AbhishekGhosh1998 commited on
Commit
89261b7
·
verified ·
1 Parent(s): d1b232d

granularity results echo/tierb @ 2026-08-26T18:29:43Z from sp-aiinfra-interactive-aghosh-b5xzj-8xb200-moe-cg-0

Browse files
Files changed (18) hide show
  1. echo/tierb/tierb_gbs8/DSV2_LITE_L24__echo__matched__seq8192__gbs8__cap1p0755208333333333/rep1/mem_samples.csv +248 -0
  2. echo/tierb/tierb_gbs8/DSV2_LITE_L24__echo__matched__seq8192__gbs8__cap1p0755208333333333/rep1/train.log +0 -0
  3. echo/tierb/tierb_gbs8/DSV2_LITE_L24__echo__matched__seq8192__gbs8__cap1p0755208333333333/rep2/mem_samples.csv +256 -0
  4. echo/tierb/tierb_gbs8/DSV2_LITE_L24__echo__matched__seq8192__gbs8__cap1p0755208333333333/rep2/run.env +61 -0
  5. echo/tierb/tierb_gbs8/DSV2_LITE_L24__echo__matched__seq8192__gbs8__cap1p0755208333333333/rep2/train.log +0 -0
  6. echo/tierb/tierb_gbs8/DSV2_LITE_L24__echo__matched__seq8192__gbs8__cap1p0755208333333333/rep3/mem_samples.csv +248 -0
  7. echo/tierb/tierb_gbs8/DSV2_LITE_L24__echo__matched__seq8192__gbs8__cap1p0755208333333333/rep3/run.env +61 -0
  8. echo/tierb/tierb_gbs8/DSV2_LITE_L24__echo__matched__seq8192__gbs8__cap1p0755208333333333/rep3/train.log +0 -0
  9. echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep1/mem_samples.csv +240 -0
  10. echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep1/run.env +61 -0
  11. echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep1/train.log +1199 -0
  12. echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep2/mem_samples.csv +224 -0
  13. echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep2/run.env +61 -0
  14. echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep2/train.log +1200 -0
  15. echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep3/mem_samples.csv +232 -0
  16. echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep3/run.env +61 -0
  17. echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep3/train.log +1198 -0
  18. echo/tierb/tierb_gbs8/results.tsv +6 -0
echo/tierb/tierb_gbs8/DSV2_LITE_L24__echo__matched__seq8192__gbs8__cap1p0755208333333333/rep1/mem_samples.csv CHANGED
@@ -6,3 +6,251 @@
6
  5, 0
7
  6, 0
8
  7, 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, 4
43
+ 2, 4
44
+ 3, 4
45
+ 4, 4
46
+ 5, 4
47
+ 6, 4
48
+ 7, 4
49
+ 0, 821
50
+ 1, 821
51
+ 2, 821
52
+ 3, 821
53
+ 4, 821
54
+ 5, 821
55
+ 6, 821
56
+ 7, 821
57
+ 0, 1955
58
+ 1, 1955
59
+ 2, 1955
60
+ 3, 1955
61
+ 4, 1955
62
+ 5, 1955
63
+ 6, 1955
64
+ 7, 1955
65
+ 0, 5821
66
+ 1, 6027
67
+ 2, 6039
68
+ 3, 5857
69
+ 4, 5881
70
+ 5, 6063
71
+ 6, 6015
72
+ 7, 6297
73
+ 0, 26775
74
+ 1, 26811
75
+ 2, 26671
76
+ 3, 26617
77
+ 4, 26579
78
+ 5, 26709
79
+ 6, 26745
80
+ 7, 26767
81
+ 0, 36451
82
+ 1, 36533
83
+ 2, 36411
84
+ 3, 36361
85
+ 4, 36371
86
+ 5, 36395
87
+ 6, 36503
88
+ 7, 36525
89
+ 0, 37785
90
+ 1, 37869
91
+ 2, 37719
92
+ 3, 37717
93
+ 4, 37717
94
+ 5, 37715
95
+ 6, 37851
96
+ 7, 37869
97
+ 0, 37873
98
+ 1, 37957
99
+ 2, 37807
100
+ 3, 37805
101
+ 4, 37805
102
+ 5, 37803
103
+ 6, 37939
104
+ 7, 37957
105
+ 0, 37875
106
+ 1, 37959
107
+ 2, 37809
108
+ 3, 37807
109
+ 4, 37807
110
+ 5, 37805
111
+ 6, 37941
112
+ 7, 37959
113
+ 0, 38821
114
+ 1, 38905
115
+ 2, 38755
116
+ 3, 38755
117
+ 4, 38755
118
+ 5, 38751
119
+ 6, 38887
120
+ 7, 38905
121
+ 0, 38821
122
+ 1, 38905
123
+ 2, 38755
124
+ 3, 38755
125
+ 4, 38755
126
+ 5, 38751
127
+ 6, 38887
128
+ 7, 38905
129
+ 0, 38827
130
+ 1, 38911
131
+ 2, 38761
132
+ 3, 38761
133
+ 4, 38761
134
+ 5, 38757
135
+ 6, 38893
136
+ 7, 38911
137
+ 0, 38829
138
+ 1, 38913
139
+ 2, 38763
140
+ 3, 38763
141
+ 4, 38763
142
+ 5, 38759
143
+ 6, 38895
144
+ 7, 38913
145
+ 0, 111963
146
+ 1, 112059
147
+ 2, 111927
148
+ 3, 111917
149
+ 4, 111879
150
+ 5, 111891
151
+ 6, 112031
152
+ 7, 112045
153
+ 0, 112443
154
+ 1, 112529
155
+ 2, 112377
156
+ 3, 112377
157
+ 4, 112377
158
+ 5, 112375
159
+ 6, 112509
160
+ 7, 112529
161
+ 0, 112443
162
+ 1, 112529
163
+ 2, 112377
164
+ 3, 112377
165
+ 4, 112377
166
+ 5, 112375
167
+ 6, 112509
168
+ 7, 112529
169
+ 0, 112443
170
+ 1, 112529
171
+ 2, 112377
172
+ 3, 112377
173
+ 4, 112377
174
+ 5, 112375
175
+ 6, 112509
176
+ 7, 112529
177
+ 0, 112443
178
+ 1, 112529
179
+ 2, 112377
180
+ 3, 112377
181
+ 4, 112377
182
+ 5, 112375
183
+ 6, 112509
184
+ 7, 112529
185
+ 0, 113127
186
+ 1, 113213
187
+ 2, 113061
188
+ 3, 113061
189
+ 4, 113061
190
+ 5, 113059
191
+ 6, 113193
192
+ 7, 113213
193
+ 0, 116353
194
+ 1, 116441
195
+ 2, 116287
196
+ 3, 116287
197
+ 4, 116287
198
+ 5, 116285
199
+ 6, 116419
200
+ 7, 116441
201
+ 0, 116957
202
+ 1, 117085
203
+ 2, 116911
204
+ 3, 116919
205
+ 4, 116943
206
+ 5, 116911
207
+ 6, 117073
208
+ 7, 117085
209
+ 0, 94701
210
+ 1, 88371
211
+ 2, 92559
212
+ 3, 92553
213
+ 4, 94697
214
+ 5, 92315
215
+ 6, 98971
216
+ 7, 92657
217
+ 0, 151295
218
+ 1, 151305
219
+ 2, 151201
220
+ 3, 151201
221
+ 4, 151201
222
+ 5, 151153
223
+ 6, 151361
224
+ 7, 151305
225
+ 0, 151295
226
+ 1, 151305
227
+ 2, 151201
228
+ 3, 151201
229
+ 4, 151201
230
+ 5, 151153
231
+ 6, 151361
232
+ 7, 151305
233
+ 0, 151295
234
+ 1, 151305
235
+ 2, 151201
236
+ 3, 151201
237
+ 4, 151201
238
+ 5, 151153
239
+ 6, 151361
240
+ 7, 151305
241
+ 0, 150347
242
+ 1, 2058
243
+ 2, 1034
244
+ 3, 150253
245
+ 4, 2314
246
+ 5, 150205
247
+ 6, 2314
248
+ 7, 150357
249
+ 0, 0
250
+ 1, 0
251
+ 2, 0
252
+ 3, 0
253
+ 4, 0
254
+ 5, 0
255
+ 6, 0
256
+ 7, 0
echo/tierb/tierb_gbs8/DSV2_LITE_L24__echo__matched__seq8192__gbs8__cap1p0755208333333333/rep1/train.log CHANGED
The diff for this file is too large to render. See raw diff
 
echo/tierb/tierb_gbs8/DSV2_LITE_L24__echo__matched__seq8192__gbs8__cap1p0755208333333333/rep2/mem_samples.csv ADDED
@@ -0,0 +1,256 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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, 2
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, 4
43
+ 2, 4
44
+ 3, 4
45
+ 4, 4
46
+ 5, 4
47
+ 6, 4
48
+ 7, 4
49
+ 0, 821
50
+ 1, 821
51
+ 2, 821
52
+ 3, 821
53
+ 4, 821
54
+ 5, 821
55
+ 6, 819
56
+ 7, 821
57
+ 0, 1537
58
+ 1, 1537
59
+ 2, 1537
60
+ 3, 1537
61
+ 4, 1537
62
+ 5, 1537
63
+ 6, 1537
64
+ 7, 1537
65
+ 0, 2471
66
+ 1, 2471
67
+ 2, 2471
68
+ 3, 2471
69
+ 4, 2471
70
+ 5, 2471
71
+ 6, 2471
72
+ 7, 2471
73
+ 0, 26621
74
+ 1, 26701
75
+ 2, 26499
76
+ 3, 26491
77
+ 4, 26523
78
+ 5, 26525
79
+ 6, 26677
80
+ 7, 26719
81
+ 0, 36435
82
+ 1, 36509
83
+ 2, 36387
84
+ 3, 36395
85
+ 4, 36381
86
+ 5, 36397
87
+ 6, 36493
88
+ 7, 36519
89
+ 0, 37785
90
+ 1, 37869
91
+ 2, 37719
92
+ 3, 37717
93
+ 4, 37717
94
+ 5, 37715
95
+ 6, 37851
96
+ 7, 37869
97
+ 0, 37873
98
+ 1, 37957
99
+ 2, 37807
100
+ 3, 37805
101
+ 4, 37805
102
+ 5, 37803
103
+ 6, 37939
104
+ 7, 37957
105
+ 0, 38117
106
+ 1, 38191
107
+ 2, 38017
108
+ 3, 38063
109
+ 4, 38021
110
+ 5, 38033
111
+ 6, 38175
112
+ 7, 38161
113
+ 0, 38821
114
+ 1, 38905
115
+ 2, 38755
116
+ 3, 38755
117
+ 4, 38755
118
+ 5, 38751
119
+ 6, 38887
120
+ 7, 38905
121
+ 0, 38821
122
+ 1, 38905
123
+ 2, 38755
124
+ 3, 38755
125
+ 4, 38755
126
+ 5, 38751
127
+ 6, 38887
128
+ 7, 38905
129
+ 0, 38829
130
+ 1, 38913
131
+ 2, 38763
132
+ 3, 38763
133
+ 4, 38763
134
+ 5, 38759
135
+ 6, 38895
136
+ 7, 38913
137
+ 0, 38829
138
+ 1, 38913
139
+ 2, 38763
140
+ 3, 38763
141
+ 4, 38763
142
+ 5, 38759
143
+ 6, 38895
144
+ 7, 38913
145
+ 0, 111991
146
+ 1, 112065
147
+ 2, 111957
148
+ 3, 111973
149
+ 4, 111917
150
+ 5, 111911
151
+ 6, 112059
152
+ 7, 112087
153
+ 0, 112443
154
+ 1, 112529
155
+ 2, 112377
156
+ 3, 112377
157
+ 4, 112377
158
+ 5, 112375
159
+ 6, 112509
160
+ 7, 112529
161
+ 0, 112443
162
+ 1, 112529
163
+ 2, 112377
164
+ 3, 112377
165
+ 4, 112377
166
+ 5, 112375
167
+ 6, 112509
168
+ 7, 112529
169
+ 0, 112443
170
+ 1, 112529
171
+ 2, 112377
172
+ 3, 112377
173
+ 4, 112377
174
+ 5, 112375
175
+ 6, 112509
176
+ 7, 112529
177
+ 0, 112443
178
+ 1, 112529
179
+ 2, 112377
180
+ 3, 112377
181
+ 4, 112377
182
+ 5, 112375
183
+ 6, 112509
184
+ 7, 112529
185
+ 0, 115705
186
+ 1, 115791
187
+ 2, 115639
188
+ 3, 115639
189
+ 4, 115639
190
+ 5, 115637
191
+ 6, 115771
192
+ 7, 115791
193
+ 0, 116353
194
+ 1, 116441
195
+ 2, 116287
196
+ 3, 116287
197
+ 4, 116287
198
+ 5, 116285
199
+ 6, 116419
200
+ 7, 116439
201
+ 0, 117041
202
+ 1, 117121
203
+ 2, 116937
204
+ 3, 116959
205
+ 4, 116925
206
+ 5, 116945
207
+ 6, 117093
208
+ 7, 117121
209
+ 0, 99081
210
+ 1, 88359
211
+ 2, 92559
212
+ 3, 94705
213
+ 4, 94607
214
+ 5, 92499
215
+ 6, 103447
216
+ 7, 92645
217
+ 0, 151295
218
+ 1, 151305
219
+ 2, 151201
220
+ 3, 151201
221
+ 4, 151201
222
+ 5, 151153
223
+ 6, 151361
224
+ 7, 151305
225
+ 0, 151295
226
+ 1, 151305
227
+ 2, 151201
228
+ 3, 151201
229
+ 4, 151201
230
+ 5, 151153
231
+ 6, 151361
232
+ 7, 151305
233
+ 0, 151295
234
+ 1, 151305
235
+ 2, 151201
236
+ 3, 151201
237
+ 4, 151201
238
+ 5, 151153
239
+ 6, 151361
240
+ 7, 151305
241
+ 0, 2314
242
+ 1, 150357
243
+ 2, 2314
244
+ 3, 150253
245
+ 4, 150253
246
+ 5, 2058
247
+ 6, 1034
248
+ 7, 150357
249
+ 0, 0
250
+ 1, 0
251
+ 2, 0
252
+ 3, 0
253
+ 4, 0
254
+ 5, 0
255
+ 6, 0
256
+ 7, 0
echo/tierb/tierb_gbs8/DSV2_LITE_L24__echo__matched__seq8192__gbs8__cap1p0755208333333333/rep2/run.env ADDED
@@ -0,0 +1,61 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ backend=echo
2
+ config=DSV2_LITE_L24
3
+ family=tierB
4
+ repo=/workspace/Megatron-MoE-Echo
5
+ branch=bracing-regions-through-kernels
6
+ commit=9a236793
7
+ num_layers=24
8
+ hidden=2048
9
+ ffn_hidden=10944
10
+ moe_ffn_hidden=1408
11
+ num_experts=64
12
+ topk=6
13
+ experts_per_rank=8
14
+ expert_params=8650752
15
+ heads=16
16
+ kv_groups=0
17
+ kv_channels=128
18
+ qk_layernorm=1
19
+ shared_ffn=2816
20
+ shared_expert_overlap=0
21
+ mla=1
22
+ dense_first_layers=1
23
+ moe_layers=23
24
+ rotary_base=10000
25
+ rotary_scaling=40
26
+ qkv_bias=0
27
+ router_balancing=seq_aux_loss
28
+ aux_loss_coeff=1e-3
29
+ router_pre_softmax=1
30
+ router_fusion=0
31
+ router_dtype=fp32
32
+ device_initiated_grouped_gemm=1
33
+ norm_epsilon=1e-6
34
+ seq_length=8192
35
+ micro_batch_size=1
36
+ global_batch_size=8
37
+ num_microbatches=1
38
+ assignments_per_rank=49152
39
+ mean_tokens_per_expert=6144
40
+ slot_policy=matched
41
+ slots_per_rank=8
42
+ capacity_mode=static
43
+ capacity=1.0755208333333333
44
+ cuda_graph=1
45
+ probe_mode=0
46
+ async_grad_reduce=0
47
+ branch_role=timing
48
+ memory_dump_mode=none
49
+ memory_snapshot_path=
50
+ cuda_graph_warmup_steps=3
51
+ recompute=none
52
+ force_balance=0
53
+ train_iters=12
54
+ ep_size=8
55
+ echo_dispatch_overlap=1
56
+ echo_dispatch_recompute=0
57
+ ultraep_quota_min_tokens=
58
+ ultraep_weight_sync_join=
59
+ profile_enable=0
60
+ kernel_trace=0
61
+ fp8_recipe=mxfp8
echo/tierb/tierb_gbs8/DSV2_LITE_L24__echo__matched__seq8192__gbs8__cap1p0755208333333333/rep2/train.log ADDED
The diff for this file is too large to render. See raw diff
 
echo/tierb/tierb_gbs8/DSV2_LITE_L24__echo__matched__seq8192__gbs8__cap1p0755208333333333/rep3/mem_samples.csv ADDED
@@ -0,0 +1,248 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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, 2
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, 4
43
+ 2, 27
44
+ 3, 4
45
+ 4, 4
46
+ 5, 4
47
+ 6, 26
48
+ 7, 4
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, 1955
58
+ 1, 1955
59
+ 2, 1955
60
+ 3, 1955
61
+ 4, 1955
62
+ 5, 1955
63
+ 6, 1955
64
+ 7, 1955
65
+ 0, 26289
66
+ 1, 26377
67
+ 2, 26223
68
+ 3, 26223
69
+ 4, 26223
70
+ 5, 26223
71
+ 6, 26355
72
+ 7, 26377
73
+ 0, 32145
74
+ 1, 32231
75
+ 2, 32081
76
+ 3, 32079
77
+ 4, 32013
78
+ 5, 32077
79
+ 6, 32145
80
+ 7, 32231
81
+ 0, 36587
82
+ 1, 36651
83
+ 2, 36541
84
+ 3, 36529
85
+ 4, 36569
86
+ 5, 36537
87
+ 6, 36683
88
+ 7, 36701
89
+ 0, 37785
90
+ 1, 37869
91
+ 2, 37719
92
+ 3, 37717
93
+ 4, 37717
94
+ 5, 37715
95
+ 6, 37851
96
+ 7, 37869
97
+ 0, 37873
98
+ 1, 37957
99
+ 2, 37807
100
+ 3, 37805
101
+ 4, 37805
102
+ 5, 37803
103
+ 6, 37939
104
+ 7, 37957
105
+ 0, 38017
106
+ 1, 38061
107
+ 2, 37935
108
+ 3, 37941
109
+ 4, 37929
110
+ 5, 37927
111
+ 6, 38065
112
+ 7, 38067
113
+ 0, 38821
114
+ 1, 38905
115
+ 2, 38755
116
+ 3, 38755
117
+ 4, 38755
118
+ 5, 38751
119
+ 6, 38887
120
+ 7, 38905
121
+ 0, 38821
122
+ 1, 38905
123
+ 2, 38755
124
+ 3, 38755
125
+ 4, 38755
126
+ 5, 38751
127
+ 6, 38887
128
+ 7, 38905
129
+ 0, 38829
130
+ 1, 38913
131
+ 2, 38763
132
+ 3, 38763
133
+ 4, 38763
134
+ 5, 38759
135
+ 6, 38895
136
+ 7, 38913
137
+ 0, 38829
138
+ 1, 38913
139
+ 2, 38763
140
+ 3, 38763
141
+ 4, 38763
142
+ 5, 38759
143
+ 6, 38895
144
+ 7, 38913
145
+ 0, 111919
146
+ 1, 111999
147
+ 2, 111869
148
+ 3, 111833
149
+ 4, 111853
150
+ 5, 111851
151
+ 6, 111981
152
+ 7, 112005
153
+ 0, 112443
154
+ 1, 112529
155
+ 2, 112377
156
+ 3, 112377
157
+ 4, 112377
158
+ 5, 112375
159
+ 6, 112509
160
+ 7, 112529
161
+ 0, 112443
162
+ 1, 112529
163
+ 2, 112377
164
+ 3, 112377
165
+ 4, 112377
166
+ 5, 112375
167
+ 6, 112509
168
+ 7, 112529
169
+ 0, 112443
170
+ 1, 112529
171
+ 2, 112377
172
+ 3, 112377
173
+ 4, 112377
174
+ 5, 112375
175
+ 6, 112509
176
+ 7, 112529
177
+ 0, 112443
178
+ 1, 112529
179
+ 2, 112377
180
+ 3, 112377
181
+ 4, 112377
182
+ 5, 112375
183
+ 6, 112509
184
+ 7, 112529
185
+ 0, 115705
186
+ 1, 115791
187
+ 2, 115639
188
+ 3, 115639
189
+ 4, 115639
190
+ 5, 115637
191
+ 6, 115771
192
+ 7, 115791
193
+ 0, 116353
194
+ 1, 116439
195
+ 2, 116289
196
+ 3, 116287
197
+ 4, 116289
198
+ 5, 116287
199
+ 6, 116419
200
+ 7, 116439
201
+ 0, 116963
202
+ 1, 117117
203
+ 2, 116965
204
+ 3, 116987
205
+ 4, 116897
206
+ 5, 116959
207
+ 6, 117053
208
+ 7, 117067
209
+ 0, 92647
210
+ 1, 88353
211
+ 2, 92363
212
+ 3, 98993
213
+ 4, 88261
214
+ 5, 92315
215
+ 6, 92713
216
+ 7, 88369
217
+ 0, 151295
218
+ 1, 151305
219
+ 2, 151201
220
+ 3, 151201
221
+ 4, 151201
222
+ 5, 151153
223
+ 6, 151361
224
+ 7, 151305
225
+ 0, 151295
226
+ 1, 151305
227
+ 2, 151201
228
+ 3, 151201
229
+ 4, 151201
230
+ 5, 151153
231
+ 6, 151361
232
+ 7, 151305
233
+ 0, 151295
234
+ 1, 151305
235
+ 2, 151201
236
+ 3, 151201
237
+ 4, 151201
238
+ 5, 151153
239
+ 6, 151361
240
+ 7, 151305
241
+ 0, 150347
242
+ 1, 2058
243
+ 2, 778
244
+ 3, 1034
245
+ 4, 150253
246
+ 5, 2058
247
+ 6, 1034
248
+ 7, 150357
echo/tierb/tierb_gbs8/DSV2_LITE_L24__echo__matched__seq8192__gbs8__cap1p0755208333333333/rep3/run.env ADDED
@@ -0,0 +1,61 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ backend=echo
2
+ config=DSV2_LITE_L24
3
+ family=tierB
4
+ repo=/workspace/Megatron-MoE-Echo
5
+ branch=bracing-regions-through-kernels
6
+ commit=9a236793
7
+ num_layers=24
8
+ hidden=2048
9
+ ffn_hidden=10944
10
+ moe_ffn_hidden=1408
11
+ num_experts=64
12
+ topk=6
13
+ experts_per_rank=8
14
+ expert_params=8650752
15
+ heads=16
16
+ kv_groups=0
17
+ kv_channels=128
18
+ qk_layernorm=1
19
+ shared_ffn=2816
20
+ shared_expert_overlap=0
21
+ mla=1
22
+ dense_first_layers=1
23
+ moe_layers=23
24
+ rotary_base=10000
25
+ rotary_scaling=40
26
+ qkv_bias=0
27
+ router_balancing=seq_aux_loss
28
+ aux_loss_coeff=1e-3
29
+ router_pre_softmax=1
30
+ router_fusion=0
31
+ router_dtype=fp32
32
+ device_initiated_grouped_gemm=1
33
+ norm_epsilon=1e-6
34
+ seq_length=8192
35
+ micro_batch_size=1
36
+ global_batch_size=8
37
+ num_microbatches=1
38
+ assignments_per_rank=49152
39
+ mean_tokens_per_expert=6144
40
+ slot_policy=matched
41
+ slots_per_rank=8
42
+ capacity_mode=static
43
+ capacity=1.0755208333333333
44
+ cuda_graph=1
45
+ probe_mode=0
46
+ async_grad_reduce=0
47
+ branch_role=timing
48
+ memory_dump_mode=none
49
+ memory_snapshot_path=
50
+ cuda_graph_warmup_steps=3
51
+ recompute=none
52
+ force_balance=0
53
+ train_iters=12
54
+ ep_size=8
55
+ echo_dispatch_overlap=1
56
+ echo_dispatch_recompute=0
57
+ ultraep_quota_min_tokens=
58
+ ultraep_weight_sync_join=
59
+ profile_enable=0
60
+ kernel_trace=0
61
+ fp8_recipe=mxfp8
echo/tierb/tierb_gbs8/DSV2_LITE_L24__echo__matched__seq8192__gbs8__cap1p0755208333333333/rep3/train.log ADDED
The diff for this file is too large to render. See raw diff
 
echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/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, 4
41
+ 0, 4
42
+ 1, 4
43
+ 2, 4
44
+ 3, 4
45
+ 4, 4
46
+ 5, 4
47
+ 6, 4
48
+ 7, 4
49
+ 0, 821
50
+ 1, 821
51
+ 2, 821
52
+ 3, 821
53
+ 4, 821
54
+ 5, 821
55
+ 6, 821
56
+ 7, 821
57
+ 0, 1803
58
+ 1, 1803
59
+ 2, 1803
60
+ 3, 1803
61
+ 4, 1803
62
+ 5, 1803
63
+ 6, 1803
64
+ 7, 1803
65
+ 0, 41093
66
+ 1, 41093
67
+ 2, 41093
68
+ 3, 41093
69
+ 4, 41093
70
+ 5, 41093
71
+ 6, 41093
72
+ 7, 41093
73
+ 0, 43905
74
+ 1, 43335
75
+ 2, 43335
76
+ 3, 44035
77
+ 4, 44035
78
+ 5, 44035
79
+ 6, 44035
80
+ 7, 44035
81
+ 0, 44559
82
+ 1, 43991
83
+ 2, 43991
84
+ 3, 44691
85
+ 4, 44691
86
+ 5, 44691
87
+ 6, 44691
88
+ 7, 44691
89
+ 0, 47239
90
+ 1, 46671
91
+ 2, 46671
92
+ 3, 47371
93
+ 4, 47371
94
+ 5, 47371
95
+ 6, 47371
96
+ 7, 47371
97
+ 0, 47519
98
+ 1, 46951
99
+ 2, 46951
100
+ 3, 47651
101
+ 4, 47651
102
+ 5, 47651
103
+ 6, 47651
104
+ 7, 47651
105
+ 0, 47917
106
+ 1, 47349
107
+ 2, 47349
108
+ 3, 48049
109
+ 4, 48049
110
+ 5, 48049
111
+ 6, 48043
112
+ 7, 48049
113
+ 0, 48553
114
+ 1, 47985
115
+ 2, 47985
116
+ 3, 48685
117
+ 4, 48685
118
+ 5, 48685
119
+ 6, 48685
120
+ 7, 48685
121
+ 0, 48799
122
+ 1, 48231
123
+ 2, 48231
124
+ 3, 48931
125
+ 4, 48931
126
+ 5, 48931
127
+ 6, 48931
128
+ 7, 48931
129
+ 0, 49933
130
+ 1, 49365
131
+ 2, 49365
132
+ 3, 50065
133
+ 4, 50065
134
+ 5, 50065
135
+ 6, 50065
136
+ 7, 50065
137
+ 0, 49933
138
+ 1, 49365
139
+ 2, 49365
140
+ 3, 50065
141
+ 4, 50065
142
+ 5, 50065
143
+ 6, 50065
144
+ 7, 50065
145
+ 0, 77603
146
+ 1, 77619
147
+ 2, 77649
148
+ 3, 77629
149
+ 4, 77631
150
+ 5, 77601
151
+ 6, 77611
152
+ 7, 77645
153
+ 0, 78077
154
+ 1, 78061
155
+ 2, 78059
156
+ 3, 78081
157
+ 4, 78081
158
+ 5, 78081
159
+ 6, 78081
160
+ 7, 78081
161
+ 0, 78427
162
+ 1, 78411
163
+ 2, 78409
164
+ 3, 78431
165
+ 4, 78431
166
+ 5, 78431
167
+ 6, 78431
168
+ 7, 78431
169
+ 0, 78427
170
+ 1, 78411
171
+ 2, 78409
172
+ 3, 78431
173
+ 4, 78431
174
+ 5, 78431
175
+ 6, 78431
176
+ 7, 78431
177
+ 0, 78427
178
+ 1, 78411
179
+ 2, 78409
180
+ 3, 78431
181
+ 4, 78431
182
+ 5, 78431
183
+ 6, 78431
184
+ 7, 78431
185
+ 0, 79529
186
+ 1, 79513
187
+ 2, 79511
188
+ 3, 79533
189
+ 4, 79545
190
+ 5, 79533
191
+ 6, 79533
192
+ 7, 79533
193
+ 0, 82709
194
+ 1, 81573
195
+ 2, 82691
196
+ 3, 82713
197
+ 4, 82993
198
+ 5, 82993
199
+ 6, 81873
200
+ 7, 81593
201
+ 0, 90017
202
+ 1, 88743
203
+ 2, 89937
204
+ 3, 89943
205
+ 4, 90169
206
+ 5, 90187
207
+ 6, 89065
208
+ 7, 88841
209
+ 0, 109745
210
+ 1, 103473
211
+ 2, 104119
212
+ 3, 105879
213
+ 4, 108941
214
+ 5, 105755
215
+ 6, 109781
216
+ 7, 103825
217
+ 0, 112291
218
+ 1, 111969
219
+ 2, 112055
220
+ 3, 112327
221
+ 4, 112327
222
+ 5, 112327
223
+ 6, 112327
224
+ 7, 112327
225
+ 0, 111397
226
+ 1, 111075
227
+ 2, 3918
228
+ 3, 111433
229
+ 4, 111433
230
+ 5, 111433
231
+ 6, 111433
232
+ 7, 111433
233
+ 0, 3662
234
+ 1, 2382
235
+ 2, 2382
236
+ 3, 2382
237
+ 4, 2382
238
+ 5, 2382
239
+ 6, 2638
240
+ 7, 111433
echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep1/run.env ADDED
@@ -0,0 +1,61 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ backend=echo
2
+ config=QWEN2_57B
3
+ family=tierB
4
+ repo=/workspace/Megatron-MoE-Echo
5
+ branch=bracing-regions-through-kernels
6
+ commit=9a236793
7
+ num_layers=10
8
+ hidden=3584
9
+ ffn_hidden=2560
10
+ moe_ffn_hidden=2560
11
+ num_experts=64
12
+ topk=8
13
+ experts_per_rank=8
14
+ expert_params=27525120
15
+ heads=28
16
+ kv_groups=4
17
+ kv_channels=
18
+ qk_layernorm=0
19
+ shared_ffn=20480
20
+ shared_expert_overlap=0
21
+ mla=0
22
+ dense_first_layers=0
23
+ moe_layers=10
24
+ rotary_base=1000000
25
+ rotary_scaling=
26
+ qkv_bias=1
27
+ router_balancing=aux_loss
28
+ aux_loss_coeff=1e-2
29
+ router_pre_softmax=1
30
+ router_fusion=0
31
+ router_dtype=fp32
32
+ device_initiated_grouped_gemm=1
33
+ norm_epsilon=1e-6
34
+ seq_length=4096
35
+ micro_batch_size=1
36
+ global_batch_size=8
37
+ num_microbatches=1
38
+ assignments_per_rank=32768
39
+ mean_tokens_per_expert=4096
40
+ slot_policy=matched
41
+ slots_per_rank=8
42
+ capacity_mode=static
43
+ capacity=1.08984375
44
+ cuda_graph=1
45
+ probe_mode=0
46
+ async_grad_reduce=0
47
+ branch_role=timing
48
+ memory_dump_mode=none
49
+ memory_snapshot_path=
50
+ cuda_graph_warmup_steps=3
51
+ recompute=none
52
+ force_balance=0
53
+ train_iters=12
54
+ ep_size=8
55
+ echo_dispatch_overlap=1
56
+ echo_dispatch_recompute=0
57
+ ultraep_quota_min_tokens=
58
+ ultraep_weight_sync_join=
59
+ profile_enable=0
60
+ kernel_trace=0
61
+ fp8_recipe=mxfp8
echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep1/train.log ADDED
@@ -0,0 +1,1199 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ==============================================================
2
+ QWEN2_57B / echo (slots/rank=8 policy=matched)
3
+ E=64 K=8 H'=2560 m=8
4
+ L=10 seq=4096 gbs=8 (1 microbatches)
5
+ cap_mode=static cap=1.08984375 cg=1 recompute=none
6
+ memory_dump=none snapshots=/workspace/tierb_study_20260826_144644/echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep1/memory_snapshots/snapshot.pickle
7
+ repo=/workspace/Megatron-MoE-Echo @ 9a236793
8
+ ==============================================================
9
+ W0826 18:23:37.850000 499283 torch/distributed/run.py:766]
10
+ W0826 18:23:37.850000 499283 torch/distributed/run.py:766] *****************************************
11
+ W0826 18:23:37.850000 499283 torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.
