AbhishekGhosh1998 commited on
Commit
999d988
·
verified ·
1 Parent(s): d03d6d5

granularity results tierb-fixed-moonep-full/ultraep/tierbtrace @ 2026-08-30T00:58:05Z from sp-aiinfra-interactive-aghosh-7jtjl-8xb200-moe-cg-0

Browse files
.gitattributes CHANGED
@@ -61,3 +61,4 @@ tierb-fixed-moonep-full/echo/tierbtrace/tierbtrace_gbs8/DSV2_LITE_L24__echo__mat
61
  tierb-fixed-moonep-full/echo/tierbtrace/tierbtrace_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/nsys/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375.nsys-rep filter=lfs diff=lfs merge=lfs -text
62
  tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__ultraep__matched__seq8192__gbs8__cap1p083984375/nsys/QWEN3_235B__ultraep__matched__seq8192__gbs8__cap1p083984375.nsys-rep filter=lfs diff=lfs merge=lfs -text
63
  tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/QWEN3_30B__ultraep__matched__seq4096__gbs8__cap1p11328125/nsys/QWEN3_30B__ultraep__matched__seq4096__gbs8__cap1p11328125.nsys-rep filter=lfs diff=lfs merge=lfs -text
 
 
61
  tierb-fixed-moonep-full/echo/tierbtrace/tierbtrace_gbs8/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375/nsys/QWEN2_57B__echo__matched__seq4096__gbs8__cap1p08984375.nsys-rep filter=lfs diff=lfs merge=lfs -text
62
  tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__ultraep__matched__seq8192__gbs8__cap1p083984375/nsys/QWEN3_235B__ultraep__matched__seq8192__gbs8__cap1p083984375.nsys-rep filter=lfs diff=lfs merge=lfs -text
63
  tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/QWEN3_30B__ultraep__matched__seq4096__gbs8__cap1p11328125/nsys/QWEN3_30B__ultraep__matched__seq4096__gbs8__cap1p11328125.nsys-rep filter=lfs diff=lfs merge=lfs -text
64
+ tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/DSV2_LITE_L24__ultraep__matched__seq8192__gbs8__cap1p0755208333333333/nsys/DSV2_LITE_L24__ultraep__matched__seq8192__gbs8__cap1p0755208333333333.nsys-rep filter=lfs diff=lfs merge=lfs -text
tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/DSV2_LITE_L24__ultraep__matched__seq8192__gbs8__cap1p0755208333333333/nsys/DSV2_LITE_L24__ultraep__matched__seq8192__gbs8__cap1p0755208333333333.nsys-rep ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e9ecafdd2599a425aa07d6af3a40f423aa730b495b0e2398b0370e076da8a0b
3
+ size 14039442
tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/DSV2_LITE_L24__ultraep__matched__seq8192__gbs8__cap1p0755208333333333/nsys/run.env ADDED
@@ -0,0 +1,70 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ backend=ultraep
2
+ config=DSV2_LITE_L24
3
+ family=tierB
4
+ repo=/workspace/Megatron-UltraEP
5
+ branch=ultraep-bracing-regions-through-kernels
6
+ commit=03d57ef7
7
+ num_layers=24
8
+ hidden=2048
9
+ ffn_hidden=10944
10
+ moe_ffn_hidden=1408
11
+ num_experts=64
12
+ topk=6
13
+ experts_per_rank=8
14
+ expert_params=8650752
15
+ heads=16
16
+ kv_groups=0
17
+ kv_channels=128
18
+ qk_layernorm=1
19
+ shared_ffn=2816
20
+ shared_expert_overlap=0
21
+ moe_token_dispatcher=flex
22
+ mla=1
23
+ dense_first_layers=1
24
+ moe_layers=23
25
+ rotary_base=10000
26
+ rotary_scaling=40
27
+ qkv_bias=0
28
+ router_balancing=seq_aux_loss
29
+ aux_loss_coeff=1e-3
30
+ router_pre_softmax=1
31
+ router_fusion=0
32
+ router_dtype=fp32
33
+ device_initiated_grouped_gemm=1
34
+ norm_epsilon=1e-6
35
+ seq_length=8192
36
+ micro_batch_size=1
37
+ global_batch_size=8
38
+ num_microbatches=1
39
+ assignments_per_rank=49152
40
+ mean_tokens_per_expert=6144
41
+ slot_policy=matched
42
+ slots_per_rank=8
43
+ capacity_mode=static
44
+ capacity=1.0755208333333333
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=768
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
tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/DSV2_LITE_L24__ultraep__matched__seq8192__gbs8__cap1p0755208333333333/nsys/train.log ADDED
The diff for this file is too large to render. See raw diff
 
tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/DSV2_LITE_L24__ultraep__matched__seq8192__gbs8__cap1p0755208333333333/nsys/verify_trace_ranks.log ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ report : /workspace/tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/DSV2_LITE_L24__ultraep__matched__seq8192__gbs8__cap1p0755208333333333/nsys/DSV2_LITE_L24__ultraep__matched__seq8192__gbs8__cap1p0755208333333333.nsys-rep
2
+ kernels : 167820
3
+ devices : 8 of 8 -> [0, 1, 2, 3, 4, 5, 6, 7]
4
+ processes: 1
5
+ markers : 20608 ultraep_trace__ kernels on devices [0, 1, 2, 3, 4, 5, 6, 7]
6
+ PASS: all ranks present and backend markers emitted
tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/QWEN2_57B__ultraep__matched__seq4096__gbs8__cap1p08984375/nsys/run.env ADDED
@@ -0,0 +1,70 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ backend=ultraep
2
+ config=QWEN2_57B
3
+ family=tierB
4
+ repo=/workspace/Megatron-UltraEP
5
+ branch=ultraep-bracing-regions-through-kernels
6
+ commit=03d57ef7
7
+ num_layers=10
8
+ hidden=3584
9
+ ffn_hidden=2560
10
+ moe_ffn_hidden=2560
11
+ num_experts=64
12
+ topk=8
13
+ experts_per_rank=8
14
+ expert_params=27525120
15
+ heads=28
16
+ kv_groups=4
17
+ kv_channels=
18
+ qk_layernorm=0
19
+ shared_ffn=20480
20
+ shared_expert_overlap=0
21
+ moe_token_dispatcher=flex
22
+ mla=0
23
+ dense_first_layers=0
24
+ moe_layers=10
25
+ rotary_base=1000000
26
+ rotary_scaling=
27
+ qkv_bias=1
28
+ router_balancing=aux_loss
29
+ aux_loss_coeff=1e-2
30
+ router_pre_softmax=1
31
+ router_fusion=0
32
+ router_dtype=fp32
33
+ device_initiated_grouped_gemm=1
34
+ norm_epsilon=1e-6
35
+ seq_length=4096
36
+ micro_batch_size=1
37
+ global_batch_size=8
38
+ num_microbatches=1
39
+ assignments_per_rank=32768
40
+ mean_tokens_per_expert=4096
41
+ slot_policy=matched
42
+ slots_per_rank=8
