
+ >>> done with dataset index builder. Compilation time: 0.035 seconds
616
+ WARNING: constraints for invoking optimized fused softmax kernel are not met. We default back to unfused kernel invocations.
617
+ > compiling and loading fused kernels ...
618
+ [rank0]:[W921 07:44:57.486836543 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.
619
+ >>> done with compiling and loading fused kernels. Compilation time: 4.083 seconds
620
+ time to initialize megatron (seconds): 7.567
621
+ [after megatron is initialized] datetime: 2026-09-21 07:45:03
622
+ building GPT model ...
623
+ /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.
624
+ warnings.warn(
625
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
626
+ warnings.warn(
627
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
628
+ warnings.warn(
629
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
630
+ warnings.warn(
631
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
632
+ warnings.warn(
633
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
634
+ warnings.warn(
635
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
636
+ warnings.warn(
637
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
638
+ warnings.warn(
639
+ > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 3480525824
640
+ 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)
641
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 9
642
+ Params for bucket 1 (360579072 elements, 360579072 padded size):
643
+ module.output_layer.weight
644
+ Params for bucket 2 (44047360 elements, 44047360 padded size):
645
+ module.decoder.final_layernorm.weight
646
+ module.decoder.layers.1.mlp.shared_experts.linear_fc2.weight
647
+ module.decoder.layers.1.mlp.shared_experts.linear_fc1.weight
648
+ Params for bucket 3 (45163520 elements, 45163520 padded size):
649
+ module.decoder.layers.1.self_attention.linear_kv_down_proj.weight
650
+ module.decoder.layers.1.self_attention.linear_kv_up_proj.layer_norm_weight
651
+ module.decoder.layers.1.self_attention.linear_q_up_proj.weight
652
+ module.decoder.layers.1.self_attention.linear_q_down_proj.weight
653
+ module.decoder.layers.1.mlp.router.weight
654
+ module.decoder.layers.1.self_attention.linear_kv_up_proj.weight
655
+ module.decoder.layers.1.pre_mlp_layernorm.weight
656
+ module.decoder.layers.1.self_attention.linear_q_up_proj.layer_norm_weight
657
+ Params for bucket 4 (58720256 elements, 58720256 padded size):
658
+ module.decoder.layers.1.self_attention.linear_proj.weight
659
+ Params for bucket 5 (132127744 elements, 132127744 padded size):
660
+ module.decoder.layers.1.input_layernorm.weight
661
+ module.decoder.layers.0.mlp.linear_fc2.weight
662
+ Params for bucket 6 (264241152 elements, 264241152 padded size):
663
+ module.decoder.layers.0.mlp.linear_fc1.weight
664
+ Params for bucket 7 (42411008 elements, 42411008 padded size):
665
+ module.decoder.layers.0.self_attention.linear_kv_down_proj.weight
666
+ module.decoder.layers.0.self_attention.linear_q_up_proj.weight
667
+ module.decoder.layers.0.self_attention.linear_kv_up_proj.layer_norm_weight
668
+ module.decoder.layers.0.self_attention.linear_q_down_proj.weight
669
+ module.decoder.layers.0.mlp.linear_fc1.layer_norm_weight
670
+ module.decoder.layers.0.self_attention.linear_kv_up_proj.weight
671
+ module.decoder.layers.0.self_attention.linear_q_up_proj.layer_norm_weight
672
+ Params for bucket 8 (58720256 elements, 58720256 padded size):
673
+ module.decoder.layers.0.self_attention.linear_proj.weight
674
+ Params for bucket 9 (360586240 elements, 360586240 padded size):
675
+ module.decoder.layers.0.input_layernorm.weight
676
+ module.embedding.word_embeddings.weight
677
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 40
678
+ Params for bucket 1 (44040192 elements, 44040192 padded size):
679
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight47
680
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight45
681
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight46
682
+ Params for bucket 2 (44040192 elements, 44040192 padded size):
683
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight44
684
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight42
685
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight43
686
+ Params for bucket 3 (44040192 elements, 44040192 padded size):
687
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight41
688
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight39
689
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight40
690
+ Params for bucket 4 (44040192 elements, 44040192 padded size):
691
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight38
692
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight36
693
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight37
694
+ Params for bucket 5 (44040192 elements, 44040192 padded size):
695
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight35
696
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight33
697
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight34
698
+ Params for bucket 6 (44040192 elements, 44040192 padded size):
699
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight32
700
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight31
701
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight30
702
+ Params for bucket 7 (44040192 elements, 44040192 padded size):
703
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight29
704
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight28
705
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight27
706
+ Params for bucket 8 (44040192 elements, 44040192 padded size):
707
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight26
708
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight25
709
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight24
710
+ Params for bucket 9 (44040192 elements, 44040192 padded size):
711
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight23
712
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight22
713
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight21
714
+ Params for bucket 10 (44040192 elements, 44040192 padded size):
715
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight20
716
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight19
717
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight18
718
+ Params for bucket 11 (44040192 elements, 44040192 padded size):
719
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight17
720
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight16
721
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight15
722
+ Params for bucket 12 (44040192 elements, 44040192 padded size):
723
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight14
724
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight13
725
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight12
726
+ Params for bucket 13 (44040192 elements, 44040192 padded size):
727
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight11
728
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight9
729
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight10
730
+ Params for bucket 14 (44040192 elements, 44040192 padded size):
731
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight8
732
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight6
733
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight7
734
+ Params for bucket 15 (44040192 elements, 44040192 padded size):
735
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight5
736
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight3
737
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight4
738
+ Params for bucket 16 (44040192 elements, 44040192 padded size):
739
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight2
740
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight0
741
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight1
742
+ Params for bucket 17 (58720256 elements, 58720256 padded size):
743
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight47
744
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight46
745
+ Params for bucket 18 (58720256 elements, 58720256 padded size):
746
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight45
747
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight44
748
+ Params for bucket 19 (58720256 elements, 58720256 padded size):
749
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight43
750
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight42
751
+ Params for bucket 20 (58720256 elements, 58720256 padded size):
752
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight41
753
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight40
754
+ Params for bucket 21 (58720256 elements, 58720256 padded size):
755
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight39
756
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight38
757
+ Params for bucket 22 (58720256 elements, 58720256 padded size):
758
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight37
759
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight36
760
+ Params for bucket 23 (58720256 elements, 58720256 padded size):
761
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight35
762
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight34
763
+ Params for bucket 24 (58720256 elements, 58720256 padded size):
764
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight33
765
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight32
766
+ Params for bucket 25 (58720256 elements, 58720256 padded size):
767
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight31
768
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight30
769
+ Params for bucket 26 (58720256 elements, 58720256 padded size):
770
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight29
771
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight28
772
+ Params for bucket 27 (58720256 elements, 58720256 padded size):
773
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight27
774
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight26
775
+ Params for bucket 28 (58720256 elements, 58720256 padded size):
776
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight25
777
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight24
778
+ Params for bucket 29 (58720256 elements, 58720256 padded size):
779
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight23
780
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight22
781
+ Params for bucket 30 (58720256 elements, 58720256 padded size):
782
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight21
783
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight20
784
+ Params for bucket 31 (58720256 elements, 58720256 padded size):
785
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight19
786
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight18
787
+ Params for bucket 32 (58720256 elements, 58720256 padded size):
788
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight17
789
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight16
790
+ Params for bucket 33 (58720256 elements, 58720256 padded size):
791
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight15
792
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight14
793
+ Params for bucket 34 (58720256 elements, 58720256 padded size):
794
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight13
795
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight12
796
+ Params for bucket 35 (58720256 elements, 58720256 padded size):
797
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight11
798
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight10
799
+ Params for bucket 36 (58720256 elements, 58720256 padded size):
800
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight9
801
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight8
802
+ Params for bucket 37 (58720256 elements, 58720256 padded size):
803
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight7
804
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight6
805
+ Params for bucket 38 (58720256 elements, 58720256 padded size):
806
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight5
807
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight4
808
+ Params for bucket 39 (58720256 elements, 58720256 padded size):
809
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight3
810
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight2
811
+ Params for bucket 40 (58720256 elements, 58720256 padded size):
812
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight1
813
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight0
814
+ 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 0x77ab89f0a240>, config_logger_dir='')
815
+ INFO:megatron.core.optimizer_param_scheduler:> learning rate decay style: cosine
816
+ [after model, optimizer, and learning rate scheduler are built] datetime: 2026-09-21 07:45:03
817
+ > building train, validation, and test datasets ...
818
+ > datasets target sizes (minimum size):
819
+ train: 96
820
+ validation: 0
821
+ test: 0
822
+ INFO:megatron.core.datasets.blended_megatron_dataset_config:Let split_matrix = [(0, 1.0), None, None]
823
+ > building train, validation, and test datasets for GPT ...
824
+ 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 0x77ab8a1de240>, 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)
825
+ INFO:megatron.core.datasets.indexed_dataset:Load the _IndexReader from /workspace/data/wikitext/wikitext103_gpt2_text_document.idx
826
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence lengths
827
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence pointers
828
+ INFO:megatron.core.datasets.indexed_dataset: Extract the document indices
829
+ INFO:megatron.core.datasets.indexed_dataset:> total number of sequences: 1165029
830
+ INFO:megatron.core.datasets.indexed_dataset:> total number of documents: 1801350
831
+ INFO:megatron.core.datasets.gpt_dataset:Load the GPTDataset train indices
832
+ INFO:megatron.core.datasets.gpt_dataset: Load the document index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-document_index.npy
833
+ INFO:megatron.core.datasets.gpt_dataset: Load the sample index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-sample_index.npy
834
+ INFO:megatron.core.datasets.gpt_dataset: Load the shuffle index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-shuffle_index.npy
835
+ INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 29073
836
+ > finished creating GPT datasets ...