12
+ W0826 18:23:37.850000 499283 torch/distributed/run.py:766] *****************************************
13
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
14
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
15
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
16
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
17
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
18
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
19
+ WARNING: one_logger package is required to enable e2e metrics tracking. please go to https://confluence.nvidia.com/display/MLWFO/Package+Repositories for details to install it
20
+ /workspace/Megatron-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
+ using world size: 8, data-parallel size: 8, context-parallel size: 1, hierarchical context-parallel sizes: None, tensor-model-parallel size: 1, pipeline-model-parallel size: 1
25
+ WARNING: overriding default arguments for tokenizer_type:GPT2BPETokenizer with tokenizer_type:GPT2BPETokenizer
26
+ Number of virtual stages per pipeline stage: None
27
+ accumulate and all-reduce gradients in fp32 for bfloat16 data type.
28
+ using torch.bfloat16 for parameters ...
29
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
30
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
31
+ ------------------------ arguments ------------------------
32
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
33
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
34
+ account_for_embedding_in_pipeline_split ......... False
35
+ account_for_loss_in_pipeline_split .............. False
36
+ accumulate_allreduce_grads_in_fp32 .............. True
37
+ adam_beta1 ...................................... 0.9
38
+ adam_beta2 ...................................... 0.999
39
+ adam_eps ........................................ 1e-08
40
+ add_bias_linear ................................. False
41
+ add_position_embedding .......................... False
42
+ add_qkv_bias .................................... True
43
+ adlr_autoresume ................................. False
44
+ adlr_autoresume_interval ........................ 1000
45
+ align_grad_reduce ............................... True
46
+ align_param_gather .............................. False
47
+ app_tag_run_name ................................ None
48
+ app_tag_run_version ............................. 0.0.0
49
+ apply_layernorm_1p .............................. False
50
+ apply_query_key_layer_scaling ................... False
51
+ apply_residual_connection_post_layernorm ........ False
52
+ apply_rope_fusion ............................... True
53
+ async_save ...................................... None
54
+ async_tensor_model_parallel_allreduce ........... True
55
+ attention_backend ............................... AttnBackend.auto
56
+ attention_dropout ............................... 0.0
57
+ attention_softmax_in_fp32 ....................... False
58
+ auto_detect_ckpt_format ......................... False
59
+ barrier_with_L1_time ............................ True
60
+ bert_binary_head ................................ True
61
+ bert_embedder_type .............................. megatron
62
+ bert_load ....................................... None
63
+ bf16 ............................................ True
64
+ bias_dropout_fusion ............................. True
65
+ bias_gelu_fusion ................................ False
66
+ bias_swiglu_fusion .............................. True
67
+ biencoder_projection_dim ........................ 0
68
+ biencoder_shared_query_context_model ............ False
69
+ block_data_path ................................. None
70
+ calc_ft_timeouts ................................ False
71
+ calculate_per_token_loss ........................ False
72
+ check_for_large_grads ........................... False
73
+ check_for_nan_in_loss_and_grad .................. False
74
+ check_for_spiky_loss ............................ False
75
+ check_weight_hash_across_dp_replicas_interval ... None
76
+ ckpt_assume_constant_structure .................. False
77
+ ckpt_convert_format ............................. None
78
+ ckpt_convert_save ............................... None
79
+ ckpt_convert_update_legacy_dist_opt_format ...... False
80
+ ckpt_format ..................................... torch_dist
81
+ ckpt_fully_parallel_load ........................ False
82
+ ckpt_fully_parallel_save ........................ True
83
+ ckpt_fully_parallel_save_deprecated ............. False
84
+ ckpt_step ....................................... None
85
+ classes_fraction ................................ 1.0
86
+ clip_grad ....................................... 1.0
87
+ clone_scatter_output_in_embedding ............... True
88
+ config_logger_dir ...............................
89
+ consumed_train_samples .......................... 0
90
+ consumed_valid_samples .......................... 0
91
+ context_parallel_size ........................... 1
92
+ cp_comm_type .................................... ['p2p']
93
+ create_attention_mask_in_dataloader ............. True
94
+ cross_entropy_fusion_impl ....................... native
95
+ cross_entropy_loss_fusion ....................... False
96
+ cuda_graph_scope ................................ full_iteration
97
+ cuda_graph_warmup_steps ......................... 3
98
+ data_args_path .................................. None
99
+ data_cache_path ................................. None
100
+ data_parallel_random_init ....................... False
101
+ data_parallel_sharding_strategy ................. no_shard
102
+ data_parallel_size .............................. 8
103
+ data_path ....................................... ['/workspace/data/wikitext/wikitext103_gpt2_text_document']
104
+ data_per_class_fraction ......................... 1.0
105
+ data_sharding ................................... True
106
+ dataloader_type ................................. single
107
+ ddp_average_in_collective ....................... False
108
+ ddp_bucket_size ................................. None
109
+ ddp_num_buckets ................................. None
110
+ ddp_pad_buckets_for_high_nccl_busbw ............. False
111
+ decoder_first_pipeline_num_layers ............... None
112
+ decoder_last_pipeline_num_layers ................ None
113
+ decoder_num_layers .............................. None
114
+ decoder_seq_length .............................. None
115
+ decoupled_lr .................................... None
116
+ decoupled_min_lr ................................ None
117
+ decrease_batch_size_if_needed ................... False
118
+ defer_embedding_wgrad_compute ................... False
119
+ delay_wgrad_compute ............................. False
120
+ deprecated_use_mcore_models ..................... True
121
+ deterministic_mode .............................. False
122
+ dino_bottleneck_size ............................ 256
123
+ dino_freeze_last_layer .......................... 1
124
+ dino_head_hidden_size ........................... 2048
125
+ dino_local_crops_number ......................... 10
126
+ dino_local_img_size ............................. 96
127
+ dino_norm_last_layer ............................ False
128
+ dino_teacher_temp ............................... 0.07
129
+ dino_warmup_teacher_temp ........................ 0.04
130
+ dino_warmup_teacher_temp_epochs ................. 30
131
+ disable_bf16_reduced_precision_matmul ........... False
132
+ disable_mamba_mem_eff_path ...................... False
133
+ disable_straggler_on_startup .................... False
134
+ dist_ckpt_format_deprecated ..................... None
135
+ dist_ckpt_strictness ............................ assume_ok_unexpected
136
+ distribute_saved_activations .................... False
137
+ distributed_backend ............................. nccl
138
+ distributed_timeout_minutes ..................... 10
139
+ embedding_init_method_std ....................... None
140
+ embedding_path .................................. None
141
+ empty_unused_memory_level ....................... 0
142
+ enable_cuda_graph ............................... True
143
+ enable_experimental ............................. False
144
+ enable_ft_package ............................... False
145
+ enable_gloo_process_groups ...................... True
146
+ enable_msc ...................................... True
147
+ enable_one_logger ............................... True
148
+ encoder_num_layers .............................. 10
149
+ encoder_seq_length .............................. 4096
150
+ end_weight_decay ................................ 0.1
151
+ eod_mask_loss ................................... False
152
+ error_injection_rate ............................ 0
153
+ error_injection_type ............................ transient_error
154
+ eval_interval ................................... 1000000
155
+ eval_iters ...................................... 0
156
+ evidence_data_path .............................. None
157
+ exit_duration_in_mins ........................... None
158
+ exit_interval ................................... None
159
+ exit_on_missing_checkpoint ...................... False
160
+ exit_signal_handler ............................. False
161
+ exp_avg_dtype ................................... torch.float32
162
+ exp_avg_sq_dtype ................................ torch.float32
163
+ expert_model_parallel_size ...................... 8
164
+ expert_tensor_parallel_size ..................... 1
165
+ export_force_local_attention .................... False
166
+ export_kd_cfg ................................... None
167
+ export_kd_teacher_ckpt_format ................... None
168
+ export_kd_teacher_load .......................... None
169
+ export_kv_cache_quant ........................... False
170
+ export_legacy_megatron .......................... False
171
+ export_model_type ............................... GPTModel
172
+ export_moe_apply_probs_on_input ................. False
173
+ export_qk_l2_norm ............................... False
174
+ export_quant_cfg ................................ None
175
+ export_real_quant_cfg ........................... None
176
+ export_te_mcore_model ........................... False
177
+ external_cuda_graph ............................. False
178
+ ffn_hidden_size ................................. 2560
179
+ finetune ........................................ False
180
+ finetune_data_split ............................. train
181
+ finetune_hf_dataset ............................. None
182
+ first_last_layers_bf16 .......................... False
183
+ flash_decode .................................... False
184
+ fp16 ............................................ False
185
+ fp16_lm_cross_entropy ........................... False
186
+ fp32_residual_connection ........................ False
187
+ fp8 ............................................. e4m3
188
+ fp8_amax_compute_algo ........................... most_recent
189
+ fp8_amax_history_len ............................ 1
190
+ fp8_interval .................................... 1
191
+ fp8_margin ...................................... 0
192
+ fp8_param_gather ................................ False
193
+ fp8_recipe ...................................... mxfp8
194
+ fp8_wgrad ....................................... True
195
+ fsdp_double_buffer .............................. False
196
+ global_batch_size ............................... 8
197
+ grad_reduce_in_bf16 ............................. False
198
+ gradient_accumulation_fusion .................... True
199
+ gradient_reduce_div_fusion ...................... True
200
+ group_query_attention ........................... True
201
+ head_lr_mult .................................... 1.0
202
+ heterogeneous_layers_config_encoded_json ........ None
203
+ heterogeneous_layers_config_path ................ None
204
+ hidden_dropout .................................. 0.0
205
+ hidden_size ..................................... 3584
206
+ hierarchical_context_parallel_sizes ............. None
207
+ high_priority_stream_groups ..................... []
208
+ hybrid_attention_ratio .......................... 0.0
209
+ hybrid_mlp_ratio ................................ 0.0
210
+ hybrid_override_pattern ......................... None
211
+ hysteresis ...................................... 2
212
+ ict_head_size ................................... None
213
+ ict_load ........................................ None
214
+ img_h ........................................... 224
215
+ img_w ........................................... 224
216
+ indexer_batch_size .............................. 128
217
+ indexer_log_interval ............................ 1000
218
+ inference_batch_times_seqlen_threshold .......... -1
219
+ inference_dynamic_batching ...................... False
220
+ inference_dynamic_batching_buffer_guaranteed_fraction 0.2
221
+ inference_dynamic_batching_buffer_overflow_factor None
222
+ inference_dynamic_batching_buffer_size_gb ....... 40.0
223
+ inference_dynamic_batching_chunk_size ........... 256
224
+ inference_dynamic_batching_max_requests_override None
225
+ inference_dynamic_batching_max_tokens_override .. None
226
+ inference_dynamic_batching_num_cuda_graphs ...... 16
227
+ inference_max_batch_size ........................ 8
228
+ inference_max_seq_length ........................ 2560
229
+ inference_rng_tracker ........................... False
230
+ init_method_std ................................. 0.01
231
+ init_method_xavier_uniform ...................... False
232
+ init_model_with_meta_device ..................... False
233
+ initial_loss_scale .............................. 4294967296
234
+ inprocess_active_world_size ..................... 8
235
+ inprocess_barrier_timeout ....................... 120
236
+ inprocess_completion_timeout .................... 120
237
+ inprocess_empty_cuda_cache ...................... False
238
+ inprocess_granularity ........................... node
239
+ inprocess_hard_timeout .......................... 90
240
+ inprocess_heartbeat_interval .................... 30
241
+ inprocess_heartbeat_timeout ..................... 60
242
+ inprocess_last_call_wait ........................ 1
243
+ inprocess_max_iterations ........................ None
244
+ inprocess_monitor_process_interval .............. 1.0
245
+ inprocess_monitor_thread_interval ............... 1.0
246
+ inprocess_progress_watchdog_interval ............ 1.0
247
+ inprocess_restart ............................... False
248
+ inprocess_soft_timeout .......................... 60
249
+ inprocess_termination_grace_time ................ 1
250
+ is_hybrid_model ................................. False
251
+ iter_per_epoch .................................. 1250
252
+ iterations_to_skip .............................. []
253
+ keep_fp8_transpose_cache_when_using_custom_fsdp . False
254
+ kitchen_config_file ............................. None
255
+ kitchen_recipe_number ........................... None
256
+ kv_channels ..................................... 128
257
+ kv_lora_rank .................................... 32
258
+ lazy_mpu_init ................................... None
259
+ load ............................................ None
260
+ load_main_params_from_ckpt ...................... None
261
+ load_model_opt_format ........................... False
262
+ local_rank ...................................... 0
263
+ log_energy ...................................... False
264
+ log_interval .................................... 1
265
+ log_loss_scale_to_tensorboard ................... True
266
+ log_memory_to_tensorboard ....................... False
267
+ log_num_zeros_in_grad ........................... False
268
+ log_params_norm ................................. False
269
+ log_progress .................................... False
270
+ log_straggler ................................... False
271
+ log_throughput .................................. True
272
+ log_timers_to_tensorboard ....................... False
273
+ log_validation_ppl_to_tensorboard ............... False
274
+ log_world_size_to_tensorboard ................... False
275
+ logging_level ................................... None
276
+ loss_scale ...................................... None
277
+ loss_scale_window ............................... 1000
278
+ lr .............................................. 0.0001
279
+ lr_decay_iters .................................. 8
280
+ lr_decay_samples ................................ None
281
+ lr_decay_style .................................. cosine
282
+ lr_warmup_fraction .............................. None
283
+ lr_warmup_init .................................. 0.0
284
+ lr_warmup_iters ................................. 5
285
+ lr_warmup_samples ............................... 0
286
+ lr_wsd_decay_iters .............................. None
287
+ lr_wsd_decay_samples ............................ None
288
+ lr_wsd_decay_style .............................. exponential
289
+ main_grads_dtype ................................ torch.float32
290
+ main_params_dtype ............................... torch.float32
291
+ make_vocab_size_divisible_by .................... 128
292
+ mamba_head_dim .................................. 64
293
+ mamba_num_groups ................................ 8
294
+ mamba_num_heads ................................. None
295
+ mamba_state_dim ................................. 128
296
+ manual_gc ....................................... False
297
+ manual_gc_eval .................................. True
298
+ manual_gc_interval .............................. 0
299
+ mask_factor ..................................... 1.0
300
+ mask_prob ....................................... 0.15
301
+ mask_type ....................................... random
302
+ masked_softmax_fusion ........................... False
303
+ max_position_embeddings ......................... 32768
304
+ max_tokens_to_oom ............................... 12000
305
+ memory_snapshot_path ............................ snapshot.pickle
306
+ merge_file ...................................... /workspace/data/wikitext/gpt2-merges.txt
307
+ micro_batch_size ................................ 1
308
+ microbatch_group_size_per_vp_stage .............. None
309
+ mid_level_dataset_surplus ....................... 0.005
310
+ min_loss_scale .................................. 1.0
311
+ min_lr .......................................... 1e-05
312
+ mlp_chunks_for_prefill .......................... 1
313
+ mmap_bin_files .................................. True
314
+ mock_data ....................................... False
315
+ moe_apply_probs_on_input ........................ False
316
+ moe_aux_loss_coeff .............................. 0.01
317
+ moe_deepep_num_sms .............................. 20
318
+ moe_echo_enable_random_offloading ............... False
319
+ moe_echo_expert_dispatch_overlap ................ True
320
+ moe_echo_expert_dispatcher_type ................. hybridep
321
+ moe_echo_recompute_expert_dispatch .............. False
322
+ moe_enable_deepep ............................... False
323
+ moe_enable_echo ................................. True
324
+ moe_enable_hybridep ............................. True
325
+ moe_expert_capacity_factor ...................... None
326
+ moe_extended_tp ................................. False
327
+ moe_ffn_hidden_size ............................. 2560
328
+ moe_grouped_gemm ................................ True
329
+ moe_hybridep_num_sms ............................ 16
330
+ moe_input_jitter_eps ............................ None
331
+ moe_layer_freq .................................. 1
332
+ moe_layer_recompute ............................. False
333
+ moe_num_echo_experts ............................ 64
334
+ moe_pad_expert_input_to_capacity ................ False
335
+ moe_per_layer_logging ........................... False
336
+ moe_permute_fusion .............................. False
337
+ moe_received_token_capacity ..................... 1.08984375
338
+ moe_router_bias_update_rate ..................... 0.001
339
+ moe_router_dtype ................................ fp32
340
+ moe_router_enable_expert_bias ................... False
341
+ moe_router_force_load_balancing ................. False
342
+ moe_router_group_topk ........................... None
343
+ moe_router_load_balancing_type .................. aux_loss
344
+ moe_router_num_groups ........................... None
345
+ moe_router_padding_for_fp8 ...................... False
346
+ moe_router_pre_softmax .......................... True
347
+ moe_router_score_function ....................... softmax
348
+ moe_router_topk ................................. 8
349
+ moe_router_topk_scaling_factor .................. None
350
+ moe_shared_expert_intermediate_size ............. 20480
351
+ moe_shared_expert_overlap ....................... False
352
+ moe_token_dispatcher_type ....................... flex
353
+ moe_token_drop_policy ........................... probs
354
+ moe_upcycling_granularity ....................... 1
355
+ moe_use_device_initiated_grouped_gemm ........... True
356
+ moe_use_legacy_grouped_gemm ..................... False
357
+ moe_use_upcycling ............................... False
358
+ moe_z_loss_coeff ................................ None
359
+ mrope_section ................................... None
360
+ mscale .......................................... 1.0
361
+ mscale_all_dim .................................. 0.0
362
+ mtp_loss_scaling_factor ......................... 0.1
363
+ mtp_num_layers .................................. None
364
+ multi_latent_attention .......................... False
365
+ nccl_all_reduce_for_prefill ..................... False
366
+ nccl_communicator_config_path ................... None
367
+ nccl_ub ......................................... False
368
+ no_load_optim ................................... None
369
+ no_load_rng ..................................... None
370
+ no_persist_layer_norm ........................... False
371
+ no_rope_freq .................................... None
372
+ no_save_optim ................................... None
373
+ no_save_rng ..................................... None
374
+ non_persistent_ckpt_type ........................ None
375
+ non_persistent_global_ckpt_dir .................. None
376
+ non_persistent_local_ckpt_algo .................. fully_parallel
377
+ non_persistent_local_ckpt_dir ................... None
378
+ non_persistent_save_interval .................... None
379
+ norm_epsilon .................................... 1e-06
380
+ normalization ................................... RMSNorm
381
+ num_attention_heads ............................. 28
382
+ num_channels .................................... 3
383
+ num_classes ..................................... 1000
384
+ num_dataset_builder_threads ..................... 1
385
+ num_distributed_optimizer_instances ............. 1
386
+ num_experts ..................................... 64
387
+ num_layers ...................................... 10
388
+ num_layers_at_end_in_bf16 ....................... 1
389
+ num_layers_at_start_in_bf16 ..................... 1
390
+ num_layers_per_virtual_pipeline_stage ........... None
391
+ num_query_groups ................................ 4
392
+ num_virtual_stages_per_pipeline_rank ............ None
393
+ num_workers ..................................... 2
394
+ object_storage_cache_path ....................... None
395
+ one_logger_async ................................ False
396
+ one_logger_project .............................. megatron-lm
397
+ one_logger_run_name ............................. None
398
+ onnx_safe ....................................... None
399
+ openai_gelu ..................................... False
400
+ optimizer ....................................... adam
401
+ optimizer_cpu_offload ........................... False
402
+ optimizer_offload_fraction ...................... 1.0
403
+ output_bert_embeddings .......................... False
404
+ overlap_cpu_optimizer_d2h_h2d ................... False
405
+ overlap_grad_reduce ............................. True
406
+ overlap_moe_expert_parallel_comm ................ False
407
+ overlap_p2p_comm ................................ False
408
+ overlap_p2p_comm_warmup_flush ................... False
409
+ overlap_param_gather ............................ True
410
+ overlap_param_gather_with_optimizer_step ........ False
411
+ override_opt_param_scheduler .................... False
412
+ params_dtype .................................... torch.bfloat16
413
+ patch_dim ....................................... 16
414
+ per_split_data_args_path ........................ None
415
+ perform_initialization .......................... True
416
+ pin_cpu_grads ................................... True
417
+ pin_cpu_params .................................. True
418
+ pipeline_model_parallel_comm_backend ............ None
419
+ pipeline_model_parallel_layout .................. None
420
+ pipeline_model_parallel_size .................... 1
421
+ position_embedding_type ......................... rope
422
+ pretrained_checkpoint ........................... None
423
+ profile ......................................... False
424
+ profile_ranks ................................... [0]
425
+ profile_step_end ................................ 12
426
+ profile_step_start .............................. 10
427
+ q_lora_rank ..................................... None
428
+ qk_head_dim ..................................... 128
429
+ qk_l2_norm ...................................... False
430
+ qk_layernorm .................................... False
431
+ qk_pos_emb_head_dim ............................. 64
432
+ query_in_block_prob ............................. 0.1
433
+ rampup_batch_size ............................... None
434
+ rank ............................................ 0
435
+ recompute_granularity ........................... None
436
+ recompute_method ................................ None
437
+ recompute_modules ............................... None
438
+ recompute_num_layers ............................ None
439
+ record_memory_history ........................... False
440
+ relative_attention_max_distance ................. 128
441
+ relative_attention_num_buckets .................. 32
442
+ replication ..................................... False
443
+ replication_factor .............................. 2
444
+ replication_jump ................................ None
445
+ rerun_mode ...................................... validate_results
446
+ reset_attention_mask ............................ False
447
+ reset_position_ids .............................. False
448
+ result_rejected_tracker_filename ................ None
449
+ retriever_report_topk_accuracies ................ []
450
+ retriever_score_scaling ......................... False
451
+ retriever_seq_length ............................ 256
452
+ retro_add_retriever ............................. False
453
+ retro_attention_gate ............................ 1
454
+ retro_cyclic_train_iters ........................ None
455
+ retro_encoder_attention_dropout ................. 0.1
456
+ retro_encoder_hidden_dropout .................... 0.1
457
+ retro_encoder_layers ............................ 2
458
+ retro_num_neighbors ............................. 2
459
+ retro_num_retrieved_chunks ...................... 2
460
+ retro_project_dir ............................... None
461
+ retro_verify_neighbor_count ..................... True
462
+ reuse_grad_buf_for_mxfp8_param_ag ............... False
463
+ rope_scaling_factor ............................. 8.0
464
+ rope_type ....................................... None
465
+ rotary_base ..................................... 1000000
466
+ rotary_interleaved .............................. False
467
+ rotary_percent .................................. 1.0
468
+ rotary_scaling_factor ........................... 1.0
469
+ rotary_seq_len_interpolation_factor ............. None
470
+ run_workload_inspector_server ................... False
471
+ sample_rate ..................................... 1.0
472
+ save ............................................ None
473
+ save_interval ................................... None
474
+ save_retain_interval ............................ None
475
+ scatter_gather_tensors_in_pipeline .............. True
476
+ seed ............................................ 1234
477
+ seq_length ...................................... 4096
478
+ sequence_parallel ............................... False
479
+ sft ............................................. False
480
+ sft_tokenizer_prompt_format ..................... nemotron-h-aligned
481
+ sgd_momentum .................................... 0.9
482
+ short_seq_prob .................................. 0.1
483
+ skip_train ...................................... False
484
+ skipped_train_samples ........................... 0
485
+ spec ............................................ None
486
+ split ........................................... 100,0,0
487
+ squared_relu .................................... False
488
+ start_weight_decay .............................. 0.1
489
+ straggler_ctrlr_port ............................ 65535
490
+ straggler_minmax_count .......................... 1
491
+ suggested_communication_unit_size ............... None
492
+ swiglu .......................................... True
493
+ swin_backbone_type .............................. tiny
494
+ symmetric_ar_type ............................... None
495
+ te_rng_tracker .................................. True
496
+ tensor_model_parallel_size ...................... 1
497
+ tensorboard_dir ................................. None
498
+ tensorboard_log_interval ........................ 1
499
+ tensorboard_queue_size .......................... 1000
500
+ test_data_path .................................. None
501
+ test_mode ....................................... False
502
+ tiktoken_num_special_tokens ..................... 1000
503
+ tiktoken_pattern ................................ None
504
+ tiktoken_special_tokens ......................... None
505
+ timing_log_level ................................ 0
506
+ timing_log_option ............................... minmax
507
+ titles_data_path ................................ None
508
+ tokenizer_model ................................. None
509
+ tokenizer_type .................................. GPT2BPETokenizer
510
+ torch_fsdp2_reshard_after_forward ............... True
511
+ tp_comm_bootstrap_backend ....................... nccl
512
+ tp_comm_bulk_dgrad .............................. True
513
+ tp_comm_bulk_wgrad .............................. True
514
+ tp_comm_overlap ................................. False
515
+ tp_comm_overlap_ag .............................. True
516
+ tp_comm_overlap_cfg ............................. None
517
+ tp_comm_overlap_rs .............................. True
518
+ tp_comm_overlap_rs_dgrad ........................ False
519
+ tp_comm_split_ag ................................ True
520
+ tp_comm_split_rs ................................ True
521
+ train_data_path ................................. None
522
+ train_iters ..................................... 12
523
+ train_samples ................................... None
524
+ train_sync_interval ............................. None
525
+ transformer_impl ................................ transformer_engine
526
+ transformer_pipeline_model_parallel_size ........ 1
527
+ untie_embeddings_and_output_weights ............. True
528
+ use_checkpoint_args ............................. False
529
+ use_checkpoint_opt_param_scheduler .............. False
530
+ use_cpu_initialization .......................... None
531
+ use_custom_fsdp ................................. False
532
+ use_dist_ckpt ................................... True
533
+ use_dist_ckpt_deprecated ........................ False
534
+ use_distributed_optimizer ....................... True
535
+ use_flash_attn .................................. False
536
+ use_legacy_models ............................... False
537
+ use_mp_args_from_checkpoint_args ................ False
538
+ use_one_sent_docs ............................... False
539
+ use_persistent_ckpt_worker ...................... False
540
+ use_precision_aware_optimizer ................... False
541
+ use_pytorch_profiler ............................ False
542
+ use_ring_exchange_p2p ........................... False
543
+ use_rope_scaling ................................ False
544
+ use_rotary_position_embeddings .................. False
545
+ use_sharp ....................................... False
546
+ use_tokenizer_model_from_checkpoint_args ........ True
547
+ use_torch_fsdp2 ................................. False
548
+ use_torch_optimizer_for_cpu_offload ............. False
549
+ use_tp_pp_dp_mapping ............................ False
550
+ v_head_dim ...................................... 128
551
+ valid_data_path ................................. None
552
+ variable_seq_lengths ............................ False
553
+ virtual_pipeline_model_parallel_size ............ None
554
+ vision_backbone_type ............................ vit
555
+ vision_pretraining .............................. False
556
+ vision_pretraining_type ......................... classify
557
+ vocab_extra_ids ................................. 0
558
+ vocab_file ...................................... /workspace/data/wikitext/gpt2-vocab.json
559
+ vocab_size ...................................... None
560
+ wandb_exp_name ..................................