43
+ capacity_mode=static
44
+ capacity=1.08984375
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=512
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
tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/QWEN2_57B__ultraep__matched__seq4096__gbs8__cap1p08984375/nsys/train.log ADDED
@@ -0,0 +1,1119 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ==============================================================
2
+ QWEN2_57B / ultraep (slots/rank=8 policy=matched)
3
+ E=64 K=8 H'=2560 m=8
4
+ L=10 seq=4096 gbs=8 (1 microbatches)
5
+ cap_mode=static cap=1.08984375 cg=1 recompute=none
6
+ memory_dump=none snapshots=/workspace/tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/QWEN2_57B__ultraep__matched__seq4096__gbs8__cap1p08984375/nsys/memory_snapshots/snapshot.pickle
7
+ repo=/workspace/Megatron-UltraEP @ 03d57ef7
8
+ ==============================================================
9
+ W0830 00:57:15.319000 354957 torch/distributed/run.py:766]
10
+ W0830 00:57:15.319000 354957 torch/distributed/run.py:766] *****************************************
11
+ W0830 00:57:15.319000 354957 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 00:57:15.319000 354957 torch/distributed/run.py:766] *****************************************
13
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
14
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
15
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
16
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
17
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
18
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
19
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
20
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
21
+ using world size: 8, data-parallel size: 8, context-parallel size: 1, hierarchical context-parallel sizes: None, tensor-model-parallel size: 1, pipeline-model-parallel size: 1
22
+ WARNING: overriding default arguments for tokenizer_type:GPT2BPETokenizer with tokenizer_type:GPT2BPETokenizer
23
+ Number of virtual stages per pipeline stage: None
24
+ accumulate and all-reduce gradients in fp32 for bfloat16 data type.
25
+ using torch.bfloat16 for parameters ...
26
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
27
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
28
+ ------------------------ arguments ------------------------
29
+ account_for_embedding_in_pipeline_split ......... False
30
+ account_for_loss_in_pipeline_split .............. False
31
+ accumulate_allreduce_grads_in_fp32 .............. True
32
+ adam_beta1 ...................................... 0.9
33
+ adam_beta2 ...................................... 0.999
34
+ adam_eps ........................................ 1e-08
35
+ add_bias_linear ................................. False
36
+ add_position_embedding .......................... False
37
+ add_qkv_bias .................................... True
38
+ adlr_autoresume ................................. False
39
+ adlr_autoresume_interval ........................ 1000
40
+ align_grad_reduce ............................... True
41
+ align_param_gather .............................. False
42
+ app_tag_run_name ................................ None
43
+ app_tag_run_version ............................. 0.0.0
44
+ apply_layernorm_1p .............................. False
45
+ apply_query_key_layer_scaling ................... False
46
+ apply_residual_connection_post_layernorm ........ False
47
+ apply_rope_fusion ............................... True
48
+ async_save ...................................... None
49
+ async_tensor_model_parallel_allreduce ........... True
50
+ attention_backend ............................... AttnBackend.auto
51
+ attention_dropout ............................... 0.0
52
+ attention_softmax_in_fp32 ....................... False
53
+ auto_detect_ckpt_format ......................... False
54
+ barrier_with_L1_time ............................ True
55
+ bert_binary_head ................................ True
56
+ bert_embedder_type .............................. megatron
57
+ bert_load ....................................... None
58
+ bf16 ............................................ True
59
+ bias_dropout_fusion ............................. True
60
+ bias_gelu_fusion ................................ False
61
+ bias_swiglu_fusion .............................. True
62
+ biencoder_projection_dim ........................ 0
63
+ biencoder_shared_query_context_model ............ False
64
+ block_data_path ................................. None
65
+ calc_ft_timeouts ................................ False
66
+ calculate_per_token_loss ........................ False
67
+ check_for_large_grads ........................... False
68
+ check_for_nan_in_loss_and_grad .................. False
69
+ check_for_spiky_loss ............................ False
70
+ check_weight_hash_across_dp_replicas_interval ... None
71
+ ckpt_assume_constant_structure .................. False
72
+ ckpt_convert_format ............................. None
73
+ ckpt_convert_save ............................... None
74
+ ckpt_convert_update_legacy_dist_opt_format ...... False
75
+ ckpt_format ..................................... torch_dist
76
+ ckpt_fully_parallel_load ........................ False
77
+ ckpt_fully_parallel_save ........................ True
78
+ ckpt_fully_parallel_save_deprecated ............. False
79
+ ckpt_step ....................................... None
80
+ classes_fraction ................................ 1.0
81
+ clip_grad ....................................... 1.0
82
+ clone_scatter_output_in_embedding ............... True
83
+ config_logger_dir ...............................