837
+ [after dataloaders are built] datetime: 2026-09-21 07:45:04
838
+ done with setup ...
839
+ training ...
840
+ Overwriting rerun_state_machine.current_iteration from -1 to 0...
841
+ (min, max) time across ranks (ms):
842
+ model-and-optimizer-setup ......................: (343.56, 349.32)
843
+ train/valid/test-data-iterators-setup ..........: (66.88, 77.87)
844
+ [before the start of training step] datetime: 2026-09-21 07:45:04
845
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
846
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
847
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
848
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
849
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
850
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
851
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
852
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
853
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
854
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
855
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
856
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
857
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
858
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
859
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
860
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
861
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
862
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
863
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
864
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
865
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
866
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
867
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
868
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
869
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
870
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
871
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
872
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
873
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
874
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
875
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
876
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
877
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
878
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
879
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
880
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
881
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
882
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
883
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
884
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
885
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
886
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
887
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
888
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
889
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
890
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
891
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
892
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
893
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
894
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
895
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
896
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
897
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
898
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
899
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
900
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
901
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
902
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
903
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
904
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
905
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
906
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
907
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
908
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
909
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
910
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
911
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
912
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
913
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
914
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
915
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
916
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
917
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
918
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
919
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
920
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
921
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
922
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
923
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
924
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
925
+ [2026-09-21 07:45:45] iteration 1/ 12 | consumed samples: 8 | elapsed time per iteration (ms): 41350.5 | throughput per GPU (TFLOP/s/GPU): 0.9 | learning rate: 2.000000E-05 | global batch size: 8 | lm loss: 1.227922E+01 | loss scale: 1.0 | grad norm: 12.616 | number of skipped iterations: 0 | number of nan iterations: 0 |
926
+ Number of parameters in transformer block in billions: 17.55
927
+ Number of parameters in embedding layers in billions: 0.72
928
+ Total number of parameters in billions: 18.28
929
+ Number of parameters in most loaded shard in billions: 18.2753
930
+ Theoretical memory footprints: weight and optimizer=130715.23 MB
931
+ [Rank 0] (after 1 iterations) memory (MB) | allocated: 53862.45263671875 | max allocated: 57609.2568359375 | reserved: 74666.0 | max reserved: 74666.0
932
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
933
+ warnings.warn(
934
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
935
+ warnings.warn(
936
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
937
+ warnings.warn(
938
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
939
+ warnings.warn(
940
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
941
+ warnings.warn(
942
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
943
+ warnings.warn(
944
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
945
+ warnings.warn(
946
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
947
+ warnings.warn(
948
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
949
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
950
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
951
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
952
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
953
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
954
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
955
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
956
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
957
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
958
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
959
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
960
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
961
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
962
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
963
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
964
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
965
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
966
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
967
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
968
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
969
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
970
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
971
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
972
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
973
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
974
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
975
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
976
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
977
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
978
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
979
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
980
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
981
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
982
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
983
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
984
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
985
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
986
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
987
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
988
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
989
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
990
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
991
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
992
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
993
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
994
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
995
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
996
+ [2026-09-21 07:45:45] iteration 2/ 12 | consumed samples: 16 | elapsed time per iteration (ms): 292.1 | throughput per GPU (TFLOP/s/GPU): 121.1 | learning rate: 4.000000E-05 | global batch size: 8 | lm loss: 1.227086E+01 | loss scale: 1.0 | grad norm: 12.533 | number of skipped iterations: 0 | number of nan iterations: 0 |
997
+ [2026-09-21 07:45:45] iteration 3/ 12 | consumed samples: 24 | elapsed time per iteration (ms): 127.7 | throughput per GPU (TFLOP/s/GPU): 276.9 | learning rate: 6.000000E-05 | global batch size: 8 | lm loss: 1.088333E+01 | loss scale: 1.0 | grad norm: 10.978 | number of skipped iterations: 0 | number of nan iterations: 0 |
998
+ Capture CUDA graph for training!!!
999
+ CUDA graph capture done!!!
1000
+ [2026-09-21 07:45:47] iteration 4/ 12 | consumed samples: 32 | elapsed time per iteration (ms): 1722.3 | throughput per GPU (TFLOP/s/GPU): 20.5 | learning rate: 8.000000E-05 | global batch size: 8 | lm loss: 1.448197E+01 | loss scale: 1.0 | grad norm: 72.852 | number of skipped iterations: 0 | number of nan iterations: 0 |
1001
+ [2026-09-21 07:45:47] iteration 5/ 12 | consumed samples: 40 | elapsed time per iteration (ms): 120.9 | throughput per GPU (TFLOP/s/GPU): 292.6 | learning rate: 1.000000E-04 | global batch size: 8 | lm loss: 1.808492E+01 | loss scale: 1.0 | grad norm: 107.033 | number of skipped iterations: 0 | number of nan iterations: 0 |
1002
+ [2026-09-21 07:45:47] iteration 6/ 12 | consumed samples: 48 | elapsed time per iteration (ms): 119.3 | throughput per GPU (TFLOP/s/GPU): 296.4 | learning rate: 7.750000E-05 | global batch size: 8 | lm loss: 1.488903E+01 | loss scale: 1.0 | grad norm: 83.729 | number of skipped iterations: 0 | number of nan iterations: 0 |
1003
+ [2026-09-21 07:45:47] iteration 7/ 12 | consumed samples: 56 | elapsed time per iteration (ms): 119.6 | throughput per GPU (TFLOP/s/GPU): 295.9 | learning rate: 3.250000E-05 | global batch size: 8 | lm loss: 1.142221E+01 | loss scale: 1.0 | grad norm: 26.935 | number of skipped iterations: 0 | number of nan iterations: 0 |