561
+ wandb_project ...................................
562
+ wandb_save_dir ..................................
563
+ weight_decay .................................... 0.1
564
+ weight_decay_incr_style ......................... constant
565
+ wgrad_deferral_limit ............................ 0
566
+ world_size ...................................... 8
567
+ yaml_cfg ........................................ None
568
+ -------------------- end of arguments ---------------------
569
+ INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 1
570
+ > building GPT2BPETokenizer tokenizer ...
571
+ > padded vocab (size: 50257) with 47 dummy tokens (new size: 50304)
572
+ WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.VALIDATE_RESULTS
573
+ > initializing torch distributed ...
574
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
575
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
576
+ [Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
577
+ [Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
578
+ [Gloo] Rank [Gloo] Rank [Gloo] Rank [Gloo] Rank 3165 is connected to is connected to is connected to is connected to 7777 peer ranks. peer ranks. peer ranks. peer ranks. Expected number of connected peer ranks is : Expected number of connected peer ranks is : Expected number of connected peer ranks is : Expected number of connected peer ranks is : 7777
579
+
580
+
581
+
582
+ [Gloo] Rank [Gloo] Rank 74 is connected to is connected to 77 peer ranks. peer ranks. Expected number of connected peer ranks is : Expected number of connected peer ranks is : 77
583
+
584
+ [Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
585
+ [Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
586
+ [Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
587
+ [Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
588
+ [Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
589
+ [Gloo] Rank [Gloo] Rank 5 is connected to 77 is connected to peer ranks. 7Expected number of connected peer ranks is : peer ranks. 7Expected number of connected peer ranks is : [Gloo] Rank
590
+ 76
591
+ is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
592
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
593
+ > initialized tensor model parallel with size 1
594
+ > initialized pipeline model parallel with size 1
595
+ > setting random seeds to 1234 ...
596
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
597
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
598
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
599
+ [Gloo] Rank 0 is connected to 0 peer ranks. [Gloo] Rank Expected number of connected peer ranks is : 00
600
+ is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
601
+ [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
602
+
603
+ [rank5]:[W826 18:23:55.923229216 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.
604
+ [rank2]:[W826 18:23:55.924199894 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.
605
+ [rank6]:[W826 18:23:55.924318434 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.
606
+ [rank1]:[W826 18:23:55.924371588 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
+ [rank4]:[W826 18:23:55.926705726 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
608
+ [rank7]:[W826 18:23:55.926744842 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
609
+ [rank3]:[W826 18:23:55.926751638 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
+ > compiling dataset index builder ...
611
+ make: Entering directory '/workspace/Megatron-MoE-Echo/megatron/core/datasets'
612
+ make: Nothing to be done for 'default'.
613
+ make: Leaving directory '/workspace/Megatron-MoE-Echo/megatron/core/datasets'
614
+ >>> done with dataset index builder. Compilation time: 0.038 seconds
615
+ WARNING: constraints for invoking optimized fused softmax kernel are not met. We default back to unfused kernel invocations.
616
+ > compiling and loading fused kernels ...
617
+ [rank0]:[W826 18:23:56.090693413 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.
618
+ >>> done with compiling and loading fused kernels. Compilation time: 3.626 seconds
619
+ time to initialize megatron (seconds): 6.914
620
+ [after megatron is initialized] datetime: 2026-08-26 18:24:01
621
+ building GPT model ...
622
+ /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.
623
+ warnings.warn(
624
+ /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.
625
+ warnings.warn(
626
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
627
+ warnings.warn(
628
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
629
+ warnings.warn(
630
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
631
+ warnings.warn(
632
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
633
+ warnings.warn(
634
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
635
+ warnings.warn(
636
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
637
+ warnings.warn(
638
+ > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 5060614656
639
+ 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)
640
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 22
641
+ Params for bucket 1 (180289536 elements, 180289536 padded size):
642
+ module.output_layer.weight
643
+ Params for bucket 2 (73403904 elements, 73403904 padded size):
644
+ module.decoder.final_layernorm.weight
645
+ module.decoder.layers.9.mlp.shared_experts.linear_fc2.weight
646
+ Params for bucket 3 (146800640 elements, 146800640 padded size):
647
+ module.decoder.layers.9.mlp.shared_experts.linear_fc1.weight
648
+ Params for bucket 4 (103001600 elements, 103001600 padded size):
649
+ module.decoder.layers.9.mlp.router.weight
650
+ module.decoder.layers.9.self_attention.linear_qkv.layer_norm_weight
651
+ module.decoder.layers.9.self_attention.linear_qkv.weight
652
+ module.decoder.layers.9.self_attention.linear_proj.weight
653
+ module.decoder.layers.9.pre_mlp_layernorm.weight
654
+ module.decoder.layers.9.self_attention.linear_qkv.bias
655
+ module.decoder.layers.8.mlp.shared_experts.linear_fc2.weight
656
+ Params for bucket 5 (146800640 elements, 146800640 padded size):
657
+ module.decoder.layers.8.mlp.shared_experts.linear_fc1.weight
658
+ Params for bucket 6 (103001600 elements, 103001600 padded size):
659
+ module.decoder.layers.8.self_attention.linear_qkv.layer_norm_weight
660
+ module.decoder.layers.8.pre_mlp_layernorm.weight
661
+ module.decoder.layers.8.self_attention.linear_qkv.bias
662
+ module.decoder.layers.8.self_attention.linear_proj.weight
663
+ module.decoder.layers.7.mlp.shared_experts.linear_fc2.weight
664
+ module.decoder.layers.8.mlp.router.weight
665
+ module.decoder.layers.8.self_attention.linear_qkv.weight
666
+ Params for bucket 7 (146800640 elements, 146800640 padded size):
667
+ module.decoder.layers.7.mlp.shared_experts.linear_fc1.weight
668
+ Params for bucket 8 (103001600 elements, 103001600 padded size):
669
+ module.decoder.layers.7.mlp.router.weight
670
+ module.decoder.layers.7.self_attention.linear_qkv.layer_norm_weight
671
+ module.decoder.layers.7.self_attention.linear_qkv.weight
672
+ module.decoder.layers.7.self_attention.linear_proj.weight
673
+ module.decoder.layers.6.mlp.shared_experts.linear_fc2.weight
674
+ module.decoder.layers.7.pre_mlp_layernorm.weight
675
+ module.decoder.layers.7.self_attention.linear_qkv.bias
676
+ Params for bucket 9 (146800640 elements, 146800640 padded size):
677
+ module.decoder.layers.6.mlp.shared_experts.linear_fc1.weight
678
+ Params for bucket 10 (103001600 elements, 103001600 padded size):
679
+ module.decoder.layers.6.pre_mlp_layernorm.weight
680
+ module.decoder.layers.6.self_attention.linear_qkv.bias
681
+ module.decoder.layers.5.mlp.shared_experts.linear_fc2.weight
682
+ module.decoder.layers.6.mlp.router.weight
683
+ module.decoder.layers.6.self_attention.linear_qkv.layer_norm_weight
684
+ module.decoder.layers.6.self_attention.linear_qkv.weight
685
+ module.decoder.layers.6.self_attention.linear_proj.weight
686
+ Params for bucket 11 (146800640 elements, 146800640 padded size):
687
+ module.decoder.layers.5.mlp.shared_experts.linear_fc1.weight
688
+ Params for bucket 12 (103001600 elements, 103001600 padded size):
689
+ module.decoder.layers.5.mlp.router.weight
690
+ module.decoder.layers.5.self_attention.linear_qkv.weight
691
+ module.decoder.layers.5.self_attention.linear_qkv.layer_norm_weight
692
+ module.decoder.layers.5.self_attention.linear_proj.weight
693
+ module.decoder.layers.5.pre_mlp_layernorm.weight
694
+ module.decoder.layers.5.self_attention.linear_qkv.bias
695
+ module.decoder.layers.4.mlp.shared_experts.linear_fc2.weight
696
+ Params for bucket 13 (146800640 elements, 146800640 padded size):
697
+ module.decoder.layers.4.mlp.shared_experts.linear_fc1.weight
698
+ Params for bucket 14 (103001600 elements, 103001600 padded size):
699
+ module.decoder.layers.4.self_attention.linear_qkv.layer_norm_weight
700
+ module.decoder.layers.4.self_attention.linear_qkv.weight
701
+ module.decoder.layers.4.self_attention.linear_proj.weight
702
+ module.decoder.layers.3.mlp.shared_experts.linear_fc2.weight
703
+ module.decoder.layers.4.mlp.router.weight
704
+ module.decoder.layers.4.self_attention.linear_qkv.bias
705
+ module.decoder.layers.4.pre_mlp_layernorm.weight
706
+ Params for bucket 15 (146800640 elements, 146800640 padded size):
707
+ module.decoder.layers.3.mlp.shared_experts.linear_fc1.weight
708
+ Params for bucket 16 (103001600 elements, 103001600 padded size):
709
+ module.decoder.layers.3.self_attention.linear_qkv.bias
710
+ module.decoder.layers.3.pre_mlp_layernorm.weight
711
+ module.decoder.layers.3.self_attention.linear_qkv.layer_norm_weight
712
+ module.decoder.layers.3.self_attention.linear_proj.weight
713
+ module.decoder.layers.3.mlp.router.weight
714
+ module.decoder.layers.3.self_attention.linear_qkv.weight
715
+ module.decoder.layers.2.mlp.shared_experts.linear_fc2.weight
716
+ Params for bucket 17 (146800640 elements, 146800640 padded size):
717
+ module.decoder.layers.2.mlp.shared_experts.linear_fc1.weight
718
+ Params for bucket 18 (103001600 elements, 103001600 padded size):
719
+ module.decoder.layers.2.mlp.router.weight
720
+ module.decoder.layers.2.self_attention.linear_qkv.layer_norm_weight
721
+ module.decoder.layers.2.self_attention.linear_qkv.bias
722
+ module.decoder.layers.2.pre_mlp_layernorm.weight
723
+ module.decoder.layers.2.self_attention.linear_proj.weight
724
+ module.decoder.layers.1.mlp.shared_experts.linear_fc2.weight
725
+ module.decoder.layers.2.self_attention.linear_qkv.weight
726
+ Params for bucket 19 (146800640 elements, 146800640 padded size):
727
+ module.decoder.layers.1.mlp.shared_experts.linear_fc1.weight
728
+ Params for bucket 20 (103001600 elements, 103001600 padded size):
729
+ module.decoder.layers.1.mlp.router.weight
730
+ module.decoder.layers.1.self_attention.linear_qkv.weight
731
+ module.decoder.layers.1.self_attention.linear_qkv.layer_norm_weight
732
+ module.decoder.layers.1.self_attention.linear_proj.weight
733
+ module.decoder.layers.0.mlp.shared_experts.linear_fc2.weight
734
+ module.decoder.layers.1.self_attention.linear_qkv.bias
735
+ module.decoder.layers.1.pre_mlp_layernorm.weight
736
+ Params for bucket 21 (146800640 elements, 146800640 padded size):
737
+ module.decoder.layers.0.mlp.shared_experts.linear_fc1.weight
738
+ Params for bucket 22 (209890816 elements, 209890816 padded size):
739
+ module.decoder.layers.0.self_attention.linear_qkv.bias
740
+ module.decoder.layers.0.pre_mlp_layernorm.weight
741
+ module.decoder.layers.0.self_attention.linear_qkv.weight
742
+ module.decoder.layers.0.mlp.router.weight
743
+ module.decoder.layers.0.self_attention.linear_qkv.layer_norm_weight
744
+ module.decoder.layers.0.self_attention.linear_proj.weight
745
+ module.embedding.word_embeddings.weight
746
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 44
747
+ Params for bucket 1 (45875200 elements, 45875200 padded size):
748
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight7
749
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight5
750
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight3
751
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight6
752
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight4
753
+ Params for bucket 2 (45875200 elements, 45875200 padded size):
754
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight2
755
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight1
756
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight0
757
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight7
758
+ Params for bucket 3 (55050240 elements, 55050240 padded size):
759
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight6
760
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight4
761
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight5
762
+ Params for bucket 4 (55050240 elements, 55050240 padded size):
763
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight3
764
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight1
765
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight2
766
+ Params for bucket 5 (45875200 elements, 45875200 padded size):
767
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight0
768
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight7
769
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight5
770
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight6
771
+ Params for bucket 6 (45875200 elements, 45875200 padded size):
772
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight1
773
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight3
774
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight4
775
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight2
776
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight0
777
+ Params for bucket 7 (55050240 elements, 55050240 padded size):
778
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight7
779
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight6
780
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight5
781
+ Params for bucket 8 (55050240 elements, 55050240 padded size):
782
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight4
783
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight3
784
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight2
785
+ Params for bucket 9 (45875200 elements, 45875200 padded size):
786
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight1
787
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight0
788
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight7
789
+ Params for bucket 10 (45875200 elements, 45875200 padded size):
790
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight6
791
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight2
792
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight4
793
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight3
794
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight5
795
+ Params for bucket 11 (55050240 elements, 55050240 padded size):
796
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight1
797
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight6
798
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight0
799
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight7
800
+ Params for bucket 12 (55050240 elements, 55050240 padded size):
801
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight5
802
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight3
803
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight4
804
+ Params for bucket 13 (55050240 elements, 55050240 padded size):
805
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight2
806
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight0
807
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight1
808
+ Params for bucket 14 (45875200 elements, 45875200 padded size):
809
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight7
810
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight3
811
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight5
812
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight6
813
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight4
814
+ Params for bucket 15 (45875200 elements, 45875200 padded size):
815
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight2
816
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight0
817
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight7
818
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight1
819
+ Params for bucket 16 (55050240 elements, 55050240 padded size):
820
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight6
821
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight5
822
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight4
823
+ Params for bucket 17 (55050240 elements, 55050240 padded size):
824
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight3
825
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight1
826
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight2
827
+ Params for bucket 18 (45875200 elements, 45875200 padded size):
828
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight0
829
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight5
830
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight6
831
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight7
832
+ Params for bucket 19 (45875200 elements, 45875200 padded size):
833
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight3
834
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight1
835
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight4
836
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight2
837
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight0
838
+ Params for bucket 20 (55050240 elements, 55050240 padded size):
839
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight7
840
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight5
841
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight6
842
+ Params for bucket 21 (55050240 elements, 55050240 padded size):
843
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight4
844
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight3
845
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight2
846
+ Params for bucket 22 (45875200 elements, 45875200 padded size):
847
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight1
848
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight7
849
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight0
850
+ Params for bucket 23 (45875200 elements, 45875200 padded size):
851
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight5
852
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight3
853
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight6
854
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight4
855
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight2
856
+ Params for bucket 24 (55050240 elements, 55050240 padded size):
857
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight1
858
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight7
859
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight6
860
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight0
861
+ Params for bucket 25 (55050240 elements, 55050240 padded size):
862
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight5
863
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight4
864
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight3
865
+ Params for bucket 26 (55050240 elements, 55050240 padded size):
866
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight2
867
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight0
868
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight1
869
+ Params for bucket 27 (45875200 elements, 45875200 padded size):
870
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight6
871
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight4
872
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight7
873
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight3
874
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight5
875
+ Params for bucket 28 (45875200 elements, 45875200 padded size):
876
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight2
877
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight7
878
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight0
879
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight1
880
+ Params for bucket 29 (55050240 elements, 55050240 padded size):
881
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight6
882
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight4
883
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight5
884
+ Params for bucket 30 (55050240 elements, 55050240 padded size):
885
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight3
886
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight1
887
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight2
888
+ Params for bucket 31 (45875200 elements, 45875200 padded size):
889
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight0
890
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight6
891
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight5
892
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight7
893
+ Params for bucket 32 (45875200 elements, 45875200 padded size):
894
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight3
895
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight1
896
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight4
897
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight2
898
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight0
899
+ Params for bucket 33 (55050240 elements, 55050240 padded size):
900
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight7
901
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight6
902
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight5
903
+ Params for bucket 34 (55050240 elements, 55050240 padded size):
904
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight4
905
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight2
906
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight3
907
+ Params for bucket 35 (45875200 elements, 45875200 padded size):
908
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight1
909
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight0
910
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight7
911
+ Params for bucket 36 (45875200 elements, 45875200 padded size):
912
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight6
913
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight2
914
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight4
915
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight3
916
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight5
917
+ Params for bucket 37 (55050240 elements, 55050240 padded size):
918
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight1
919
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight6
920
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight0
921
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight7
922
+ Params for bucket 38 (55050240 elements, 55050240 padded size):
923
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight5
924
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight3
925
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight4
926
+ Params for bucket 39 (55050240 elements, 55050240 padded size):
927
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight2
928
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight0
929
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight1
930
+ Params for bucket 40 (45875200 elements, 45875200 padded size):
931
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight4
932
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight6
933
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight3
934
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight7
935
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight5
936
+ Params for bucket 41 (45875200 elements, 45875200 padded size):
937
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight2
938
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight1
939
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight0
940
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight7
941
+ Params for bucket 42 (55050240 elements, 55050240 padded size):
942
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight6
943
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight5
944
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight4
945
+ Params for bucket 43 (55050240 elements, 55050240 padded size):
946
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight3
947
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight2
948
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight1
949
+ Params for bucket 44 (18350080 elements, 18350080 padded size):
950
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight0
951
+ 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 0x7cbf89478470>, config_logger_dir='')
952
+ INFO:megatron.core.optimizer_param_scheduler:> learning rate decay style: cosine
953
+ [after model, optimizer, and learning rate scheduler are built] datetime: 2026-08-26 18:24:02
954
+ > building train, validation, and test datasets ...
955
+ > datasets target sizes (minimum size):
956
+ train: 96
957
+ validation: 0
958
+ test: 0
959
+ INFO:megatron.core.datasets.blended_megatron_dataset_config:Let split_matrix = [(0, 1.0), None, None]
960
+ > building train, validation, and test datasets for GPT ...
961
+ 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 0x7cbf896a6ae0>, 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)
962
+ INFO:megatron.core.datasets.indexed_dataset:Load the _IndexReader from /workspace/data/wikitext/wikitext103_gpt2_text_document.idx
963
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence lengths
964
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence pointers
965
+ INFO:megatron.core.datasets.indexed_dataset: Extract the document indices
966
+ INFO:megatron.core.datasets.indexed_dataset:> total number of sequences: 1165029
967
+ INFO:megatron.core.datasets.indexed_dataset:> total number of documents: 1801350
968
+ INFO:megatron.core.datasets.gpt_dataset:Load the GPTDataset train indices
969
+ INFO:megatron.core.datasets.gpt_dataset: Load the document index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-document_index.npy
970
+ INFO:megatron.core.datasets.gpt_dataset: Load the sample index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-sample_index.npy
971
+ INFO:megatron.core.datasets.gpt_dataset: Load the shuffle index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-shuffle_index.npy
972
+ INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 29073
973
+ > finished creating GPT datasets ...
974
+ [after dataloaders are built] datetime: 2026-08-26 18:24:02
975
+ done with setup ...
976
+ training ...
977
+ Overwriting rerun_state_machine.current_iteration from -1 to 0...
978
+ (min, max) time across ranks (ms):
979
+ model-and-optimizer-setup ......................: (370.39, 376.11)
980
+ train/valid/test-data-iterators-setup ..........: (61.27, 78.86)
981
+ [before the start of training step] datetime: 2026-08-26 18:24:02
982
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: 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:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
985
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
986
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: 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:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
989
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
990
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: 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:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
993
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
994
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: 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
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
997
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
998
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
999
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1000
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1001
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1002
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1003
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1004
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1005
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1006
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1007
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1008
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1009
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1010
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1011
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1012
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1013
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1014
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1015
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1016
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1017
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1018
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1019
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1020
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1021
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1022
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1023
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1024
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1025
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1026
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1027
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1028
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1029
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1030
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1031
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1032
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1033
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1034
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1035
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1036
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1037
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1038
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1039
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1040
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1041
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1042
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1043
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1044
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1045
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1046
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1047
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1048
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1049
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1050
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1051
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1052
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1053
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1054
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1055
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1056
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1057
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1058
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1059
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1060
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1061
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1062
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1063
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1064
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1065
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1066
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1067
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1068
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1069
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1070
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1071
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1072
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1073
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1074
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1075
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1076
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1077
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1078
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1079
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1080
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1081
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1082
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1083
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1084
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1085
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1086
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1087
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1088
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1089
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1090
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1091
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1092
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1093
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1094
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1095
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1096
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1097
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1098
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1099
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1100
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1101
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1102
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1103
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1104
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1105
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1106
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1107
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1108
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1109
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1110
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1111
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1112
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1113
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1114
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1115
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1116
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1117
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1118
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1119
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1120
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1121
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1122
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1123
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1124
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1125
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1126
+ [2026-08-26 18:24:39] iteration 1/ 12 | consumed samples: 8 | elapsed time per iteration (ms): 37654.1 | throughput per GPU (TFLOP/s/GPU): 3.3 | learning rate: 2.000000E-05 | global batch size: 8 | lm loss: 1.097931E+01 | loss scale: 1.0 | grad norm: 29.879 | number of skipped iterations: 0 | number of nan iterations: 0 |
1127
+ Number of parameters in transformer block in billions: 20.11
1128
+ Number of parameters in embedding layers in billions: 0.36
1129
+ Total number of parameters in billions: 20.47
1130
+ Number of parameters in most loaded shard in billions: 20.4724
1131
+ Theoretical memory footprints: weight and optimizer=146430.14 MB
1132
+ [Rank 0] (after 1 iterations) memory (MB) | allocated: 68733.94384765625 | max allocated: 70397.458984375 | reserved: 81416.0 | max reserved: 81416.0
1133
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1134
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1135
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1136
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1137
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1138
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1139
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1140
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1141
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1142
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1143
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1144
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1145
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1146
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1147
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1148
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1149
+ [2026-08-26 18:24:40] iteration 2/ 12 | consumed samples: 16 | elapsed time per iteration (ms): 430.9 | throughput per GPU (TFLOP/s/GPU): 286.6 | learning rate: 4.000000E-05 | global batch size: 8 | lm loss: 1.097906E+01 | loss scale: 1.0 | grad norm: 30.625 | number of skipped iterations: 0 | number of nan iterations: 0 |
1150
+ [2026-08-26 18:24:40] iteration 3/ 12 | consumed samples: 24 | elapsed time per iteration (ms): 227.6 | throughput per GPU (TFLOP/s/GPU): 542.5 | learning rate: 6.000000E-05 | global batch size: 8 | lm loss: 9.538181E+00 | loss scale: 1.0 | grad norm: 32.388 | number of skipped iterations: 0 | number of nan iterations: 0 |
1151
+ Capture CUDA graph for training!!!