84
+ consumed_train_samples .......................... 0
85
+ consumed_valid_samples .......................... 0
86
+ context_parallel_size ........................... 1
87
+ cp_comm_type .................................... ['p2p']
88
+ create_attention_mask_in_dataloader ............. True
89
+ cross_entropy_fusion_impl ....................... native
90
+ cross_entropy_loss_fusion ....................... False
91
+ cuda_graph_scope ................................ full_iteration
92
+ cuda_graph_warmup_steps ......................... 3
93
+ data_args_path .................................. None
94
+ data_cache_path ................................. None
95
+ data_parallel_random_init ....................... False
96
+ data_parallel_sharding_strategy ................. no_shard
97
+ data_parallel_size .............................. 8
98
+ data_path ....................................... ['/workspace/data/wikitext/wikitext103_gpt2_text_document']
99
+ data_per_class_fraction ......................... 1.0
100
+ data_sharding ................................... True
101
+ dataloader_type ................................. single
102
+ ddp_average_in_collective ....................... False
103
+ ddp_bucket_size ................................. None
104
+ ddp_num_buckets ................................. None
105
+ ddp_pad_buckets_for_high_nccl_busbw ............. False
106
+ decoder_first_pipeline_num_layers ............... None
107
+ decoder_last_pipeline_num_layers ................ None
108
+ decoder_num_layers .............................. None
109
+ decoder_seq_length .............................. None
110
+ decoupled_lr .................................... None
111
+ decoupled_min_lr ................................ None
112
+ decrease_batch_size_if_needed ................... False
113
+ defer_embedding_wgrad_compute ................... False
114
+ delay_wgrad_compute ............................. False
115
+ deprecated_use_mcore_models ..................... True
116
+ deterministic_mode .............................. False
117
+ dino_bottleneck_size ............................ 256
118
+ dino_freeze_last_layer .......................... 1
119
+ dino_head_hidden_size ........................... 2048
120
+ dino_local_crops_number ......................... 10
121
+ dino_local_img_size ............................. 96
122
+ dino_norm_last_layer ............................ False
123
+ dino_teacher_temp ............................... 0.07
124
+ dino_warmup_teacher_temp ........................ 0.04
125
+ dino_warmup_teacher_temp_epochs ................. 30
126
+ disable_bf16_reduced_precision_matmul ........... False
127
+ disable_mamba_mem_eff_path ...................... False
128
+ disable_straggler_on_startup .................... False
129
+ dist_ckpt_format_deprecated ..................... None
130
+ dist_ckpt_strictness ............................ assume_ok_unexpected
131
+ distribute_saved_activations .................... False
132
+ distributed_backend ............................. nccl
133
+ distributed_timeout_minutes ..................... 10
134
+ embedding_init_method_std ....................... None
135
+ embedding_path .................................. None
136
+ empty_unused_memory_level ....................... 0
137
+ enable_cuda_graph ............................... True
138
+ enable_experimental ............................. False
139
+ enable_ft_package ............................... False
140
+ enable_gloo_process_groups ...................... True
141
+ enable_msc ...................................... True
142
+ enable_one_logger ............................... True
143
+ encoder_num_layers .............................. 10
144
+ encoder_seq_length .............................. 4096
145
+ end_weight_decay ................................ 0.1
146
+ eod_mask_loss ................................... False
147
+ error_injection_rate ............................ 0
148
+ error_injection_type ............................ transient_error
149
+ eval_interval ................................... 1000000
150
+ eval_iters ...................................... 0
151
+ evidence_data_path .............................. None
152
+ exit_duration_in_mins ........................... None
153
+ exit_interval ................................... None
154
+ exit_on_missing_checkpoint ...................... False
155
+ exit_signal_handler ............................. False
156
+ exp_avg_dtype ................................... torch.float32
157
+ exp_avg_sq_dtype ................................ torch.float32
158
+ expert_model_parallel_size ...................... 8
159
+ expert_tensor_parallel_size ..................... 1
160
+ export_force_local_attention .................... False
161
+ export_kd_cfg ................................... None
162
+ export_kd_teacher_ckpt_format ................... None
163
+ export_kd_teacher_load .......................... None
164
+ export_kv_cache_quant ........................... False
165
+ export_legacy_megatron .......................... False
166
+ export_model_type ............................... GPTModel
167
+ export_moe_apply_probs_on_input ................. False
168
+ export_qk_l2_norm ............................... False
169
+ export_quant_cfg ................................ None
170
+ export_real_quant_cfg ........................... None
171
+ export_te_mcore_model ........................... False
172
+ external_cuda_graph ............................. False
173
+ ffn_hidden_size ................................. 2560
174
+ finetune ........................................ False
175
+ finetune_data_split ............................. train
176
+ finetune_hf_dataset ............................. None
177
+ first_last_layers_bf16 .......................... False
178
+ flash_decode .................................... False
179
+ fp16 ............................................ False
180
+ fp16_lm_cross_entropy ........................... False
181
+ fp32_residual_connection ........................ False
182
+ fp8 ............................................. e4m3
183
+ fp8_amax_compute_algo ........................... most_recent
184
+ fp8_amax_history_len ............................ 1
185
+ fp8_interval .................................... 1
186
+ fp8_margin ...................................... 0
187
+ fp8_param_gather ................................ False
188
+ fp8_recipe ...................................... mxfp8
189
+ fp8_wgrad ....................................... True
190
+ fsdp_double_buffer .............................. False
191
+ global_batch_size ............................... 8
192
+ grad_reduce_in_bf16 ............................. False
193
+ gradient_accumulation_fusion .................... True
194
+ gradient_reduce_div_fusion ...................... True
195
+ group_query_attention ........................... True
196
+ head_lr_mult .................................... 1.0
197
+ heterogeneous_layers_config_encoded_json ........ None
198
+ heterogeneous_layers_config_path ................ None
199
+ hidden_dropout .................................. 0.0
200
+ hidden_size ..................................... 3584
201
+ hierarchical_context_parallel_sizes ............. None
202
+ high_priority_stream_groups ..................... []
