AbhishekGhosh1998 commited on
Commit
e85c195
·
verified ·
1 Parent(s): e95c7d1

granularity results tiera-rerun-fixed-moonep @ 2026-08-30T09:13:36Z from sp-aiinfra-interactive-aghosh-7jtjl-8xb200-moe-cg-0 (part 3)

Browse files
.gitattributes CHANGED
@@ -76,3 +76,4 @@ tiera-rerun-fixed-moonep/echo/tierbtrace/tierbtrace_gbs16/G3__echo__matched__seq
76
  tiera-rerun-fixed-moonep/echo/tierbtrace/tierbtrace_gbs16/G4__echo__matched__seq4096__gbs16__cap1p095703125/nsys/G4__echo__matched__seq4096__gbs16__cap1p095703125.nsys-rep filter=lfs diff=lfs merge=lfs -text
77
  tiera-rerun-fixed-moonep/ultraep/tierbtrace/tierbtrace_gbs16/G1__ultraep__matched__seq4096__gbs16__cap2p40625/nsys/G1__ultraep__matched__seq4096__gbs16__cap2p40625.nsys-rep filter=lfs diff=lfs merge=lfs -text
78
  tiera-rerun-fixed-moonep/ultraep/tierbtrace/tierbtrace_gbs16/G2__ultraep__matched__seq4096__gbs16__cap1p08203125/nsys/G2__ultraep__matched__seq4096__gbs16__cap1p08203125.nsys-rep filter=lfs diff=lfs merge=lfs -text
 
 
76
  tiera-rerun-fixed-moonep/echo/tierbtrace/tierbtrace_gbs16/G4__echo__matched__seq4096__gbs16__cap1p095703125/nsys/G4__echo__matched__seq4096__gbs16__cap1p095703125.nsys-rep filter=lfs diff=lfs merge=lfs -text
77
  tiera-rerun-fixed-moonep/ultraep/tierbtrace/tierbtrace_gbs16/G1__ultraep__matched__seq4096__gbs16__cap2p40625/nsys/G1__ultraep__matched__seq4096__gbs16__cap2p40625.nsys-rep filter=lfs diff=lfs merge=lfs -text
78
  tiera-rerun-fixed-moonep/ultraep/tierbtrace/tierbtrace_gbs16/G2__ultraep__matched__seq4096__gbs16__cap1p08203125/nsys/G2__ultraep__matched__seq4096__gbs16__cap1p08203125.nsys-rep filter=lfs diff=lfs merge=lfs -text
79
+ tiera-rerun-fixed-moonep/ultraep/tierbtrace/tierbtrace_gbs16/G3__ultraep__matched__seq4096__gbs16__cap1p072265625/nsys/G3__ultraep__matched__seq4096__gbs16__cap1p072265625.nsys-rep filter=lfs diff=lfs merge=lfs -text
tiera-rerun-fixed-moonep/ultraep/tierbtrace/tierbtrace_gbs16/G2__ultraep__matched__seq4096__gbs16__cap1p08203125/nsys/train.log CHANGED
@@ -1130,3 +1130,47 @@ Capture range ended in the application.
1130
  Generating '/tmp/nsys-report-f1d3.qdstrm'
1131
 
1132
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1130
  Generating '/tmp/nsys-report-f1d3.qdstrm'
1131
 
1132
 
1133
+ Collection stop is not allowed in this state.
1134
+ W0830 09:09:54.499000 910350 torch/distributed/elastic/agent/server/api.py:719] Received 15 death signal, shutting down workers
1135
+ W0830 09:09:54.500000 910350 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 910461 closing signal SIGTERM
1136
+ Generated:
1137
+ /workspace/tiera-rerun-fixed-moonep/ultraep/tierbtrace/tierbtrace_gbs16/G2__ultraep__matched__seq4096__gbs16__cap1p08203125/nsys/G2__ultraep__matched__seq4096__gbs16__cap1p08203125.nsys-rep
1138
+ W0830 09:09:54.506000 910350 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 910462 closing signal SIGTERM
1139
+ W0830 09:09:54.515000 910350 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 910463 closing signal SIGTERM
1140
+ W0830 09:09:54.516000 910350 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 910464 closing signal SIGTERM
1141
+ W0830 09:09:54.523000 910350 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 910465 closing signal SIGTERM
1142
+ W0830 09:09:54.526000 910350 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 910466 closing signal SIGTERM
1143
+ W0830 09:09:54.545000 910350 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 910467 closing signal SIGTERM
1144
+ W0830 09:09:54.547000 910350 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 910468 closing signal SIGTERM
1145
+ Traceback (most recent call last):
1146
+ File "/usr/local/bin/torchrun", line 8, in <module>
1147
+ sys.exit(main())
1148
+ ^^^^^^
1149
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper
1150
+ return f(*args, **kwargs)
1151
+ ^^^^^^^^^^^^^^^^^^
1152
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 892, in main
1153
+ run(args)
1154
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 883, in run
1155
+ elastic_launch(
1156
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 139, in __call__
1157
+ return launch_agent(self._config, self._entrypoint, list(args))
1158
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1159
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 261, in launch_agent
1160
+ result = agent.run()
1161
+ ^^^^^^^^^^^
1162
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/metrics/api.py", line 138, in wrapper
1163
+ result = f(*args, **kwargs)
1164
+ ^^^^^^^^^^^^^^^^^^
1165
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 711, in run
1166
+ result = self._invoke_run(role)
1167
+ ^^^^^^^^^^^^^^^^^^^^^^
1168
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 870, in _invoke_run
1169
+ time.sleep(monitor_interval)
1170
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/api.py", line 84, in _terminate_process_handler
1171
+ raise SignalException(f"Process {os.getpid()} got signal: {sigval}", sigval=sigval)
1172
+ torch.distributed.elastic.multiprocessing.api.SignalException: Process 910350 got signal: 15
1173
+ The target application terminated. One or more process it created re-parented.
1174
+ Waiting for termination of re-parented processes.
1175
+ Use the `--wait` option to modify this behavior.