1152
+ CUDA graph capture done!!!
1153
+ [2026-08-26 18:24:42] iteration 4/ 12 | consumed samples: 32 | elapsed time per iteration (ms): 1760.2 | throughput per GPU (TFLOP/s/GPU): 70.2 | learning rate: 8.000000E-05 | global batch size: 8 | lm loss: 1.109149E+01 | loss scale: 1.0 | grad norm: 107.126 | number of skipped iterations: 0 | number of nan iterations: 0 |
1154
+ [2026-08-26 18:24:42] iteration 5/ 12 | consumed samples: 40 | elapsed time per iteration (ms): 181.6 | throughput per GPU (TFLOP/s/GPU): 680.0 | learning rate: 1.000000E-04 | global batch size: 8 | lm loss: 9.997242E+00 | loss scale: 1.0 | grad norm: 26.217 | number of skipped iterations: 0 | number of nan iterations: 0 |
1155
+ [2026-08-26 18:24:42] iteration 6/ 12 | consumed samples: 48 | elapsed time per iteration (ms): 180.5 | throughput per GPU (TFLOP/s/GPU): 684.2 | learning rate: 7.750000E-05 | global batch size: 8 | lm loss: 1.049742E+01 | loss scale: 1.0 | grad norm: 8.721 | number of skipped iterations: 0 | number of nan iterations: 0 |
1156
+ [2026-08-26 18:24:42] iteration 7/ 12 | consumed samples: 56 | elapsed time per iteration (ms): 179.4 | throughput per GPU (TFLOP/s/GPU): 688.4 | learning rate: 3.250000E-05 | global batch size: 8 | lm loss: 9.440645E+00 | loss scale: 1.0 | grad norm: 5.389 | number of skipped iterations: 0 | number of nan iterations: 0 |
1157
+ [2026-08-26 18:24:42] iteration 8/ 12 | consumed samples: 64 | elapsed time per iteration (ms): 178.5 | throughput per GPU (TFLOP/s/GPU): 691.9 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.962455E+00 | loss scale: 1.0 | grad norm: 4.550 | number of skipped iterations: 0 | number of nan iterations: 0 |
1158
+ [2026-08-26 18:24:43] iteration 9/ 12 | consumed samples: 72 | elapsed time per iteration (ms): 178.3 | throughput per GPU (TFLOP/s/GPU): 692.5 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.746603E+00 | loss scale: 1.0 | grad norm: 4.297 | number of skipped iterations: 0 | number of nan iterations: 0 |
1159
+ [2026-08-26 18:24:43] iteration 10/ 12 | consumed samples: 80 | elapsed time per iteration (ms): 178.7 | throughput per GPU (TFLOP/s/GPU): 690.9 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.686147E+00 | loss scale: 1.0 | grad norm: 4.422 | number of skipped iterations: 0 | number of nan iterations: 0 |
1160
+ [2026-08-26 18:24:43] iteration 11/ 12 | consumed samples: 88 | elapsed time per iteration (ms): 177.4 | throughput per GPU (TFLOP/s/GPU): 696.0 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.625414E+00 | loss scale: 1.0 | grad norm: 4.347 | number of skipped iterations: 0 | number of nan iterations: 0 |
1161
+ [2026-08-26 18:24:43] iteration 12/ 12 | consumed samples: 96 | elapsed time per iteration (ms): 178.8 | throughput per GPU (TFLOP/s/GPU): 690.8 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.549947E+00 | loss scale: 1.0 | grad norm: 4.284 | number of skipped iterations: 0 | number of nan iterations: 0 |
1162
+ [after training is done] datetime: 2026-08-26 18:24:43
1163
+ [rank0]:[W826 18:24:44.908108663 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())
1164
+ W0826 18:24:46.539000 499283 torch/distributed/elastic/agent/server/api.py:719] Received 15 death signal, shutting down workers
1165
+ W0826 18:24:46.540000 499283 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 499370 closing signal SIGTERM
1166
+ W0826 18:24:46.540000 499283 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 499371 closing signal SIGTERM
1167
+ W0826 18:24:46.541000 499283 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 499373 closing signal SIGTERM
1168
+ W0826 18:24:46.541000 499283 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 499374 closing signal SIGTERM
1169
+ W0826 18:24:46.541000 499283 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 499375 closing signal SIGTERM
1170
+ W0826 18:24:46.541000 499283 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 499376 closing signal SIGTERM
1171
+ W0826 18:24:46.542000 499283 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 499377 closing signal SIGTERM
1172
+ Traceback (most recent call last):
1173
+ File "/usr/local/bin/torchrun", line 8, in <module>
1174
+ sys.exit(main())
1175
+ ^^^^^^
1176
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper
1177
+ return f(*args, **kwargs)
1178
+ ^^^^^^^^^^^^^^^^^^
1179
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 892, in main
1180
+ run(args)
1181
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 883, in run
1182
+ elastic_launch(
1183
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 139, in __call__
1184
+ return launch_agent(self._config, self._entrypoint, list(args))
1185
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1186
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 261, in launch_agent
1187
+ result = agent.run()
1188
+ ^^^^^^^^^^^
1189
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/metrics/api.py", line 138, in wrapper
1190
+ result = f(*args, **kwargs)
1191
+ ^^^^^^^^^^^^^^^^^^
1192
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 711, in run
1193
+ result = self._invoke_run(role)
1194
+ ^^^^^^^^^^^^^^^^^^^^^^
1195
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 870, in _invoke_run
1196
+ time.sleep(monitor_interval)
1197
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/api.py", line 84, in _terminate_process_handler
1198
+ raise SignalException(f"Process {os.getpid()} got signal: {sigval}", sigval=sigval)
1199
+ torch.distributed.elastic.multiprocessing.api.SignalException: Process 499283 got signal: 15
echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep2/mem_samples.csv ADDED
@@ -0,0 +1,224 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ 0, 0
2
+ 1, 0
3
+ 2, 0
4
+ 3, 0
5
+ 4, 0
6
+ 5, 0
7
+ 6, 0
8
+ 7, 0
9
+ 0, 0
10
+ 1, 0
11
+ 2, 0
12
+ 3, 0
13
+ 4, 0
14
+ 5, 0
15
+ 6, 0
16
+ 7, 0
17
+ 0, 2
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, 4
43
+ 2, 629
44
+ 3, 4
45
+ 4, 4
46
+ 5, 4
47
+ 6, 4
48
+ 7, 4
49
+ 0, 857
50
+ 1, 853
51
+ 2, 851
52
+ 3, 851
53
+ 4, 853
54
+ 5, 859
55
+ 6, 853
56
+ 7, 857
57
+ 0, 1803
58
+ 1, 1803
59
+ 2, 1803
60
+ 3, 1803
61
+ 4, 1803
62
+ 5, 1803
63
+ 6, 1803
64
+ 7, 1803
65
+ 0, 43693
66
+ 1, 43107
67
+ 2, 43149
68
+ 3, 43821
69
+ 4, 43827
70
+ 5, 43849
71
+ 6, 43803
72
+ 7, 43857
73
+ 0, 44169
74
+ 1, 43601
75
+ 2, 43601
76
+ 3, 44301
77
+ 4, 44301
78
+ 5, 44301
79
+ 6, 44301
80
+ 7, 44301
81
+ 0, 47239
82
+ 1, 46671
83
+ 2, 46671
84
+ 3, 47371
85
+ 4, 47371
86
+ 5, 47371
87
+ 6, 47371
88
+ 7, 47371
89
+ 0, 47519
90
+ 1, 46951
91
+ 2, 46951
92
+ 3, 47651
93
+ 4, 47651
94
+ 5, 47651
95
+ 6, 47651
96
+ 7, 47651
97
+ 0, 47827
98
+ 1, 47225
99
+ 2, 47227
100
+ 3, 47947
101
+ 4, 47919
102
+ 5, 47917
103
+ 6, 47957
104
+ 7, 47939
105
+ 0, 48553
106
+ 1, 47985
107
+ 2, 47985
108
+ 3, 48685
109
+ 4, 48685
110
+ 5, 48685
111
+ 6, 48685
112
+ 7, 48685
113
+ 0, 48799
114
+ 1, 48231
115
+ 2, 48231
116
+ 3, 48931
117
+ 4, 48931
118
+ 5, 48931
119
+ 6, 48931
120
+ 7, 48931
121
+ 0, 49933
122
+ 1, 49365
123
+ 2, 49365
124
+ 3, 50065
125
+ 4, 50065
126
+ 5, 50065
127
+ 6, 50065
128
+ 7, 50065
129
+ 0, 49933
130
+ 1, 49365
131
+ 2, 49365
132
+ 3, 50065
133
+ 4, 50065
134
+ 5, 50065
135
+ 6, 50065
136
+ 7, 50065
137
+ 0, 77673
138
+ 1, 77661
139
+ 2, 77657
140
+ 3, 77655
141
+ 4, 77677
142
+ 5, 77649
143
+ 6, 77663
144
+ 7, 77661
145
+ 0, 78077
146
+ 1, 78061
147
+ 2, 78059
148
+ 3, 78081
149
+ 4, 78081
150
+ 5, 78081
151
+ 6, 78081
152
+ 7, 78081
153
+ 0, 78427
154
+ 1, 78411
155
+ 2, 78409
156
+ 3, 78431
157
+ 4, 78431
158
+ 5, 78431
159
+ 6, 78431
160
+ 7, 78431
161
+ 0, 78427
162
+ 1, 78411
163
+ 2, 78409
164
+ 3, 78431
165
+ 4, 78431
166
+ 5, 78431
167
+ 6, 78431
168
+ 7, 78431
169
+ 0, 78427
170
+ 1, 78411
171
+ 2, 78409
172
+ 3, 78431
173
+ 4, 78431
174
+ 5, 78431
175
+ 6, 78431
176
+ 7, 78431
177
+ 0, 78883
178
+ 1, 78889
179
+ 2, 78899
180
+ 3, 78863
181
+ 4, 78913
182
+ 5, 78877
183
+ 6, 78925
184
+ 7, 78889
185
+ 0, 82149
186
+ 1, 81573
187
+ 2, 82691
188
+ 3, 82153
189
+ 4, 81593
190
+ 5, 82153
191
+ 6, 81593
192
+ 7, 82153
193
+ 0, 89407
194
+ 1, 88701
195
+ 2, 89933
196
+ 3, 89331
197
+ 4, 88775
198
+ 5, 89337
199
+ 6, 88829
200
+ 7, 89303
201
+ 0, 110305
202
+ 1, 109423
203
+ 2, 110909
204
+ 3, 111181
205
+ 4, 112327
206
+ 5, 111181
207
+ 6, 112321
208
+ 7, 112301
209
+ 0, 112291
210
+ 1, 111969
211
+ 2, 112055
212
+ 3, 112327
213
+ 4, 112327
214
+ 5, 112327
215
+ 6, 112327
216
+ 7, 112327
217
+ 0, 3662
218
+ 1, 2638
219
+ 2, 111161
220
+ 3, 3662
221
+ 4, 2638
222
+ 5, 111433
223
+ 6, 111433
224
+ 7, 111433
echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep2/run.env ADDED
@@ -0,0 +1,61 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ backend=echo
2
+ config=QWEN2_57B
3
+ family=tierB
4
+ repo=/workspace/Megatron-MoE-Echo
5
+ branch=bracing-regions-through-kernels
6
+ commit=9a236793
7
+ num_layers=10
8
+ hidden=3584
9
+ ffn_hidden=2560
10
+ moe_ffn_hidden=2560
11
+ num_experts=64
12
+ topk=8
13
+ experts_per_rank=8
14
+ expert_params=27525120
15
+ heads=28
16
+ kv_groups=4
17
+ kv_channels=
18
+ qk_layernorm=0
19
+ shared_ffn=20480
20
+ shared_expert_overlap=0
21
+ mla=0
22
+ dense_first_layers=0
23
+ moe_layers=10
24
+ rotary_base=1000000
25
+ rotary_scaling=
26
+ qkv_bias=1
27
+ router_balancing=aux_loss
28
+ aux_loss_coeff=1e-2
29
+ router_pre_softmax=1
30
+ router_fusion=0
31
+ router_dtype=fp32
32
+ device_initiated_grouped_gemm=1
33
+ norm_epsilon=1e-6
34
+ seq_length=4096
35
+ micro_batch_size=1
36
+ global_batch_size=8
37
+ num_microbatches=1
38
+ assignments_per_rank=32768
39
+ mean_tokens_per_expert=4096
40
+ slot_policy=matched
41
+ slots_per_rank=8
42
+ capacity_mode=static
43
+ capacity=1.08984375
44
+ cuda_graph=1
45
+ probe_mode=0
46
+ async_grad_reduce=0
47
+ branch_role=timing
48
+ memory_dump_mode=none
49
+ memory_snapshot_path=
50
+ cuda_graph_warmup_steps=3
51
+ recompute=none
52
+ force_balance=0
53
+ train_iters=12
54
+ ep_size=8
55
+ echo_dispatch_overlap=1
56
+ echo_dispatch_recompute=0
57
+ ultraep_quota_min_tokens=
58
+ ultraep_weight_sync_join=
59
+ profile_enable=0
60
+ kernel_trace=0
61
+ fp8_recipe=mxfp8
echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep2/train.log ADDED
@@ -0,0 +1,1200 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ==============================================================
2
+ QWEN2_57B / echo (slots/rank=8 policy=matched)
3
+ E=64 K=8 H'=2560 m=8
4
+ L=10 seq=4096 gbs=8 (1 microbatches)
5
+ cap_mode=static cap=1.08984375 cg=1 recompute=none
6
+ memory_dump=none snapshots=/workspace/tierb_study_20260826_144644/echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep2/memory_snapshots/snapshot.pickle
7
+ repo=/workspace/Megatron-MoE-Echo @ 9a236793
8
+ ==============================================================
9
+ W0826 18:24:52.059000 502750 torch/distributed/run.py:766]
10
+ W0826 18:24:52.059000 502750 torch/distributed/run.py:766] *****************************************
11
+ W0826 18:24:52.059000 502750 torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.
12
+ W0826 18:24:52.059000 502750 torch/distributed/run.py:766] *****************************************
13
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
14
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
15
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
16
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
17
+ WARNING: one_logger package is required to enable e2e metrics tracking. please go to https://confluence.nvidia.com/display/MLWFO/Package+Repositories for details to install it
18
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
19
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
20
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
21
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
22
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
23
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
24
+ using world size: 8, data-parallel size: 8, context-parallel size: 1, hierarchical context-parallel sizes: None, tensor-model-parallel size: 1, pipeline-model-parallel size: 1
25
+ WARNING: overriding default arguments for tokenizer_type:GPT2BPETokenizer with tokenizer_type:GPT2BPETokenizer
26
+ Number of virtual stages per pipeline stage: None
27
+ accumulate and all-reduce gradients in fp32 for bfloat16 data type.
28
+ using torch.bfloat16 for parameters ...
29
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
30
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
31
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
32
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
33
+ ------------------------ arguments ------------------------
34
+ account_for_embedding_in_pipeline_split ......... False
35
+ account_for_loss_in_pipeline_split .............. False
36
+ accumulate_allreduce_grads_in_fp32 .............. True
37
+ adam_beta1 ...................................... 0.9
38
+ adam_beta2 ...................................... 0.999
39
+ adam_eps ........................................ 1e-08
40
+ add_bias_linear ................................. False
41
+ add_position_embedding .......................... False
42
+ add_qkv_bias .................................... True
43
+ adlr_autoresume ................................. False
44
+ adlr_autoresume_interval ........................ 1000
45
+ align_grad_reduce ............................... True
46
+ align_param_gather .............................. False
47
+ app_tag_run_name ................................ None
48
+ app_tag_run_version ............................. 0.0.0
49
+ apply_layernorm_1p .............................. False
50
+ apply_query_key_layer_scaling ................... False
51
+ apply_residual_connection_post_layernorm ........ False
52
+ apply_rope_fusion ............................... True
53
+ async_save ...................................... None
54
+ async_tensor_model_parallel_allreduce ........... True
55
+ attention_backend ............................... AttnBackend.auto
56
+ attention_dropout ............................... 0.0
57
+ attention_softmax_in_fp32 ....................... False
58
+ auto_detect_ckpt_format ......................... False
59
+ barrier_with_L1_time ............................ True
60
+ bert_binary_head ................................ True
61
+ bert_embedder_type .............................. megatron
62
+ bert_load ....................................... None
63
+ bf16 ............................................ True
64
+ bias_dropout_fusion ............................. True
65
+ bias_gelu_fusion ................................ False
66
+ bias_swiglu_fusion .............................. True
67
+ biencoder_projection_dim ........................ 0
68
+ biencoder_shared_query_context_model ............ False
69
+ block_data_path ................................. None
70
+ calc_ft_timeouts ................................ False
71
+ calculate_per_token_loss ........................ False
72
+ check_for_large_grads ........................... False
73
+ check_for_nan_in_loss_and_grad .................. False
74
+ check_for_spiky_loss ............................ False
75
+ check_weight_hash_across_dp_replicas_interval ... None
76
+ ckpt_assume_constant_structure .................. False
77
+ ckpt_convert_format ............................. None
78
+ ckpt_convert_save ............................... None
79
+ ckpt_convert_update_legacy_dist_opt_format ...... False
80
+ ckpt_format ..................................... torch_dist
81
+ ckpt_fully_parallel_load ........................ False
82
+ ckpt_fully_parallel_save ........................ True
83
+ ckpt_fully_parallel_save_deprecated ............. False
84
+ ckpt_step ....................................... None
85
+ classes_fraction ................................ 1.0
86
+ clip_grad ....................................... 1.0
87
+ clone_scatter_output_in_embedding ............... True
88
+ config_logger_dir ...............................
89
+ consumed_train_samples .......................... 0
90
+ consumed_valid_samples .......................... 0
91
+ context_parallel_size ........................... 1
92
+ cp_comm_type .................................... ['p2p']
93
+ create_attention_mask_in_dataloader ............. True
94
+ cross_entropy_fusion_impl ....................... native
95
+ cross_entropy_loss_fusion ....................... False
96
+ cuda_graph_scope ................................ full_iteration
97
+ cuda_graph_warmup_steps ......................... 3
98
+ data_args_path .................................. None
99
+ data_cache_path ................................. None
100
+ data_parallel_random_init ....................... False
101
+ data_parallel_sharding_strategy ................. no_shard
102
+ data_parallel_size .............................. 8
103
+ data_path ....................................... ['/workspace/data/wikitext/wikitext103_gpt2_text_document']
104
+ data_per_class_fraction ......................... 1.0
105
+ data_sharding ................................... True
106
+ dataloader_type ................................. single
107
+ ddp_average_in_collective ....................... False
108
+ ddp_bucket_size ................................. None
109
+ ddp_num_buckets ................................. None
110
+ ddp_pad_buckets_for_high_nccl_busbw ............. False
111
+ decoder_first_pipeline_num_layers ............... None
112
+ decoder_last_pipeline_num_layers ................ None
113
+ decoder_num_layers .............................. None
114
+ decoder_seq_length .............................. None
115
+ decoupled_lr .................................... None
116
+ decoupled_min_lr ................................ None
117
+ decrease_batch_size_if_needed ................... False
118
+ defer_embedding_wgrad_compute ................... False
119
+ delay_wgrad_compute ............................. False
120
+ deprecated_use_mcore_models ..................... True
121
+ deterministic_mode .............................. False
122
+ dino_bottleneck_size ............................ 256
123
+ dino_freeze_last_layer .......................... 1
124
+ dino_head_hidden_size ........................... 2048
125
+ dino_local_crops_number ......................... 10
126
+ dino_local_img_size ............................. 96
127
+ dino_norm_last_layer ............................ False
128
+ dino_teacher_temp ............................... 0.07
129
+ dino_warmup_teacher_temp ........................ 0.04
130
+ dino_warmup_teacher_temp_epochs ................. 30
131
+ disable_bf16_reduced_precision_matmul ........... False
132
+ disable_mamba_mem_eff_path ...................... False
133
+ disable_straggler_on_startup .................... False
134
+ dist_ckpt_format_deprecated ..................... None
135
+ dist_ckpt_strictness ............................ assume_ok_unexpected
136
+ distribute_saved_activations .................... False
137
+ distributed_backend ............................. nccl
138
+ distributed_timeout_minutes ..................... 10
139
+ embedding_init_method_std ....................... None
140
+ embedding_path .................................. None
141
+ empty_unused_memory_level ....................... 0
142
+ enable_cuda_graph ............................... True
143
+ enable_experimental ............................. False
144
+ enable_ft_package ............................... False
145
+ enable_gloo_process_groups ...................... True
146
+ enable_msc ...................................... True
147
+ enable_one_logger ............................... True
148
+ encoder_num_layers .............................. 10
149
+ encoder_seq_length .............................. 4096
150
+ end_weight_decay ................................ 0.1
151
+ eod_mask_loss ................................... False
152
+ error_injection_rate ............................ 0
153
+ error_injection_type ............................ transient_error
154
+ eval_interval ................................... 1000000
155
+ eval_iters ...................................... 0
156
+ evidence_data_path .............................. None
157
+ exit_duration_in_mins ........................... None
158
+ exit_interval ................................... None
159
+ exit_on_missing_checkpoint ...................... False
160
+ exit_signal_handler ............................. False
161
+ exp_avg_dtype ................................... torch.float32
162
+ exp_avg_sq_dtype ................................ torch.float32
163
+ expert_model_parallel_size ...................... 8
164
+ expert_tensor_parallel_size ..................... 1
165
+ export_force_local_attention .................... False
166
+ export_kd_cfg ................................... None
167
+ export_kd_teacher_ckpt_format ................... None
168
+ export_kd_teacher_load .......................... None
169
+ export_kv_cache_quant ........................... False
170
+ export_legacy_megatron .......................... False
171
+ export_model_type ............................... GPTModel
172
+ export_moe_apply_probs_on_input ................. False
173
+ export_qk_l2_norm ............................... False
174
+ export_quant_cfg ................................ None
175
+ export_real_quant_cfg ........................... None
176
+ export_te_mcore_model ........................... False
177
+ external_cuda_graph ............................. False
178
+ ffn_hidden_size ................................. 2560
179
+ finetune ........................................ False
180
+ finetune_data_split ............................. train
181
+ finetune_hf_dataset ............................. None
182
+ first_last_layers_bf16 .......................... False
183
+ flash_decode .................................... False
184
+ fp16 ............................................ False
185
+ fp16_lm_cross_entropy ........................... False
186
+ fp32_residual_connection ........................ False
187
+ fp8 ............................................. e4m3
188
+ fp8_amax_compute_algo ........................... most_recent
189
+ fp8_amax_history_len ............................ 1
190
+ fp8_interval .................................... 1
191
+ fp8_margin ...................................... 0
192
+ fp8_param_gather ................................ False
193
+ fp8_recipe ...................................... mxfp8
194
+ fp8_wgrad ....................................... True
195
+ fsdp_double_buffer .............................. False
196
+ global_batch_size ............................... 8
197
+ grad_reduce_in_bf16 ............................. False
198
+ gradient_accumulation_fusion .................... True
199
+ gradient_reduce_div_fusion ...................... True
200
+ group_query_attention ........................... True
201
+ head_lr_mult .................................... 1.0
202
+ heterogeneous_layers_config_encoded_json ........ None
203
+ heterogeneous_layers_config_path ................ None
204
+ hidden_dropout .................................. 0.0
205
+ hidden_size ..................................... 3584
206
+ hierarchical_context_parallel_sizes ............. None
207
+ high_priority_stream_groups ..................... []
208
+ hybrid_attention_ratio .......................... 0.0
209
+ hybrid_mlp_ratio ................................ 0.0
210
+ hybrid_override_pattern ......................... None
211
+ hysteresis ...................................... 2
212
+ ict_head_size ................................... None
213
+ ict_load ........................................ None
214
+ img_h ........................................... 224
215
+ img_w ........................................... 224
216
+ indexer_batch_size .............................. 128
217
+ indexer_log_interval ............................ 1000
218
+ inference_batch_times_seqlen_threshold .......... -1
219
+ inference_dynamic_batching ...................... False
220
+ inference_dynamic_batching_buffer_guaranteed_fraction 0.2
221
+ inference_dynamic_batching_buffer_overflow_factor None
222
+ inference_dynamic_batching_buffer_size_gb ....... 40.0
223
+ inference_dynamic_batching_chunk_size ........... 256
224
+ inference_dynamic_batching_max_requests_override None
225
+ inference_dynamic_batching_max_tokens_override .. None
226
+ inference_dynamic_batching_num_cuda_graphs ...... 16
227
+ inference_max_batch_size ........................ 8
228
+ inference_max_seq_length ........................ 2560
229
+ inference_rng_tracker ........................... False
230
+ init_method_std ................................. 0.01
231
+ init_method_xavier_uniform ...................... False
232
+ init_model_with_meta_device ..................... False
233
+ initial_loss_scale .............................. 4294967296
234
+ inprocess_active_world_size ..................... 8
235
+ inprocess_barrier_timeout ....................... 120
236
+ inprocess_completion_timeout .................... 120
237
+ inprocess_empty_cuda_cache ...................... False
238
+ inprocess_granularity ........................... node
239
+ inprocess_hard_timeout .......................... 90
240
+ inprocess_heartbeat_interval .................... 30
241
+ inprocess_heartbeat_timeout ..................... 60
242
+ inprocess_last_call_wait ........................ 1
243
+ inprocess_max_iterations ........................ None
244
+ inprocess_monitor_process_interval .............. 1.0
245
+ inprocess_monitor_thread_interval ............... 1.0
246
+ inprocess_progress_watchdog_interval ............ 1.0
247
+ inprocess_restart ............................... False
248
+ inprocess_soft_timeout .......................... 60
249
+ inprocess_termination_grace_time ................ 1
250
+ is_hybrid_model ................................. False
251
+ iter_per_epoch .................................. 1250
252
+ iterations_to_skip .............................. []
253
+ keep_fp8_transpose_cache_when_using_custom_fsdp . False
254
+ kitchen_config_file ............................. None
255
+ kitchen_recipe_number ........................... None
256
+ kv_channels ..................................... 128
257
+ kv_lora_rank .................................... 32
258
+ lazy_mpu_init ................................... None
259
+ load ............................................ None
260
+ load_main_params_from_ckpt ...................... None
261
+ load_model_opt_format ........................... False
262
+ local_rank ...................................... 0
263
+ log_energy ...................................... False
264
+ log_interval .................................... 1
265
+ log_loss_scale_to_tensorboard ................... True
266
+ log_memory_to_tensorboard ....................... False
267
+ log_num_zeros_in_grad ........................... False
268
+ log_params_norm ................................. False
269
+ log_progress .................................... False
270
+ log_straggler ................................... False
271
+ log_throughput .................................. True
272
+ log_timers_to_tensorboard ....................... False
273
+ log_validation_ppl_to_tensorboard ............... False
274
+ log_world_size_to_tensorboard ................... False
275
+ logging_level ................................... None
276
+ loss_scale ...................................... None
277
+ loss_scale_window ............................... 1000
278
+ lr .............................................. 0.0001
279
+ lr_decay_iters .................................. 8
280
+ lr_decay_samples ................................ None
281
+ lr_decay_style .................................. cosine
282
+ lr_warmup_fraction .............................. None
283
+ lr_warmup_init .................................. 0.0
284
+ lr_warmup_iters ................................. 5
285
+ lr_warmup_samples ............................... 0
286
+ lr_wsd_decay_iters .............................. None
287
+ lr_wsd_decay_samples ............................ None
288
+ lr_wsd_decay_style .............................. exponential
289
+ main_grads_dtype ................................ torch.float32
290
+ main_params_dtype ............................... torch.float32
291
+ make_vocab_size_divisible_by .................... 128
292
+ mamba_head_dim .................................. 64
293
+ mamba_num_groups ................................ 8
294
+ mamba_num_heads ................................. None
295
+ mamba_state_dim ................................. 128
296
+ manual_gc ....................................... False
297
+ manual_gc_eval .................................. True
298
+ manual_gc_interval .............................. 0
299
+ mask_factor ..................................... 1.0
300
+ mask_prob ....................................... 0.15
301
+ mask_type ....................................... random
302
+ masked_softmax_fusion ........................... False
303
+ max_position_embeddings ......................... 32768
304
+ max_tokens_to_oom ............................... 12000
305
+ memory_snapshot_path ............................ snapshot.pickle
306
+ merge_file ...................................... /workspace/data/wikitext/gpt2-merges.txt
307
+ micro_batch_size ................................ 1
308
+ microbatch_group_size_per_vp_stage .............. None
309
+ mid_level_dataset_surplus ....................... 0.005
310
+ min_loss_scale .................................. 1.0
311
+ min_lr .......................................... 1e-05
312
+ mlp_chunks_for_prefill .......................... 1
313
+ mmap_bin_files .................................. True
314
+ mock_data ....................................... False
315
+ moe_apply_probs_on_input ........................ False
316
+ moe_aux_loss_coeff .............................. 0.01
317
+ moe_deepep_num_sms .............................. 20
318
+ moe_echo_enable_random_offloading ............... False
319
+ moe_echo_expert_dispatch_overlap ................ True
320
+ moe_echo_expert_dispatcher_type ................. hybridep
321
+ moe_echo_recompute_expert_dispatch .............. False
322
+ moe_enable_deepep ............................... False
323
+ moe_enable_echo ................................. True
324
+ moe_enable_hybridep ............................. True
325
+ moe_expert_capacity_factor ...................... None
326
+ moe_extended_tp ................................. False
327
+ moe_ffn_hidden_size ............................. 2560
328
+ moe_grouped_gemm ................................ True
329
+ moe_hybridep_num_sms ............................ 16
330
+ moe_input_jitter_eps ............................ None
331
+ moe_layer_freq .................................. 1
332
+ moe_layer_recompute ............................. False
333
+ moe_num_echo_experts ............................ 64
334
+ moe_pad_expert_input_to_capacity ................ False
335
+ moe_per_layer_logging ........................... False
336
+ moe_permute_fusion .............................. False
337
+ moe_received_token_capacity ..................... 1.08984375
338
+ moe_router_bias_update_rate ..................... 0.001
339
+ moe_router_dtype ................................ fp32
340
+ moe_router_enable_expert_bias ................... False
341
+ moe_router_force_load_balancing ................. False
342
+ moe_router_group_topk ........................... None
343
+ moe_router_load_balancing_type .................. aux_loss
344
+ moe_router_num_groups ........................... None
345
+ moe_router_padding_for_fp8 ...................... False
346
+ moe_router_pre_softmax .......................... True
347
+ moe_router_score_function ....................... softmax
348
+ moe_router_topk ................................. 8
349
+ moe_router_topk_scaling_factor .................. None
350
+ moe_shared_expert_intermediate_size ............. 20480
351
+ moe_shared_expert_overlap ....................... False
352
+ moe_token_dispatcher_type ....................... flex
353
+ moe_token_drop_policy ........................... probs
354
+ moe_upcycling_granularity ....................... 1
355
+ moe_use_device_initiated_grouped_gemm ........... True
356
+ moe_use_legacy_grouped_gemm ..................... False
357
+ moe_use_upcycling ............................... False
358
+ moe_z_loss_coeff ................................ None
359
+ mrope_section ................................... None
360
+ mscale .......................................... 1.0
361
+ mscale_all_dim .................................. 0.0
362
+ mtp_loss_scaling_factor ......................... 0.1
363
+ mtp_num_layers .................................. None
364
+ multi_latent_attention .......................... False
365
+ nccl_all_reduce_for_prefill ..................... False
366
+ nccl_communicator_config_path ................... None
367
+ nccl_ub ......................................... False
368
+ no_load_optim ................................... None
369
+ no_load_rng ..................................... None
370
+ no_persist_layer_norm ........................... False
371
+ no_rope_freq .................................... None
372
+ no_save_optim ................................... None
373
+ no_save_rng ..................................... None
374
+ non_persistent_ckpt_type ........................ None
375
+ non_persistent_global_ckpt_dir .................. None
376
+ non_persistent_local_ckpt_algo .................. fully_parallel
377
+ non_persistent_local_ckpt_dir ................... None
378
+ non_persistent_save_interval .................... None
379
+ norm_epsilon .................................... 1e-06
380
+ normalization ................................... RMSNorm
381
+ num_attention_heads ............................. 28
382
+ num_channels .................................... 3
383
+ num_classes ..................................... 1000
384
+ num_dataset_builder_threads ..................... 1
385
+ num_distributed_optimizer_instances ............. 1
386
+ num_experts ..................................... 64
387
+ num_layers ...................................... 10
388
+ num_layers_at_end_in_bf16 ....................... 1
389
+ num_layers_at_start_in_bf16 ..................... 1
390
+ num_layers_per_virtual_pipeline_stage ........... None
391
+ num_query_groups ................................ 4
392
+ num_virtual_stages_per_pipeline_rank ............ None
393
+ num_workers ..................................... 2
394
+ object_storage_cache_path ....................... None
395
+ one_logger_async ................................ False
396
+ one_logger_project .............................. megatron-lm
397
+ one_logger_run_name ............................. None
398
+ onnx_safe ....................................... None
399
+ openai_gelu ..................................... False
400