1004
+ [2026-09-21 07:45:48] iteration 8/ 12 | consumed samples: 64 | elapsed time per iteration (ms): 120.1 | throughput per GPU (TFLOP/s/GPU): 294.5 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 1.163188E+01 | loss scale: 1.0 | grad norm: 12.747 | number of skipped iterations: 0 | number of nan iterations: 0 |
1005
+ [2026-09-21 07:45:48] iteration 9/ 12 | consumed samples: 72 | elapsed time per iteration (ms): 119.1 | throughput per GPU (TFLOP/s/GPU): 297.0 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 1.140652E+01 | loss scale: 1.0 | grad norm: 14.164 | number of skipped iterations: 0 | number of nan iterations: 0 |
1006
+ [2026-09-21 07:45:48] iteration 10/ 12 | consumed samples: 80 | elapsed time per iteration (ms): 118.3 | throughput per GPU (TFLOP/s/GPU): 298.9 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 1.102892E+01 | loss scale: 1.0 | grad norm: 9.158 | number of skipped iterations: 0 | number of nan iterations: 0 |
1007
+ [2026-09-21 07:45:48] iteration 11/ 12 | consumed samples: 88 | elapsed time per iteration (ms): 122.5 | throughput per GPU (TFLOP/s/GPU): 288.9 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 1.082034E+01 | loss scale: 1.0 | grad norm: 9.517 | number of skipped iterations: 0 | number of nan iterations: 0 |
1008
+ [2026-09-21 07:45:48] iteration 12/ 12 | consumed samples: 96 | elapsed time per iteration (ms): 119.3 | throughput per GPU (TFLOP/s/GPU): 296.5 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 1.056247E+01 | loss scale: 1.0 | grad norm: 7.645 | number of skipped iterations: 0 | number of nan iterations: 0 |
1009
+ [after training is done] datetime: 2026-09-21 07:45:48
1010
+ W0921 07:45:49.058000 504994 torch/distributed/elastic/agent/server/api.py:719] Received 15 death signal, shutting down workers
1011
+ W0921 07:45:49.058000 504994 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 505098 closing signal SIGTERM
1012
+ W0921 07:45:49.059000 504994 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 505099 closing signal SIGTERM
1013
+ W0921 07:45:49.060000 504994 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 505100 closing signal SIGTERM
1014
+ W0921 07:45:49.062000 504994 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 505101 closing signal SIGTERM
1015
+ W0921 07:45:49.063000 504994 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 505102 closing signal SIGTERM
1016
+ W0921 07:45:49.064000 504994 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 505103 closing signal SIGTERM
1017
+ W0921 07:45:49.065000 504994 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 505104 closing signal SIGTERM
1018
+ W0921 07:45:49.066000 504994 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 505105 closing signal SIGTERM
1019
+ Traceback (most recent call last):
1020
+ File "/usr/local/bin/torchrun", line 8, in <module>
1021
+ sys.exit(main())
1022
+ ^^^^^^
1023
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper
1024
+ return f(*args, **kwargs)
1025
+ ^^^^^^^^^^^^^^^^^^
1026
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 892, in main
1027
+ run(args)
1028
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 883, in run
1029
+ elastic_launch(
1030
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 139, in __call__
1031
+ return launch_agent(self._config, self._entrypoint, list(args))
1032
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1033
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 261, in launch_agent
1034
+ result = agent.run()
1035
+ ^^^^^^^^^^^
1036
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/metrics/api.py", line 138, in wrapper
1037
+ result = f(*args, **kwargs)
1038
+ ^^^^^^^^^^^^^^^^^^
1039
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 711, in run
1040
+ result = self._invoke_run(role)
1041
+ ^^^^^^^^^^^^^^^^^^^^^^
1042
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 870, in _invoke_run
1043
+ time.sleep(monitor_interval)
1044
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/api.py", line 84, in _terminate_process_handler
1045
+ raise SignalException(f"Process {os.getpid()} got signal: {sigval}", sigval=sigval)
1046
+ torch.distributed.elastic.multiprocessing.api.SignalException: Process 504994 got signal: 15
granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep3/mem_samples.csv ADDED
@@ -0,0 +1,240 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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, 4
42
+ 1, 629
43
+ 2, 629
44
+ 3, 629
45
+ 4, 629
46
+ 5, 629
47
+ 6, 629
48
+ 7, 629
49
+ 0, 941
50
+ 1, 907
51
+ 2, 907
52
+ 3, 917
53
+ 4, 919
54
+ 5, 921
55
+ 6, 921
56
+ 7, 933
57
+ 0, 1927
58
+ 1, 1927
59
+ 2, 1927
60
+ 3, 1927
61
+ 4, 1927
62
+ 5, 1927
63
+ 6, 1927
64
+ 7, 1927
65
+ 0, 36229
66
+ 1, 36219
67
+ 2, 36205
68
+ 3, 36231
69
+ 4, 36237
70
+ 5, 36245
71
+ 6, 36279
72
+ 7, 36261
73
+ 0, 45317
74
+ 1, 45309
75
+ 2, 45313
76
+ 3, 45295
77
+ 4, 45369
78
+ 5, 45321
79
+ 6, 45327
80
+ 7, 45303
81
+ 0, 73339
82
+ 1, 73339
83
+ 2, 73339
84
+ 3, 73339
85
+ 4, 73339
86
+ 5, 73339
87
+ 6, 73339
88
+ 7, 73339
89
+ 0, 76315
90
+ 1, 76315
91
+ 2, 76315
92
+ 3, 76315
93
+ 4, 76315
94
+ 5, 76315
95
+ 6, 76315
96
+ 7, 76315
97
+ 0, 77659
98
+ 1, 77659
99
+ 2, 77659
100
+ 3, 77659
101
+ 4, 77659
102
+ 5, 77659
103
+ 6, 77659
104
+ 7, 77659
105
+ 0, 78881
106
+ 1, 78881
107
+ 2, 78881
108
+ 3, 78881
109
+ 4, 78881
110
+ 5, 78881
111
+ 6, 78881
112
+ 7, 78881
113
+ 0, 78881
114
+ 1, 78881
115
+ 2, 78881
116
+ 3, 78881
117
+ 4, 78881
118
+ 5, 78881
119
+ 6, 78881
120
+ 7, 78881
121
+ 0, 78881
122
+ 1, 78881
123
+ 2, 78881
124
+ 3, 78881
125
+ 4, 78881
126
+ 5, 78881
127
+ 6, 78881
128
+ 7, 78881
129
+ 0, 81891
130
+ 1, 81893
131
+ 2, 81891
132
+ 3, 81891
133
+ 4, 81891
134
+ 5, 81891
135
+ 6, 81905
136
+ 7, 81891
137
+ 0, 81897
138
+ 1, 81899
139
+ 2, 81897
140
+ 3, 81897
141
+ 4, 81897
142
+ 5, 81897
143
+ 6, 81911
144
+ 7, 81897
145
+ 0, 83395
146
+ 1, 83397
147
+ 2, 83395
148
+ 3, 83395
149
+ 4, 83395
150
+ 5, 83395
151
+ 6, 83409
152
+ 7, 83395
153
+ 0, 83397
154
+ 1, 83397
155
+ 2, 83397
156
+ 3, 83397
157
+ 4, 83397
158
+ 5, 83397
159
+ 6, 83411
160
+ 7, 83397
161
+ 0, 91415
162
+ 1, 91415
163
+ 2, 91415
164
+ 3, 91415
165
+ 4, 91415
166
+ 5, 91303
167
+ 6, 91317
168
+ 7, 91415
169
+ 0, 91415
170
+ 1, 91415
171
+ 2, 91415
172
+ 3, 91415
173
+ 4, 91415
174
+ 5, 91303
175
+ 6, 91317
176
+ 7, 91415
177
+ 0, 92759
178
+ 1, 92759
179
+ 2, 92759
180
+ 3, 92759
181
+ 4, 92759
182
+ 5, 92647
183
+ 6, 92661
184
+ 7, 92759
185
+ 0, 104683
186
+ 1, 104665
187
+ 2, 104691
188
+ 3, 104683
189
+ 4, 104693
190
+ 5, 104557
191
+ 6, 104579
192
+ 7, 104687
193
+ 0, 105429
194
+ 1, 105425
195
+ 2, 105451
196
+ 3, 105449
197
+ 4, 105435
198
+ 5, 105383
199
+ 6, 105397
200
+ 7, 105457
201
+ 0, 106199
202
+ 1, 106189
203
+ 2, 106245
204
+ 3, 106181
205
+ 4, 106241
206
+ 5, 106117
207
+ 6, 106185
208
+ 7, 106279
209
+ 0, 124605
210
+ 1, 124607
211
+ 2, 124605
212
+ 3, 124591
213
+ 4, 124591
214
+ 5, 124591
215
+ 6, 124591
216
+ 7, 124591
217
+ 0, 132177
218
+ 1, 132179
219
+ 2, 132163
220
+ 3, 132219
221
+ 4, 132219
222
+ 5, 132203
223
+ 6, 131991
224
+ 7, 132219
225
+ 0, 132177
226
+ 1, 132179
227
+ 2, 132163
228
+ 3, 132219
229
+ 4, 132219
230
+ 5, 132203
231
+ 6, 131991
232
+ 7, 132219
233
+ 0, 24723
234
+ 1, 24726
235
+ 2, 25110
236
+ 3, 131643
237
+ 4, 26646
238
+ 5, 24726
239
+ 6, 24726
240
+ 7, 24726
granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep3/run.env ADDED
@@ -0,0 +1,102 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ backend=echo
2
+ config=KIMI_K2_L2
3
+ family=tierB
4
+ source_model=Kimi-K2
5
+ standard_num_layers=61
6
+ base_rung=
7
+ repo=/workspace/Megatron-MoE-Echo
8
+ branch=bracing-regions-through-kernels
9
+ commit=9a236793
10
+ num_layers=2
11
+ hidden=7168
12
+ ffn_hidden=18432
13
+ moe_ffn_hidden=2048
14
+ num_experts=384
15
+ topk=8
16
+ experts_per_rank=48
17
+ expert_params=44040192
18
+ heads=64
19
+ kv_groups=64
20
+ kv_channels=64
21
+ qk_layernorm=1
22
+ shared_ffn=2048
23
+ shared_expert_overlap=0
24
+ moe_token_dispatcher=flex
25
+ mla=1
26
+ q_lora_rank=1536
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=1
32
+ moe_layers=1
33
+ rotary_base=50000
34
+ rotary_percent=1.0
35
+ rotary_scaling=32
36
+ mscale=1.0
37
+ mscale_all_dim=1.0
38
+ qkv_bias=0
39
+ router_balancing=seq_aux_loss
40
+ aux_loss_coeff=0
41
+ router_pre_softmax=0
42
+ router_fusion=0
43
+ router_dtype=fp32
44
+ router_groups=1
45
+ router_group_topk=1
46
+ router_score_function=sigmoid
47
+ router_expert_bias=1
48
+ router_bias_update_rate=1e-3
49
+ router_topk_scaling_factor=2.827
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-6
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=682
73
+ slot_policy=matched
74
+ slots_per_rank=48
75
+ capacity_mode=static
76
+ capacity=1.17578125
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=
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/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep3/train.log ADDED
@@ -0,0 +1,1047 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [07:45:54] WARN: bracing-regions-through-kernels lacks --moe-router-fusion; KIMI_K2_L2 runs unfused (timing only, routing identical)
2
+ ==============================================================
3
+ KIMI_K2_L2 / echo (slots/rank=48 policy=matched)
4
+ E=384 K=8 H'=2048 m=48
5
+ L=2 seq=4096 gbs=8 (1 microbatches)
6
+ cap_mode=static cap=1.17578125 cg=1 recompute=none
7
+ memory_dump=none snapshots=/workspace/granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep3/memory_snapshots/snapshot.pickle
8
+ repo=/workspace/Megatron-MoE-Echo @ 9a236793
9
+ ==============================================================
10
+ W0921 07:45:55.614000 508525 torch/distributed/run.py:766]
11
+ W0921 07:45:55.614000 508525 torch/distributed/run.py:766] *****************************************
12
+ W0921 07:45:55.614000 508525 torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.
13
+ W0921 07:45:55.614000 508525 torch/distributed/run.py:766] *****************************************
14
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
15
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
16
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
17
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
18
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
19
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
20
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
21
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
22
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
23
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
24
+ 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
25
+ using world size: 8, data-parallel size: 8, context-parallel size: 1, hierarchical context-parallel sizes: None, tensor-model-parallel size: 1, pipeline-model-parallel size: 1
26
+ WARNING: overriding default arguments for tokenizer_type:GPT2BPETokenizer with tokenizer_type:GPT2BPETokenizer
27
+ Number of virtual stages per pipeline stage: None
28
+ accumulate and all-reduce gradients in fp32 for bfloat16 data type.
29
+ using torch.bfloat16 for parameters ...