1176
+ Session is not found: gran_G2__ultraep__matched__seq4096__gbs16__cap1p08203125_903224
tiera-rerun-fixed-moonep/ultraep/tierbtrace/tierbtrace_gbs16/G2__ultraep__matched__seq4096__gbs16__cap1p08203125/nsys/verify_trace_ranks.log ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ report : /workspace/tiera-rerun-fixed-moonep/ultraep/tierbtrace/tierbtrace_gbs16/G2__ultraep__matched__seq4096__gbs16__cap1p08203125/nsys/G2__ultraep__matched__seq4096__gbs16__cap1p08203125.nsys-rep
2
+ kernels : 143382
3
+ devices : 8 of 8 -> [0, 1, 2, 3, 4, 5, 6, 7]
4
+ processes: 1
5
+ markers : 26624 ultraep_trace__ kernels on devices [0, 1, 2, 3, 4, 5, 6, 7]
6
+ PASS: all ranks present and backend markers emitted
tiera-rerun-fixed-moonep/ultraep/tierbtrace/tierbtrace_gbs16/G3__ultraep__matched__seq4096__gbs16__cap1p072265625/nsys/G3__ultraep__matched__seq4096__gbs16__cap1p072265625.nsys-rep ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b39f354228ec2f8f860ce525fe38ade321a87bd67a1968c8a1c28eae6c1eadd1
3
+ size 12455065
tiera-rerun-fixed-moonep/ultraep/tierbtrace/tierbtrace_gbs16/G3__ultraep__matched__seq4096__gbs16__cap1p072265625/nsys/run.env ADDED
@@ -0,0 +1,70 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ backend=ultraep
2
+ config=G3
3
+ family=tierA
4
+ repo=/workspace/Megatron-UltraEP
5
+ branch=ultraep-bracing-regions-through-kernels
6
+ commit=03d57ef7
7
+ num_layers=16
8
+ hidden=4096
9
+ ffn_hidden=14336
10
+ moe_ffn_hidden=1792
11
+ num_experts=64
12
+ topk=16
13
+ experts_per_rank=8
14
+ expert_params=22020096
15
+ heads=32
16
+ kv_groups=8
17
+ kv_channels=
18
+ qk_layernorm=0
19
+ shared_ffn=0
20
+ shared_expert_overlap=0
21
+ moe_token_dispatcher=flex
22
+ mla=0
23
+ dense_first_layers=0
24
+ moe_layers=16
25
+ rotary_base=1000000
26
+ rotary_scaling=
27
+ qkv_bias=0
28
+ router_balancing=aux_loss
29
+ aux_loss_coeff=1e-2
30
+ router_pre_softmax=0
31
+ router_fusion=0
32
+ router_dtype=fp32
33
+ device_initiated_grouped_gemm=1
34
+ norm_epsilon=1e-5
35
+ seq_length=4096
36
+ micro_batch_size=1
37
+ global_batch_size=16
38
+ num_microbatches=2
39
+ assignments_per_rank=65536
40
+ mean_tokens_per_expert=8192
41
+ slot_policy=matched
42
+ slots_per_rank=8
43
+ capacity_mode=static
44
+ capacity=1.072265625
45
+ cuda_graph=1
46
+ probe_mode=0
47
+ async_grad_reduce=0
48
+ branch_role=timing
49
+ memory_dump_mode=none
50
+ memory_snapshot_path=
51
+ cuda_graph_warmup_steps=3
52
+ recompute=none
53
+ force_balance=0
54
+ train_iters=10
55
+ ep_size=8
56
+ echo_dispatch_overlap=1
57
+ echo_dispatch_recompute=0
58
+ ultraep_quota_min_tokens=1024
59
+ ultraep_weight_sync_join=
60
+ profile_enable=1
61
+ kernel_trace=1
62
+ fp8_mode=mxfp8
63
+ fp8_recipe=mxfp8
64
+ training_seed=1234
65
+ learning_rate=1e-4
66
+ min_learning_rate=1.0e-5
67
+ lr_warmup_iters=5
68
+ state_dump_enable=0
69
+ state_dump_dir=
70
+ moonep_assert_no_pending_reduce=0
tiera-rerun-fixed-moonep/ultraep/tierbtrace/tierbtrace_gbs16/G3__ultraep__matched__seq4096__gbs16__cap1p072265625/nsys/train.log ADDED
@@ -0,0 +1,1268 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ==============================================================
2
+ G3 / ultraep (slots/rank=8 policy=matched)
3
+ E=64 K=16 H'=1792 m=8
4
+ L=16 seq=4096 gbs=16 (2 microbatches)
5
+ cap_mode=static cap=1.072265625 cg=1 recompute=none
6
+ memory_dump=none snapshots=/workspace/tiera-rerun-fixed-moonep/ultraep/tierbtrace/tierbtrace_gbs16/G3__ultraep__matched__seq4096__gbs16__cap1p072265625/nsys/memory_snapshots/snapshot.pickle
7
+ repo=/workspace/Megatron-UltraEP @ 03d57ef7
8
+ ==============================================================
9
+ W0830 09:10:19.658000 917683 torch/distributed/run.py:766]
10
+ W0830 09:10:19.658000 917683 torch/distributed/run.py:766] *****************************************
11
+ W0830 09:10:19.658000 917683 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
+ W0830 09:10:19.658000 917683 torch/distributed/run.py:766] *****************************************
13
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
14
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
15
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
16
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
17
+ 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
+ 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-UltraEP/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 .................................... False
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 .............................. 16
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 ................................. 14336
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 ............................... 16
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 ..................................... 4096
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_capacity_monitor ............................ True
310
+ moe_deepep_num_sms .............................. 20
311
+ moe_echo_enable_random_offloading ............... False
312
+ moe_echo_expert_dispatch_overlap ................ False
313
+ moe_echo_expert_dispatcher_type ................. hybridep
314
+ moe_echo_recompute_expert_dispatch .............. False
315
+ moe_enable_deepep ............................... False
316
+ moe_enable_echo ................................. False
317
+ moe_enable_hybridep ............................. True
318
+ moe_enable_ultraep .............................. True
319
+ moe_expert_capacity_factor ...................... None
320
+ moe_extended_tp ................................. False
321
+ moe_ffn_hidden_size ............................. 1792
322
+ moe_grouped_gemm ................................ True
323
+ moe_hybridep_num_sms ............................ 16
324
+ moe_input_jitter_eps ............................ None
325
+ moe_layer_freq .................................. 1
326
+ moe_layer_recompute ............................. False
327
+ moe_num_echo_experts ............................ None
328
+ moe_num_redundant_experts_per_rank .............. 8
329
+ moe_pad_expert_input_to_capacity ................ False
330
+ moe_per_layer_logging ........................... False
331
+ moe_permute_fusion .............................. False
332
+ moe_received_token_capacity ..................... 1.072265625
333
+ moe_router_bias_update_rate ..................... 0.001
334
+ moe_router_dtype ................................ fp32
335
+ moe_router_enable_expert_bias ................... False
336
+ moe_router_force_load_balancing ................. False
337
+ moe_router_group_topk ........................... None
338
+ moe_router_load_balancing_type .................. aux_loss
339
+ moe_router_num_groups ........................... None
340
+ moe_router_padding_for_fp8 ...................... False
341
+ moe_router_pre_softmax .......................... False
342
+ moe_router_score_function ....................... softmax
343
+ moe_router_topk ................................. 16
344
+ moe_router_topk_scaling_factor .................. None
345
+ moe_shared_expert_intermediate_size ............. None
346
+ moe_shared_expert_overlap ....................... False
347
+ moe_token_dispatcher_type ....................... flex
348
+ moe_token_drop_policy ........................... probs
349
+ moe_ultraep_max_microbatches .................... None
350
+ moe_ultraep_overlap_weight_sync ................. True
351
+ moe_ultraep_weight_sync_join .................... before_dispatch
352
+ moe_upcycling_granularity ....................... 1
353
+ moe_use_device_initiated_grouped_gemm ........... True
354
+ moe_use_legacy_grouped_gemm ..................... False
355
+ moe_use_upcycling ............................... False
356
+ moe_z_loss_coeff ................................ None
357
+ mrope_section ................................... None
358
+ mscale .......................................... 1.0
359
+ mscale_all_dim .................................. 0.0
360
+ mtp_loss_scaling_factor ......................... 0.1
361
+ mtp_num_layers .................................. None
362
+ multi_latent_attention .......................... False
363
+ nccl_all_reduce_for_prefill ..................... False
364
+ nccl_communicator_config_path ................... None
365
+ nccl_ub ......................................... False
366
+ no_load_optim ................................... None
367
+ no_load_rng ..................................... None
368
+ no_persist_layer_norm ........................... False
369
+ no_rope_freq .................................... None
370
+ no_save_optim ................................... None
371
+ no_save_rng ..................................... None
372
+ non_persistent_ckpt_type ........................ None
373