+ optimizer ....................................... adam
401
+ optimizer_cpu_offload ........................... False
402
+ optimizer_offload_fraction ...................... 1.0
403
+ output_bert_embeddings .......................... False
404
+ overlap_cpu_optimizer_d2h_h2d ................... False
405
+ overlap_grad_reduce ............................. True
406
+ overlap_moe_expert_parallel_comm ................ False
407
+ overlap_p2p_comm ................................ False
408
+ overlap_p2p_comm_warmup_flush ................... False
409
+ overlap_param_gather ............................ True
410
+ overlap_param_gather_with_optimizer_step ........ False
411
+ override_opt_param_scheduler .................... False
412
+ params_dtype .................................... torch.bfloat16
413
+ patch_dim ....................................... 16
414
+ per_split_data_args_path ........................ None
415
+ perform_initialization .......................... True
416
+ pin_cpu_grads ................................... True
417
+ pin_cpu_params .................................. True
418
+ pipeline_model_parallel_comm_backend ............ None
419
+ pipeline_model_parallel_layout .................. None
420
+ pipeline_model_parallel_size .................... 1
421
+ position_embedding_type ......................... rope
422
+ pretrained_checkpoint ........................... None
423
+ profile ......................................... False
424
+ profile_ranks ................................... [0]
425
+ profile_step_end ................................ 12
426
+ profile_step_start .............................. 10
427
+ q_lora_rank ..................................... None
428
+ qk_head_dim ..................................... 128
429
+ qk_l2_norm ...................................... False
430
+ qk_layernorm .................................... False
431
+ qk_pos_emb_head_dim ............................. 64
432
+ query_in_block_prob ............................. 0.1
433
+ rampup_batch_size ............................... None
434
+ rank ............................................ 0
435
+ recompute_granularity ........................... None
436
+ recompute_method ................................ None
437
+ recompute_modules ............................... None
438
+ recompute_num_layers ............................ None
439
+ record_memory_history ........................... False
440
+ relative_attention_max_distance ................. 128
441
+ relative_attention_num_buckets .................. 32
442
+ replication ..................................... False
443
+ replication_factor .............................. 2
444
+ replication_jump ................................ None
445
+ rerun_mode ...................................... validate_results
446
+ reset_attention_mask ............................ False
447
+ reset_position_ids .............................. False
448
+ result_rejected_tracker_filename ................ None
449
+ retriever_report_topk_accuracies ................ []
450
+ retriever_score_scaling ......................... False
451
+ retriever_seq_length ............................ 256
452
+ retro_add_retriever ............................. False
453
+ retro_attention_gate ............................ 1
454
+ retro_cyclic_train_iters ........................ None
455
+ retro_encoder_attention_dropout ................. 0.1
456
+ retro_encoder_hidden_dropout .................... 0.1
457
+ retro_encoder_layers ............................ 2
458
+ retro_num_neighbors ............................. 2
459
+ retro_num_retrieved_chunks ...................... 2
460
+ retro_project_dir ............................... None
461
+ retro_verify_neighbor_count ..................... True
462
+ reuse_grad_buf_for_mxfp8_param_ag ............... False
463
+ rope_scaling_factor ............................. 8.0
464
+ rope_type ....................................... None
465
+ rotary_base ..................................... 1000000
466
+ rotary_interleaved .............................. False
467
+ rotary_percent .................................. 1.0
468
+ rotary_scaling_factor ........................... 1.0
469
+ rotary_seq_len_interpolation_factor ............. None
470
+ run_workload_inspector_server ................... False
471
+ sample_rate ..................................... 1.0
472
+ save ............................................ None
473
+ save_interval ................................... None
474
+ save_retain_interval ............................ None
475
+ scatter_gather_tensors_in_pipeline .............. True
476
+ seed ............................................ 1234
477
+ seq_length ...................................... 4096
478
+ sequence_parallel ............................... False
479
+ sft ............................................. False
480
+ sft_tokenizer_prompt_format ..................... nemotron-h-aligned
481
+ sgd_momentum .................................... 0.9
482
+ short_seq_prob .................................. 0.1
483
+ skip_train ...................................... False
484
+ skipped_train_samples ........................... 0
485
+ spec ............................................ None
486
+ split ........................................... 100,0,0
487
+ squared_relu .................................... False
488
+ start_weight_decay .............................. 0.1
489
+ straggler_ctrlr_port ............................ 65535
490
+ straggler_minmax_count .......................... 1
491
+ suggested_communication_unit_size ............... None
492
+ swiglu .......................................... True
493
+ swin_backbone_type .............................. tiny
494
+ symmetric_ar_type ............................... None
495
+ te_rng_tracker .................................. True
496
+ tensor_model_parallel_size ...................... 1
497
+ tensorboard_dir ................................. None
498
+ tensorboard_log_interval ........................ 1
499
+ tensorboard_queue_size .......................... 1000
500
+ test_data_path .................................. None
501
+ test_mode ....................................... False
502
+ tiktoken_num_special_tokens ..................... 1000
503
+ tiktoken_pattern ................................ None
504
+ tiktoken_special_tokens ......................... None
505
+ timing_log_level ................................ 0
506
+ timing_log_option ............................... minmax
507
+ titles_data_path ................................ None
508
+ tokenizer_model ................................. None
509
+ tokenizer_type .................................. GPT2BPETokenizer
510
+ torch_fsdp2_reshard_after_forward ............... True
511
+ tp_comm_bootstrap_backend ....................... nccl
512
+ tp_comm_bulk_dgrad .............................. True
513
+ tp_comm_bulk_wgrad .............................. True
514
+ tp_comm_overlap ................................. False
515
+ tp_comm_overlap_ag .............................. True
516
+ tp_comm_overlap_cfg ............................. None
517
+ tp_comm_overlap_rs .............................. True
518
+ tp_comm_overlap_rs_dgrad ........................ False
519
+ tp_comm_split_ag ................................ True
520
+ tp_comm_split_rs ................................ True
521
+ train_data_path ................................. None
522
+ train_iters ..................................... 12
523
+ train_samples ................................... None
524
+ train_sync_interval ............................. None
525
+ transformer_impl ................................ transformer_engine
526
+ transformer_pipeline_model_parallel_size ........ 1
527
+ untie_embeddings_and_output_weights ............. True
528
+ use_checkpoint_args ............................. False
529
+ use_checkpoint_opt_param_scheduler .............. False
530
+ use_cpu_initialization .......................... None
531
+ use_custom_fsdp ................................. False
532
+ use_dist_ckpt ................................... True
533
+ use_dist_ckpt_deprecated ........................ False
534
+ use_distributed_optimizer ....................... True
535
+ use_flash_attn .................................. False
536
+ use_legacy_models ............................... False
537
+ use_mp_args_from_checkpoint_args ................ False
538
+ use_one_sent_docs ............................... False
539
+ use_persistent_ckpt_worker ...................... False
540
+ use_precision_aware_optimizer ................... False
541
+ use_pytorch_profiler ............................ False
542
+ use_ring_exchange_p2p ........................... False
543
+ use_rope_scaling ................................ False
544
+ use_rotary_position_embeddings .................. False
545
+ use_sharp ....................................... False
546
+ use_tokenizer_model_from_checkpoint_args ........ True
547
+ use_torch_fsdp2 ................................. False
548
+ use_torch_optimizer_for_cpu_offload ............. False
549
+ use_tp_pp_dp_mapping ............................ False
550
+ v_head_dim ...................................... 128
551
+ valid_data_path ................................. None
552
+ variable_seq_lengths ............................ False
553
+ virtual_pipeline_model_parallel_size ............ None
554
+ vision_backbone_type ............................ vit
555
+ vision_pretraining .............................. False
556
+ vision_pretraining_type ......................... classify
557
+ vocab_extra_ids ................................. 0
558
+ vocab_file ...................................... /workspace/data/wikitext/gpt2-vocab.json
559
+ vocab_size ...................................... None
560
+ wandb_exp_name ..................................
561
+ wandb_project ...................................
562
+ wandb_save_dir ..................................
563
+ weight_decay .................................... 0.1
564
+ weight_decay_incr_style ......................... constant
565
+ wgrad_deferral_limit ............................ 0
566
+ world_size ...................................... 8
567
+ yaml_cfg ........................................ None
568
+ -------------------- end of arguments ---------------------
569
+ INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 1
570
+ > building GPT2BPETokenizer tokenizer ...
571
+ /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
+ > padded vocab (size: 50257) with 47 dummy tokens (new size: 50304)
574
+ WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.VALIDATE_RESULTS
575
+ > initializing torch distributed ...
576
+ [Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7[Gloo] Rank
577
+ 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
578
+ [Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
579
+ [Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
580
+ [Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
581
+ [Gloo] Rank 5 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
582
+ [Gloo] Rank [Gloo] Rank 6 is connected to 77 is connected to peer ranks. 7Expected number of connected peer ranks is : peer ranks. 7Expected number of connected peer ranks is :
583
+ 7
584
+ [Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
585
+ [Gloo] Rank [Gloo] Rank [Gloo] Rank 23 is connected to 5 is connected to 7 is connected to 7 peer ranks. 7 peer ranks. Expected number of connected peer ranks is : peer ranks. Expected number of connected peer ranks is : 7Expected number of connected peer ranks is : 7
586
+ 7
587
+
588
+ [Gloo] Rank [Gloo] Rank 1 is connected to 67 is connected to peer ranks. 7Expected number of connected peer ranks is : peer ranks. 7Expected number of connected peer ranks is :
589
+ 7
590
+ [Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
591
+ [Gloo] Rank 7 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
592
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
593
+ > initialized tensor model parallel with size 1
594
+ > initialized pipeline model parallel with size 1
595
+ > setting random seeds to 1234 ...
596
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
597
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
598
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
599
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
600
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
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
+ [rank5]:[W826 18:25:10.560747491 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.
604
+ [rank2]:[W826 18:25:10.560748031 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.
605
+ [rank3]:[W826 18:25:10.562223435 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.
606
+ [rank6]:[W826 18:25:10.568096140 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
+ [rank4]:[W826 18:25:10.568096628 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
608
+ [rank7]:[W826 18:25:10.570484045 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
609
+ [rank1]:[W826 18:25:10.571444131 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 1] using GPU 1 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
610
+ > compiling dataset index builder ...
611
+ make: Entering directory '/workspace/Megatron-MoE-Echo/megatron/core/datasets'
612
+ make: Nothing to be done for 'default'.
613
+ make: Leaving directory '/workspace/Megatron-MoE-Echo/megatron/core/datasets'
614
+ >>> done with dataset index builder. Compilation time: 0.034 seconds
615
+ WARNING: constraints for invoking optimized fused softmax kernel are not met. We default back to unfused kernel invocations.
616
+ > compiling and loading fused kernels ...
617
+ [rank0]:[W826 18:25:10.717747320 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.
618
+ >>> done with compiling and loading fused kernels. Compilation time: 3.640 seconds
619
+ time to initialize megatron (seconds): 6.656
620
+ [after megatron is initialized] datetime: 2026-08-26 18:25:16
621
+ building GPT model ...
622
+ /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.
623
+ warnings.warn(
624
+ /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.
625
+ warnings.warn(
626
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
627
+ warnings.warn(
628
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
629
+ warnings.warn(
630
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
631
+ warnings.warn(
632
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
633
+ warnings.warn(
634
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
635
+ warnings.warn(
636
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
637
+ warnings.warn(
638
+ > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 5060614656
639
+ 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)
640
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 22
641
+ Params for bucket 1 (180289536 elements, 180289536 padded size):
642
+ module.output_layer.weight
643
+ Params for bucket 2 (73403904 elements, 73403904 padded size):
644
+ module.decoder.final_layernorm.weight
645
+ module.decoder.layers.9.mlp.shared_experts.linear_fc2.weight
646
+ Params for bucket 3 (146800640 elements, 146800640 padded size):
647
+ module.decoder.layers.9.mlp.shared_experts.linear_fc1.weight
648
+ Params for bucket 4 (103001600 elements, 103001600 padded size):
649
+ module.decoder.layers.9.pre_mlp_layernorm.weight
650
+ module.decoder.layers.9.self_attention.linear_qkv.bias
651
+ module.decoder.layers.9.mlp.router.weight
652
+ module.decoder.layers.9.self_attention.linear_qkv.weight
653
+ module.decoder.layers.9.self_attention.linear_qkv.layer_norm_weight
654
+ module.decoder.layers.9.self_attention.linear_proj.weight
655
+ module.decoder.layers.8.mlp.shared_experts.linear_fc2.weight
656
+ Params for bucket 5 (146800640 elements, 146800640 padded size):
657
+ module.decoder.layers.8.mlp.shared_experts.linear_fc1.weight
658
+ Params for bucket 6 (103001600 elements, 103001600 padded size):
659
+ module.decoder.layers.8.mlp.router.weight
660
+ module.decoder.layers.8.self_attention.linear_qkv.weight
661
+ module.decoder.layers.7.mlp.shared_experts.linear_fc2.weight
662
+ module.decoder.layers.8.self_attention.linear_qkv.layer_norm_weight
663
+ module.decoder.layers.8.pre_mlp_layernorm.weight
664
+ module.decoder.layers.8.self_attention.linear_qkv.bias
665
+ module.decoder.layers.8.self_attention.linear_proj.weight
666
+ Params for bucket 7 (146800640 elements, 146800640 padded size):
667
+ module.decoder.layers.7.mlp.shared_experts.linear_fc1.weight
668
+ Params for bucket 8 (103001600 elements, 103001600 padded size):
669
+ module.decoder.layers.7.mlp.router.weight
670
+ module.decoder.layers.7.self_attention.linear_qkv.layer_norm_weight
671
+ module.decoder.layers.7.self_attention.linear_qkv.weight
672
+ module.decoder.layers.7.self_attention.linear_proj.weight
673
+ module.decoder.layers.7.self_attention.linear_qkv.bias
674
+ module.decoder.layers.7.pre_mlp_layernorm.weight
675
+ module.decoder.layers.6.mlp.shared_experts.linear_fc2.weight
676
+ Params for bucket 9 (146800640 elements, 146800640 padded size):
677
+ module.decoder.layers.6.mlp.shared_experts.linear_fc1.weight
678
+ Params for bucket 10 (103001600 elements, 103001600 padded size):
679
+ module.decoder.layers.6.mlp.router.weight
680
+ module.decoder.layers.6.self_attention.linear_qkv.layer_norm_weight
681
+ module.decoder.layers.6.self_attention.linear_qkv.weight
682
+ module.decoder.layers.6.self_attention.linear_proj.weight
683
+ module.decoder.layers.5.mlp.shared_experts.linear_fc2.weight
684
+ module.decoder.layers.6.pre_mlp_layernorm.weight
685
+ module.decoder.layers.6.self_attention.linear_qkv.bias
686
+ Params for bucket 11 (146800640 elements, 146800640 padded size):
687
+ module.decoder.layers.5.mlp.shared_experts.linear_fc1.weight
688
+ Params for bucket 12 (103001600 elements, 103001600 padded size):
689
+ module.decoder.layers.5.mlp.router.weight
690
+ module.decoder.layers.5.self_attention.linear_qkv.layer_norm_weight
691
+ module.decoder.layers.5.self_attention.linear_qkv.weight
692
+ module.decoder.layers.5.self_attention.linear_proj.weight
693
+ module.decoder.layers.4.mlp.shared_experts.linear_fc2.weight
694
+ module.decoder.layers.5.self_attention.linear_qkv.bias
695
+ module.decoder.layers.5.pre_mlp_layernorm.weight
696
+ Params for bucket 13 (146800640 elements, 146800640 padded size):
697
+ module.decoder.layers.4.mlp.shared_experts.linear_fc1.weight
698
+ Params for bucket 14 (103001600 elements, 103001600 padded size):
699
+ module.decoder.layers.4.mlp.router.weight
700
+ module.decoder.layers.4.self_attention.linear_qkv.bias
701
+ module.decoder.layers.4.pre_mlp_layernorm.weight
702
+ module.decoder.layers.4.self_attention.linear_qkv.layer_norm_weight
703
+ module.decoder.layers.4.self_attention.linear_proj.weight
704
+ module.decoder.layers.3.mlp.shared_experts.linear_fc2.weight
705
+ module.decoder.layers.4.self_attention.linear_qkv.weight
706
+ Params for bucket 15 (146800640 elements, 146800640 padded size):
707
+ module.decoder.layers.3.mlp.shared_experts.linear_fc1.weight
708
+ Params for bucket 16 (103001600 elements, 103001600 padded size):
709
+ module.decoder.layers.3.self_attention.linear_qkv.layer_norm_weight
710
+ module.decoder.layers.3.pre_mlp_layernorm.weight
711
+ module.decoder.layers.3.self_attention.linear_qkv.bias
712
+ module.decoder.layers.3.self_attention.linear_proj.weight
713
+ module.decoder.layers.2.mlp.shared_experts.linear_fc2.weight
714
+ module.decoder.layers.3.mlp.router.weight
715
+ module.decoder.layers.3.self_attention.linear_qkv.weight
716
+ Params for bucket 17 (146800640 elements, 146800640 padded size):
717
+ module.decoder.layers.2.mlp.shared_experts.linear_fc1.weight
718
+ Params for bucket 18 (103001600 elements, 103001600 padded size):
719
+ module.decoder.layers.2.pre_mlp_layernorm.weight
720
+ module.decoder.layers.2.self_attention.linear_qkv.layer_norm_weight
721
+ module.decoder.layers.2.self_attention.linear_qkv.bias
722
+ module.decoder.layers.2.self_attention.linear_proj.weight
723
+ module.decoder.layers.1.mlp.shared_experts.linear_fc2.weight
724
+ module.decoder.layers.2.mlp.router.weight
725
+ module.decoder.layers.2.self_attention.linear_qkv.weight
726
+ Params for bucket 19 (146800640 elements, 146800640 padded size):
727
+ module.decoder.layers.1.mlp.shared_experts.linear_fc1.weight
728
+ Params for bucket 20 (103001600 elements, 103001600 padded size):
729
+ module.decoder.layers.1.mlp.router.weight
730
+ module.decoder.layers.1.pre_mlp_layernorm.weight
731
+ module.decoder.layers.1.self_attention.linear_qkv.bias
732
+ module.decoder.layers.1.self_attention.linear_qkv.layer_norm_weight
733
+ module.decoder.layers.1.self_attention.linear_qkv.weight
734
+ module.decoder.layers.1.self_attention.linear_proj.weight
735
+ module.decoder.layers.0.mlp.shared_experts.linear_fc2.weight
736
+ Params for bucket 21 (146800640 elements, 146800640 padded size):
737
+ module.decoder.layers.0.mlp.shared_experts.linear_fc1.weight
738
+ Params for bucket 22 (209890816 elements, 209890816 padded size):
739
+ module.decoder.layers.0.pre_mlp_layernorm.weight
740
+ module.decoder.layers.0.self_attention.linear_qkv.weight
741
+ module.decoder.layers.0.self_attention.linear_qkv.bias
742
+ module.decoder.layers.0.self_attention.linear_proj.weight
743
+ module.decoder.layers.0.mlp.router.weight
744
+ module.decoder.layers.0.self_attention.linear_qkv.layer_norm_weight
745
+ module.embedding.word_embeddings.weight
746
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 44
747
+ Params for bucket 1 (45875200 elements, 45875200 padded size):
748
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight7
749
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight3
750
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight5
751
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight4
752
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight6
753
+ Params for bucket 2 (45875200 elements, 45875200 padded size):
754
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight2
755
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight0
756
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight7
757
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight1
758
+ Params for bucket 3 (55050240 elements, 55050240 padded size):
759
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight6
760
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight4
761
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight5
762
+ Params for bucket 4 (55050240 elements, 55050240 padded size):
763
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight3
764
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight1
765
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight2
766
+ Params for bucket 5 (45875200 elements, 45875200 padded size):
767
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight0
768
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight5
769
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight6
770
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight7
771
+ Params for bucket 6 (45875200 elements, 45875200 padded size):
772
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight3
773
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight1
774
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight4
775
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight2
776
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight0
777
+ Params for bucket 7 (55050240 elements, 55050240 padded size):
778
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight7
779
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight5
780
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight6
781
+ Params for bucket 8 (55050240 elements, 55050240 padded size):
782
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight4
783
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight2
784
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight3
785
+ Params for bucket 9 (45875200 elements, 45875200 padded size):
786
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight1
787
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight7
788
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight0
789
+ Params for bucket 10 (45875200 elements, 45875200 padded size):
790
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight3
791
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight5
792
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight6
793
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight2
794
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight4
795
+ Params for bucket 11 (55050240 elements, 55050240 padded size):
796
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight1
797
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight0
798
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight6
799
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight7
800
+ Params for bucket 12 (55050240 elements, 55050240 padded size):
801
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight5
802
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight3
803
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight4
804
+ Params for bucket 13 (55050240 elements, 55050240 padded size):
805
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight2
806
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight0
807
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight1
808
+ Params for bucket 14 (45875200 elements, 45875200 padded size):
809
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight4
810
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight6
811
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight7
812
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight3
813
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight5
814
+ Params for bucket 15 (45875200 elements, 45875200 padded size):
815
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight2
816
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight1
817
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight7
818
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight0
819
+ Params for bucket 16 (55050240 elements, 55050240 padded size):
820
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight6
821
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight4
822
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight5
823
+ Params for bucket 17 (55050240 elements, 55050240 padded size):
824
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight3
825
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight1
826
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight2
827
+ Params for bucket 18 (45875200 elements, 45875200 padded size):
828
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight0
829
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight6
830
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight7
831
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight5
832
+ Params for bucket 19 (45875200 elements, 45875200 padded size):
833
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight4
834
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight2
835
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight0
836
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight3
837
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight1
838
+ Params for bucket 20 (55050240 elements, 55050240 padded size):
839
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight7
840
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight5
841
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight6
842
+ Params for bucket 21 (55050240 elements, 55050240 padded size):
843
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight4
844
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight2
845
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight3
846
+ Params for bucket 22 (45875200 elements, 45875200 padded size):
847
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight1
848
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight0
849
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight7
850
+ Params for bucket 23 (45875200 elements, 45875200 padded size):
851
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight6
852
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight4
853
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight2
854
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight5
855
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight3
856
+ Params for bucket 24 (55050240 elements, 55050240 padded size):
857
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight1
858
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight7
859
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight6
860
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight0
861
+ Params for bucket 25 (55050240 elements, 55050240 padded size):
862
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight5
863
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight4
864
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight3
865
+ Params for bucket 26 (55050240 elements, 55050240 padded size):
866
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight2
867
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight1
868
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight0
869
+ Params for bucket 27 (45875200 elements, 45875200 padded size):
870
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight6
871
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight4
872
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight7
873
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight3
874
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight5
875
+ Params for bucket 28 (45875200 elements, 45875200 padded size):
876
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight2
877
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight7
878
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight0
879
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight1
880
+ Params for bucket 29 (55050240 elements, 55050240 padded size):
881
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight6
882
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight4
883
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight5
884
+ Params for bucket 30 (55050240 elements, 55050240 padded size):
885
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight3
886
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight1
887
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight2
888
+ Params for bucket 31 (45875200 elements, 45875200 padded size):
889
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight0
890
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight6
891
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight7
892
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight5
893
+ Params for bucket 32 (45875200 elements, 45875200 padded size):
894
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight4
895
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight0
896
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight2
897
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight3
898
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight1
899
+ Params for bucket 33 (55050240 elements, 55050240 padded size):
900
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight7
901
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight6
902
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight5
903
+ Params for bucket 34 (55050240 elements, 55050240 padded size):
904
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight4
905
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight3
906
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight2
907
+ Params for bucket 35 (45875200 elements, 45875200 padded size):
908
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight1
909
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight0
910
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight7
911
+ Params for bucket 36 (45875200 elements, 45875200 padded size):
912
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight6
913
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight2
914
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight4
915
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight3
916
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight5
917
+ Params for bucket 37 (55050240 elements, 55050240 padded size):
918
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight1
919
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight6
920
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight0
921
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight7
922
+ Params for bucket 38 (55050240 elements, 55050240 padded size):
923
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight5
924
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight3
925
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight4
926
+ Params for bucket 39 (55050240 elements, 55050240 padded size):
927
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight2
928
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight0
929
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight1
930
+ Params for bucket 40 (45875200 elements, 45875200 padded size):
931
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight6
932
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight4
933
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight3
934
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight7
935
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight5
936
+ Params for bucket 41 (45875200 elements, 45875200 padded size):
937
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight2
938
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight1
939
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight0
940
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight7
941
+ Params for bucket 42 (55050240 elements, 55050240 padded size):
942
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight6
943
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight4
944
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight5
945
+ Params for bucket 43 (55050240 elements, 55050240 padded size):
946
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight3
947
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight1
948
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight2
949
+ Params for bucket 44 (18350080 elements, 18350080 padded size):
950
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight0
951
+ 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 0x7ccb642dfec0>, config_logger_dir='')
952
+ INFO:megatron.core.optimizer_param_scheduler:> learning rate decay style: cosine
953
+ [after model, optimizer, and learning rate scheduler are built] datetime: 2026-08-26 18:25:16
954
+ > building train, validation, and test datasets ...