30
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
31
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
32
+ ------------------------ arguments ------------------------
33
+ account_for_embedding_in_pipeline_split ......... False
34
+ account_for_loss_in_pipeline_split .............. False
35
+ accumulate_allreduce_grads_in_fp32 .............. True
36
+ adam_beta1 ...................................... 0.9
37
+ adam_beta2 ...................................... 0.999
38
+ adam_eps ........................................ 1e-08
39
+ add_bias_linear ................................. False
40
+ add_position_embedding .......................... False
41
+ add_qkv_bias .................................... False
42
+ adlr_autoresume ................................. False
43
+ adlr_autoresume_interval ........................ 1000
44
+ align_grad_reduce ............................... True
45
+ align_param_gather .............................. False
46
+ app_tag_run_name ................................ None
47
+ app_tag_run_version ............................. 0.0.0
48
+ apply_layernorm_1p .............................. False
49
+ apply_query_key_layer_scaling ................... False
50
+ apply_residual_connection_post_layernorm ........ False
51
+ apply_rope_fusion ............................... False
52
+ async_save ...................................... None
53
+ async_tensor_model_parallel_allreduce ........... True
54
+ attention_backend ............................... AttnBackend.auto
55
+ attention_dropout ............................... 0.0
56
+ attention_softmax_in_fp32 ....................... False
57
+ auto_detect_ckpt_format ......................... False
58
+ barrier_with_L1_time ............................ True
59
+ bert_binary_head ................................ True
60
+ bert_embedder_type .............................. megatron
61
+ bert_load ....................................... None
62
+ bf16 ............................................ True
63
+ bias_dropout_fusion ............................. True
64
+ bias_gelu_fusion ................................ False
65
+ bias_swiglu_fusion .............................. True
66
+ biencoder_projection_dim ........................ 0
67
+ biencoder_shared_query_context_model ............ False
68
+ block_data_path ................................. None
69
+ calc_ft_timeouts ................................ False
70
+ calculate_per_token_loss ........................ False
71
+ check_for_large_grads ........................... False
72
+ check_for_nan_in_loss_and_grad .................. False
73
+ check_for_spiky_loss ............................ False
74
+ check_weight_hash_across_dp_replicas_interval ... None
75
+ ckpt_assume_constant_structure .................. False
76
+ ckpt_convert_format ............................. None
77
+ ckpt_convert_save ............................... None
78
+ ckpt_convert_update_legacy_dist_opt_format ...... False
79
+ ckpt_format ..................................... torch_dist
80
+ ckpt_fully_parallel_load ........................ False
81
+ ckpt_fully_parallel_save ........................ True
82
+ ckpt_fully_parallel_save_deprecated ............. False
83
+ ckpt_step ....................................... None
84
+ classes_fraction ................................ 1.0
85
+ clip_grad ....................................... 1.0
86
+ clone_scatter_output_in_embedding ............... True
87
+ config_logger_dir ...............................
88
+ consumed_train_samples .......................... 0
89
+ consumed_valid_samples .......................... 0
90
+ context_parallel_size ........................... 1
91
+ cp_comm_type .................................... ['p2p']
92
+ create_attention_mask_in_dataloader ............. True
93
+ cross_entropy_fusion_impl ....................... native
94
+ cross_entropy_loss_fusion ....................... False
95
+ cuda_graph_scope ................................ full_iteration
96
+ cuda_graph_warmup_steps ......................... 3
97
+ data_args_path .................................. None
98
+ data_cache_path ................................. None
99
+ data_parallel_random_init ....................... False
100
+ data_parallel_sharding_strategy ................. no_shard
101
+ data_parallel_size .............................. 8
102
+ data_path ....................................... ['/workspace/data/wikitext/wikitext103_gpt2_text_document']
103
+ data_per_class_fraction ......................... 1.0
104
+ data_sharding ................................... True
105
+ dataloader_type ................................. single
106
+ ddp_average_in_collective ....................... False
107
+ ddp_bucket_size ................................. None
108
+ ddp_num_buckets ................................. None
109
+ ddp_pad_buckets_for_high_nccl_busbw ............. False
110
+ decoder_first_pipeline_num_layers ............... None
111
+ decoder_last_pipeline_num_layers ................ None
112
+ decoder_num_layers .............................. None
113
+ decoder_seq_length .............................. None
114
+ decoupled_lr .................................... None
115
+ decoupled_min_lr ................................ None
116
+ decrease_batch_size_if_needed ................... False
117
+ defer_embedding_wgrad_compute ................... False
118
+ delay_wgrad_compute ............................. False
119
+ deprecated_use_mcore_models ..................... True
120
+ deterministic_mode .............................. False
121
+ dino_bottleneck_size ............................ 256
122
+ dino_freeze_last_layer .......................... 1
123
+ dino_head_hidden_size ........................... 2048
124
+ dino_local_crops_number ......................... 10
125
+ dino_local_img_size ............................. 96
126
+ dino_norm_last_layer ............................ False
127
+ dino_teacher_temp ............................... 0.07
128
+ dino_warmup_teacher_temp ........................ 0.04
129
+ dino_warmup_teacher_temp_epochs ................. 30
130
+ disable_bf16_reduced_precision_matmul ........... False
131
+ disable_mamba_mem_eff_path ...................... False
132
+ disable_straggler_on_startup .................... False
133
+ dist_ckpt_format_deprecated ..................... None
134
+ dist_ckpt_strictness ............................ assume_ok_unexpected
135
+ distribute_saved_activations .................... False
136
+ distributed_backend ............................. nccl
137
+ distributed_timeout_minutes ..................... 10
138
+ embedding_init_method_std ....................... None
139
+ embedding_path .................................. None
140
+ empty_unused_memory_level ....................... 0
141
+ enable_cuda_graph ............................... True
142
+ enable_experimental ............................. False
143
+ enable_ft_package ............................... False
144
+ enable_gloo_process_groups ...................... True
145
+ enable_msc ...................................... True
146
+ enable_one_logger ............................... True
147
+ encoder_num_layers .............................. 2
148
+ encoder_seq_length .............................. 4096
149
+ end_weight_decay ................................ 0.1
150
+ eod_mask_loss ................................... False
151
+ error_injection_rate ............................ 0
152
+ error_injection_type ............................ transient_error
153
+ eval_interval ................................... 1000000
154
+ eval_iters ...................................... 0
155
+ evidence_data_path .............................. None
156
+ exit_duration_in_mins ........................... None
157
+ exit_interval ................................... None
158
+ exit_on_missing_checkpoint ...................... False
159
+ exit_signal_handler ............................. False
160
+ exp_avg_dtype ................................... torch.float32
161
+ exp_avg_sq_dtype ................................ torch.float32
162
+ expert_model_parallel_size ...................... 8
163
+ expert_tensor_parallel_size ..................... 1
164
+ export_force_local_attention .................... False
165
+ export_kd_cfg ................................... None
166
+ export_kd_teacher_ckpt_format ................... None
167
+ export_kd_teacher_load .......................... None
168
+ export_kv_cache_quant ........................... False
169
+ export_legacy_megatron .......................... False
170
+ export_model_type ............................... GPTModel
171
+ export_moe_apply_probs_on_input ................. False
172
+ export_qk_l2_norm ............................... False
173
+ export_quant_cfg ................................ None
174
+ export_real_quant_cfg ........................... None
175
+ export_te_mcore_model ........................... False
176
+ external_cuda_graph ............................. False
177
+ ffn_hidden_size ................................. 18432
178
+ finetune ........................................ False
179
+ finetune_data_split ............................. train
180
+ finetune_hf_dataset ............................. None
181
+ first_last_layers_bf16 .......................... False
182
+ flash_decode .................................... False
183
+ fp16 ............................................ False
184
+ fp16_lm_cross_entropy ........................... False
185
+ fp32_residual_connection ........................ False
186
+ fp8 ............................................. e4m3
187
+ fp8_amax_compute_algo ........................... most_recent
188
+ fp8_amax_history_len ............................ 1
189
+ fp8_interval .................................... 1
190
+ fp8_margin ...................................... 0
191
+ fp8_param_gather ................................ False
192
+ fp8_recipe ...................................... mxfp8
193
+ fp8_wgrad ....................................... True
194
+ fsdp_double_buffer .............................. False
195
+ global_batch_size ............................... 8
196
+ grad_reduce_in_bf16 ............................. False
197
+ gradient_accumulation_fusion .................... True
198
+ gradient_reduce_div_fusion ...................... True
199
+ group_query_attention ........................... False
200
+ head_lr_mult .................................... 1.0
201
+ heterogeneous_layers_config_encoded_json ........ None
202
+ heterogeneous_layers_config_path ................ None
203
+ hidden_dropout .................................. 0.0
204
+ hidden_size ..................................... 7168
205
+ hierarchical_context_parallel_sizes ............. None
206
+ high_priority_stream_groups ..................... []
207
+ hybrid_attention_ratio .......................... 0.0
208
+ hybrid_mlp_ratio ................................ 0.0
209
+ hybrid_override_pattern ......................... None
210
+ hysteresis ...................................... 2
211
+ ict_head_size ................................... None
212
+ ict_load ........................................ None
213
+ img_h ........................................... 224
214
+ img_w ........................................... 224
215
+ indexer_batch_size .............................. 128
216
+ indexer_log_interval ............................ 1000
217
+ inference_batch_times_seqlen_threshold .......... -1
218
+ inference_dynamic_batching ...................... False
219
+ inference_dynamic_batching_buffer_guaranteed_fraction 0.2
220
+ inference_dynamic_batching_buffer_overflow_factor None
221
+ inference_dynamic_batching_buffer_size_gb ....... 40.0
222
+ inference_dynamic_batching_chunk_size ........... 256
223
+ inference_dynamic_batching_max_requests_override None
224
+ inference_dynamic_batching_max_tokens_override .. None
225
+ inference_dynamic_batching_num_cuda_graphs ...... 16
226
+ inference_max_batch_size ........................ 8
227
+ inference_max_seq_length ........................ 2560
228
+ inference_rng_tracker ........................... False
229