+ non_persistent_global_ckpt_dir .................. None
374
+ non_persistent_local_ckpt_algo .................. fully_parallel
375
+ non_persistent_local_ckpt_dir ................... None
376
+ non_persistent_save_interval .................... None
377
+ norm_epsilon .................................... 1e-05
378
+ normalization ................................... RMSNorm
379
+ num_attention_heads ............................. 32
380
+ num_channels .................................... 3
381
+ num_classes ..................................... 1000
382
+ num_dataset_builder_threads ..................... 1
383
+ num_distributed_optimizer_instances ............. 1
384
+ num_experts ..................................... 64
385
+ num_layers ...................................... 16
386
+ num_layers_at_end_in_bf16 ....................... 1
387
+ num_layers_at_start_in_bf16 ..................... 1
388
+ num_layers_per_virtual_pipeline_stage ........... None
389
+ num_query_groups ................................ 8
390
+ num_virtual_stages_per_pipeline_rank ............ None
391
+ num_workers ..................................... 2
392
+ object_storage_cache_path ....................... None
393
+ one_logger_async ................................ False
394
+ one_logger_project .............................. megatron-lm
395
+ one_logger_run_name ............................. None
396
+ onnx_safe ....................................... None
397
+ openai_gelu ..................................... False
398
+ optimizer ....................................... adam
399
+ optimizer_cpu_offload ........................... False
400
+ optimizer_offload_fraction ...................... 1.0
401
+ output_bert_embeddings .......................... False
402
+ overlap_cpu_optimizer_d2h_h2d ................... False
403
+ overlap_grad_reduce ............................. True
404
+ overlap_moe_expert_parallel_comm ................ False
405
+ overlap_p2p_comm ................................ False
406
+ overlap_p2p_comm_warmup_flush ................... False
407
+ overlap_param_gather ............................ True
408
+ overlap_param_gather_with_optimizer_step ........ False
409
+ override_opt_param_scheduler .................... False
410
+ params_dtype .................................... torch.bfloat16
411
+ patch_dim ....................................... 16
412
+ per_split_data_args_path ........................ None
413
+ perform_initialization .......................... True
414
+ pin_cpu_grads ................................... True
415
+ pin_cpu_params .................................. True
416
+ pipeline_model_parallel_comm_backend ............ None
417
+ pipeline_model_parallel_layout .................. None
418
+ pipeline_model_parallel_size .................... 1
419
+ position_embedding_type ......................... rope
420
+ pretrained_checkpoint ........................... None
421
+ profile ......................................... True
422
+ profile_ranks ................................... [0, 1, 2, 3, 4, 5, 6, 7]
423
+ profile_step_end ................................ 8
424
+ profile_step_start .............................. 6
425
+ q_lora_rank ..................................... None
426
+ qk_head_dim ..................................... 128
427
+ qk_l2_norm ...................................... False
428
+ qk_layernorm .................................... False
429
+ qk_pos_emb_head_dim ............................. 64
430
+ query_in_block_prob ............................. 0.1
431
+ rampup_batch_size ............................... None
432
+ rank ............................................ 0
433
+ recompute_granularity ........................... None
434
+ recompute_method ................................ None
435
+ recompute_modules ............................... None
436
+ recompute_num_layers ............................ None
437
+ record_memory_history ........................... False
438
+ relative_attention_max_distance ................. 128
439
+ relative_attention_num_buckets .................. 32
440
+ replication ..................................... False
441
+ replication_factor .............................. 2
442
+ replication_jump ................................ None
443
+ rerun_mode ...................................... validate_results
444
+ reset_attention_mask ............................ False
445
+ reset_position_ids .............................. False
446
+ result_rejected_tracker_filename ................ None
447
+ retriever_report_topk_accuracies ................ []
448
+ retriever_score_scaling ......................... False
449
+ retriever_seq_length ............................ 256
450
+ retro_add_retriever ............................. False
451
+ retro_attention_gate ............................ 1
452
+ retro_cyclic_train_iters ........................ None
453
+ retro_encoder_attention_dropout ................. 0.1
454
+ retro_encoder_hidden_dropout .................... 0.1
455
+ retro_encoder_layers ............................ 2
456
+ retro_num_neighbors ............................. 2
457
+ retro_num_retrieved_chunks ...................... 2
458
+ retro_project_dir ............................... None
459
+ retro_verify_neighbor_count ..................... True
460
+ reuse_grad_buf_for_mxfp8_param_ag ............... False
461
+ rope_scaling_factor ............................. 8.0
462
+ rope_type ....................................... None
463
+ rotary_base ..................................... 1000000
464
+ rotary_interleaved .............................. False
465
+ rotary_percent .................................. 1.0
466
+ rotary_scaling_factor ........................... 1.0
467
+ rotary_seq_len_interpolation_factor ............. None
468
+ run_workload_inspector_server ................... False
469
+ sample_rate ..................................... 1.0
470
+ save ............................................ None
471
+ save_interval ................................... None
472
+ save_retain_interval ............................ None
473
+ scatter_gather_tensors_in_pipeline .............. True
474
+ seed ............................................ 1234
475
+ seq_length ...................................... 4096
476
+ sequence_parallel ............................... False
477
+ sft ............................................. False
478
+ sft_tokenizer_prompt_format ..................... nemotron-h-aligned
479
+ sgd_momentum .................................... 0.9
480
+ short_seq_prob .................................. 0.1
481
+ skip_train ...................................... False
482
+ skipped_train_samples ........................... 0
483
+ spec ............................................ None
484
+ split ........................................... 100,0,0
485
+ squared_relu .................................... False
486
+ start_weight_decay .............................. 0.1
487
+ straggler_ctrlr_port ............................ 65535
488
+ straggler_minmax_count .......................... 1
489
+ suggested_communication_unit_size ............... None
490
+ swiglu .......................................... True
491
+ swin_backbone_type .............................. tiny
492
+ symmetric_ar_type ............................... None
493
+ te_rng_tracker .................................. True
494
+ tensor_model_parallel_size ...................... 1
495
+ tensorboard_dir ................................. None
496
+ tensorboard_log_interval ........................ 1
497
+ tensorboard_queue_size .......................... 1000
498
+ test_data_path .................................. None
499
+ test_mode ....................................... False
500
+ tiktoken_num_special_tokens ..................... 1000
501
+ tiktoken_pattern ................................ None
502
+ tiktoken_special_tokens ......................... None
503
+ timing_log_level ................................ 0
504
+ timing_log_option ............................... minmax
505
+ titles_data_path ................................ None
506
+ tokenizer_model ................................. None
507
+ tokenizer_type .................................. GPT2BPETokenizer
508
+ torch_fsdp2_reshard_after_forward ............... True
509
+ tp_comm_bootstrap_backend ....................... nccl
510
+ tp_comm_bulk_dgrad .............................. True
511
+ tp_comm_bulk_wgrad .............................. True
512
+ tp_comm_overlap ................................. False
513
+ tp_comm_overlap_ag .............................. True
514
+ tp_comm_overlap_cfg ............................. None
515
+ tp_comm_overlap_rs .............................. True
516
+ tp_comm_overlap_rs_dgrad ........................ False
517