955
+ > datasets target sizes (minimum size):
956
+ train: 96
957
+ validation: 0
958
+ test: 0
959
+ INFO:megatron.core.datasets.blended_megatron_dataset_config:Let split_matrix = [(0, 1.0), None, None]
960
+ > building train, validation, and test datasets for GPT ...
961
+ 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 0x7ccb64382db0>, 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)
962
+ INFO:megatron.core.datasets.indexed_dataset:Load the _IndexReader from /workspace/data/wikitext/wikitext103_gpt2_text_document.idx
963
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence lengths
964
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence pointers
965
+ INFO:megatron.core.datasets.indexed_dataset: Extract the document indices
966
+ INFO:megatron.core.datasets.indexed_dataset:> total number of sequences: 1165029
967
+ INFO:megatron.core.datasets.indexed_dataset:> total number of documents: 1801350
968
+ INFO:megatron.core.datasets.gpt_dataset:Load the GPTDataset train indices
969
+ INFO:megatron.core.datasets.gpt_dataset: Load the document index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-document_index.npy
970
+ INFO:megatron.core.datasets.gpt_dataset: Load the sample index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-sample_index.npy
971
+ INFO:megatron.core.datasets.gpt_dataset: Load the shuffle index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-shuffle_index.npy
972
+ INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 29073
973
+ > finished creating GPT datasets ...
974
+ [after dataloaders are built] datetime: 2026-08-26 18:25:16
975
+ done with setup ...
976
+ training ...
977
+ Overwriting rerun_state_machine.current_iteration from -1 to 0...
978
+ (min, max) time across ranks (ms):
979
+ model-and-optimizer-setup ......................: (329.75, 334.86)
980
+ train/valid/test-data-iterators-setup ..........: (58.47, 75.27)
981
+ [before the start of training step] datetime: 2026-08-26 18:25:16
982
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: 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:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
985
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
986
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: 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:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
989
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
990
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: 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:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
993
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
994
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: 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
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
997
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
998
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
999
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1000
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1001
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1002
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1003
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1004
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1005
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1006
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1007
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1008
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1009
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1010
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1011
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1012
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1013
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1014
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1015
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1016
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1017
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1018
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1019
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1020
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1021
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1022
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1023
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1024
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1025
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1026
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1027
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1028
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1029
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1030
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1031
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1032
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1033
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1034
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1035
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1036
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1037
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1038
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1039
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1040
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1041
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1042
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1043
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1044
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1045
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1046
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1047
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1048
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1049
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1050
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1051
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1052
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1053
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1054
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1055
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1056
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1057
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1058
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1059
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1060
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1061
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1062
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1063
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1064
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1065
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1066
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1067
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1068
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1069
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1070
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1071
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1072
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1073
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1074
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1075
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1076
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1077
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1078
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1079
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1080
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1081
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1082
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1083
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1084
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1085
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1086
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1087
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1088
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1089
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1090
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1091
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1092
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1093
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1094
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1095
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1096
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1097
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1098
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1099
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1100
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1101
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1102
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1103
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1104
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1105
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1106
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1107
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1108
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1109
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1110
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1111
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1112
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1113
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1114
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1115
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1116
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1117
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1118
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1119
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1120
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1121
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1122
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1123
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1124
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1125
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1126
+ [2026-08-26 18:25:54] iteration 1/ 12 | consumed samples: 8 | elapsed time per iteration (ms): 37967.5 | throughput per GPU (TFLOP/s/GPU): 3.3 | learning rate: 2.000000E-05 | global batch size: 8 | lm loss: 1.097931E+01 | loss scale: 1.0 | grad norm: 29.866 | number of skipped iterations: 0 | number of nan iterations: 0 |
1127
+ Number of parameters in transformer block in billions: 20.11
1128
+ Number of parameters in embedding layers in billions: 0.36
1129
+ Total number of parameters in billions: 20.47
1130
+ Number of parameters in most loaded shard in billions: 20.4724
1131
+ Theoretical memory footprints: weight and optimizer=146430.14 MB
1132
+ [Rank 0] (after 1 iterations) memory (MB) | allocated: 68733.94384765625 | max allocated: 70397.458984375 | reserved: 80856.0 | max reserved: 80856.0
1133
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1134
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1135
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1136
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1137
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1138
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1139
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1140
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1141
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1142
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1143
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1144
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1145
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1146
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1147
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1148
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1149
+ [2026-08-26 18:25:55] iteration 2/ 12 | consumed samples: 16 | elapsed time per iteration (ms): 340.9 | throughput per GPU (TFLOP/s/GPU): 362.2 | learning rate: 4.000000E-05 | global batch size: 8 | lm loss: 1.097906E+01 | loss scale: 1.0 | grad norm: 30.626 | number of skipped iterations: 0 | number of nan iterations: 0 |
1150
+ [2026-08-26 18:25:55] iteration 3/ 12 | consumed samples: 24 | elapsed time per iteration (ms): 228.8 | throughput per GPU (TFLOP/s/GPU): 539.8 | learning rate: 6.000000E-05 | global batch size: 8 | lm loss: 9.538590E+00 | loss scale: 1.0 | grad norm: 32.403 | number of skipped iterations: 0 | number of nan iterations: 0 |
1151
+ Capture CUDA graph for training!!!
1152
+ CUDA graph capture done!!!
1153
+ [2026-08-26 18:25:57] iteration 4/ 12 | consumed samples: 32 | elapsed time per iteration (ms): 1743.0 | throughput per GPU (TFLOP/s/GPU): 70.8 | learning rate: 8.000000E-05 | global batch size: 8 | lm loss: 1.108131E+01 | loss scale: 1.0 | grad norm: 106.897 | number of skipped iterations: 0 | number of nan iterations: 0 |
1154
+ [2026-08-26 18:25:57] iteration 5/ 12 | consumed samples: 40 | elapsed time per iteration (ms): 184.5 | throughput per GPU (TFLOP/s/GPU): 669.3 | learning rate: 1.000000E-04 | global batch size: 8 | lm loss: 1.000114E+01 | loss scale: 1.0 | grad norm: 26.488 | number of skipped iterations: 0 | number of nan iterations: 0 |
1155
+ [2026-08-26 18:25:57] iteration 6/ 12 | consumed samples: 48 | elapsed time per iteration (ms): 180.3 | throughput per GPU (TFLOP/s/GPU): 684.8 | learning rate: 7.750000E-05 | global batch size: 8 | lm loss: 1.049881E+01 | loss scale: 1.0 | grad norm: 8.722 | number of skipped iterations: 0 | number of nan iterations: 0 |
1156
+ [2026-08-26 18:25:57] iteration 7/ 12 | consumed samples: 56 | elapsed time per iteration (ms): 178.8 | throughput per GPU (TFLOP/s/GPU): 690.8 | learning rate: 3.250000E-05 | global batch size: 8 | lm loss: 9.445707E+00 | loss scale: 1.0 | grad norm: 5.501 | number of skipped iterations: 0 | number of nan iterations: 0 |
1157
+ [2026-08-26 18:25:57] iteration 8/ 12 | consumed samples: 64 | elapsed time per iteration (ms): 178.6 | throughput per GPU (TFLOP/s/GPU): 691.6 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.971989E+00 | loss scale: 1.0 | grad norm: 4.568 | number of skipped iterations: 0 | number of nan iterations: 0 |
1158
+ [2026-08-26 18:25:57] iteration 9/ 12 | consumed samples: 72 | elapsed time per iteration (ms): 178.7 | throughput per GPU (TFLOP/s/GPU): 691.2 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.757829E+00 | loss scale: 1.0 | grad norm: 4.317 | number of skipped iterations: 0 | number of nan iterations: 0 |
1159
+ [2026-08-26 18:25:58] iteration 10/ 12 | consumed samples: 80 | elapsed time per iteration (ms): 178.6 | throughput per GPU (TFLOP/s/GPU): 691.6 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.698141E+00 | loss scale: 1.0 | grad norm: 4.447 | number of skipped iterations: 0 | number of nan iterations: 0 |
1160
+ [2026-08-26 18:25:58] iteration 11/ 12 | consumed samples: 88 | elapsed time per iteration (ms): 177.2 | throughput per GPU (TFLOP/s/GPU): 696.9 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.636470E+00 | loss scale: 1.0 | grad norm: 4.370 | number of skipped iterations: 0 | number of nan iterations: 0 |
1161
+ [2026-08-26 18:25:58] iteration 12/ 12 | consumed samples: 96 | elapsed time per iteration (ms): 178.5 | throughput per GPU (TFLOP/s/GPU): 692.0 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.561048E+00 | loss scale: 1.0 | grad norm: 4.308 | number of skipped iterations: 0 | number of nan iterations: 0 |
1162
+ [after training is done] datetime: 2026-08-26 18:25:58
1163
+ [rank0]:[W826 18:25:59.530637261 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())
1164
+ W0826 18:26:00.746000 502750 torch/distributed/elastic/agent/server/api.py:719] Received 15 death signal, shutting down workers
1165
+ W0826 18:26:00.746000 502750 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 502862 closing signal SIGTERM
1166
+ W0826 18:26:00.748000 502750 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 502863 closing signal SIGTERM
1167
+ W0826 18:26:00.749000 502750 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 502864 closing signal SIGTERM
1168
+ W0826 18:26:00.750000 502750 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 502865 closing signal SIGTERM
1169
+ W0826 18:26:00.750000 502750 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 502866 closing signal SIGTERM
1170
+ W0826 18:26:00.750000 502750 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 502867 closing signal SIGTERM
1171
+ W0826 18:26:00.751000 502750 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 502868 closing signal SIGTERM
1172
+ W0826 18:26:00.752000 502750 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 502869 closing signal SIGTERM
1173
+ Traceback (most recent call last):
1174
+ File "/usr/local/bin/torchrun", line 8, in <module>
1175
+ sys.exit(main())
1176
+ ^^^^^^
1177
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper
1178
+ return f(*args, **kwargs)
1179
+ ^^^^^^^^^^^^^^^^^^
1180
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 892, in main
1181
+ run(args)
1182
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 883, in run
1183
+ elastic_launch(
1184
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 139, in __call__
1185
+ return launch_agent(self._config, self._entrypoint, list(args))
1186
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1187
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 261, in launch_agent
1188
+ result = agent.run()
1189
+ ^^^^^^^^^^^
1190
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/metrics/api.py", line 138, in wrapper
1191
+ result = f(*args, **kwargs)
1192
+ ^^^^^^^^^^^^^^^^^^
1193
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 711, in run
1194
+ result = self._invoke_run(role)
1195
+ ^^^^^^^^^^^^^^^^^^^^^^
1196
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 870, in _invoke_run
1197
+ time.sleep(monitor_interval)
1198
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/api.py", line 84, in _terminate_process_handler
1199
+ raise SignalException(f"Process {os.getpid()} got signal: {sigval}", sigval=sigval)
1200
+ torch.distributed.elastic.multiprocessing.api.SignalException: Process 502750 got signal: 15
echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep3/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, 2
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, 821
50
+ 1, 821
51
+ 2, 821
52
+ 3, 821
53
+ 4, 821
54
+ 5, 821
55
+ 6, 821
56
+ 7, 821
57
+ 0, 1803
58
+ 1, 1803
59
+ 2, 1803
60
+ 3, 1803
61
+ 4, 1803
62
+ 5, 1803
63
+ 6, 1803
64
+ 7, 1803
65
+ 0, 43735
66
+ 1, 43185
67
+ 2, 43161
68
+ 3, 43875
69
+ 4, 43887
70
+ 5, 43931
71
+ 6, 43885
72
+ 7, 43859
73
+ 0, 44557
74
+ 1, 43989
75
+ 2, 43989
76
+ 3, 44689
77
+ 4, 44689
78
+ 5, 44689
79
+ 6, 44689
80
+ 7, 44689
81
+ 0, 47239
82
+ 1, 46671
83
+ 2, 46671
84
+ 3, 47371
85
+ 4, 47371
86
+ 5, 47371
87
+ 6, 47371
88
+ 7, 47371
89
+ 0, 47519
90
+ 1, 46951
91
+ 2, 46951
92
+ 3, 47651
93
+ 4, 47651
94
+ 5, 47651
95
+ 6, 47651
96
+ 7, 47651
97
+ 0, 47917
98
+ 1, 47349
99
+ 2, 47349
100
+ 3, 48043
101
+ 4, 48049
102
+ 5, 48015
103
+ 6, 48037
104
+ 7, 48049
105
+ 0, 48553
106
+ 1, 47985
107
+ 2, 47985
108
+ 3, 48685
109
+ 4, 48685
110
+ 5, 48685
111
+ 6, 48685
112
+ 7, 48685
113
+ 0, 48799
114
+ 1, 48231
115
+ 2, 48231
116
+ 3, 48931
117
+ 4, 48931
118
+ 5, 48931
119
+ 6, 48931
120
+ 7, 48931
121
+ 0, 49933
122
+ 1, 49365
123
+ 2, 49365
124
+ 3, 50065
125
+ 4, 50065
126
+ 5, 50065
127
+ 6, 50065
128
+ 7, 50065
129
+ 0, 49933
130
+ 1, 49365
131
+ 2, 49365
132
+ 3, 50065
133
+ 4, 50065
134
+ 5, 50065
135
+ 6, 50065
136
+ 7, 50065
137
+ 0, 77611
138
+ 1, 77601
139
+ 2, 77575
140
+ 3, 77639
141
+ 4, 77595
142
+ 5, 77611
143
+ 6, 77615
144
+ 7, 77613
145
+ 0, 78077
146
+ 1, 78061
147
+ 2, 78059
148
+ 3, 78081
149
+ 4, 78081
150
+ 5, 78081
151
+ 6, 78081
152
+ 7, 78081
153
+ 0, 78427
154
+ 1, 78411
155
+ 2, 78409
156
+ 3, 78431
157
+ 4, 78431
158
+ 5, 78431
159
+ 6, 78431
160
+ 7, 78431
161
+ 0, 78427
162
+ 1, 78411
163
+ 2, 78409
164
+ 3, 78431
165
+ 4, 78431
166
+ 5, 78431
167
+ 6, 78431
168
+ 7, 78431
169
+ 0, 78427
170
+ 1, 78411
171
+ 2, 78409
172
+ 3, 78431
173
+ 4, 78431
174
+ 5, 78431
175
+ 6, 78431
176
+ 7, 78431
177
+ 0, 79541
178
+ 1, 79525
179
+ 2, 79523
180
+ 3, 79545
181
+ 4, 79545
182
+ 5, 79545
183
+ 6, 79545
184
+ 7, 79545
185
+ 0, 82257
186
+ 1, 82557
187
+ 2, 82167
188
+ 3, 81917
189
+ 4, 82477
190
+ 5, 82897
191
+ 6, 81881
192
+ 7, 82125
193
+ 0, 89439
194
+ 1, 89589
195
+ 2, 89375
196
+ 3, 88861
197
+ 4, 89411
198
+ 5, 89673
199
+ 6, 88831
200
+ 7, 89087
201
+ 0, 112265
202
+ 1, 111663
203
+ 2, 112055
204
+ 3, 112321
205
+ 4, 112607
206
+ 5, 112607
207
+ 6, 112327
208
+ 7, 112301
209
+ 0, 112291
210
+ 1, 111969
211
+ 2, 112055
212
+ 3, 112327
213
+ 4, 112607
214
+ 5, 112607
215
+ 6, 112327
216
+ 7, 112327
217
+ 0, 111397
218
+ 1, 3662
219
+ 2, 2382
220
+ 3, 2638
221
+ 4, 111713
222
+ 5, 3918
223
+ 6, 111433
224
+ 7, 111433
225
+ 0, 0
226
+ 1, 0
227
+ 2, 0
228
+ 3, 0
229
+ 4, 0
230
+ 5, 0
231
+ 6, 0
232
+ 7, 0
echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep3/run.env ADDED
@@ -0,0 +1,61 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ backend=echo
2
+ config=QWEN2_57B
3
+ family=tierB
4
+ repo=/workspace/Megatron-MoE-Echo
5
+ branch=bracing-regions-through-kernels
6
+ commit=9a236793
7
+ num_layers=10
8
+ hidden=3584
9
+ ffn_hidden=2560
10
+ moe_ffn_hidden=2560
11
+ num_experts=64
12
+ topk=8
13
+ experts_per_rank=8
14
+ expert_params=27525120
15
+ heads=28
16
+ kv_groups=4
17
+ kv_channels=
18
+ qk_layernorm=0
19
+ shared_ffn=20480
20
+ shared_expert_overlap=0
21
+ mla=0
22
+ dense_first_layers=0
23
+ moe_layers=10
24
+ rotary_base=1000000
25
+ rotary_scaling=
26
+ qkv_bias=1
27
+ router_balancing=aux_loss
28
+ aux_loss_coeff=1e-2
29
+ router_pre_softmax=1
30
+ router_fusion=0
31
+ router_dtype=fp32
32
+ device_initiated_grouped_gemm=1
33
+ norm_epsilon=1e-6
34
+ seq_length=4096
35
+ micro_batch_size=1
36
+ global_batch_size=8
37
+ num_microbatches=1
38
+ assignments_per_rank=32768
39
+ mean_tokens_per_expert=4096
40
+ slot_policy=matched
41
+ slots_per_rank=8
42
+ capacity_mode=static
43
+ capacity=1.08984375
44
+ cuda_graph=1
45
+ probe_mode=0
46
+ async_grad_reduce=0
47
+ branch_role=timing
48
+ memory_dump_mode=none
49
+ memory_snapshot_path=
50
+ cuda_graph_warmup_steps=3
51
+ recompute=none
52
+ force_balance=0
53
+ train_iters=12
54
+ ep_size=8
55
+ echo_dispatch_overlap=1
56
+ echo_dispatch_recompute=0
57
+ ultraep_quota_min_tokens=
58
+ ultraep_weight_sync_join=
59
+ profile_enable=0
60
+ kernel_trace=0
61
+ fp8_recipe=mxfp8
echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep3/train.log ADDED
@@ -0,0 +1,1198 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ==============================================================
2
+ QWEN2_57B / echo (slots/rank=8 policy=matched)
3
+ E=64 K=8 H'=2560 m=8
4
+ L=10 seq=4096 gbs=8 (1 microbatches)
5
+ cap_mode=static cap=1.08984375 cg=1 recompute=none
6
+ memory_dump=none snapshots=/workspace/tierb_study_20260826_144644/echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep3/memory_snapshots/snapshot.pickle
7
+ repo=/workspace/Megatron-MoE-Echo @ 9a236793
8
+ ==============================================================
9
+ W0826 18:26:06.263000 506268 torch/distributed/run.py:766]
10
+ W0826 18:26:06.263000 506268 torch/distributed/run.py:766] *****************************************
11
+ W0826 18:26:06.263000 506268 torch/distributed/run.py:766] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.
12
+ W0826 18:26:06.263000 506268 torch/distributed/run.py:766] *****************************************
13
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
14
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
15
+ WARNING: one_logger package is required to enable e2e metrics tracking. please go to https://confluence.nvidia.com/display/MLWFO/Package+Repositories for details to install it
16
+ /workspace/Megatron-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
+ 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
19
+ WARNING: overriding default arguments for tokenizer_type:GPT2BPETokenizer with tokenizer_type:GPT2BPETokenizer
20
+ Number of virtual stages per pipeline stage: None
21
+ accumulate and all-reduce gradients in fp32 for bfloat16 data type.
22
+ using torch.bfloat16 for parameters ...
23
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
24
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
25
+ ------------------------ arguments ------------------------
26
+ account_for_embedding_in_pipeline_split ......... False
27
+ account_for_loss_in_pipeline_split .............. False
28
+ accumulate_allreduce_grads_in_fp32 .............. True
29
+ adam_beta1 ...................................... 0.9
30
+ adam_beta2 ...................................... 0.999
31
+ adam_eps ........................................ 1e-08
32
+ add_bias_linear ................................. False
33
+ add_position_embedding .......................... False
34
+ add_qkv_bias .................................... True
35
+ adlr_autoresume ................................. False
36
+ adlr_autoresume_interval ........................ 1000
37
+ align_grad_reduce ............................... True
38
+ align_param_gather .............................. False
39
+ app_tag_run_name ................................ None
40
+ app_tag_run_version ............................. 0.0.0
41
+ apply_layernorm_1p .............................. False
42
+ apply_query_key_layer_scaling ................... False
43
+ apply_residual_connection_post_layernorm ........ False
44
+ apply_rope_fusion ............................... True
45
+ async_save ...................................... None
46
+ async_tensor_model_parallel_allreduce ........... True
47
+ attention_backend ............................... AttnBackend.auto
48
+ attention_dropout ............................... 0.0
49
+ attention_softmax_in_fp32 ....................... False
50
+ auto_detect_ckpt_format ......................... False
51
+ barrier_with_L1_time ............................ True
52
+ bert_binary_head ................................ True
53
+ bert_embedder_type .............................. megatron
54
+ bert_load ....................................... None
55
+ bf16 ............................................ True
56
+ bias_dropout_fusion ............................. True
57
+ bias_gelu_fusion ................................ False
58
+ bias_swiglu_fusion .............................. True
59
+ biencoder_projection_dim ........................ 0
60
+ biencoder_shared_query_context_model ............ False
61
+ block_data_path ................................. None
62
+ calc_ft_timeouts ................................ False
63
+ calculate_per_token_loss ........................ False
64
+ check_for_large_grads ........................... False
65
+ check_for_nan_in_loss_and_grad .................. False
66
+ check_for_spiky_loss ............................ False
67
+ check_weight_hash_across_dp_replicas_interval ... None
68
+ ckpt_assume_constant_structure .................. False
69
+ ckpt_convert_format ............................. None
70
+ ckpt_convert_save ............................... None
71
+ ckpt_convert_update_legacy_dist_opt_format ...... False
72
+ ckpt_format ..................................... torch_dist
73
+ ckpt_fully_parallel_load ........................ False
74
+ ckpt_fully_parallel_save ........................ True
75
+ ckpt_fully_parallel_save_deprecated ............. False
76
+ ckpt_step ....................................... None
77
+ classes_fraction ................................ 1.0
78
+ clip_grad ....................................... 1.0
79
+ clone_scatter_output_in_embedding ............... True
80
+ config_logger_dir ...............................