203
+ hybrid_attention_ratio .......................... 0.0
204
+ hybrid_mlp_ratio ................................ 0.0
205
+ hybrid_override_pattern ......................... None
206
+ hysteresis ...................................... 2
207
+ ict_head_size ................................... None
208
+ ict_load ........................................ None
209
+ img_h ........................................... 224
210
+ img_w ........................................... 224
211
+ indexer_batch_size .............................. 128
212
+ indexer_log_interval ............................ 1000
213
+ inference_batch_times_seqlen_threshold .......... -1
214
+ inference_dynamic_batching ...................... False
215
+ inference_dynamic_batching_buffer_guaranteed_fraction 0.2
216
+ inference_dynamic_batching_buffer_overflow_factor None
217
+ inference_dynamic_batching_buffer_size_gb ....... 40.0
218
+ inference_dynamic_batching_chunk_size ........... 256
219
+ inference_dynamic_batching_max_requests_override None
220
+ inference_dynamic_batching_max_tokens_override .. None
221
+ inference_dynamic_batching_num_cuda_graphs ...... 16
222
+ inference_max_batch_size ........................ 8
223
+ inference_max_seq_length ........................ 2560
224
+ inference_rng_tracker ........................... False
225
+ init_method_std ................................. 0.01
226
+ init_method_xavier_uniform ...................... False
227
+ init_model_with_meta_device ..................... False
228
+ initial_loss_scale .............................. 4294967296
229
+ inprocess_active_world_size ..................... 8
230
+ inprocess_barrier_timeout ....................... 120
231
+ inprocess_completion_timeout .................... 120
232
+ inprocess_empty_cuda_cache ...................... False
233
+ inprocess_granularity ........................... node
234
+ inprocess_hard_timeout .......................... 90
235
+ inprocess_heartbeat_interval .................... 30
236
+ inprocess_heartbeat_timeout ..................... 60
237
+ inprocess_last_call_wait ........................ 1
238
+ inprocess_max_iterations ........................ None
239
+ inprocess_monitor_process_interval .............. 1.0
240
+ inprocess_monitor_thread_interval ............... 1.0
241
+ inprocess_progress_watchdog_interval ............ 1.0
242
+ inprocess_restart ............................... False
243
+ inprocess_soft_timeout .......................... 60
244
+ inprocess_termination_grace_time ................ 1
245
+ is_hybrid_model ................................. False
246
+ iter_per_epoch .................................. 1250
247
+ iterations_to_skip .............................. []
248
+ keep_fp8_transpose_cache_when_using_custom_fsdp . False
249
+ kitchen_config_file ............................. None
250
+ kitchen_recipe_number ........................... None
251
+ kv_channels ..................................... 128
252
+ kv_lora_rank .................................... 32
253
+ lazy_mpu_init ................................... None
254
+ load ............................................ None
255
+ load_main_params_from_ckpt ...................... None
256
+ load_model_opt_format ........................... False
257
+ local_rank ...................................... 0
258
+ log_energy ...................................... False
259
+ log_interval .................................... 1
260
+ log_loss_scale_to_tensorboard ................... True
261
+ log_memory_to_tensorboard ....................... False
262
+ log_num_zeros_in_grad ........................... False
263
+ log_params_norm ................................. False
264
+ log_progress .................................... False
265
+ log_straggler ................................... False
266
+ log_throughput .................................. True
267
+ log_timers_to_tensorboard ....................... False
268
+ log_validation_ppl_to_tensorboard ............... False
269
+ log_world_size_to_tensorboard ................... False
270
+ logging_level ................................... None
271
+ loss_scale ...................................... None
272
+ loss_scale_window ............................... 1000
273
+ lr .............................................. 0.0001
274
+ lr_decay_iters .................................. 8
275
+ lr_decay_samples ................................ None
276
+ lr_decay_style .................................. cosine
277
+ lr_warmup_fraction .............................. None
278
+ lr_warmup_init .................................. 0.0
279
+ lr_warmup_iters ................................. 5
280
+ lr_warmup_samples ............................... 0
281
+ lr_wsd_decay_iters .............................. None
282
+ lr_wsd_decay_samples ............................ None
283
+ lr_wsd_decay_style .............................. exponential
284
+ main_grads_dtype ................................ torch.float32
285
+ main_params_dtype ............................... torch.float32
286
+ make_vocab_size_divisible_by .................... 128
287
+ mamba_head_dim .................................. 64
288
+ mamba_num_groups ................................ 8
289
+ mamba_num_heads ................................. None
290
+ mamba_state_dim ................................. 128
291
+ manual_gc ....................................... False
292
+ manual_gc_eval .................................. True
293
+ manual_gc_interval .............................. 0
294
+ mask_factor ..................................... 1.0
295
+ mask_prob ....................................... 0.15
296
+ mask_type ....................................... random
297
+ masked_softmax_fusion ........................... False
298
+ max_position_embeddings ......................... 32768
299
+ max_tokens_to_oom ............................... 12000
300
+ memory_snapshot_path ............................ snapshot.pickle
301
+ merge_file ...................................... /workspace/data/wikitext/gpt2-merges.txt
302
+ micro_batch_size ................................ 1
303
+ microbatch_group_size_per_vp_stage .............. None
304
+ mid_level_dataset_surplus ....................... 0.005
305
+ min_loss_scale .................................. 1.0
306
+ min_lr .......................................... 1e-05
307
+ mlp_chunks_for_prefill .......................... 1
308
+ mmap_bin_files .................................. True
309
+ mock_data ....................................... False
310
+ moe_apply_probs_on_input ........................ False
311
+ moe_aux_loss_coeff .............................. 0.01
312
+ moe_capacity_monitor ............................ True
313
+ moe_deepep_num_sms .............................. 20
314
+ moe_echo_enable_random_offloading ............... False
315
+ moe_echo_expert_dispatch_overlap ................ False
316
+ moe_echo_expert_dispatcher_type ................. hybridep
317
+ moe_echo_recompute_expert_dispatch .............. False
318
+ moe_enable_deepep ............................... False
319
+ moe_enable_echo ................................. False
320
+ moe_enable_hybridep ............................. True
321
+ moe_enable_ultraep .............................. True
322
+ moe_expert_capacity_factor ...................... None