+ init_method_std ................................. 0.02
230
+ init_method_xavier_uniform ...................... False
231
+ init_model_with_meta_device ..................... False
232
+ initial_loss_scale .............................. 4294967296
233
+ inprocess_active_world_size ..................... 8
234
+ inprocess_barrier_timeout ....................... 120
235
+ inprocess_completion_timeout .................... 120
236
+ inprocess_empty_cuda_cache ...................... False
237
+ inprocess_granularity ........................... node
238
+ inprocess_hard_timeout .......................... 90
239
+ inprocess_heartbeat_interval .................... 30
240
+ inprocess_heartbeat_timeout ..................... 60
241
+ inprocess_last_call_wait ........................ 1
242
+ inprocess_max_iterations ........................ None
243
+ inprocess_monitor_process_interval .............. 1.0
244
+ inprocess_monitor_thread_interval ............... 1.0
245
+ inprocess_progress_watchdog_interval ............ 1.0
246
+ inprocess_restart ............................... False
247
+ inprocess_soft_timeout .......................... 60
248
+ inprocess_termination_grace_time ................ 1
249
+ is_hybrid_model ................................. False
250
+ iter_per_epoch .................................. 1250
251
+ iterations_to_skip .............................. []
252
+ keep_fp8_transpose_cache_when_using_custom_fsdp . False
253
+ kitchen_config_file ............................. None
254
+ kitchen_recipe_number ........................... None
255
+ kv_channels ..................................... 64
256
+ kv_lora_rank .................................... 512
257
+ lazy_mpu_init ................................... None
258
+ load ............................................ None
259
+ load_main_params_from_ckpt ...................... None
260
+ load_model_opt_format ........................... False
261
+ local_rank ...................................... 0
262
+ log_energy ...................................... False
263
+ log_interval .................................... 1
264
+ log_loss_scale_to_tensorboard ................... True
265
+ log_memory_to_tensorboard ....................... False
266
+ log_num_zeros_in_grad ........................... False
267
+ log_params_norm ................................. False
268
+ log_progress .................................... False
269
+ log_straggler ................................... False
270
+ log_throughput .................................. True
271
+ log_timers_to_tensorboard ....................... False
272
+ log_validation_ppl_to_tensorboard ............... False
273
+ log_world_size_to_tensorboard ................... False
274
+ logging_level ................................... None
275
+ loss_scale ...................................... None
276
+ loss_scale_window ............................... 1000
277
+ lr .............................................. 0.0001
278
+ lr_decay_iters .................................. 8
279
+ lr_decay_samples ................................ None
280
+ lr_decay_style .................................. cosine
281
+ lr_warmup_fraction .............................. None
282
+ lr_warmup_init .................................. 0.0
283
+ lr_warmup_iters ................................. 5
284
+ lr_warmup_samples ............................... 0
285
+ lr_wsd_decay_iters .............................. None
286
+ lr_wsd_decay_samples ............................ None
287
+ lr_wsd_decay_style .............................. exponential
288
+ main_grads_dtype ................................ torch.float32
289
+ main_params_dtype ............................... torch.float32
290
+ make_vocab_size_divisible_by .................... 128
291
+ mamba_head_dim .................................. 64
292
+ mamba_num_groups ................................ 8
293
+ mamba_num_heads ................................. None
294
+ mamba_state_dim ................................. 128
295
+ manual_gc ....................................... False
296
+ manual_gc_eval .................................. True
297
+ manual_gc_interval .............................. 0
298
+ mask_factor ..................................... 1.0
299
+ mask_prob ....................................... 0.15
300
+ mask_type ....................................... random
301
+ masked_softmax_fusion ........................... False
302
+ max_position_embeddings ......................... 4096
303
+ max_tokens_to_oom ............................... 12000
304
+ memory_snapshot_path ............................ snapshot.pickle
305
+ merge_file ...................................... /workspace/data/wikitext/gpt2-merges.txt
306
+ micro_batch_size ................................ 1
307
+ microbatch_group_size_per_vp_stage .............. None
308
+ mid_level_dataset_surplus ....................... 0.005
309
+ min_loss_scale .................................. 1.0
310
+ min_lr .......................................... 1e-05
311
+ mlp_chunks_for_prefill .......................... 1
312
+ mmap_bin_files .................................. True
313
+ mock_data ....................................... False
314
+ moe_apply_probs_on_input ........................ False
315
+ moe_aux_loss_coeff .............................. 0.0
316
+ moe_deepep_num_sms .............................. 20
317
+ moe_echo_enable_random_offloading ............... False
318
+ moe_echo_expert_dispatch_overlap ................ True
319
+ moe_echo_expert_dispatcher_type ................. hybridep
320
+ moe_echo_recompute_expert_dispatch .............. False
321
+ moe_enable_deepep ............................... False
322
+ moe_enable_echo ................................. True
323
+ moe_enable_hybridep ............................. True
324
+ moe_expert_capacity_factor ...................... None
325
+ moe_extended_tp ................................. False
326
+ moe_ffn_hidden_size ............................. 2048
327
+ moe_grouped_gemm ................................ True
328
+ moe_hybridep_num_sms ............................ 16
329
+ moe_input_jitter_eps ............................ None
330
+ moe_layer_freq .................................. [0, 1]
331
+ moe_layer_recompute ............................. False
332
+ moe_num_echo_experts ............................ 384
333
+ moe_pad_expert_input_to_capacity ................ False
334
+ moe_per_layer_logging ........................... False
335
+ moe_permute_fusion .............................. False
336
+ moe_received_token_capacity ..................... 1.17578125
337
+ moe_router_bias_update_rate ..................... 0.001
338
+ moe_router_dtype ................................ fp32
339
+ moe_router_enable_expert_bias ................... True
340
+ moe_router_force_load_balancing ................. False
341
+ moe_router_group_topk ........................... 1
342
+ moe_router_load_balancing_type .................. seq_aux_loss
343
+ moe_router_num_groups ........................... 1
344
+ moe_router_padding_for_fp8 ...................... False
345
+ moe_router_pre_softmax .......................... False
346
+ moe_router_score_function ....................... sigmoid
347
+ moe_router_topk ................................. 8
348
+ moe_router_topk_scaling_factor .................. 2.827
349
+ moe_shared_expert_intermediate_size ............. 2048
350
+ moe_shared_expert_overlap ....................... False
351
+ moe_token_dispatcher_type ....................... flex
352
+ moe_token_drop_policy ........................... probs
353
+ moe_upcycling_granularity ....................... 1
354
+ moe_use_device_initiated_grouped_gemm ........... True
355
+ moe_use_legacy_grouped_gemm ..................... False
356
+ moe_use_upcycling ............................... False
357
+ moe_z_loss_coeff ................................ None
358
+ mrope_section ................................... None
359
+ mscale .......................................... 1.0
360
+ mscale_all_dim .................................. 1.0
361
+ mtp_loss_scaling_factor ......................... 0.1
362
+ mtp_num_layers .................................. None
363
+ multi_latent_attention .......................... True
364
+ nccl_all_reduce_for_prefill ..................... False
365
+ nccl_communicator_config_path ................... None
366
+ nccl_ub ......................................... False
367
+ no_load_optim ................................... None
368
+ no_load_rng ..................................... None
369
+ no_persist_layer_norm ........................... False
370
+ no_rope_freq .................................... None
371
+ no_save_optim ................................... None
372
+ no_save_rng ..................................... None
373
+ non_persistent_ckpt_type ........................ None
374
+ non_persistent_global_ckpt_dir .................. None
375
+ non_persistent_local_ckpt_algo .................. fully_parallel
376
+ non_persistent_local_ckpt_dir ................... None
377
+ non_persistent_save_interval .................... None
378
+ norm_epsilon .................................... 1e-06
379
+ normalization ................................... RMSNorm
380
+ num_attention_heads ............................. 64
381
+ num_channels .................................... 3
382
+ num_classes ..................................... 1000
383
+ num_dataset_builder_threads ..................... 1
384
+ num_distributed_optimizer_instances ............. 1
385
+ num_experts ..................................... 384
386
+ num_layers ...................................... 2
387
+ num_layers_at_end_in_bf16 ....................... 1
388
+ num_layers_at_start_in_bf16 ..................... 1
389
+ num_layers_per_virtual_pipeline_stage ........... None
390
+ num_query_groups ................................ 1
391
+ num_virtual_stages_per_pipeline_rank ............ None
392
+ num_workers ..................................... 2
393
+ object_storage_cache_path ....................... None
394
+ one_logger_async ................................ False
395
+ one_logger_project .............................. megatron-lm
396
+ one_logger_run_name ............................. None
397
+ onnx_safe ....................................... None
398
+ openai_gelu ..................................... False
399
+ optimizer ....................................... adam
400
+ optimizer_cpu_offload ........................... False
401
+ optimizer_offload_fraction ...................... 1.0
402
+ output_bert_embeddings .......................... False
403
+ overlap_cpu_optimizer_d2h_h2d ................... False
404
+ overlap_grad_reduce ............................. True
405
+ overlap_moe_expert_parallel_comm ................ False
406
+ overlap_p2p_comm ................................ False
407
+ overlap_p2p_comm_warmup_flush ................... False
408
+ overlap_param_gather ............................ True
409
+ overlap_param_gather_with_optimizer_step ........ False
410
+ override_opt_param_scheduler .................... False
411
+ params_dtype .................................... torch.bfloat16
412
+ patch_dim ....................................... 16
413
+ per_split_data_args_path ........................ None
414
+ perform_initialization .......................... True
415
+ pin_cpu_grads ................................... True
416
+ pin_cpu_params .................................. True
417
+ pipeline_model_parallel_comm_backend ............ None
418
+ pipeline_model_parallel_layout .................. None
419
+ pipeline_model_parallel_size .................... 1
420
+ position_embedding_type ......................... rope
421
+ pretrained_checkpoint ........................... None
422
+ profile ......................................... False
423
+ profile_ranks ................................... [0]
424
+ profile_step_end ................................ 12
425
+ profile_step_start .............................. 10
426
+ q_lora_rank ..................................... 1536
427
+ qk_head_dim ..................................... 128
428