81
+ consumed_train_samples .......................... 0
82
+ consumed_valid_samples .......................... 0
83
+ context_parallel_size ........................... 1
84
+ cp_comm_type .................................... ['p2p']
85
+ create_attention_mask_in_dataloader ............. True
86
+ cross_entropy_fusion_impl ....................... native
87
+ cross_entropy_loss_fusion ....................... False
88
+ cuda_graph_scope ................................ full_iteration
89
+ cuda_graph_warmup_steps ......................... 3
90
+ data_args_path .................................. None
91
+ data_cache_path ................................. None
92
+ data_parallel_random_init ....................... False
93
+ data_parallel_sharding_strategy ................. no_shard
94
+ data_parallel_size .............................. 8
95
+ data_path ....................................... ['/workspace/data/wikitext/wikitext103_gpt2_text_document']
96
+ data_per_class_fraction ......................... 1.0
97
+ data_sharding ................................... True
98
+ dataloader_type ................................. single
99
+ ddp_average_in_collective ....................... False
100
+ ddp_bucket_size ................................. None
101
+ ddp_num_buckets ................................. None
102
+ ddp_pad_buckets_for_high_nccl_busbw ............. False
103
+ decoder_first_pipeline_num_layers ............... None
104
+ decoder_last_pipeline_num_layers ................ None
105
+ decoder_num_layers .............................. None
106
+ decoder_seq_length .............................. None
107
+ decoupled_lr .................................... None
108
+ decoupled_min_lr ................................ None
109
+ decrease_batch_size_if_needed ................... False
110
+ defer_embedding_wgrad_compute ................... False
111
+ delay_wgrad_compute ............................. False
112
+ deprecated_use_mcore_models ..................... True
113
+ deterministic_mode .............................. False
114
+ dino_bottleneck_size ............................ 256
115
+ dino_freeze_last_layer .......................... 1
116
+ dino_head_hidden_size ........................... 2048
117
+ dino_local_crops_number ......................... 10
118
+ dino_local_img_size ............................. 96
119
+ dino_norm_last_layer ............................ False
120
+ dino_teacher_temp ............................... 0.07
121
+ dino_warmup_teacher_temp ........................ 0.04
122
+ dino_warmup_teacher_temp_epochs ................. 30
123
+ disable_bf16_reduced_precision_matmul ........... False
124
+ disable_mamba_mem_eff_path ...................... False
125
+ disable_straggler_on_startup .................... False
126
+ dist_ckpt_format_deprecated ..................... None
127
+ dist_ckpt_strictness ............................ assume_ok_unexpected
128
+ distribute_saved_activations .................... False
129
+ distributed_backend ............................. nccl
130
+ distributed_timeout_minutes ..................... 10
131
+ embedding_init_method_std ....................... None
132
+ embedding_path .................................. None
133
+ empty_unused_memory_level ....................... 0
134
+ enable_cuda_graph ............................... True
135
+ enable_experimental ............................. False
136
+ enable_ft_package ............................... False
137
+ enable_gloo_process_groups ...................... True
138
+ enable_msc ...................................... True
139
+ enable_one_logger ............................... True
140
+ encoder_num_layers .............................. 10
141
+ encoder_seq_length .............................. 4096
142
+ end_weight_decay ................................ 0.1
143
+ eod_mask_loss ................................... False
144
+ error_injection_rate ............................ 0
145
+ error_injection_type ............................ transient_error
146
+ eval_interval ................................... 1000000
147
+ eval_iters ...................................... 0
148
+ evidence_data_path .............................. None
149
+ exit_duration_in_mins ........................... None
150
+ exit_interval ................................... None
151
+ exit_on_missing_checkpoint ...................... False
152
+ exit_signal_handler ............................. False
153
+ exp_avg_dtype ................................... torch.float32
154
+ exp_avg_sq_dtype ................................ torch.float32
155
+ expert_model_parallel_size ...................... 8
156
+ expert_tensor_parallel_size ..................... 1
157
+ export_force_local_attention .................... False
158
+ export_kd_cfg ................................... None
159
+ export_kd_teacher_ckpt_format ................... None
160
+ export_kd_teacher_load .......................... None
161
+ export_kv_cache_quant ........................... False
162
+ export_legacy_megatron .......................... False
163
+ export_model_type ............................... GPTModel
164
+ export_moe_apply_probs_on_input ................. False
165
+ export_qk_l2_norm ............................... False
166
+ export_quant_cfg ................................ None
167
+ export_real_quant_cfg ........................... None
168
+ export_te_mcore_model ........................... False
169
+ external_cuda_graph ............................. False
170
+ ffn_hidden_size ................................. 2560
171
+ finetune ........................................ False
172
+ finetune_data_split ............................. train
173
+ finetune_hf_dataset ............................. None
174
+ first_last_layers_bf16 .......................... False
175
+ flash_decode .................................... False
176
+ fp16 ............................................ False
177
+ fp16_lm_cross_entropy ........................... False
178
+ fp32_residual_connection ........................ False
179
+ fp8 ............................................. e4m3
180
+ fp8_amax_compute_algo ........................... most_recent
181
+ fp8_amax_history_len ............................ 1
182
+ fp8_interval .................................... 1
183
+ fp8_margin ...................................... 0
184
+ fp8_param_gather ................................ False
185
+ fp8_recipe ...................................... mxfp8
186
+ fp8_wgrad ....................................... True
187
+ fsdp_double_buffer .............................. False
188
+ global_batch_size ............................... 8
189
+ grad_reduce_in_bf16 ............................. False
190
+ gradient_accumulation_fusion .................... True
191
+ gradient_reduce_div_fusion ...................... True
192
+ group_query_attention ........................... True
193
+ head_lr_mult .................................... 1.0
194
+ heterogeneous_layers_config_encoded_json ........ None
195
+ heterogeneous_layers_config_path ................ None
196
+ hidden_dropout .................................. 0.0
197
+ hidden_size ..................................... 3584
198
+ hierarchical_context_parallel_sizes ............. None
199
+ high_priority_stream_groups ..................... []
200
+ hybrid_attention_ratio .......................... 0.0
201
+ hybrid_mlp_ratio ................................ 0.0
202
+ hybrid_override_pattern ......................... None
203
+ hysteresis ...................................... 2
204
+ ict_head_size ................................... None
205
+ ict_load ........................................ None
206
+ img_h ........................................... 224
207
+ img_w ........................................... 224
208
+ indexer_batch_size .............................. 128
209
+ indexer_log_interval ............................ 1000
210
+ inference_batch_times_seqlen_threshold .......... -1
211
+ inference_dynamic_batching ...................... False
212
+ inference_dynamic_batching_buffer_guaranteed_fraction 0.2
213
+ inference_dynamic_batching_buffer_overflow_factor None
214
+ inference_dynamic_batching_buffer_size_gb ....... 40.0
215
+ inference_dynamic_batching_chunk_size ........... 256
216
+ inference_dynamic_batching_max_requests_override None
217
+ inference_dynamic_batching_max_tokens_override .. None
218
+ inference_dynamic_batching_num_cuda_graphs ...... 16
219
+ inference_max_batch_size ........................ 8
220
+ inference_max_seq_length ........................ 2560
221
+ inference_rng_tracker ........................... False
222
+ init_method_std ................................. 0.01
223
+ init_method_xavier_uniform ...................... False
224
+ init_model_with_meta_device ..................... False
225
+ initial_loss_scale .............................. 4294967296
226
+ inprocess_active_world_size ..................... 8
227
+ inprocess_barrier_timeout ....................... 120
228
+ inprocess_completion_timeout .................... 120
229
+ inprocess_empty_cuda_cache ...................... False
230
+ inprocess_granularity ........................... node
231
+ inprocess_hard_timeout .......................... 90
232
+ inprocess_heartbeat_interval .................... 30
233
+ inprocess_heartbeat_timeout ..................... 60
234
+ inprocess_last_call_wait ........................ 1
235
+ inprocess_max_iterations ........................ None
236
+ inprocess_monitor_process_interval .............. 1.0
237
+ inprocess_monitor_thread_interval ............... 1.0
238
+ inprocess_progress_watchdog_interval ............ 1.0
239
+ inprocess_restart ............................... False
240
+ inprocess_soft_timeout .......................... 60
241
+ inprocess_termination_grace_time ................ 1
242
+ is_hybrid_model ................................. False
243
+ iter_per_epoch .................................. 1250
244
+ iterations_to_skip .............................. []
245
+ keep_fp8_transpose_cache_when_using_custom_fsdp . False
246
+ kitchen_config_file ............................. None
247
+ kitchen_recipe_number ........................... None
248
+ kv_channels ..................................... 128
249
+ kv_lora_rank .................................... 32
250
+ lazy_mpu_init ................................... None
251
+ load ............................................ None
252
+ load_main_params_from_ckpt ...................... None
253
+ load_model_opt_format ........................... False
254
+ local_rank ...................................... 0
255
+ log_energy ...................................... False
256
+ log_interval .................................... 1
257
+ log_loss_scale_to_tensorboard ................... True
258
+ log_memory_to_tensorboard ....................... False
259
+ log_num_zeros_in_grad ........................... False
260
+ log_params_norm ................................. False
261
+ log_progress .................................... False
262
+ log_straggler ................................... False
263
+ log_throughput .................................. True
264
+ log_timers_to_tensorboard ....................... False
265
+ log_validation_ppl_to_tensorboard ............... False
266
+ log_world_size_to_tensorboard ................... False
267
+ logging_level ................................... None
268
+ loss_scale ...................................... None
269
+ loss_scale_window ............................... 1000
270
+ lr .............................................. 0.0001
271
+ lr_decay_iters .................................. 8
272
+ lr_decay_samples ................................ None
273
+ lr_decay_style .................................. cosine
274
+ lr_warmup_fraction .............................. None
275
+ lr_warmup_init .................................. 0.0
276
+ lr_warmup_iters ................................. 5
277
+ lr_warmup_samples ............................... 0
278
+ lr_wsd_decay_iters .............................. None
279
+ lr_wsd_decay_samples ............................ None
280
+ lr_wsd_decay_style .............................. exponential
281
+ main_grads_dtype ................................ torch.float32
282
+ main_params_dtype ............................... torch.float32
283
+ make_vocab_size_divisible_by .................... 128
284
+ mamba_head_dim .................................. 64
285
+ mamba_num_groups ................................ 8
286
+ mamba_num_heads ................................. None
287
+ mamba_state_dim ................................. 128
288
+ manual_gc ....................................... False
289
+ manual_gc_eval .................................. True
290
+ manual_gc_interval .............................. 0
291
+ mask_factor ..................................... 1.0
292
+ mask_prob ....................................... 0.15
293
+ mask_type ....................................... random
294
+ masked_softmax_fusion ........................... False
295
+ max_position_embeddings ......................... 32768
296
+ max_tokens_to_oom ............................... 12000
297
+ memory_snapshot_path ............................ snapshot.pickle
298
+ merge_file ...................................... /workspace/data/wikitext/gpt2-merges.txt
299
+ micro_batch_size ................................ 1
300
+ microbatch_group_size_per_vp_stage .............. None
301
+ mid_level_dataset_surplus ....................... 0.005
302
+ min_loss_scale .................................. 1.0
303
+ min_lr .......................................... 1e-05
304
+ mlp_chunks_for_prefill .......................... 1
305
+ mmap_bin_files .................................. True
306
+ mock_data ....................................... False
307
+ moe_apply_probs_on_input ........................ False
308
+ moe_aux_loss_coeff .............................. 0.01
309
+ moe_deepep_num_sms .............................. 20
310
+ moe_echo_enable_random_offloading ............... False
311
+ moe_echo_expert_dispatch_overlap ................ True
312
+ moe_echo_expert_dispatcher_type ................. hybridep
313
+ moe_echo_recompute_expert_dispatch .............. False
314
+ moe_enable_deepep ............................... False
315
+ moe_enable_echo ................................. True
316
+ moe_enable_hybridep ............................. True
317
+ moe_expert_capacity_factor ...................... None
318
+ moe_extended_tp ................................. False
319
+ moe_ffn_hidden_size ............................. 2560
320
+ moe_grouped_gemm ................................ True
321
+ moe_hybridep_num_sms ............................ 16
322
+ moe_input_jitter_eps ............................ None
323
+ moe_layer_freq .................................. 1
324
+ moe_layer_recompute ............................. False
325
+ moe_num_echo_experts ............................ 64
326
+ moe_pad_expert_input_to_capacity ................ False
327
+ moe_per_layer_logging ........................... False
328
+ moe_permute_fusion .............................. False
329
+ moe_received_token_capacity ..................... 1.08984375
330
+ moe_router_bias_update_rate ..................... 0.001
331
+ moe_router_dtype ................................ fp32
332
+ moe_router_enable_expert_bias ................... False
333
+ moe_router_force_load_balancing ................. False
334
+ moe_router_group_topk ........................... None
335
+ moe_router_load_balancing_type .................. aux_loss
336
+ moe_router_num_groups ........................... None
337
+ moe_router_padding_for_fp8 ...................... False
338
+ moe_router_pre_softmax .......................... True
339
+ moe_router_score_function ....................... softmax
340
+ moe_router_topk ................................. 8
341
+ moe_router_topk_scaling_factor .................. None
342
+ moe_shared_expert_intermediate_size ............. 20480
343
+ moe_shared_expert_overlap ....................... False
344
+ moe_token_dispatcher_type ....................... flex
345
+ moe_token_drop_policy ........................... probs
346
+ moe_upcycling_granularity ....................... 1
347
+ moe_use_device_initiated_grouped_gemm ........... True
348
+ moe_use_legacy_grouped_gemm ..................... False
349
+ moe_use_upcycling ............................... False
350
+ moe_z_loss_coeff ................................ None
351
+ mrope_section ................................... None
352
+ mscale .......................................... 1.0
353
+ mscale_all_dim .................................. 0.0
354
+ mtp_loss_scaling_factor ......................... 0.1
355
+ mtp_num_layers .................................. None
356
+ multi_latent_attention .......................... False
357
+ nccl_all_reduce_for_prefill ..................... False
358
+ nccl_communicator_config_path ................... None
359
+ nccl_ub ......................................... False
360
+ no_load_optim ................................... None
361
+ no_load_rng ..................................... None
362
+ no_persist_layer_norm ........................... False
363
+ no_rope_freq .................................... None
364
+ no_save_optim ................................... None
365
+ no_save_rng ..................................... None
366
+ non_persistent_ckpt_type ........................ None
367
+ non_persistent_global_ckpt_dir .................. None
368
+ non_persistent_local_ckpt_algo .................. fully_parallel
369
+ non_persistent_local_ckpt_dir ................... None
370
+ non_persistent_save_interval .................... None
371
+ norm_epsilon .................................... 1e-06
372
+ normalization ................................... RMSNorm
373
+ num_attention_heads ............................. 28
374
+ num_channels .................................... 3
375
+ num_classes ..................................... 1000
376
+ num_dataset_builder_threads ..................... 1
377
+ num_distributed_optimizer_instances ............. 1
378
+ num_experts ..................................... 64
379
+ num_layers ...................................... 10
380
+ num_layers_at_end_in_bf16 ....................... 1
381
+ num_layers_at_start_in_bf16 ..................... 1
382
+ num_layers_per_virtual_pipeline_stage ........... None
383
+ num_query_groups ................................ 4
384
+ num_virtual_stages_per_pipeline_rank ............ None
385
+ num_workers ..................................... 2
386
+ object_storage_cache_path ....................... None
387
+ one_logger_async ................................ False
388
+ one_logger_project .............................. megatron-lm
389
+ one_logger_run_name ............................. None
390
+ onnx_safe ....................................... None
391
+ openai_gelu ..................................... False
392
+ optimizer ....................................... adam
393
+ optimizer_cpu_offload ........................... False
394
+ optimizer_offload_fraction ...................... 1.0
395
+ output_bert_embeddings .......................... False
396
+ overlap_cpu_optimizer_d2h_h2d ................... False
397
+ overlap_grad_reduce ............................. True
398
+ overlap_moe_expert_parallel_comm ................ False
399
+ overlap_p2p_comm ................................ False
400
+ overlap_p2p_comm_warmup_flush ................... False
401
+ overlap_param_gather ............................ True
402
+ overlap_param_gather_with_optimizer_step ........ False
403
+ override_opt_param_scheduler .................... False
404
+ params_dtype .................................... torch.bfloat16
405
+ patch_dim ....................................... 16
406
+ per_split_data_args_path ........................ None
407
+ perform_initialization .......................... True
408
+ pin_cpu_grads ................................... True
409
+ pin_cpu_params .................................. True
410
+ pipeline_model_parallel_comm_backend ............ None
411
+ pipeline_model_parallel_layout .................. None
412
+ pipeline_model_parallel_size .................... 1
413
+ position_embedding_type ......................... rope
414
+ pretrained_checkpoint ........................... None
415
+ profile ......................................... False
416
+ profile_ranks ................................... [0]
417
+ profile_step_end ................................ 12
418
+ profile_step_start .............................. 10
419
+ q_lora_rank ..................................... None
420
+ qk_head_dim ..................................... 128
421
+ qk_l2_norm ...................................... False
422
+ qk_layernorm .................................... False
423
+ qk_pos_emb_head_dim ............................. 64
424
+ query_in_block_prob ............................. 0.1
425
+ rampup_batch_size ............................... None
426
+ rank ............................................ 0
427
+ recompute_granularity ........................... None
428
+ recompute_method ................................ None
429
+ recompute_modules ............................... None
430
+ recompute_num_layers ............................ None
431
+ record_memory_history ........................... False
432
+ relative_attention_max_distance ................. 128
433
+ relative_attention_num_buckets .................. 32
434
+ replication ..................................... False
435
+ replication_factor .............................. 2
436
+ replication_jump ................................ None
437
+ rerun_mode ...................................... validate_results
438
+ reset_attention_mask ............................ False
439
+ reset_position_ids .............................. False
440
+ result_rejected_tracker_filename ................ None
441
+ retriever_report_topk_accuracies ................ []
442
+ retriever_score_scaling ......................... False
443
+ retriever_seq_length ............................ 256
444
+ retro_add_retriever ............................. False
445
+ retro_attention_gate ............................ 1
446
+ retro_cyclic_train_iters ........................ None
447
+ retro_encoder_attention_dropout ................. 0.1
448
+ retro_encoder_hidden_dropout .................... 0.1
449
+ retro_encoder_layers ............................ 2
450
+ retro_num_neighbors ............................. 2
451
+ retro_num_retrieved_chunks ...................... 2
452
+ retro_project_dir ............................... None
453
+ retro_verify_neighbor_count ..................... True
454
+ reuse_grad_buf_for_mxfp8_param_ag ............... False
455
+ rope_scaling_factor ............................. 8.0
456
+ rope_type ....................................... None
457
+ rotary_base ..................................... 1000000
458
+ rotary_interleaved .............................. False
459
+ rotary_percent .................................. 1.0
460
+ rotary_scaling_factor ........................... 1.0
461
+ rotary_seq_len_interpolation_factor ............. None
462
+ run_workload_inspector_server ................... False
463
+ sample_rate ..................................... 1.0
464
+ save ............................................ None
465
+ save_interval ................................... None
466
+ save_retain_interval ............................ None
467
+ scatter_gather_tensors_in_pipeline .............. True
468
+ seed ............................................ 1234
469
+ seq_length ...................................... 4096
470
+ sequence_parallel ............................... False
471
+ sft ............................................. False
472
+ sft_tokenizer_prompt_format ..................... nemotron-h-aligned
473
+ sgd_momentum .................................... 0.9
474
+ short_seq_prob .................................. 0.1
475
+ skip_train ...................................... False
476
+ skipped_train_samples ........................... 0
477
+ spec ............................................ None
478
+ split ........................................... 100,0,0
479
+ squared_relu .................................... False
480
+ start_weight_decay .............................. 0.1
481
+ straggler_ctrlr_port ............................ 65535
482
+ straggler_minmax_count .......................... 1
483
+ suggested_communication_unit_size ............... None
484
+ swiglu .......................................... True
485
+ swin_backbone_type .............................. tiny
486
+ symmetric_ar_type ............................... None
487
+ te_rng_tracker .................................. True
488
+ tensor_model_parallel_size ...................... 1
489
+ tensorboard_dir ................................. None
490
+ tensorboard_log_interval ........................ 1
491
+ tensorboard_queue_size .......................... 1000
492
+ test_data_path .................................. None
493
+ test_mode ....................................... False
494
+ tiktoken_num_special_tokens ..................... 1000
495
+ tiktoken_pattern ................................ None
496
+ tiktoken_special_tokens ......................... None
497
+ timing_log_level ................................ 0
498
+ timing_log_option ............................... minmax
499
+ titles_data_path ................................ None
500
+ tokenizer_model ................................. None
501
+ tokenizer_type .................................. GPT2BPETokenizer
502
+ torch_fsdp2_reshard_after_forward ............... True
503
+ tp_comm_bootstrap_backend ....................... nccl
504
+ tp_comm_bulk_dgrad .............................. True
505
+ tp_comm_bulk_wgrad .............................. True
506
+ tp_comm_overlap ................................. False
507
+ tp_comm_overlap_ag .............................. True
508
+ tp_comm_overlap_cfg ............................. None
509
+ tp_comm_overlap_rs .............................. True
510
+ tp_comm_overlap_rs_dgrad ........................ False
511
+ tp_comm_split_ag ................................ True
512
+ tp_comm_split_rs ................................ True
513
+ train_data_path ................................. None
514
+ train_iters ..................................... 12
515
+ train_samples ................................... None
516
+ train_sync_interval ............................. None
517
+ transformer_impl ................................ transformer_engine
518
+ transformer_pipeline_model_parallel_size ........ 1
519
+ untie_embeddings_and_output_weights ............. True
520
+ use_checkpoint_args ............................. False
521
+ use_checkpoint_opt_param_scheduler .............. False
522
+ use_cpu_initialization .......................... None
523
+ use_custom_fsdp ................................. False
524
+ use_dist_ckpt ................................... True
525
+ use_dist_ckpt_deprecated ........................ False
526
+ use_distributed_optimizer ....................... True
527
+ use_flash_attn .................................. False
528
+ use_legacy_models ............................... False
529
+ use_mp_args_from_checkpoint_args ................ False
530
+ use_one_sent_docs ............................... False
531
+ use_persistent_ckpt_worker ...................... False
532
+ use_precision_aware_optimizer ................... False
533
+ use_pytorch_profiler ............................ False
534
+ use_ring_exchange_p2p ........................... False
535
+ use_rope_scaling ................................ False
536
+ use_rotary_position_embeddings .................. False
537
+ use_sharp ....................................... False
538
+ use_tokenizer_model_from_checkpoint_args ........ True
539
+ use_torch_fsdp2 ................................. False
540
+ use_torch_optimizer_for_cpu_offload ............. False
541
+ use_tp_pp_dp_mapping ............................ False
542
+ v_head_dim ...................................... 128
543
+ valid_data_path ................................. None
544
+ variable_seq_lengths ............................ False
545
+ virtual_pipeline_model_parallel_size ............ None
546
+ vision_backbone_type ............................ vit
547
+ vision_pretraining .............................. False
548
+ vision_pretraining_type ......................... classify
549
+ vocab_extra_ids ................................. 0
550
+ vocab_file ...................................... /workspace/data/wikitext/gpt2-vocab.json
551
+ vocab_size ...................................... None
552
+ wandb_exp_name ..................................
553
+ wandb_project ...................................
554
+ wandb_save_dir ..................................
555
+ weight_decay .................................... 0.1
556
+ weight_decay_incr_style ......................... constant
557
+ wgrad_deferral_limit ............................ 0
558
+ world_size ...................................... 8
559
+ yaml_cfg ........................................ None
560
+ -------------------- end of arguments ---------------------
561
+ INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 1
562
+ > building GPT2BPETokenizer tokenizer ...
563
+ > padded vocab (size: 50257) with 47 dummy tokens (new size: 50304)
564
+ WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.VALIDATE_RESULTS
565
+ > initializing torch distributed ...
566
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
567
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
568
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
569
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
570
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
571
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
572
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
573
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
574
+ /workspace/Megatron-MoE-Echo/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
575
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
576
+ [Gloo] Rank [Gloo] Rank 0 is connected to 57 is connected to peer ranks. 7Expected number of connected peer ranks is : peer ranks. 7Expected number of connected peer ranks is : 7
577
+
578
+ [Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
579
+ [Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
580
+ [Gloo] Rank 7 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
581
+ [Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
582
+ [Gloo] Rank 6 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
583
+ [Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
584
+ [Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
585
+ [Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
586
+ [Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
587
+ [Gloo] Rank 2 is connected to [Gloo] Rank 7 peer ranks. 6Expected number of connected peer ranks is : is connected to 77[Gloo] Rank
588
+ peer ranks. Expected number of connected peer ranks is : 75
589
+ 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 7 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
592
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
593
+ > initialized tensor model parallel with size 1
594
+ > initialized pipeline model parallel with size 1
595
+ > setting random seeds to 1234 ...
596
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
597
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
598
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
599
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
600
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
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
+ [rank6]:[W826 18:26:24.730725652 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.
604
+ [rank4]:[W826 18:26:24.730737921 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
+ [rank1]:[W826 18:26:24.730786372 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.
606
+ [rank3]:[W826 18:26:24.730817009 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.
607
+ [rank2]:[W826 18:26:24.730895996 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 2] using GPU 2 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
608
+ [rank5]:[W826 18:26:24.731111982 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
609
+ [rank7]:[W826 18:26:24.731964744 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
610
+ > compiling dataset index builder ...
611
+ make: Entering directory '/workspace/Megatron-MoE-Echo/megatron/core/datasets'
612
+ make: Nothing to be done for 'default'.
613
+ make: Leaving directory '/workspace/Megatron-MoE-Echo/megatron/core/datasets'
614
+ >>> done with dataset index builder. Compilation time: 0.037 seconds
615
+ WARNING: constraints for invoking optimized fused softmax kernel are not met. We default back to unfused kernel invocations.
616
+ > compiling and loading fused kernels ...
617
+ [rank0]:[W826 18:26:24.885887979 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.
618
+ >>> done with compiling and loading fused kernels. Compilation time: 3.534 seconds
619
+ time to initialize megatron (seconds): 6.657
620
+ [after megatron is initialized] datetime: 2026-08-26 18:26:30
621
+ building GPT model ...
622
+ /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.
623
+ warnings.warn(
624
+ /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.