323
+ moe_extended_tp ................................. False
324
+ moe_ffn_hidden_size ............................. 2560
325
+ moe_grouped_gemm ................................ True
326
+ moe_hybridep_num_sms ............................ 16
327
+ moe_input_jitter_eps ............................ None
328
+ moe_layer_freq .................................. 1
329
+ moe_layer_recompute ............................. False
330
+ moe_num_echo_experts ............................ None
331
+ moe_num_redundant_experts_per_rank .............. 8
332
+ moe_pad_expert_input_to_capacity ................ False
333
+ moe_per_layer_logging ........................... False
334
+ moe_permute_fusion .............................. False
335
+ moe_received_token_capacity ..................... 1.08984375
336
+ moe_router_bias_update_rate ..................... 0.001
337
+ moe_router_dtype ................................ fp32
338
+ moe_router_enable_expert_bias ................... False
339
+ moe_router_force_load_balancing ................. False
340
+ moe_router_group_topk ........................... None
341
+ moe_router_load_balancing_type .................. aux_loss
342
+ moe_router_num_groups ........................... None
343
+ moe_router_padding_for_fp8 ...................... False
344
+ moe_router_pre_softmax .......................... True
345
+ moe_router_score_function ....................... softmax
346
+ moe_router_topk ................................. 8
347
+ moe_router_topk_scaling_factor .................. None
348
+ moe_shared_expert_intermediate_size ............. 20480
349
+ moe_shared_expert_overlap ....................... False
350
+ moe_token_dispatcher_type ....................... flex
351
+ moe_token_drop_policy ........................... probs
352
+ moe_ultraep_max_microbatches .................... None
353
+ moe_ultraep_overlap_weight_sync ................. True
354
+ moe_ultraep_weight_sync_join .................... before_dispatch
355
+ moe_upcycling_granularity ....................... 1
356
+ moe_use_device_initiated_grouped_gemm ........... True
357
+ moe_use_legacy_grouped_gemm ..................... False
358
+ moe_use_upcycling ............................... False
359
+ moe_z_loss_coeff ................................ None
360
+ mrope_section ................................... None
361
+ mscale .......................................... 1.0
362
+ mscale_all_dim .................................. 0.0
363
+ mtp_loss_scaling_factor ......................... 0.1
364
+ mtp_num_layers .................................. None
365
+ multi_latent_attention .......................... False
366
+ nccl_all_reduce_for_prefill ..................... False
367
+ nccl_communicator_config_path ................... None
368
+ nccl_ub ......................................... False
369
+ no_load_optim ................................... None
370
+ no_load_rng ..................................... None
371
+ no_persist_layer_norm ........................... False
372
+ no_rope_freq .................................... None
373
+ no_save_optim ................................... None
374
+ no_save_rng ..................................... None
375
+ non_persistent_ckpt_type ........................ None
376
+ non_persistent_global_ckpt_dir .................. None
377
+ non_persistent_local_ckpt_algo .................. fully_parallel
378
+ non_persistent_local_ckpt_dir ................... None
379
+ non_persistent_save_interval .................... None
380
+ norm_epsilon .................................... 1e-06
381
+ normalization ................................... RMSNorm
382
+ num_attention_heads ............................. 28
383
+ num_channels .................................... 3
384
+ num_classes ..................................... 1000
385
+ num_dataset_builder_threads ..................... 1
386
+ num_distributed_optimizer_instances ............. 1
387
+ num_experts ..................................... 64
388
+ num_layers ...................................... 10
389
+ num_layers_at_end_in_bf16 ....................... 1
390
+ num_layers_at_start_in_bf16 ..................... 1
391
+ num_layers_per_virtual_pipeline_stage ........... None
392
+ num_query_groups ................................ 4
393
+ num_virtual_stages_per_pipeline_rank ............ None
394
+ num_workers ..................................... 2
395
+ object_storage_cache_path ....................... None
396
+ one_logger_async ................................ False
397
+ one_logger_project .............................. megatron-lm
398
+ one_logger_run_name ............................. None
399
+ onnx_safe ....................................... None
400
+ openai_gelu ..................................... False
401
+ optimizer ....................................... adam
402
+ optimizer_cpu_offload ........................... False
403
+ optimizer_offload_fraction ...................... 1.0
404
+ output_bert_embeddings .......................... False
405
+ overlap_cpu_optimizer_d2h_h2d ................... False
406
+ overlap_grad_reduce ............................. True
407
+ overlap_moe_expert_parallel_comm ................ False
408
+ overlap_p2p_comm ................................ False
409
+ overlap_p2p_comm_warmup_flush ................... False
410
+ overlap_param_gather ............................ True
411
+ overlap_param_gather_with_optimizer_step ........ False
412
+ override_opt_param_scheduler .................... False
413
+ params_dtype .................................... torch.bfloat16
414
+ patch_dim ....................................... 16
415
+ per_split_data_args_path ........................ None
416
+ perform_initialization .......................... True
417
+ pin_cpu_grads ................................... True
418
+ pin_cpu_params .................................. True
419
+ pipeline_model_parallel_comm_backend ............ None
420
+ pipeline_model_parallel_layout .................. None
421
+ pipeline_model_parallel_size .................... 1
422
+ position_embedding_type ......................... rope
423
+ pretrained_checkpoint ........................... None
424
+ profile ......................................... True
425
+ profile_ranks ................................... [0, 1, 2, 3, 4, 5, 6, 7]
426
+ profile_step_end ................................ 8
427
+ profile_step_start .............................. 6
428
+ q_lora_rank ..................................... None
429
+ qk_head_dim ..................................... 128
430
+ qk_l2_norm ...................................... False
431
+ qk_layernorm .................................... False
432
+ qk_pos_emb_head_dim ............................. 64
433
+ query_in_block_prob ............................. 0.1
434
+ rampup_batch_size ............................... None
435
+ rank ............................................ 0
436
+ recompute_granularity ........................... None
437
+ recompute_method ................................ None
438
+ recompute_modules ............................... None
439
+ recompute_num_layers ............................ None
440
+ record_memory_history ........................... False
441
+ relative_attention_max_distance ................. 128
442
+ relative_attention_num_buckets .................. 32