+ qk_l2_norm ...................................... False
429
+ qk_layernorm .................................... True
430
+ qk_pos_emb_head_dim ............................. 64
431
+ query_in_block_prob ............................. 0.1
432
+ rampup_batch_size ............................... None
433
+ rank ............................................ 0
434
+ recompute_granularity ........................... None
435
+ recompute_method ................................ None
436
+ recompute_modules ............................... None
437
+ recompute_num_layers ............................ None
438
+ record_memory_history ........................... False
439
+ relative_attention_max_distance ................. 128
440
+ relative_attention_num_buckets .................. 32
441
+ replication ..................................... False
442
+ replication_factor .............................. 2
443
+ replication_jump ................................ None
444
+ rerun_mode ...................................... validate_results
445
+ reset_attention_mask ............................ False
446
+ reset_position_ids .............................. False
447
+ result_rejected_tracker_filename ................ None
448
+ retriever_report_topk_accuracies ................ []
449
+ retriever_score_scaling ......................... False
450
+ retriever_seq_length ............................ 256
451
+ retro_add_retriever ............................. False
452
+ retro_attention_gate ............................ 1
453
+ retro_cyclic_train_iters ........................ None
454
+ retro_encoder_attention_dropout ................. 0.1
455
+ retro_encoder_hidden_dropout .................... 0.1
456
+ retro_encoder_layers ............................ 2
457
+ retro_num_neighbors ............................. 2
458
+ retro_num_retrieved_chunks ...................... 2
459
+ retro_project_dir ............................... None
460
+ retro_verify_neighbor_count ..................... True
461
+ reuse_grad_buf_for_mxfp8_param_ag ............... False
462
+ rope_scaling_factor ............................. 8.0
463
+ rope_type ....................................... None
464
+ rotary_base ..................................... 50000
465
+ rotary_interleaved .............................. False
466
+ rotary_percent .................................. 1.0
467
+ rotary_scaling_factor ........................... 32.0
468
+ rotary_seq_len_interpolation_factor ............. None
469
+ run_workload_inspector_server ................... False
470
+ sample_rate ..................................... 1.0
471
+ save ............................................ None
472
+ save_interval ................................... None
473
+ save_retain_interval ............................ None
474
+ scatter_gather_tensors_in_pipeline .............. True
475
+ seed ............................................ 1234
476
+ seq_length ...................................... 4096
477
+ sequence_parallel ............................... False
478
+ sft ............................................. False
479
+ sft_tokenizer_prompt_format ..................... nemotron-h-aligned
480
+ sgd_momentum .................................... 0.9
481
+ short_seq_prob .................................. 0.1
482
+ skip_train ...................................... False
483
+ skipped_train_samples ........................... 0
484
+ spec ............................................ None
485
+ split ........................................... 100,0,0
486
+ squared_relu .................................... False
487
+ start_weight_decay .............................. 0.1
488
+ straggler_ctrlr_port ............................ 65535
489
+ straggler_minmax_count .......................... 1
490
+ suggested_communication_unit_size ............... None
491
+ swiglu .......................................... True
492
+ swin_backbone_type .............................. tiny
493
+ symmetric_ar_type ............................... None
494
+ te_rng_tracker .................................. True
495
+ tensor_model_parallel_size ...................... 1
496
+ tensorboard_dir ................................. None
497
+ tensorboard_log_interval ........................ 1
498
+ tensorboard_queue_size .......................... 1000
499
+ test_data_path .................................. None
500
+ test_mode ....................................... False
501
+ tiktoken_num_special_tokens ..................... 1000
502
+ tiktoken_pattern ................................ None
503
+ tiktoken_special_tokens ......................... None
504
+ timing_log_level ................................ 0
505
+ timing_log_option ............................... minmax
506
+ titles_data_path ................................ None
507
+ tokenizer_model ................................. None
508
+ tokenizer_type .................................. GPT2BPETokenizer
509
+ torch_fsdp2_reshard_after_forward ............... True
510
+ tp_comm_bootstrap_backend ....................... nccl
511
+ tp_comm_bulk_dgrad .............................. True
512
+ tp_comm_bulk_wgrad .............................. True
513
+ tp_comm_overlap ................................. False
514
+ tp_comm_overlap_ag .............................. True
515
+ tp_comm_overlap_cfg ............................. None
516
+ tp_comm_overlap_rs .............................. True
517
+ tp_comm_overlap_rs_dgrad ........................ False
518
+ tp_comm_split_ag ................................ True
519
+ tp_comm_split_rs ................................ True
520
+ train_data_path ................................. None
521
+ train_iters ..................................... 12
522
+ train_samples ................................... None
523
+ train_sync_interval ............................. None
524
+ transformer_impl ................................ transformer_engine
525
+ transformer_pipeline_model_parallel_size ........ 1
526
+ untie_embeddings_and_output_weights ............. True
527
+ use_checkpoint_args ............................. False
528
+ use_checkpoint_opt_param_scheduler .............. False
529
+ use_cpu_initialization .......................... None
530
+ use_custom_fsdp ................................. False
531
+ use_dist_ckpt ................................... True
532
+ use_dist_ckpt_deprecated ........................ False
533
+ use_distributed_optimizer ....................... True
534
+ use_flash_attn .................................. False
535
+ use_legacy_models ............................... False
536
+ use_mp_args_from_checkpoint_args ................ False
537
+ use_one_sent_docs ............................... False
538
+ use_persistent_ckpt_worker ...................... False
539
+ use_precision_aware_optimizer ................... False
540
+ use_pytorch_profiler ............................ False
541
+ use_ring_exchange_p2p ........................... False
542
+ use_rope_scaling ................................ False
543
+ use_rotary_position_embeddings .................. False
544
+ use_sharp ....................................... False
545
+ use_tokenizer_model_from_checkpoint_args ........ True
546
+ use_torch_fsdp2 ................................. False
547
+ use_torch_optimizer_for_cpu_offload ............. False
548
+ use_tp_pp_dp_mapping ............................ False
549
+ v_head_dim ...................................... 128
550
+ valid_data_path ................................. None
551
+ variable_seq_lengths ............................ False
552
+ virtual_pipeline_model_parallel_size ............ None
553
+ vision_backbone_type ............................ vit
554
+ vision_pretraining .............................. False
555
+ vision_pretraining_type ......................... classify
556
+ vocab_extra_ids ................................. 0
557
+ vocab_file ...................................... /workspace/data/wikitext/gpt2-vocab.json
558
+ vocab_size ...................................... None
559
+ wandb_exp_name ..................................
560
+ wandb_project ...................................
561
+ wandb_save_dir ..................................
562
+ weight_decay .................................... 0.1
563
+ weight_decay_incr_style ......................... constant
564
+ wgrad_deferral_limit ............................ 0
565
+ world_size ...................................... 8
566
+ yaml_cfg ........................................ None
567
+ -------------------- end of arguments ---------------------
568
+ INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 1
569
+ > building GPT2BPETokenizer tokenizer ...
570
+ > padded vocab (size: 50257) with 47 dummy tokens (new size: 50304)
571
+ WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.VALIDATE_RESULTS
572
+ > initializing torch distributed ...
573
+ /workspace/Megatron-MoE-Echo/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
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
576
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
577
+ [Gloo] Rank [Gloo] Rank [Gloo] Rank 30[Gloo] Rank 5 is connected to is connected to 1 is connected to 77 is connected to 7 peer ranks. peer ranks. 7 peer ranks. Expected number of connected peer ranks is : Expected number of connected peer ranks is : peer ranks. Expected number of connected peer ranks is : 77Expected number of connected peer ranks is : 7
578
+
579
+ 7
580
+
581
+ [Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
582
+ [Gloo] Rank 6 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
583
+ [Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7[Gloo] Rank
584
+ 7 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
585
+ [Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
586
+ [Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
587
+ [Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
588
+ [Gloo] Rank 5 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
589
+ [Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7[Gloo] Rank
590
+ 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
591
+ [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 :
592
+ 7
593
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
594
+ > initialized tensor model parallel with size 1
595
+ > initialized pipeline model parallel with size 1
596
+ > setting random seeds to 1234 ...
597
+ [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
598
+
599
+ [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
600
+
601
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
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
+ [rank4]:[W921 07:46:12.576918168 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.
605
+ [rank5]:[W921 07:46:12.576936106 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.
606
+ [rank7]:[W921 07:46:12.576952022 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.
607
+ [rank2]:[W921 07:46:12.580279696 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
608
+ [rank1]:[W921 07:46:12.580340827 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.
609
+ [rank3]:[W921 07:46:12.582458446 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
610
+ [rank6]:[W921 07:46:12.582495363 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
611
+ > compiling dataset index builder ...
612
+ make: Entering directory '/workspace/Megatron-MoE-Echo/megatron/core/datasets'
613
+ make: Nothing to be done for 'default'.
614
+ make: Leaving directory '/workspace/Megatron-MoE-Echo/megatron/core/datasets'
615
+ >>> done with dataset index builder. Compilation time: 0.035 seconds
616
+ WARNING: constraints for invoking optimized fused softmax kernel are not met. We default back to unfused kernel invocations.
617
+ > compiling and loading fused kernels ...
618
+ [rank0]:[W921 07:46:12.738774827 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.
619
+ >>> done with compiling and loading fused kernels. Compilation time: 3.725 seconds
620
+ time to initialize megatron (seconds): 7.078
621
+ [after megatron is initialized] datetime: 2026-09-21 07:46:18
622
+ building GPT model ...
623
+ /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.