625
+ warnings.warn(
626
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
627
+ warnings.warn(
628
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
629
+ warnings.warn(
630
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
631
+ warnings.warn(
632
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
633
+ warnings.warn(
634
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
635
+ warnings.warn(
636
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
637
+ warnings.warn(
638
+ > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 5060614656
639
+ 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)
640
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 22
641
+ Params for bucket 1 (180289536 elements, 180289536 padded size):
642
+ module.output_layer.weight
643
+ Params for bucket 2 (73403904 elements, 73403904 padded size):
644
+ module.decoder.final_layernorm.weight
645
+ module.decoder.layers.9.mlp.shared_experts.linear_fc2.weight
646
+ Params for bucket 3 (146800640 elements, 146800640 padded size):
647
+ module.decoder.layers.9.mlp.shared_experts.linear_fc1.weight
648
+ Params for bucket 4 (103001600 elements, 103001600 padded size):
649
+ module.decoder.layers.9.mlp.router.weight
650
+ module.decoder.layers.9.self_attention.linear_qkv.weight
651
+ module.decoder.layers.9.self_attention.linear_qkv.layer_norm_weight
652
+ module.decoder.layers.9.self_attention.linear_proj.weight
653
+ module.decoder.layers.9.self_attention.linear_qkv.bias
654
+ module.decoder.layers.9.pre_mlp_layernorm.weight
655
+ module.decoder.layers.8.mlp.shared_experts.linear_fc2.weight
656
+ Params for bucket 5 (146800640 elements, 146800640 padded size):
657
+ module.decoder.layers.8.mlp.shared_experts.linear_fc1.weight
658
+ Params for bucket 6 (103001600 elements, 103001600 padded size):
659
+ module.decoder.layers.8.self_attention.linear_qkv.layer_norm_weight
660
+ module.decoder.layers.8.self_attention.linear_qkv.weight
661
+ module.decoder.layers.7.mlp.shared_experts.linear_fc2.weight
662
+ module.decoder.layers.8.mlp.router.weight
663
+ module.decoder.layers.8.pre_mlp_layernorm.weight
664
+ module.decoder.layers.8.self_attention.linear_qkv.bias
665
+ module.decoder.layers.8.self_attention.linear_proj.weight
666
+ Params for bucket 7 (146800640 elements, 146800640 padded size):
667
+ module.decoder.layers.7.mlp.shared_experts.linear_fc1.weight
668
+ Params for bucket 8 (103001600 elements, 103001600 padded size):
669
+ module.decoder.layers.7.self_attention.linear_qkv.bias
670
+ module.decoder.layers.7.self_attention.linear_qkv.layer_norm_weight
671
+ module.decoder.layers.7.pre_mlp_layernorm.weight
672
+ module.decoder.layers.7.self_attention.linear_proj.weight
673
+ module.decoder.layers.6.mlp.shared_experts.linear_fc2.weight
674
+ module.decoder.layers.7.mlp.router.weight
675
+ module.decoder.layers.7.self_attention.linear_qkv.weight
676
+ Params for bucket 9 (146800640 elements, 146800640 padded size):
677
+ module.decoder.layers.6.mlp.shared_experts.linear_fc1.weight
678
+ Params for bucket 10 (103001600 elements, 103001600 padded size):
679
+ module.decoder.layers.6.mlp.router.weight
680
+ module.decoder.layers.6.self_attention.linear_qkv.weight
681
+ module.decoder.layers.6.self_attention.linear_qkv.layer_norm_weight
682
+ module.decoder.layers.6.self_attention.linear_proj.weight
683
+ module.decoder.layers.5.mlp.shared_experts.linear_fc2.weight
684
+ module.decoder.layers.6.self_attention.linear_qkv.bias
685
+ module.decoder.layers.6.pre_mlp_layernorm.weight
686
+ Params for bucket 11 (146800640 elements, 146800640 padded size):
687
+ module.decoder.layers.5.mlp.shared_experts.linear_fc1.weight
688
+ Params for bucket 12 (103001600 elements, 103001600 padded size):
689
+ module.decoder.layers.5.mlp.router.weight
690
+ module.decoder.layers.5.self_attention.linear_qkv.layer_norm_weight
691
+ module.decoder.layers.5.self_attention.linear_qkv.weight
692
+ module.decoder.layers.5.self_attention.linear_proj.weight
693
+ module.decoder.layers.5.pre_mlp_layernorm.weight
694
+ module.decoder.layers.5.self_attention.linear_qkv.bias
695
+ module.decoder.layers.4.mlp.shared_experts.linear_fc2.weight
696
+ Params for bucket 13 (146800640 elements, 146800640 padded size):
697
+ module.decoder.layers.4.mlp.shared_experts.linear_fc1.weight
698
+ Params for bucket 14 (103001600 elements, 103001600 padded size):
699
+ module.decoder.layers.4.self_attention.linear_qkv.weight
700
+ module.decoder.layers.4.self_attention.linear_proj.weight
701
+ module.decoder.layers.4.mlp.router.weight
702
+ module.decoder.layers.4.self_attention.linear_qkv.layer_norm_weight
703
+ module.decoder.layers.4.self_attention.linear_qkv.bias
704
+ module.decoder.layers.4.pre_mlp_layernorm.weight
705
+ module.decoder.layers.3.mlp.shared_experts.linear_fc2.weight
706
+ Params for bucket 15 (146800640 elements, 146800640 padded size):
707
+ module.decoder.layers.3.mlp.shared_experts.linear_fc1.weight
708
+ Params for bucket 16 (103001600 elements, 103001600 padded size):
709
+ module.decoder.layers.3.mlp.router.weight
710
+ module.decoder.layers.3.self_attention.linear_qkv.layer_norm_weight
711
+ module.decoder.layers.3.self_attention.linear_qkv.weight
712
+ module.decoder.layers.3.self_attention.linear_proj.weight
713
+ module.decoder.layers.3.self_attention.linear_qkv.bias
714
+ module.decoder.layers.3.pre_mlp_layernorm.weight
715
+ module.decoder.layers.2.mlp.shared_experts.linear_fc2.weight
716
+ Params for bucket 17 (146800640 elements, 146800640 padded size):
717
+ module.decoder.layers.2.mlp.shared_experts.linear_fc1.weight
718
+ Params for bucket 18 (103001600 elements, 103001600 padded size):
719
+ module.decoder.layers.2.self_attention.linear_qkv.bias
720
+ module.decoder.layers.2.self_attention.linear_qkv.layer_norm_weight
721
+ module.decoder.layers.2.pre_mlp_layernorm.weight
722
+ module.decoder.layers.2.self_attention.linear_proj.weight
723
+ module.decoder.layers.1.mlp.shared_experts.linear_fc2.weight
724
+ module.decoder.layers.2.mlp.router.weight
725
+ module.decoder.layers.2.self_attention.linear_qkv.weight
726
+ Params for bucket 19 (146800640 elements, 146800640 padded size):
727
+ module.decoder.layers.1.mlp.shared_experts.linear_fc1.weight
728
+ Params for bucket 20 (103001600 elements, 103001600 padded size):
729
+ module.decoder.layers.1.mlp.router.weight
730
+ module.decoder.layers.1.self_attention.linear_qkv.bias
731
+ module.decoder.layers.1.self_attention.linear_qkv.layer_norm_weight
732
+ module.decoder.layers.1.self_attention.linear_proj.weight
733
+ module.decoder.layers.0.mlp.shared_experts.linear_fc2.weight
734
+ module.decoder.layers.1.self_attention.linear_qkv.weight
735
+ module.decoder.layers.1.pre_mlp_layernorm.weight
736
+ Params for bucket 21 (146800640 elements, 146800640 padded size):
737
+ module.decoder.layers.0.mlp.shared_experts.linear_fc1.weight
738
+ Params for bucket 22 (209890816 elements, 209890816 padded size):
739
+ module.decoder.layers.0.self_attention.linear_qkv.layer_norm_weight
740
+ module.decoder.layers.0.self_attention.linear_qkv.bias
741
+ module.decoder.layers.0.self_attention.linear_qkv.weight
742
+ module.decoder.layers.0.pre_mlp_layernorm.weight
743
+ module.decoder.layers.0.self_attention.linear_proj.weight
744
+ module.embedding.word_embeddings.weight
745
+ module.decoder.layers.0.mlp.router.weight
746
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 44
747
+ Params for bucket 1 (45875200 elements, 45875200 padded size):
748
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight4
749
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight6
750
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight7
751
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight5
752
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight3
753
+ Params for bucket 2 (45875200 elements, 45875200 padded size):
754
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight2
755
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight7
756
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight1
757
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight0
758
+ Params for bucket 3 (55050240 elements, 55050240 padded size):
759
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight6
760
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight4
761
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight5
762
+ Params for bucket 4 (55050240 elements, 55050240 padded size):
763
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight3
764
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight2
765
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight1
766
+ Params for bucket 5 (45875200 elements, 45875200 padded size):
767
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight0
768
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight5
769
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight7
770
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight6
771
+ Params for bucket 6 (45875200 elements, 45875200 padded size):
772
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight1
773
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight3
774
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight4
775
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight0
776
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight2
777
+ Params for bucket 7 (55050240 elements, 55050240 padded size):
778
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight7
779
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight5
780
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight6
781
+ Params for bucket 8 (55050240 elements, 55050240 padded size):
782
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight4
783
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight2
784
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight3
785
+ Params for bucket 9 (45875200 elements, 45875200 padded size):
786
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight1
787
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight7
788
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight0
789
+ Params for bucket 10 (45875200 elements, 45875200 padded size):
790
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight3
791
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight5
792
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight6
793
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight2
794
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight4
795
+ Params for bucket 11 (55050240 elements, 55050240 padded size):
796
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight1
797
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight6
798
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight0
799
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight7
800
+ Params for bucket 12 (55050240 elements, 55050240 padded size):
801
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight5
802
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight3
803
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight4
804
+ Params for bucket 13 (55050240 elements, 55050240 padded size):
805
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight2
806
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight0
807
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight1
808
+ Params for bucket 14 (45875200 elements, 45875200 padded size):
809
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight6
810
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight4
811
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight7
812
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight3
813
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight5
814
+ Params for bucket 15 (45875200 elements, 45875200 padded size):
815
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight2
816
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight7
817
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight0
818
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight1
819
+ Params for bucket 16 (55050240 elements, 55050240 padded size):
820
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight6
821
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight5
822
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight4
823
+ Params for bucket 17 (55050240 elements, 55050240 padded size):
824
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight3
825
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight2
826
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight1
827
+ Params for bucket 18 (45875200 elements, 45875200 padded size):
828
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight0
829
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight7
830
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight5
831
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight6
832
+ Params for bucket 19 (45875200 elements, 45875200 padded size):
833
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight4
834
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight0
835
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight2
836
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight1
837
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight3
838
+ Params for bucket 20 (55050240 elements, 55050240 padded size):
839
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight7
840
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight6
841
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight5
842
+ Params for bucket 21 (55050240 elements, 55050240 padded size):
843
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight4
844
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight3
845
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight2
846
+ Params for bucket 22 (45875200 elements, 45875200 padded size):
847
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight1
848
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight0
849
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight7
850
+ Params for bucket 23 (45875200 elements, 45875200 padded size):
851
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight3
852
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight5
853
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight6
854
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight4
855
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight2
856
+ Params for bucket 24 (55050240 elements, 55050240 padded size):
857
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight1
858
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight0
859
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight7
860
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight6
861
+ Params for bucket 25 (55050240 elements, 55050240 padded size):
862
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight5
863
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight3
864
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight4
865
+ Params for bucket 26 (55050240 elements, 55050240 padded size):
866
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight2
867
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight0
868
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight1
869
+ Params for bucket 27 (45875200 elements, 45875200 padded size):
870
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight4
871
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight6
872
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight7
873
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight5
874
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight3
875
+ Params for bucket 28 (45875200 elements, 45875200 padded size):
876
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight2
877
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight0
878
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight1
879
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight7
880
+ Params for bucket 29 (55050240 elements, 55050240 padded size):
881
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight6
882
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight4
883
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight5
884
+ Params for bucket 30 (55050240 elements, 55050240 padded size):
885
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight3
886
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight1
887
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight2
888
+ Params for bucket 31 (45875200 elements, 45875200 padded size):
889
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight0
890
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight7
891
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight6
892
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight5
893
+ Params for bucket 32 (45875200 elements, 45875200 padded size):
894
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight3
895
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight1
896
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight4
897
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight0
898
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight2
899
+ Params for bucket 33 (55050240 elements, 55050240 padded size):
900
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight7
901
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight6
902
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight5
903
+ Params for bucket 34 (55050240 elements, 55050240 padded size):
904
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight4
905
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight3
906
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight2
907
+ Params for bucket 35 (45875200 elements, 45875200 padded size):
908
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight1
909
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight0
910
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight7
911
+ Params for bucket 36 (45875200 elements, 45875200 padded size):
912
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight6
913
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight4
914
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight2
915
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight5
916
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight3
917
+ Params for bucket 37 (55050240 elements, 55050240 padded size):
918
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight1
919
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight0
920
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight7
921
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight6
922
+ Params for bucket 38 (55050240 elements, 55050240 padded size):
923
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight5
924
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight3
925
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight4
926
+ Params for bucket 39 (55050240 elements, 55050240 padded size):
927
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight2
928
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight0
929
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight1
930
+ Params for bucket 40 (45875200 elements, 45875200 padded size):
931
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight7
932
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight4
933
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight3
934
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight6
935
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight5
936
+ Params for bucket 41 (45875200 elements, 45875200 padded size):
937
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight2
938
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight1
939
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight0
940
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight7
941
+ Params for bucket 42 (55050240 elements, 55050240 padded size):
942
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight6
943
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight5
944
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight4
945
+ Params for bucket 43 (55050240 elements, 55050240 padded size):
946
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight3
947
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight2
948
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight1
949
+ Params for bucket 44 (18350080 elements, 18350080 padded size):
950
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight0
951
+ 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 0x77639b5902f0>, config_logger_dir='')
952
+ INFO:megatron.core.optimizer_param_scheduler:> learning rate decay style: cosine
953
+ [after model, optimizer, and learning rate scheduler are built] datetime: 2026-08-26 18:26:30
954
+ > building train, validation, and test datasets ...
955
+ > datasets target sizes (minimum size):
956
+ train: 96
957
+ validation: 0
958
+ test: 0
959
+ INFO:megatron.core.datasets.blended_megatron_dataset_config:Let split_matrix = [(0, 1.0), None, None]
960
+ > building train, validation, and test datasets for GPT ...
961
+ 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 0x77639c3999a0>, 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)
962
+ INFO:megatron.core.datasets.indexed_dataset:Load the _IndexReader from /workspace/data/wikitext/wikitext103_gpt2_text_document.idx
963
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence lengths
964
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence pointers
965
+ INFO:megatron.core.datasets.indexed_dataset: Extract the document indices
966
+ INFO:megatron.core.datasets.indexed_dataset:> total number of sequences: 1165029
967
+ INFO:megatron.core.datasets.indexed_dataset:> total number of documents: 1801350
968
+ INFO:megatron.core.datasets.gpt_dataset:Load the GPTDataset train indices
969
+ INFO:megatron.core.datasets.gpt_dataset: Load the document index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-document_index.npy
970
+ INFO:megatron.core.datasets.gpt_dataset: Load the sample index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-sample_index.npy
971
+ INFO:megatron.core.datasets.gpt_dataset: Load the shuffle index from 466b63c71b42dae59ebdc25db6336051-GPTDataset-train-shuffle_index.npy
972
+ INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 29073
973
+ > finished creating GPT datasets ...
974
+ [after dataloaders are built] datetime: 2026-08-26 18:26:30
975
+ done with setup ...
976
+ (min, max) time across ranks (ms):
977
+ model-and-optimizer-setup ......................: (329.96, 335.85)
978
+ train/valid/test-data-iterators-setup ..........: (61.94, 71.42)
979
+ training ...
980
+ Overwriting rerun_state_machine.current_iteration from -1 to 0...
981
+ [before the start of training step] datetime: 2026-08-26 18:26:30
982
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: 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:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
985
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
986
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: 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:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
989
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
990
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: 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:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
993
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
994
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: 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
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
997
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
998
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
999
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1000
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1001
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1002
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1003
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1004
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1005
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1006
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1007
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1008
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1009
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1010
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1011
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1012
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1013
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1014
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1015
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1016
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1017
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1018
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1019
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1020
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1021
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1022
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1023
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1024
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1025
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1026
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1027
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1028
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1029
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1030
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1031
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1032
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1033
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1034
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1035
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1036
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1037
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1038
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1039
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1040
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1041
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1042
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1043
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1044
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1045
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1046
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1047
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1048
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1049
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1050
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1051
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1052
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1053
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1054
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1055
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1056
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1057
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1058
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1059
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1060
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1061
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1062
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1063
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1064
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1065
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1066
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1067
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1068
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1069
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1070
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1071
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1072
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1073
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1074
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1075
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1076
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1077
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1078
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1079
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1080
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1081
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1082
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1083
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1084
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1085
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1086
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1087
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1088
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1089
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1090
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1091
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1092
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1093
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1094
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1095
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1096
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1097
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1098
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1099
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1100
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1101
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1102
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1103
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1104
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1105
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1106
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1107
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1108
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1109
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1110
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1111
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1112
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1113
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1114
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1115
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1116
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1117
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1118
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1119
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1120
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1121
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1122
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1123
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1124
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1125
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1126
+ [2026-08-26 18:27:08] iteration 1/ 12 | consumed samples: 8 | elapsed time per iteration (ms): 37376.1 | throughput per GPU (TFLOP/s/GPU): 3.3 | learning rate: 2.000000E-05 | global batch size: 8 | lm loss: 1.097931E+01 | loss scale: 1.0 | grad norm: 29.874 | number of skipped iterations: 0 | number of nan iterations: 0 |Number of parameters in transformer block in billions: 20.11
1127
+
1128
+ Number of parameters in embedding layers in billions: 0.36
1129
+ Total number of parameters in billions: 20.47
1130
+ Number of parameters in most loaded shard in billions: 20.4724
1131
+ Theoretical memory footprints: weight and optimizer=146430.14 MB
1132
+ [Rank 0] (after 1 iterations) memory (MB) | allocated: 68733.94384765625 | max allocated: 70397.458984375 | reserved: 80856.0 | max reserved: 80856.0
1133
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1134
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1135
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1136
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1137
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1138
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1139
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1140
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1141
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1142
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1143
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1144
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1145
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1146
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1147
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1148
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1149
+ [2026-08-26 18:27:08] iteration 2/ 12 | consumed samples: 16 | elapsed time per iteration (ms): 341.8 | throughput per GPU (TFLOP/s/GPU): 361.3 | learning rate: 4.000000E-05 | global batch size: 8 | lm loss: 1.097906E+01 | loss scale: 1.0 | grad norm: 30.624 | number of skipped iterations: 0 | number of nan iterations: 0 |
1150
+ [2026-08-26 18:27:08] iteration 3/ 12 | consumed samples: 24 | elapsed time per iteration (ms): 237.8 | throughput per GPU (TFLOP/s/GPU): 519.2 | learning rate: 6.000000E-05 | global batch size: 8 | lm loss: 9.538183E+00 | loss scale: 1.0 | grad norm: 32.372 | number of skipped iterations: 0 | number of nan iterations: 0 |
1151
+ Capture CUDA graph for training!!!
1152
+ CUDA graph capture done!!!
1153
+ [2026-08-26 18:27:10] iteration 4/ 12 | consumed samples: 32 | elapsed time per iteration (ms): 1711.7 | throughput per GPU (TFLOP/s/GPU): 72.1 | learning rate: 8.000000E-05 | global batch size: 8 | lm loss: 1.109221E+01 | loss scale: 1.0 | grad norm: 107.125 | number of skipped iterations: 0 | number of nan iterations: 0 |
1154
+ [2026-08-26 18:27:10] iteration 5/ 12 | consumed samples: 40 | elapsed time per iteration (ms): 184.2 | throughput per GPU (TFLOP/s/GPU): 670.5 | learning rate: 1.000000E-04 | global batch size: 8 | lm loss: 9.997550E+00 | loss scale: 1.0 | grad norm: 26.225 | number of skipped iterations: 0 | number of nan iterations: 0 |
1155
+ [2026-08-26 18:27:10] iteration 6/ 12 | consumed samples: 48 | elapsed time per iteration (ms): 180.0 | throughput per GPU (TFLOP/s/GPU): 686.2 | learning rate: 7.750000E-05 | global batch size: 8 | lm loss: 1.049700E+01 | loss scale: 1.0 | grad norm: 8.715 | number of skipped iterations: 0 | number of nan iterations: 0 |
1156
+ [2026-08-26 18:27:10] iteration 7/ 12 | consumed samples: 56 | elapsed time per iteration (ms): 178.5 | throughput per GPU (TFLOP/s/GPU): 691.9 | learning rate: 3.250000E-05 | global batch size: 8 | lm loss: 9.440020E+00 | loss scale: 1.0 | grad norm: 5.380 | number of skipped iterations: 0 | number of nan iterations: 0 |
1157
+ [2026-08-26 18:27:11] iteration 8/ 12 | consumed samples: 64 | elapsed time per iteration (ms): 178.4 | throughput per GPU (TFLOP/s/GPU): 692.1 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.961490E+00 | loss scale: 1.0 | grad norm: 4.547 | number of skipped iterations: 0 | number of nan iterations: 0 |
1158
+ [2026-08-26 18:27:11] iteration 9/ 12 | consumed samples: 72 | elapsed time per iteration (ms): 177.9 | throughput per GPU (TFLOP/s/GPU): 694.1 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.745807E+00 | loss scale: 1.0 | grad norm: 4.295 | number of skipped iterations: 0 | number of nan iterations: 0 |
1159
+ [2026-08-26 18:27:11] iteration 10/ 12 | consumed samples: 80 | elapsed time per iteration (ms): 180.7 | throughput per GPU (TFLOP/s/GPU): 683.5 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.685427E+00 | loss scale: 1.0 | grad norm: 4.421 | number of skipped iterations: 0 | number of nan iterations: 0 |
1160
+ [2026-08-26 18:27:11] iteration 11/ 12 | consumed samples: 88 | elapsed time per iteration (ms): 177.0 | throughput per GPU (TFLOP/s/GPU): 697.5 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.624660E+00 | loss scale: 1.0 | grad norm: 4.346 | number of skipped iterations: 0 | number of nan iterations: 0 |
1161
+ [2026-08-26 18:27:11] iteration 12/ 12 | consumed samples: 96 | elapsed time per iteration (ms): 178.4 | throughput per GPU (TFLOP/s/GPU): 692.0 | learning rate: 1.000000E-05 | global batch size: 8 | lm loss: 8.548999E+00 | loss scale: 1.0 | grad norm: 4.281 | number of skipped iterations: 0 | number of nan iterations: 0 |
1162
+ [after training is done] datetime: 2026-08-26 18:27:11
1163
+ [rank0]:[W826 18:27:13.179426812 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())
1164
+ W0826 18:27:14.938000 506268 torch/distributed/elastic/agent/server/api.py:719] Received 15 death signal, shutting down workers
1165
+ W0826 18:27:14.938000 506268 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 506355 closing signal SIGTERM
1166
+ W0826 18:27:14.939000 506268 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 506356 closing signal SIGTERM
1167
+ W0826 18:27:14.939000 506268 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 506359 closing signal SIGTERM
1168
+ W0826 18:27:14.939000 506268 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 506360 closing signal SIGTERM
1169
+ W0826 18:27:14.939000 506268 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 506361 closing signal SIGTERM
1170
+ W0826 18:27:14.940000 506268 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 506362 closing signal SIGTERM
1171
+ Traceback (most recent call last):
1172
+ File "/usr/local/bin/torchrun", line 8, in <module>
1173
+ sys.exit(main())
1174
+ ^^^^^^
1175
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper
1176
+ return f(*args, **kwargs)
1177
+ ^^^^^^^^^^^^^^^^^^
1178
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 892, in main
1179
+ run(args)
1180
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 883, in run
1181
+ elastic_launch(
1182
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 139, in __call__
1183
+ return launch_agent(self._config, self._entrypoint, list(args))
1184
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1185
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 261, in launch_agent
1186
+ result = agent.run()
1187
+ ^^^^^^^^^^^
1188
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/metrics/api.py", line 138, in wrapper
1189
+ result = f(*args, **kwargs)
1190
+ ^^^^^^^^^^^^^^^^^^
1191
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 711, in run
1192
+ result = self._invoke_run(role)
1193
+ ^^^^^^^^^^^^^^^^^^^^^^
1194
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 870, in _invoke_run
1195
+ time.sleep(monitor_interval)
1196
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/api.py", line 84, in _terminate_process_handler
1197
+ raise SignalException(f"Process {os.getpid()} got signal: {sigval}", sigval=sigval)
1198
+ torch.distributed.elastic.multiprocessing.api.SignalException: Process 506268 got signal: 15
echo/tierb/tierb_gbs8/results.tsv CHANGED
@@ -5,3 +5,9 @@ QWEN3_235B echo matched 16 8192 8 1.083984375 none 3 OK 311.93 310.00 146683 /wo
5
  QWEN3_30B echo matched 16 4096 8 1.11328125 none 1 OK 389.39 385.50 143605 /workspace/tierb_study_20260826_144644/echo/tierb/tierb_gbs8/QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125/rep1
6
  QWEN3_30B echo matched 16 4096 8 1.11328125 none 2 OK 389.86 385.20 143413 /workspace/tierb_study_20260826_144644/echo/tierb/tierb_gbs8/QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125/rep2
7
  QWEN3_30B echo matched 16 4096 8 1.11328125 none 3 OK 389.72 385.20 143509 /workspace/tierb_study_20260826_144644/echo/tierb/tierb_gbs8/QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125/rep3
 
 
 
 
 
 
 
5
  QWEN3_30B echo matched 16 4096 8 1.11328125 none 1 OK 389.39 385.50 143605 /workspace/tierb_study_20260826_144644/echo/tierb/tierb_gbs8/QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125/rep1
6
  QWEN3_30B echo matched 16 4096 8 1.11328125 none 2 OK 389.86 385.20 143413 /workspace/tierb_study_20260826_144644/echo/tierb/tierb_gbs8/QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125/rep2
7
  QWEN3_30B echo matched 16 4096 8 1.11328125 none 3 OK 389.72 385.20 143509 /workspace/tierb_study_20260826_144644/echo/tierb/tierb_gbs8/QWEN3_30B__echo__matched__seq4096__gbs8__cap1p11328125/rep3
8
+ DSV2_LITE_L24 echo matched 8 8192 8 1.0755208333333333 none 1 OK 587.49 584.70 151361 /workspace/tierb_study_20260826_144644/echo/tierb/tierb_gbs8/DSV2_LITE_L24__echo__matched__seq8192__gbs8__cap1p0755208333333333/rep1
9
+ DSV2_LITE_L24 echo matched 8 8192 8 1.0755208333333333 none 2 OK 587.32 584.70 151361 /workspace/tierb_study_20260826_144644/echo/tierb/tierb_gbs8/DSV2_LITE_L24__echo__matched__seq8192__gbs8__cap1p0755208333333333/rep2
10
+ DSV2_LITE_L24 echo matched 8 8192 8 1.0755208333333333 none 3 OK 587.66 584.70 151361 /workspace/tierb_study_20260826_144644/echo/tierb/tierb_gbs8/DSV2_LITE_L24__echo__matched__seq8192__gbs8__cap1p0755208333333333/rep3
11
+ QWEN2_57B echo matched 8 4096 8 1.08984375 none 1 OK 178.80 177.40 112327 /workspace/tierb_study_20260826_144644/echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep1
12
+ QWEN2_57B echo matched 8 4096 8 1.08984375 none 2 OK 178.74 177.20 112327 /workspace/tierb_study_20260826_144644/echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep2
13
+ QWEN2_57B echo matched 8 4096 8 1.08984375 none 3 OK 178.72 177.00 112607 /workspace/tierb_study_20260826_144644/echo/tierb/tierb_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/rep3