443
+ replication ..................................... False
444
+ replication_factor .............................. 2
445
+ replication_jump ................................ None
446
+ rerun_mode ...................................... validate_results
447
+ reset_attention_mask ............................ False
448
+ reset_position_ids .............................. False
449
+ result_rejected_tracker_filename ................ None
450
+ retriever_report_topk_accuracies ................ []
451
+ retriever_score_scaling ......................... False
452
+ retriever_seq_length ............................ 256
453
+ retro_add_retriever ............................. False
454
+ retro_attention_gate ............................ 1
455
+ retro_cyclic_train_iters ........................ None
456
+ retro_encoder_attention_dropout ................. 0.1
457
+ retro_encoder_hidden_dropout .................... 0.1
458
+ retro_encoder_layers ............................ 2
459
+ retro_num_neighbors ............................. 2
460
+ retro_num_retrieved_chunks ...................... 2
461
+ retro_project_dir ............................... None
462
+ retro_verify_neighbor_count ..................... True
463
+ reuse_grad_buf_for_mxfp8_param_ag ............... False
464
+ rope_scaling_factor ............................. 8.0
465
+ rope_type ....................................... None
466
+ rotary_base ..................................... 1000000
467
+ rotary_interleaved .............................. False
468
+ rotary_percent .................................. 1.0
469
+ rotary_scaling_factor ........................... 1.0
470
+ rotary_seq_len_interpolation_factor ............. None
471
+ run_workload_inspector_server ................... False
472
+ sample_rate ..................................... 1.0
473
+ save ............................................ None
474
+ save_interval ................................... None
475
+ save_retain_interval ............................ None
476
+ scatter_gather_tensors_in_pipeline .............. True
477
+ seed ............................................ 1234
478
+ seq_length ...................................... 4096
479
+ sequence_parallel ............................... False
480
+ sft ............................................. False
481
+ sft_tokenizer_prompt_format ..................... nemotron-h-aligned
482
+ sgd_momentum .................................... 0.9
483
+ short_seq_prob .................................. 0.1
484
+ skip_train ...................................... False
485
+ skipped_train_samples ........................... 0
486
+ spec ............................................ None
487
+ split ........................................... 100,0,0
488
+ squared_relu .................................... False
489
+ start_weight_decay .............................. 0.1
490
+ straggler_ctrlr_port ............................ 65535
491
+ straggler_minmax_count .......................... 1
492
+ suggested_communication_unit_size ............... None
493
+ swiglu .......................................... True
494
+ swin_backbone_type .............................. tiny
495
+ symmetric_ar_type ............................... None
496
+ te_rng_tracker .................................. True
497
+ tensor_model_parallel_size ...................... 1
498
+ tensorboard_dir ................................. None
499
+ tensorboard_log_interval ........................ 1
500
+ tensorboard_queue_size .......................... 1000
501
+ test_data_path .................................. None
502
+ test_mode ....................................... False
503
+ tiktoken_num_special_tokens ..................... 1000
504
+ tiktoken_pattern ................................ None
505
+ tiktoken_special_tokens ......................... None
506
+ timing_log_level ................................ 0
507
+ timing_log_option ............................... minmax
508
+ titles_data_path ................................ None
509
+ tokenizer_model ................................. None
510
+ tokenizer_type .................................. GPT2BPETokenizer
511
+ torch_fsdp2_reshard_after_forward ............... True
512
+ tp_comm_bootstrap_backend ....................... nccl
513
+ tp_comm_bulk_dgrad .............................. True
514
+ tp_comm_bulk_wgrad .............................. True
515
+ tp_comm_overlap ................................. False
516
+ tp_comm_overlap_ag .............................. True
517
+ tp_comm_overlap_cfg ............................. None
518
+ tp_comm_overlap_rs .............................. True
519
+ tp_comm_overlap_rs_dgrad ........................ False
520
+ tp_comm_split_ag ................................ True
521
+ tp_comm_split_rs ................................ True
522
+ train_data_path ................................. None
523
+ train_iters ..................................... 10
524
+ train_samples ................................... None
525
+ train_sync_interval ............................. None
526
+ transformer_impl ................................ transformer_engine
527
+ transformer_pipeline_model_parallel_size ........ 1
528
+ untie_embeddings_and_output_weights ............. True
529
+ use_checkpoint_args ............................. False
530
+ use_checkpoint_opt_param_scheduler .............. False
531
+ use_cpu_initialization .......................... None
532
+ use_custom_fsdp ................................. False
533
+ use_dist_ckpt ................................... True
534
+ use_dist_ckpt_deprecated ........................ False
535
+ use_distributed_optimizer ....................... True
536
+ use_flash_attn .................................. False
537
+ use_legacy_models ............................... False
538
+ use_mp_args_from_checkpoint_args ................ False
539
+ use_one_sent_docs ............................... False
540
+ use_persistent_ckpt_worker ...................... False
541
+ use_precision_aware_optimizer ................... False
542
+ use_pytorch_profiler ............................ False
543
+ use_ring_exchange_p2p ........................... False
544
+ use_rope_scaling ................................ False
545
+ use_rotary_position_embeddings .................. False
546
+ use_sharp ....................................... False
547
+ use_tokenizer_model_from_checkpoint_args ........ True
548
+ use_torch_fsdp2 ................................. False
549
+ use_torch_optimizer_for_cpu_offload ............. False
550
+ use_tp_pp_dp_mapping ............................ False
551
+ v_head_dim ...................................... 128
552
+ valid_data_path ................................. None
553
+ variable_seq_lengths ............................ False
554
+ virtual_pipeline_model_parallel_size ............ None
555
+ vision_backbone_type ............................ vit
556
+ vision_pretraining .............................. False
557
+ vision_pretraining_type ......................... classify
558
+ vocab_extra_ids ................................. 0
559
+ vocab_file ...................................... /workspace/data/wikitext/gpt2-vocab.json
560
+ vocab_size ...................................... None
561
+ wandb_exp_name ..................................