+ tp_comm_split_ag ................................ True
518
+ tp_comm_split_rs ................................ True
519
+ train_data_path ................................. None
520
+ train_iters ..................................... 10
521
+ train_samples ................................... None
522
+ train_sync_interval ............................. None
523
+ transformer_impl ................................ transformer_engine
524
+ transformer_pipeline_model_parallel_size ........ 1
525
+ untie_embeddings_and_output_weights ............. True
526
+ use_checkpoint_args ............................. False
527
+ use_checkpoint_opt_param_scheduler .............. False
528
+ use_cpu_initialization .......................... None
529
+ use_custom_fsdp ................................. False
530
+ use_dist_ckpt ................................... True
531
+ use_dist_ckpt_deprecated ........................ False
532
+ use_distributed_optimizer ....................... True
533
+ use_flash_attn .................................. False
534
+ use_legacy_models ............................... False
535
+ use_mp_args_from_checkpoint_args ................ False
536
+ use_one_sent_docs ............................... False
537
+ use_persistent_ckpt_worker ...................... False
538
+ use_precision_aware_optimizer ................... False
539
+ use_pytorch_profiler ............................ False
540
+ use_ring_exchange_p2p ........................... False
541
+ use_rope_scaling ................................ False
542
+ use_rotary_position_embeddings .................. False
543
+ use_sharp ....................................... False
544
+ use_tokenizer_model_from_checkpoint_args ........ True
545
+ use_torch_fsdp2 ................................. False
546
+ use_torch_optimizer_for_cpu_offload ............. False
547
+ use_tp_pp_dp_mapping ............................ False
548
+ v_head_dim ...................................... 128
549
+ valid_data_path ................................. None
550
+ variable_seq_lengths ............................ False
551
+ virtual_pipeline_model_parallel_size ............ None
552
+ vision_backbone_type ............................ vit
553
+ vision_pretraining .............................. False
554
+ vision_pretraining_type ......................... classify
555
+ vocab_extra_ids ................................. 0
556
+ vocab_file ...................................... /workspace/data/wikitext/gpt2-vocab.json
557
+ vocab_size ...................................... None
558
+ wandb_exp_name ..................................
559
+ wandb_project ...................................
560
+ wandb_save_dir ..................................
561
+ weight_decay .................................... 0.1
562
+ weight_decay_incr_style ......................... constant
563
+ wgrad_deferral_limit ............................ 0
564
+ world_size ...................................... 8
565
+ yaml_cfg ........................................ None
566
+ -------------------- end of arguments ---------------------
567
+ INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 2
568
+ > building GPT2BPETokenizer tokenizer ...
569
+ > padded vocab (size: 50257) with 47 dummy tokens (new size: 50304)
570
+ WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.VALIDATE_RESULTS
571
+ > initializing torch distributed ...
572
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
573
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
574
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
575
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
576
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
577
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
578
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
579
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
580
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
581
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
582
+ [Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
583
+ [Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
584
+ [Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
585
+ [Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
586
+ [Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
587
+ [Gloo] Rank 5 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
588
+ [Gloo] Rank 7 is connected to 7 peer ranks. [Gloo] Rank Expected number of connected peer ranks is : 76
589
+ is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
590
+ [Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
591
+ [Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
592
+ [Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
593
+ [Gloo] Rank 5 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
594
+ [Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
595
+ [Gloo] Rank 6 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
596
+ [Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
597
+ [Gloo] Rank 7 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
598
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
599
+ [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
+ > initialized tensor model parallel with size 1
603
+ > initialized pipeline model parallel with size 1
604
+ > setting random seeds to 1234 ...
605
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
606
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
607
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
608
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
609
+ [rank5]:[W830 09:10:33.560063371 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 5] using GPU 5 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
610
+ [rank2]:[W830 09:10:33.560228756 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.
611
+ [rank4]:[W830 09:10:33.561043575 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.
612
+ [rank7]:[W830 09:10:33.561119539 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
613
+ [rank3]:[W830 09:10:33.561260220 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 3] using GPU 3 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
614
+ [rank1]:[W830 09:10:33.561269856 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.
615
+ [rank6]:[W830 09:10:33.563458086 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 6] using GPU 6 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
616
+ > compiling dataset index builder ...
617
+ make: Entering directory '/workspace/Megatron-UltraEP/megatron/core/datasets'
618
+ make: Nothing to be done for 'default'.
619
+ make: Leaving directory '/workspace/Megatron-UltraEP/megatron/core/datasets'
620
+ >>> done with dataset index builder. Compilation time: 0.418 seconds
621
+ WARNING: constraints for invoking optimized fused softmax kernel are not met. We default back to unfused kernel invocations.
622
+ > compiling and loading fused kernels ...
623
+ [rank0]:[W830 09:10:33.162692656 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 0] using GPU 0 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
624
+ >>> done with compiling and loading fused kernels. Compilation time: 3.654 seconds
625
+ time to initialize megatron (seconds): 9.566
626
+ [after megatron is initialized] datetime: 2026-08-30 09:10:39
627
+ building GPT model ...
628
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
629
+ warnings.warn(
630
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
631
+ warnings.warn(
632
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
633
+ warnings.warn(
634
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
635
+ warnings.warn(
636
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
637
+ warnings.warn(
638
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
639
+ warnings.warn(
640
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
641
+ warnings.warn(
642
+ /usr/local/lib/python3.12/dist-packages/transformer_engine/pytorch/cpu_offload.py:708: DeprecationWarning: Offloading weights is deprecated. Using offload_weights=True does not have any effect.
643
+ warnings.warn(
644
+ [INFO] Use MNNVL fabric: False
645
+ [WARN] MAX_NUM_NVL_PEERS is not set. Using detected value 8.
646
+ NVSHMEM v3.7.2
647
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
648
+
649
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
650
+
651
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
652
+
653
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
654
+
655
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
656
+
657
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
658
+
659
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
660
+
661
+ /dvs/p4/build/sw/rel/gpgpu/toolkit/r12.9/main_nvshmem/src/modules/transport/ibrc/ibrc.cpp:1754: error status: 7 (NVSHMEMX_ERROR_INTERNAL) Device enumeration failed.