624
+ warnings.warn(
625
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
626
+ warnings.warn(
627
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
628
+ warnings.warn(
629
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
630
+ warnings.warn(
631
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
632
+ warnings.warn(
633
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
634
+ warnings.warn(
635
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
636
+ warnings.warn(
637
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
638
+ warnings.warn(
639
+ > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 3480525824
640
+ 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)
641
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 9
642
+ Params for bucket 1 (360579072 elements, 360579072 padded size):
643
+ module.output_layer.weight
644
+ Params for bucket 2 (44047360 elements, 44047360 padded size):
645
+ module.decoder.final_layernorm.weight
646
+ module.decoder.layers.1.mlp.shared_experts.linear_fc1.weight
647
+ module.decoder.layers.1.mlp.shared_experts.linear_fc2.weight
648
+ Params for bucket 3 (45163520 elements, 45163520 padded size):
649
+ module.decoder.layers.1.mlp.router.weight
650
+ module.decoder.layers.1.self_attention.linear_kv_up_proj.layer_norm_weight
651
+ module.decoder.layers.1.pre_mlp_layernorm.weight
652
+ module.decoder.layers.1.self_attention.linear_q_up_proj.layer_norm_weight
653
+ module.decoder.layers.1.self_attention.linear_kv_up_proj.weight
654
+ module.decoder.layers.1.self_attention.linear_kv_down_proj.weight
655
+ module.decoder.layers.1.self_attention.linear_q_up_proj.weight
656
+ module.decoder.layers.1.self_attention.linear_q_down_proj.weight
657
+ Params for bucket 4 (58720256 elements, 58720256 padded size):
658
+ module.decoder.layers.1.self_attention.linear_proj.weight
659
+ Params for bucket 5 (132127744 elements, 132127744 padded size):
660
+ module.decoder.layers.1.input_layernorm.weight
661
+ module.decoder.layers.0.mlp.linear_fc2.weight
662
+ Params for bucket 6 (264241152 elements, 264241152 padded size):
663
+ module.decoder.layers.0.mlp.linear_fc1.weight
664
+ Params for bucket 7 (42411008 elements, 42411008 padded size):
665
+ module.decoder.layers.0.self_attention.linear_kv_up_proj.layer_norm_weight
666
+ module.decoder.layers.0.self_attention.linear_q_up_proj.weight
667
+ module.decoder.layers.0.self_attention.linear_kv_down_proj.weight
668
+ module.decoder.layers.0.mlp.linear_fc1.layer_norm_weight
669
+ module.decoder.layers.0.self_attention.linear_kv_up_proj.weight
670
+ module.decoder.layers.0.self_attention.linear_q_up_proj.layer_norm_weight
671
+ module.decoder.layers.0.self_attention.linear_q_down_proj.weight
672
+ Params for bucket 8 (58720256 elements, 58720256 padded size):
673
+ module.decoder.layers.0.self_attention.linear_proj.weight
674
+ Params for bucket 9 (360586240 elements, 360586240 padded size):
675
+ module.decoder.layers.0.input_layernorm.weight
676
+ module.embedding.word_embeddings.weight
677
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 40
678
+ Params for bucket 1 (44040192 elements, 44040192 padded size):
679
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight47
680
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight46
681
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight45
682
+ Params for bucket 2 (44040192 elements, 44040192 padded size):
683
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight44
684
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight42
685
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight43
686
+ Params for bucket 3 (44040192 elements, 44040192 padded size):
687
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight41
688
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight39
689
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight40
690
+ Params for bucket 4 (44040192 elements, 44040192 padded size):
691
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight38
692
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight36
693
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight37
694
+ Params for bucket 5 (44040192 elements, 44040192 padded size):
695
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight35
696
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight33
697
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight34
698
+ Params for bucket 6 (44040192 elements, 44040192 padded size):
699
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight32
700
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight30
701
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight31
702
+ Params for bucket 7 (44040192 elements, 44040192 padded size):
703
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight29
704
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight27
705
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight28
706
+ Params for bucket 8 (44040192 elements, 44040192 padded size):
707
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight26
708
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight24
709
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight25
710
+ Params for bucket 9 (44040192 elements, 44040192 padded size):
711
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight23
712
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight21
713
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight22
714
+ Params for bucket 10 (44040192 elements, 44040192 padded size):
715
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight20
716
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight18
717
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight19
718
+ Params for bucket 11 (44040192 elements, 44040192 padded size):
719
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight17
720
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight16
721
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight15
722
+ Params for bucket 12 (44040192 elements, 44040192 padded size):
723
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight14
724
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight13
725
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight12
726
+ Params for bucket 13 (44040192 elements, 44040192 padded size):
727
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight11
728
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight10
729
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight9
730
+ Params for bucket 14 (44040192 elements, 44040192 padded size):
731
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight8
732
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight7
733
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight6
734
+ Params for bucket 15 (44040192 elements, 44040192 padded size):
735
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight5
736
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight4
737
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight3
738
+ Params for bucket 16 (44040192 elements, 44040192 padded size):
739
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight2
740
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight1
741
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight0
742
+ Params for bucket 17 (58720256 elements, 58720256 padded size):
743
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight47
744
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight46
745
+ Params for bucket 18 (58720256 elements, 58720256 padded size):
746
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight45
747
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight44
748
+ Params for bucket 19 (58720256 elements, 58720256 padded size):
749
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight43
750
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight42
751
+ Params for bucket 20 (58720256 elements, 58720256 padded size):
752
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight41
753
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight40
754
+ Params for bucket 21 (58720256 elements, 58720256 padded size):
755
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight39
756
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight38
757
+ Params for bucket 22 (58720256 elements, 58720256 padded size):
758
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight37
759
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight36
760
+ Params for bucket 23 (58720256 elements, 58720256 padded size):
761
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight35
762
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight34
763
+ Params for bucket 24 (58720256 elements, 58720256 padded size):
764
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight33
765
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight32
766
+ Params for bucket 25 (58720256 elements, 58720256 padded size):
767
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight31
768
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight30
769
+ Params for bucket 26 (58720256 elements, 58720256 padded size):
770
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight29
771
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight28
772
+ Params for bucket 27 (58720256 elements, 58720256 padded size):
773
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight27
774
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight26
775
+ Params for bucket 28 (58720256 elements, 58720256 padded size):
776
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight25
777
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight24
778
+ Params for bucket 29 (58720256 elements, 58720256 padded size):
779
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight23
780
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight22
781
+ Params for bucket 30 (58720256 elements, 58720256 padded size):
782
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight21
783
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight20
784
+ Params for bucket 31 (58720256 elements, 58720256 padded size):
785
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight19
786
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight18
787
+ Params for bucket 32 (58720256 elements, 58720256 padded size):
788
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight17
789
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight16
790
+ Params for bucket 33 (58720256 elements, 58720256 padded size):
791
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight15
792
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight14
793
+ Params for bucket 34 (58720256 elements, 58720256 padded size):
794
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight13
795
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight12
796
+ Params for bucket 35 (58720256 elements, 58720256 padded size):
797
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight11
798
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight10
799
+ Params for bucket 36 (58720256 elements, 58720256 padded size):
800
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight9
801
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight8
802
+ Params for bucket 37 (58720256 elements, 58720256 padded size):
803
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight7
804
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight6
805
+ Params for bucket 38 (58720256 elements, 58720256 padded size):
806
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight5
807
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight4
808
+ Params for bucket 39 (58720256 elements, 58720256 padded size):
809
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight3
810
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight2
811
+ Params for bucket 40 (58720256 elements, 58720256 padded size):
812
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight1
813
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight0
814
+ 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 0x78160246e2d0>, config_logger_dir='')
815
+ INFO:megatron.core.optimizer_param_scheduler:> learning rate decay style: cosine
816
+ [after model, optimizer, and learning rate scheduler are built] datetime: 2026-09-21 07:46:18
817
+ > building train, validation, and test datasets ...
818
+ > datasets target sizes (minimum size):
819
+ train: 96
820
+ validation: 0
821
+ test: 0
822
+ INFO:megatron.core.datasets.blended_megatron_dataset_config:Let split_matrix = [(0, 1.0), None, None]
823
+ > building train, validation, and test datasets for GPT ...
824
+ 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 0x781603186cf0>, 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)
825
+ INFO:megatron.core.datasets.indexed_dataset:Load the _IndexReader from /workspace/data/wikitext/wikitext103_gpt2_text_document.idx
826
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence lengths
827
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence pointers
828
+ INFO:megatron.core.datasets.indexed_dataset: Extract the document indices
829
+ INFO:megatron.core.datasets.indexed_dataset:> total number of sequences: 1165029
830
+ INFO:megatron.core.datasets.indexed_dataset:> total number of documents: 1801350
831
+ INFO:megatron.core.datasets.gpt_dataset:Load the GPTDataset train indices
832
+ INFO:megatron.core.datasets.gpt_dataset: Load the document index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-document_index.npy
833
+ INFO:megatron.core.datasets.gpt_dataset: Load the sample index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-sample_index.npy
834
+ INFO:megatron.core.datasets.gpt_dataset: Load the shuffle index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-shuffle_index.npy
835
+ INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 29073
836
+ > finished creating GPT datasets ...
837
+ [after dataloaders are built] datetime: 2026-09-21 07:46:18
838
+ done with setup ...
839
+ (min, max) time across ranks (ms):
840
+ model-and-optimizer-setup ......................: (335.08, 338.66)
841
+ train/valid/test-data-iterators-setup ..........: (64.34, 82.36)
842
+ training ...
843
+ Overwriting rerun_state_machine.current_iteration from -1 to 0...