562
+ wandb_project ...................................
563
+ wandb_save_dir ..................................
564
+ weight_decay .................................... 0.1
565
+ weight_decay_incr_style ......................... constant
566
+ wgrad_deferral_limit ............................ 0
567
+ world_size ...................................... 8
568
+ yaml_cfg ........................................ None
569
+ -------------------- end of arguments ---------------------
570
+ INFO:megatron.core.num_microbatches_calculator:setting number of microbatches to constant 1
571
+ > building GPT2BPETokenizer tokenizer ...
572
+ /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
+ 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
575
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
576
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
577
+ > padded vocab (size: 50257) with 47 dummy tokens (new size: 50304)
578
+ WARNING:megatron.core.rerun_state_machine:RerunStateMachine initialized in mode RerunMode.VALIDATE_RESULTS
579
+ > initializing torch distributed ...
580
+ /workspace/Megatron-UltraEP/megatron/training/arguments.py:831: UserWarning: Disabling sequence parallelism because tensor model parallelism is disabled
581
+ warnings.warn("Disabling sequence parallelism because tensor model parallelism is disabled")
582
+ [Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
583
+ [Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
584
+ [Gloo] Rank 1 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 [Gloo] Rank 76 is connected to is connected to 77 peer ranks. peer ranks. Expected number of connected peer ranks is : Expected number of connected peer ranks is : 77
589
+
590
+ [Gloo] Rank 7 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
591
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
592
+ [Gloo] Rank 0 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
593
+ [Gloo] Rank 1 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
594
+ [Gloo] Rank 3 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
595
+ [Gloo] Rank 4 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
596
+ [Gloo] Rank 5 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
597
+ [Gloo] Rank 6 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
598
+ [Gloo] Rank 2 is connected to 7 peer ranks. Expected number of connected peer ranks is : 7
599
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
600
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
601
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
602
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
603
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
604
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
605
+ > initialized tensor model parallel with size 1
606
+ > initialized pipeline model parallel with size 1
607
+ > setting random seeds to 1234 ...
608
+ [Gloo] Rank 0 is connected to 0 peer ranks. Expected number of connected peer ranks is : 0
609
+ [rank7]:[W830 00:57:28.143210010 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 7] using GPU 7 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
610
+ [rank4]:[W830 00:57:28.145125649 ProcessGroupNCCL.cpp:4789] [PG ID 0 PG GUID 0 Rank 4] using GPU 4 as device used by this process is currently unknown. This can potentially cause a hang if this rank to GPU mapping is incorrect. You can specify device_id in init_process_group() to force use of a particular device.
611
+ [rank3]:[W830 00:57:28.149510865 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.
612
+ [rank1]:[W830 00:57:28.150618952 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.
613
+ [rank6]:[W830 00:57:28.150643894 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.
614
+ [rank5]:[W830 00:57:28.150748761 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.
615
+ [rank2]:[W830 00:57:28.150775610 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.
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: 0.427 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 00:57:29.752319455 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: 3.633 seconds
625
+ time to initialize megatron (seconds): 9.090
626
+ [after megatron is initialized] datetime: 2026-08-30 00:57:35
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): 7262624256
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: 22
667
+ Params for bucket 1 (180289536 elements, 180289536 padded size):
668
+ module.output_layer.weight
669
+ Params for bucket 2 (73403904 elements, 73403904 padded size):
670
+ module.decoder.final_layernorm.weight
671
+ module.decoder.layers.9.mlp.shared_experts.linear_fc2.weight
672
+ Params for bucket 3 (146800640 elements, 146800640 padded size):
673
+ module.decoder.layers.9.mlp.shared_experts.linear_fc1.weight
674
+ Params for bucket 4 (103001600 elements, 103001600 padded size):
675
+ module.decoder.layers.9.pre_mlp_layernorm.weight
676
+ module.decoder.layers.9.self_attention.linear_qkv.layer_norm_weight
677
+ module.decoder.layers.9.self_attention.linear_qkv.bias
678
+ module.decoder.layers.9.mlp.router.weight
679
+ module.decoder.layers.9.self_attention.linear_qkv.weight
680
+ module.decoder.layers.9.self_attention.linear_proj.weight
681
+ module.decoder.layers.8.mlp.shared_experts.linear_fc2.weight
682
+ Params for bucket 5 (146800640 elements, 146800640 padded size):
683
+ module.decoder.layers.8.mlp.shared_experts.linear_fc1.weight
684
+ Params for bucket 6 (103001600 elements, 103001600 padded size):
685
+ module.decoder.layers.8.pre_mlp_layernorm.weight
686
+ module.decoder.layers.8.self_attention.linear_qkv.layer_norm_weight
687
+ module.decoder.layers.8.self_attention.linear_qkv.bias
688
+ module.decoder.layers.8.mlp.router.weight
689
+ module.decoder.layers.8.self_attention.linear_qkv.weight
690
+ module.decoder.layers.8.self_attention.linear_proj.weight
691
+ module.decoder.layers.7.mlp.shared_experts.linear_fc2.weight
692
+ Params for bucket 7 (146800640 elements, 146800640 padded size):
693
+ module.decoder.layers.7.mlp.shared_experts.linear_fc1.weight
694
+ Params for bucket 8 (103001600 elements, 103001600 padded size):
695
+ module.decoder.layers.7.self_attention.linear_qkv.bias
696