662
+
663
+ UltraEP: NVL domain size <= 8; forcing weight_sync plan mode to direct (was adaptive).
664
+ > number of parameters on (tensor, pipeline) model parallel rank (0, 0): 6724653056
665
+ INFO:megatron.core.distributed.distributed_data_parallel:Setting up DistributedDataParallel with config DistributedDataParallelConfig(grad_reduce_in_fp32=True, overlap_grad_reduce=True, overlap_param_gather=True, align_param_gather=False, use_distributed_optimizer=True, num_distributed_optimizer_instances=1, check_for_nan_in_grad=False, check_for_large_grads=False, bucket_size=40000000, pad_buckets_for_high_nccl_busbw=False, average_in_collective=False, fp8_param_gather=False, reuse_grad_buf_for_mxfp8_param_ag=False, use_custom_fsdp=False, data_parallel_sharding_strategy='no_shard', gradient_reduce_div_fusion=True, suggested_communication_unit_size=None, preserve_fp32_weights=True, keep_fp8_transpose_cache_when_using_custom_fsdp=False, nccl_ub=False, fsdp_double_buffer=False)
666
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 18
667
+ Params for bucket 1 (206045184 elements, 206045184 padded size):
668
+ module.output_layer.weight
669
+ Params for bucket 2 (42217472 elements, 42217472 padded size):
670
+ module.decoder.layers.15.self_attention.linear_qkv.weight
671
+ module.decoder.layers.15.pre_mlp_layernorm.weight
672
+ module.decoder.final_layernorm.weight
673
+ module.decoder.layers.15.self_attention.linear_qkv.layer_norm_weight
674
+ module.decoder.layers.15.mlp.router.weight
675
+ module.decoder.layers.15.self_attention.linear_proj.weight
676
+ Params for bucket 3 (42213376 elements, 42213376 padded size):
677
+ module.decoder.layers.14.self_attention.linear_qkv.weight
678
+ module.decoder.layers.14.pre_mlp_layernorm.weight
679
+ module.decoder.layers.14.mlp.router.weight
680
+ module.decoder.layers.14.self_attention.linear_proj.weight
681
+ module.decoder.layers.14.self_attention.linear_qkv.layer_norm_weight
682
+ Params for bucket 4 (42213376 elements, 42213376 padded size):
683
+ module.decoder.layers.13.mlp.router.weight
684
+ module.decoder.layers.13.self_attention.linear_proj.weight
685
+ module.decoder.layers.13.self_attention.linear_qkv.layer_norm_weight
686
+ module.decoder.layers.13.pre_mlp_layernorm.weight
687
+ module.decoder.layers.13.self_attention.linear_qkv.weight
688
+ Params for bucket 5 (42213376 elements, 42213376 padded size):
689
+ module.decoder.layers.12.mlp.router.weight
690
+ module.decoder.layers.12.self_attention.linear_proj.weight
691
+ module.decoder.layers.12.self_attention.linear_qkv.layer_norm_weight
692
+ module.decoder.layers.12.pre_mlp_layernorm.weight
693
+ module.decoder.layers.12.self_attention.linear_qkv.weight
694
+ Params for bucket 6 (42213376 elements, 42213376 padded size):
695
+ module.decoder.layers.11.mlp.router.weight
696
+ module.decoder.layers.11.self_attention.linear_qkv.layer_norm_weight
697
+ module.decoder.layers.11.self_attention.linear_proj.weight
698
+ module.decoder.layers.11.self_attention.linear_qkv.weight
699
+ module.decoder.layers.11.pre_mlp_layernorm.weight
700
+ Params for bucket 7 (42213376 elements, 42213376 padded size):
701
+ module.decoder.layers.10.self_attention.linear_qkv.weight
702
+ module.decoder.layers.10.pre_mlp_layernorm.weight
703
+ module.decoder.layers.10.mlp.router.weight
704
+ module.decoder.layers.10.self_attention.linear_proj.weight
705
+ module.decoder.layers.10.self_attention.linear_qkv.layer_norm_weight
706
+ Params for bucket 8 (42213376 elements, 42213376 padded size):
707
+ module.decoder.layers.9.mlp.router.weight
708
+ module.decoder.layers.9.self_attention.linear_proj.weight
709
+ module.decoder.layers.9.self_attention.linear_qkv.layer_norm_weight
710
+ module.decoder.layers.9.self_attention.linear_qkv.weight
711
+ module.decoder.layers.9.pre_mlp_layernorm.weight
712
+ Params for bucket 9 (42213376 elements, 42213376 padded size):
713
+ module.decoder.layers.8.pre_mlp_layernorm.weight
714
+ module.decoder.layers.8.self_attention.linear_qkv.weight
715
+ module.decoder.layers.8.mlp.router.weight
716
+ module.decoder.layers.8.self_attention.linear_proj.weight
717
+ module.decoder.layers.8.self_attention.linear_qkv.layer_norm_weight
718
+ Params for bucket 10 (42213376 elements, 42213376 padded size):
719
+ module.decoder.layers.7.mlp.router.weight
720
+ module.decoder.layers.7.self_attention.linear_qkv.layer_norm_weight
721
+ module.decoder.layers.7.self_attention.linear_proj.weight
722
+ module.decoder.layers.7.pre_mlp_layernorm.weight
723
+ module.decoder.layers.7.self_attention.linear_qkv.weight
724
+ Params for bucket 11 (42213376 elements, 42213376 padded size):
725
+ module.decoder.layers.6.mlp.router.weight
726
+ module.decoder.layers.6.self_attention.linear_qkv.layer_norm_weight
727
+ module.decoder.layers.6.self_attention.linear_proj.weight
728
+ module.decoder.layers.6.self_attention.linear_qkv.weight
729
+ module.decoder.layers.6.pre_mlp_layernorm.weight
730
+ Params for bucket 12 (42213376 elements, 42213376 padded size):
731
+ module.decoder.layers.5.self_attention.linear_proj.weight
732
+ module.decoder.layers.5.pre_mlp_layernorm.weight
733
+ module.decoder.layers.5.self_attention.linear_qkv.weight
734
+ module.decoder.layers.5.mlp.router.weight
735
+ module.decoder.layers.5.self_attention.linear_qkv.layer_norm_weight
736
+ Params for bucket 13 (42213376 elements, 42213376 padded size):
737
+ module.decoder.layers.4.self_attention.linear_qkv.weight
738
+ module.decoder.layers.4.pre_mlp_layernorm.weight
739
+ module.decoder.layers.4.mlp.router.weight
740
+ module.decoder.layers.4.self_attention.linear_qkv.layer_norm_weight
741
+ module.decoder.layers.4.self_attention.linear_proj.weight
742
+ Params for bucket 14 (42213376 elements, 42213376 padded size):
743
+ module.decoder.layers.3.pre_mlp_layernorm.weight
744
+ module.decoder.layers.3.self_attention.linear_qkv.weight
745
+ module.decoder.layers.3.mlp.router.weight
746
+ module.decoder.layers.3.self_attention.linear_proj.weight
747
+ module.decoder.layers.3.self_attention.linear_qkv.layer_norm_weight
748
+ Params for bucket 15 (42213376 elements, 42213376 padded size):
749
+ module.decoder.layers.2.mlp.router.weight
750
+ module.decoder.layers.2.self_attention.linear_proj.weight
751
+ module.decoder.layers.2.self_attention.linear_qkv.layer_norm_weight
752
+ module.decoder.layers.2.pre_mlp_layernorm.weight
753
+ module.decoder.layers.2.self_attention.linear_qkv.weight
754
+ Params for bucket 16 (42213376 elements, 42213376 padded size):
755
+ module.decoder.layers.1.pre_mlp_layernorm.weight
756
+ module.decoder.layers.1.self_attention.linear_qkv.weight
757
+ module.decoder.layers.1.mlp.router.weight
758
+ module.decoder.layers.1.self_attention.linear_proj.weight
759
+ module.decoder.layers.1.self_attention.linear_qkv.layer_norm_weight
760
+ Params for bucket 17 (42213376 elements, 42213376 padded size):
761
+ module.decoder.layers.0.mlp.router.weight
762
+ module.decoder.layers.0.self_attention.linear_qkv.weight
763
+ module.decoder.layers.0.pre_mlp_layernorm.weight
764
+ module.decoder.layers.0.self_attention.linear_qkv.layer_norm_weight
765
+ module.decoder.layers.0.self_attention.linear_proj.weight
766
+ Params for bucket 18 (206045184 elements, 206045184 padded size):
767
+ module.embedding.word_embeddings.weight
768
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 64
769
+ Params for bucket 1 (44040192 elements, 44040192 padded size):
770
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight4
771
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight6
772
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight2
773
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight7
774
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight5
775
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight3
776
+ Params for bucket 2 (44040192 elements, 44040192 padded size):
777
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight1
778
+ module.decoder.layers.15.mlp.experts.linear_fc2.weight0
779
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight7
780
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight6
781
+ Params for bucket 3 (44040192 elements, 44040192 padded size):
782
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight5
783
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight4
784
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight3
785
+ Params for bucket 4 (44040192 elements, 44040192 padded size):
786
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight2
787
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight1
788
+ module.decoder.layers.15.mlp.experts.linear_fc1.weight0
789
+ Params for bucket 5 (44040192 elements, 44040192 padded size):
790
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight6
791
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight4
792
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight2
793
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight7
794
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight5
795
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight3
796
+ Params for bucket 6 (44040192 elements, 44040192 padded size):
797
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight1
798
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight6
799
+ module.decoder.layers.14.mlp.experts.linear_fc2.weight0
800
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight7
801
+ Params for bucket 7 (44040192 elements, 44040192 padded size):
802
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight5
803
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight3
804
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight4
805
+ Params for bucket 8 (44040192 elements, 44040192 padded size):
806
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight2
807
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight0
808
+ module.decoder.layers.14.mlp.experts.linear_fc1.weight1
809
+ Params for bucket 9 (44040192 elements, 44040192 padded size):
810
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight7