844
+ [before the start of training step] datetime: 2026-09-21 07:46:18
845
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
846
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
847
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
848
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
849
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
850
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
851
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
852
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
853
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
854
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
855
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
856
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
857
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
858
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
859
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
860
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
861
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
862
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
863
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
864
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
865
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
866
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
867
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
868
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
869
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
870
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
871
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
872
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
873
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
874
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
875
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
876
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
877
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
878
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
879
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
880
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
881
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
882
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
883
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
884
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
885
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
886
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
887
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
888
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
889
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
890
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
891
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
892
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
893
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
894
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
895
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
896
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
897
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
898
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
899
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
900
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
901
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
902
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
903
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
904
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
905
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
906
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
907
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
908
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
909
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
910
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
911
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
912
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
913
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
914
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
915
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
916
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
917
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
918
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
919
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
920
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
921
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
922
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
923
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
924
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
925
+ [2026-09-21 07:46:59] iteration 1/ 12 | consumed samples: 8 | elapsed time per iteration (ms): 40959.5 | throughput per GPU (TFLOP/s/GPU): 0.9 | learning rate: 2.000000E-05 | global batch size: 8 | lm loss: 1.227922E+01 | loss scale: 1.0 | grad norm: 12.616 | number of skipped iterations: 0 | number of nan iterations: 0 |Number of parameters in transformer block in billions: 17.55
926
+
927
+ Number of parameters in embedding layers in billions: 0.72
928
+ Total number of parameters in billions: 18.28
929
+ Number of parameters in most loaded shard in billions: 18.2753
930
+ Theoretical memory footprints: weight and optimizer=130715.23 MB
931
+ [Rank 0] (after 1 iterations) memory (MB) | allocated: 53862.45263671875 | max allocated: 57609.2568359375 | reserved: 74666.0 | max reserved: 74666.0
932
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
933
+ warnings.warn(
934
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
935
+ warnings.warn(
936
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
937
+ warnings.warn(
938
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
939
+ warnings.warn(
940
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
941
+ warnings.warn(
942
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
943
+ warnings.warn(
944
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
945
+ warnings.warn(
946
+ /workspace/Megatron-MoE-Echo/megatron/core/distributed/param_and_grad_buffer.py:282: UserWarning: The next bucket's parameter all-gather operation has already been dispatched. This may be caused by a mismatch between the order of parameter registration and forward pass execution, which will hurt the communication-computation overlap performance.
947
+ warnings.warn(
948
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
949
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
950
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
951
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
952
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
953
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
954
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
955
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
956
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
957
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
958
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
959
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
960
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
961
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
962
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
963
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
964
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
965
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
966
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
967
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
968
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
969
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
970
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
971
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
972
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
973
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
974
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
975
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
976
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
977
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
978
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
979
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
980
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
981
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
982
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
983
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
984
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
985
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
986
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
987
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
988
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
989
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
990
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
991
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
992
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
993
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
994
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
995
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
996
+ [2026-09-21 07:47:00] iteration 2/ 12 | consumed samples: 16 | elapsed time per iteration (ms): 283.8 | throughput per GPU (TFLOP/s/GPU): 124.6 | learning rate: 4.000000E-05 | global batch size: 8 | lm loss: 1.227086E+01 | loss scale: 1.0 | grad norm: 12.533 | number of skipped iterations: 0 | number of nan iterations: 0 |
997
+ [2026-09-21 07:47:00] iteration 3/ 12 | consumed samples: 24 | elapsed time per iteration (ms): 128.2 | throughput per GPU (TFLOP/s/GPU): 276.0 | learning rate: 6.000000E-05 | global batch size: 8 | lm loss: 1.088333E+01 | loss scale: 1.0 | grad norm: 10.978 | number of skipped iterations: 0 | number of nan iterations: 0 |
998
+ Capture CUDA graph for training!!!
999
+ CUDA graph capture done!!!
1000
+ [2026-09-21 07:47:02] iteration 4/ 12 | consumed samples: 32 | elapsed time per iteration (ms): 1711.4 | throughput per GPU (TFLOP/s/GPU): 20.7 | learning rate: 8.000000E-05 | global batch size: 8 | lm loss: 1.448197E+01 | loss scale: 1.0 | grad norm: 72.852 | number of skipped iterations: 0 | number of nan iterations: 0 |
1001
+ [2026-09-21 07:47:02] iteration 5/ 12 | consumed samples: 40 | elapsed time per iteration (ms): 119.1 | throughput per GPU (TFLOP/s/GPU): 297.1 | learning rate: 1.000000E-04 | global batch size: 8 | lm loss: 1.808492E+01 | loss scale: 1.0 | grad norm: 107.033 | number of skipped iterations: 0 | number of nan iterations: 0 |
1002
+ [2026-09-21 07:47:02] iteration 6/ 12 | consumed samples: 48 | elapsed time per iteration (ms): 120.2 | throughput per GPU (TFLOP/s/GPU): 294.4 | learning rate: 7.750000E-05 | global batch size: 8 | lm loss: 1.488903E+01 | loss scale: 1.0 | grad norm: 83.729 | number of skipped iterations: 0 | number of nan iterations: 0 |
1003
+ [2026-09-21 07:47:02] iteration 7/ 12 | consumed samples: 56 | elapsed time per iteration (ms): 119.9 | throughput per GPU (TFLOP/s/GPU): 295.0 | learning rate: 3.250000E-05 | global batch size: 8 | lm loss: 1.142221E+01 | loss scale: 1.0 | grad norm: 26.935 | number of skipped iterations: 0 | number of nan iterations: 0 |
1004
+ [2026-09-21 07:47:02] iteration 8/ 12 | consumed samples: 64 | elapsed time per iteration (ms): 119.0 | throughput per GPU (TFLOP/s/GPU): 297.3 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 1.163188E+01 | loss scale: 1.0 | grad norm: 12.747 | number of skipped iterations: 0 | number of nan iterations: 0 |
1005
+ [2026-09-21 07:47:02] iteration 9/ 12 | consumed samples: 72 | elapsed time per iteration (ms): 119.4 | throughput per GPU (TFLOP/s/GPU): 296.3 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 1.140652E+01 | loss scale: 1.0 | grad norm: 14.164 | number of skipped iterations: 0 | number of nan iterations: 0 |
1006
+ [2026-09-21 07:47:02] iteration 10/ 12 | consumed samples: 80 | elapsed time per iteration (ms): 119.0 | throughput per GPU (TFLOP/s/GPU): 297.3 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 1.102892E+01 | loss scale: 1.0 | grad norm: 9.158 | number of skipped iterations: 0 | number of nan iterations: 0 |
1007
+ [2026-09-21 07:47:02] iteration 11/ 12 | consumed samples: 88 | elapsed time per iteration (ms): 120.4 | throughput per GPU (TFLOP/s/GPU): 293.9 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 1.082034E+01 | loss scale: 1.0 | grad norm: 9.517 | number of skipped iterations: 0 | number of nan iterations: 0 |
1008
+ [2026-09-21 07:47:03] iteration 12/ 12 | consumed samples: 96 | elapsed time per iteration (ms): 119.8 | throughput per GPU (TFLOP/s/GPU): 295.3 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 1.056247E+01 | loss scale: 1.0 | grad norm: 7.645 | number of skipped iterations: 0 | number of nan iterations: 0 |
1009
+ [after training is done] datetime: 2026-09-21 07:47:03
1010
+ [rank0]:[W921 07:47:04.144391025 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())
1011
+ W0921 07:47:04.281000 508525 torch/distributed/elastic/agent/server/api.py:719] Received 15 death signal, shutting down workers
1012
+ W0921 07:47:04.282000 508525 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 508629 closing signal SIGTERM
1013
+ W0921 07:47:04.282000 508525 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 508630 closing signal SIGTERM
1014
+ W0921 07:47:04.283000 508525 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 508631 closing signal SIGTERM
1015
+ W0921 07:47:04.285000 508525 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 508632 closing signal SIGTERM
1016
+ W0921 07:47:04.286000 508525 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 508633 closing signal SIGTERM
1017
+ W0921 07:47:04.287000 508525 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 508634 closing signal SIGTERM
1018
+ W0921 07:47:04.287000 508525 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 508635 closing signal SIGTERM
1019
+ W0921 07:47:04.289000 508525 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 508636 closing signal SIGTERM
1020
+ Traceback (most recent call last):
1021
+ File "/usr/local/bin/torchrun", line 8, in <module>
1022
+ sys.exit(main())
1023
+ ^^^^^^
1024
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper
1025
+ return f(*args, **kwargs)
1026
+ ^^^^^^^^^^^^^^^^^^
1027
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 892, in main
1028
+ run(args)
1029
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 883, in run
1030
+ elastic_launch(
1031
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 139, in __call__
1032
+ return launch_agent(self._config, self._entrypoint, list(args))
1033
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1034
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 261, in launch_agent
1035
+ result = agent.run()
1036
+ ^^^^^^^^^^^
1037
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/metrics/api.py", line 138, in wrapper
1038
+ result = f(*args, **kwargs)
1039
+ ^^^^^^^^^^^^^^^^^^
1040
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 711, in run
1041
+ result = self._invoke_run(role)
1042
+ ^^^^^^^^^^^^^^^^^^^^^^
1043
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 870, in _invoke_run
1044
+ time.sleep(monitor_interval)
1045
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/api.py", line 84, in _terminate_process_handler
1046
+ raise SignalException(f"Process {os.getpid()} got signal: {sigval}", sigval=sigval)
1047
+ torch.distributed.elastic.multiprocessing.api.SignalException: Process 508525 got signal: 15
granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/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
+ KIMI_K2_L2 echo matched 48 4096 8 1.17578125 none 1 OK 119.40 118.30 132219 /workspace/granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep1
3
+ KIMI_K2_L2 echo matched 48 4096 8 1.17578125 none 2 OK 119.57 118.30 132219 /workspace/granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep2
4
+ KIMI_K2_L2 echo matched 48 4096 8 1.17578125 none 3 OK 119.60 118.30 132219 /workspace/granularity_next_b200/tierb_models/attempts/KIMI_K2_L2/echo/tierb/tierb_gbs8/KIMI_K2_L2__echo__matched__seq4096__gbs8__cap1p17578125/rep3