811
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight5
812
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight3
813
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight6
814
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight4
815
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight2
816
+ Params for bucket 10 (44040192 elements, 44040192 padded size):
817
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight1
818
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight6
819
+ module.decoder.layers.13.mlp.experts.linear_fc2.weight0
820
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight7
821
+ Params for bucket 11 (44040192 elements, 44040192 padded size):
822
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight5
823
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight3
824
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight4
825
+ Params for bucket 12 (44040192 elements, 44040192 padded size):
826
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight2
827
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight0
828
+ module.decoder.layers.13.mlp.experts.linear_fc1.weight1
829
+ Params for bucket 13 (44040192 elements, 44040192 padded size):
830
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight7
831
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight3
832
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight5
833
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight4
834
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight6
835
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight2
836
+ Params for bucket 14 (44040192 elements, 44040192 padded size):
837
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight1
838
+ module.decoder.layers.12.mlp.experts.linear_fc2.weight0
839
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight7
840
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight6
841
+ Params for bucket 15 (44040192 elements, 44040192 padded size):
842
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight5
843
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight3
844
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight4
845
+ Params for bucket 16 (44040192 elements, 44040192 padded size):
846
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight2
847
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight0
848
+ module.decoder.layers.12.mlp.experts.linear_fc1.weight1
849
+ Params for bucket 17 (44040192 elements, 44040192 padded size):
850
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight7
851
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight5
852
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight3
853
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight6
854
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight4
855
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight2
856
+ Params for bucket 18 (44040192 elements, 44040192 padded size):
857
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight1
858
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight6
859
+ module.decoder.layers.11.mlp.experts.linear_fc2.weight0
860
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight7
861
+ Params for bucket 19 (44040192 elements, 44040192 padded size):
862
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight5
863
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight4
864
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight3
865
+ Params for bucket 20 (44040192 elements, 44040192 padded size):
866
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight2
867
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight1
868
+ module.decoder.layers.11.mlp.experts.linear_fc1.weight0
869
+ Params for bucket 21 (44040192 elements, 44040192 padded size):
870
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight7
871
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight3
872
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight5
873
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight6
874
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight4
875
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight2
876
+ Params for bucket 22 (44040192 elements, 44040192 padded size):
877
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight1
878
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight6
879
+ module.decoder.layers.10.mlp.experts.linear_fc2.weight0
880
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight7
881
+ Params for bucket 23 (44040192 elements, 44040192 padded size):
882
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight5
883
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight3
884
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight4
885
+ Params for bucket 24 (44040192 elements, 44040192 padded size):
886
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight2
887
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight0
888
+ module.decoder.layers.10.mlp.experts.linear_fc1.weight1
889
+ Params for bucket 25 (44040192 elements, 44040192 padded size):
890
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight7
891
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight5
892
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight3
893
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight6
894
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight4
895
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight2
896
+ Params for bucket 26 (44040192 elements, 44040192 padded size):
897
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight1
898
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight6
899
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight0
900
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight7
901
+ Params for bucket 27 (44040192 elements, 44040192 padded size):
902
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight5
903
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight3
904
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight4
905
+ Params for bucket 28 (44040192 elements, 44040192 padded size):
906
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight2
907
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight0
908
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight1
909
+ Params for bucket 29 (44040192 elements, 44040192 padded size):
910
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight6
911
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight4
912
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight2
913
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight7
914
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight5
915
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight3
916
+ Params for bucket 30 (44040192 elements, 44040192 padded size):
917
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight1
918
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight6
919
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight0
920
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight7
921
+ Params for bucket 31 (44040192 elements, 44040192 padded size):
922
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight5
923
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight3
924
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight4
925
+ Params for bucket 32 (44040192 elements, 44040192 padded size):
926
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight2
927
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight1
928
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight0
929
+ Params for bucket 33 (44040192 elements, 44040192 padded size):
930
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight7
931
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight3
932
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight5
933
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight6
934
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight4
935
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight2
936
+ Params for bucket 34 (44040192 elements, 44040192 padded size):
937
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight1
938
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight0
939
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight7
940
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight6
941
+ Params for bucket 35 (44040192 elements, 44040192 padded size):
942
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight5
943
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight3
944
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight4
945
+ Params for bucket 36 (44040192 elements, 44040192 padded size):
946
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight2
947
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight0
948
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight1
949
+ Params for bucket 37 (44040192 elements, 44040192 padded size):
950
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight7
951
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight3
952
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight5
953
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight6
954
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight4
955
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight2
956
+ Params for bucket 38 (44040192 elements, 44040192 padded size):
957
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight1
958
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight0
959
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight7
960
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight6
961
+ Params for bucket 39 (44040192 elements, 44040192 padded size):
962
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight5
963
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight4
964
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight3
965
+ Params for bucket 40 (44040192 elements, 44040192 padded size):
966
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight2
967
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight1
968
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight0
969
+ Params for bucket 41 (44040192 elements, 44040192 padded size):
970
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight7
971
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight5
972
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight3
973
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight6
974
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight4
975
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight2
976
+ Params for bucket 42 (44040192 elements, 44040192 padded size):
977
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight1
978
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight0
979
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight6
980
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight7
981
+ Params for bucket 43 (44040192 elements, 44040192 padded size):
982
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight5
983
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight3
984
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight4
985
+ Params for bucket 44 (44040192 elements, 44040192 padded size):
986
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight2
987
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight0
988
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight1
989
+ Params for bucket 45 (44040192 elements, 44040192 padded size):
990
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight4
991
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight6
992
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight2
993
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight7
994
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight5
995
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight3
996
+ Params for bucket 46 (44040192 elements, 44040192 padded size):
997
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight1
998
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight7
999
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight6
1000
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight0
1001
+ Params for bucket 47 (44040192 elements, 44040192 padded size):
1002
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight5
1003
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight4
1004
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight3
1005
+ Params for bucket 48 (44040192 elements, 44040192 padded size):
1006
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight2
1007
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight1
1008
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight0
1009
+ Params for bucket 49 (44040192 elements, 44040192 padded size):
1010
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight6
1011
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight4
1012
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight2
1013
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight7
1014
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight5
1015
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight3
1016
+ Params for bucket 50 (44040192 elements, 44040192 padded size):
1017
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight1
1018
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight6
1019
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight0
1020
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight7
1021
+ Params for bucket 51 (44040192 elements, 44040192 padded size):
1022
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight5
1023
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight3
1024
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight4
1025
+ Params for bucket 52 (44040192 elements, 44040192 padded size):
1026
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight2
1027
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight1
1028
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight0
1029
+ Params for bucket 53 (44040192 elements, 44040192 padded size):
1030
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight4
1031
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight6
1032
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight2
1033
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight7
1034
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight5
1035
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight3
1036
+ Params for bucket 54 (44040192 elements, 44040192 padded size):
1037
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight1
1038
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight0
1039
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight6
1040
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight7
1041
+ Params for bucket 55 (44040192 elements, 44040192 padded size):
1042
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight5
1043
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight3
1044
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight4
1045
+ Params for bucket 56 (44040192 elements, 44040192 padded size):
1046
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight2
1047
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight0
1048
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight1
1049
+ Params for bucket 57 (44040192 elements, 44040192 padded size):
1050
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight6
1051
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight4
1052
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight2
1053
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight7
1054
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight5
1055
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight3
1056
+ Params for bucket 58 (44040192 elements, 44040192 padded size):
1057
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight1
1058
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight6
1059
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight0
1060
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight7
1061
+ Params for bucket 59 (44040192 elements, 44040192 padded size):
1062
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight5
1063
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight3
1064
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight4
1065
+ Params for bucket 60 (44040192 elements, 44040192 padded size):
1066
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight2
1067
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight0
1068
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight1
1069
+ Params for bucket 61 (44040192 elements, 44040192 padded size):
1070
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight4
1071
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight6
1072
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight2
1073
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight7
1074
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight5
1075
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight3
1076
+ Params for bucket 62 (44040192 elements, 44040192 padded size):
1077
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight1
1078
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight6
1079
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight0
1080
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight7
1081
+ Params for bucket 63 (44040192 elements, 44040192 padded size):
1082
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight5
1083
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight4
1084
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight3
1085
+ Params for bucket 64 (44040192 elements, 44040192 padded size):
1086
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight2
1087
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight0
1088
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight1
1089
+ 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 0x72c917d6d310>, config_logger_dir='')
1090
+ INFO:megatron.core.optimizer_param_scheduler:> learning rate decay style: cosine
1091
+ [after model, optimizer, and learning rate scheduler are built] datetime: 2026-08-30 09:10:43
1092
+ > building train, validation, and test datasets ...
1093
+ > datasets target sizes (minimum size):
1094
+ train: 160
1095
+ validation: 0
1096
+ test: 0
1097
+ INFO:megatron.core.datasets.blended_megatron_dataset_config:Let split_matrix = [(0, 1.0), None, None]
1098
+ > building train, validation, and test datasets for GPT ...
1099
+ INFO:megatron.core.datasets.blended_megatron_dataset_builder:Building GPTDataset splits with sizes=(160, 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 0x72c918f7e240>, 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)
1100
+ INFO:megatron.core.datasets.indexed_dataset:Load the _IndexReader from /workspace/data/wikitext/wikitext103_gpt2_text_document.idx
1101
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence lengths
1102
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence pointers
1103
+ INFO:megatron.core.datasets.indexed_dataset: Extract the document indices
1104
+ INFO:megatron.core.datasets.indexed_dataset:> total number of sequences: 1165029
1105
+ INFO:megatron.core.datasets.indexed_dataset:> total number of documents: 1801350
1106
+ INFO:megatron.core.datasets.gpt_dataset:Load the GPTDataset train indices
1107
+ INFO:megatron.core.datasets.gpt_dataset: Load the document index from 68684a515e118253a8572c4517cf2ffe-GPTDataset-train-document_index.npy
1108
+ INFO:megatron.core.datasets.gpt_dataset: Load the sample index from 68684a515e118253a8572c4517cf2ffe-GPTDataset-train-sample_index.npy
1109
+ INFO:megatron.core.datasets.gpt_dataset: Load the shuffle index from 68684a515e118253a8572c4517cf2ffe-GPTDataset-train-shuffle_index.npy
1110
+ INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 29073
1111
+ > finished creating GPT datasets ...
1112
+ [after dataloaders are built] datetime: 2026-08-30 09:10:43
1113
+ done with setup ...
1114
+ (min, max) time across ranks (ms):
1115
+ model-and-optimizer-setup ......................: (3546.39, 3549.29)
1116
+ train/valid/test-data-iterators-setup ..........: (91.31, 126.56)
1117
+ training ...
1118
+ Overwriting rerun_state_machine.current_iteration from -1 to 0...
1119
+ [before the start of training step] datetime: 2026-08-30 09:10:43
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:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1123
+ warnings.warn("The use_host_meta 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
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1127
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1128
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1129
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1130
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1131
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1132
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1133
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1134
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1135
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1136
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1137
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1138
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1139
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1140
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1141
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1142
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1143
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1144
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1145
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1146
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1147
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1148
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1149
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1150
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1151
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1152
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1153
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1154
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1155
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1156
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1157
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1158
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1159
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1160
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1161
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1162
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1163
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1164
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1165
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1166
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1167
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1168
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1169
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1170
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1171
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1172
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1173
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1174
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1175
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1176
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1177
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1178
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1179
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1180
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1181
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1182
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1183
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1184
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1185
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1186
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1187
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1188
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1189
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1190
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1191
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1192
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1193
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1194
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1195
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1196
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1197
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1198
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1199
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1200
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1201
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1202
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1203
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1204
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1205
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1206
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1207
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1208
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1209
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1210
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1211
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1212
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1213
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1214
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1215
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1216
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1217
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1218
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1219
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1220
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1221
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1222
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1223
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1224
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1225
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1226
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1227
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1228
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1229
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1230
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1231
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1232
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1233
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1234
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1235
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1236
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1237
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1238
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1239
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1240
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1241
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1242
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1243
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1244
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1245
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1246
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1247
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1248
+ [2026-08-30 09:11:13] iteration 1/ 10 | consumed samples: 16 | elapsed time per iteration (ms): 30142.3 | throughput per GPU (TFLOP/s/GPU): 11.1 | learning rate: 2.000000E-05 | global batch size: 16 | lm loss: 1.102352E+01 | loss scale: 1.0 | grad norm: 44.191 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 67072 | min safe capacity: 1.0234 |
1249
+ Number of parameters in transformer block in billions: 23.22
1250
+ Number of parameters in embedding layers in billions: 0.41
1251
+ Total number of parameters in billions: 23.63
1252
+ Number of parameters in most loaded shard in billions: 23.6320
1253
+ Theoretical memory footprints: weight and optimizer=169029.43 MB
1254
+ [Rank 0] (after 1 iterations) memory (MB) | allocated: 58124.50390625 | max allocated: 76960.7197265625 | reserved: 79598.0 | max reserved: 79598.0
1255
+ [2026-08-30 09:11:14] iteration 2/ 10 | consumed samples: 32 | elapsed time per iteration (ms): 1075.6 | throughput per GPU (TFLOP/s/GPU): 309.9 | learning rate: 4.000000E-05 | global batch size: 16 | lm loss: 1.101566E+01 | loss scale: 1.0 | grad norm: 45.942 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 67328 | min safe capacity: 1.0273 |
1256
+ [2026-08-30 09:11:15] iteration 3/ 10 | consumed samples: 48 | elapsed time per iteration (ms): 518.0 | throughput per GPU (TFLOP/s/GPU): 643.6 | learning rate: 6.000000E-05 | global batch size: 16 | lm loss: 1.051925E+01 | loss scale: 1.0 | grad norm: 215.079 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 67328 | min safe capacity: 1.0273 |
1257
+ Capture CUDA graph for training!!!
1258
+ CUDA graph capture done!!!
1259
+ [2026-08-30 09:11:18] iteration 4/ 10 | consumed samples: 64 | elapsed time per iteration (ms): 3135.9 | throughput per GPU (TFLOP/s/GPU): 106.3 | learning rate: 8.000000E-05 | global batch size: 16 | lm loss: 1.316051E+01 | loss scale: 1.0 | grad norm: 38.031 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 67328 | min safe capacity: 1.0273 |
1260
+ [2026-08-30 09:11:18] iteration 5/ 10 | consumed samples: 80 | elapsed time per iteration (ms): 449.6 | throughput per GPU (TFLOP/s/GPU): 741.6 | learning rate: 1.000000E-04 | global batch size: 16 | lm loss: 1.038140E+01 | loss scale: 1.0 | grad norm: 16.240 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 67328 | min safe capacity: 1.0273 |
1261
+ [2026-08-30 09:11:19] iteration 6/ 10 | consumed samples: 96 | elapsed time per iteration (ms): 468.9 | throughput per GPU (TFLOP/s/GPU): 711.1 | learning rate: 7.750000E-05 | global batch size: 16 | lm loss: 1.038037E+01 | loss scale: 1.0 | grad norm: 32.701 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 67328 | min safe capacity: 1.0273 |
1262
+ [2026-08-30 09:11:22] iteration 7/ 10 | consumed samples: 112 | elapsed time per iteration (ms): 3241.4 | throughput per GPU (TFLOP/s/GPU): 102.9 | learning rate: 3.250000E-05 | global batch size: 16 | lm loss: 1.004429E+01 | loss scale: 1.0 | grad norm: 7.314 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 67328 | min safe capacity: 1.0273 |
1263
+ [2026-08-30 09:11:23] iteration 8/ 10 | consumed samples: 128 | elapsed time per iteration (ms): 470.4 | throughput per GPU (TFLOP/s/GPU): 708.7 | learning rate: 1.000000E-05 | global batch size: 16 | lm loss: 9.520119E+00 | loss scale: 1.0 | grad norm: 5.761 | number of skipped iterations: 0 | number of nan iterations: 0 | peak recv tokens: 67328 | min safe capacity: 1.0273 |
1264
+ Capture range started in the application.
1265
+ Capture range ended in the application.
1266
+ Generating '/tmp/nsys-report-ed6a.qdstrm'
1267
+
1268
+
tiera-rerun-fixed-moonep/ultraep/tierbtrace/tierbtrace_gbs16/trace_results.tsv CHANGED
@@ -1,2 +1,3 @@
1
  model backend capacity status report async_overlap_json
2
  G1 ultraep 2.40625 OK /workspace/tiera-rerun-fixed-moonep/ultraep/tierbtrace/tierbtrace_gbs16/G1__ultraep__matched__seq4096__gbs16__cap2p40625/nsys/G1__ultraep__matched__seq4096__gbs16__cap2p40625.nsys-rep
 
 
1
  model backend capacity status report async_overlap_json
2
  G1 ultraep 2.40625 OK /workspace/tiera-rerun-fixed-moonep/ultraep/tierbtrace/tierbtrace_gbs16/G1__ultraep__matched__seq4096__gbs16__cap2p40625/nsys/G1__ultraep__matched__seq4096__gbs16__cap2p40625.nsys-rep
3
+ G2 ultraep 1.08203125 OK /workspace/tiera-rerun-fixed-moonep/ultraep/tierbtrace/tierbtrace_gbs16/G2__ultraep__matched__seq4096__gbs16__cap1p08203125/nsys/G2__ultraep__matched__seq4096__gbs16__cap1p08203125.nsys-rep