+ module.decoder.layers.7.pre_mlp_layernorm.weight
697
+ module.decoder.layers.7.self_attention.linear_qkv.layer_norm_weight
698
+ module.decoder.layers.7.mlp.router.weight
699
+ module.decoder.layers.7.self_attention.linear_qkv.weight
700
+ module.decoder.layers.7.self_attention.linear_proj.weight
701
+ module.decoder.layers.6.mlp.shared_experts.linear_fc2.weight
702
+ Params for bucket 9 (146800640 elements, 146800640 padded size):
703
+ module.decoder.layers.6.mlp.shared_experts.linear_fc1.weight
704
+ Params for bucket 10 (103001600 elements, 103001600 padded size):
705
+ module.decoder.layers.6.pre_mlp_layernorm.weight
706
+ module.decoder.layers.6.self_attention.linear_qkv.layer_norm_weight
707
+ module.decoder.layers.6.self_attention.linear_qkv.bias
708
+ module.decoder.layers.6.mlp.router.weight
709
+ module.decoder.layers.6.self_attention.linear_qkv.weight
710
+ module.decoder.layers.6.self_attention.linear_proj.weight
711
+ module.decoder.layers.5.mlp.shared_experts.linear_fc2.weight
712
+ Params for bucket 11 (146800640 elements, 146800640 padded size):
713
+ module.decoder.layers.5.mlp.shared_experts.linear_fc1.weight
714
+ Params for bucket 12 (103001600 elements, 103001600 padded size):
715
+ module.decoder.layers.5.self_attention.linear_qkv.layer_norm_weight
716
+ module.decoder.layers.5.pre_mlp_layernorm.weight
717
+ module.decoder.layers.5.self_attention.linear_qkv.bias
718
+ module.decoder.layers.5.mlp.router.weight
719
+ module.decoder.layers.5.self_attention.linear_qkv.weight
720
+ module.decoder.layers.5.self_attention.linear_proj.weight
721
+ module.decoder.layers.4.mlp.shared_experts.linear_fc2.weight
722
+ Params for bucket 13 (146800640 elements, 146800640 padded size):
723
+ module.decoder.layers.4.mlp.shared_experts.linear_fc1.weight
724
+ Params for bucket 14 (103001600 elements, 103001600 padded size):
725
+ module.decoder.layers.4.self_attention.linear_qkv.layer_norm_weight
726
+ module.decoder.layers.4.self_attention.linear_qkv.bias
727
+ module.decoder.layers.4.pre_mlp_layernorm.weight
728
+ module.decoder.layers.4.mlp.router.weight
729
+ module.decoder.layers.4.self_attention.linear_qkv.weight
730
+ module.decoder.layers.4.self_attention.linear_proj.weight
731
+ module.decoder.layers.3.mlp.shared_experts.linear_fc2.weight
732
+ Params for bucket 15 (146800640 elements, 146800640 padded size):
733
+ module.decoder.layers.3.mlp.shared_experts.linear_fc1.weight
734
+ Params for bucket 16 (103001600 elements, 103001600 padded size):
735
+ module.decoder.layers.3.mlp.router.weight
736
+ module.decoder.layers.3.self_attention.linear_qkv.weight
737
+ module.decoder.layers.3.self_attention.linear_proj.weight
738
+ module.decoder.layers.2.mlp.shared_experts.linear_fc2.weight
739
+ module.decoder.layers.3.pre_mlp_layernorm.weight
740
+ module.decoder.layers.3.self_attention.linear_qkv.layer_norm_weight
741
+ module.decoder.layers.3.self_attention.linear_qkv.bias
742
+ Params for bucket 17 (146800640 elements, 146800640 padded size):
743
+ module.decoder.layers.2.mlp.shared_experts.linear_fc1.weight
744
+ Params for bucket 18 (103001600 elements, 103001600 padded size):
745
+ module.decoder.layers.2.mlp.router.weight
746
+ module.decoder.layers.2.self_attention.linear_qkv.weight
747
+ module.decoder.layers.2.self_attention.linear_proj.weight
748
+ module.decoder.layers.1.mlp.shared_experts.linear_fc2.weight
749
+ module.decoder.layers.2.pre_mlp_layernorm.weight
750
+ module.decoder.layers.2.self_attention.linear_qkv.bias
751
+ module.decoder.layers.2.self_attention.linear_qkv.layer_norm_weight
752
+ Params for bucket 19 (146800640 elements, 146800640 padded size):
753
+ module.decoder.layers.1.mlp.shared_experts.linear_fc1.weight
754
+ Params for bucket 20 (103001600 elements, 103001600 padded size):
755
+ module.decoder.layers.1.pre_mlp_layernorm.weight
756
+ module.decoder.layers.1.self_attention.linear_qkv.bias
757
+ module.decoder.layers.1.self_attention.linear_qkv.layer_norm_weight
758
+ module.decoder.layers.1.mlp.router.weight
759
+ module.decoder.layers.1.self_attention.linear_qkv.weight
760
+ module.decoder.layers.1.self_attention.linear_proj.weight
761
+ module.decoder.layers.0.mlp.shared_experts.linear_fc2.weight
762
+ Params for bucket 21 (146800640 elements, 146800640 padded size):
763
+ module.decoder.layers.0.mlp.shared_experts.linear_fc1.weight
764
+ Params for bucket 22 (209890816 elements, 209890816 padded size):
765
+ module.decoder.layers.0.mlp.router.weight
766
+ module.decoder.layers.0.pre_mlp_layernorm.weight
767
+ module.decoder.layers.0.self_attention.linear_qkv.layer_norm_weight
768
+ module.decoder.layers.0.self_attention.linear_qkv.weight
769
+ module.decoder.layers.0.self_attention.linear_proj.weight
770
+ module.decoder.layers.0.self_attention.linear_qkv.bias
771
+ module.embedding.word_embeddings.weight
772
+ INFO:megatron.core.distributed.param_and_grad_buffer:Number of buckets for gradient all-reduce / reduce-scatter: 44
773
+ Params for bucket 1 (45875200 elements, 45875200 padded size):
774
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight4
775
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight6
776
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight7
777
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight3
778
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight5
779
+ Params for bucket 2 (45875200 elements, 45875200 padded size):
780
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight2
781
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight1
782
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight7
783
+ module.decoder.layers.9.mlp.experts.linear_fc2.weight0
784
+ Params for bucket 3 (55050240 elements, 55050240 padded size):
785
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight6
786
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight4
787
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight5
788
+ Params for bucket 4 (55050240 elements, 55050240 padded size):
789
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight3
790
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight1
791
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight2
792
+ Params for bucket 5 (45875200 elements, 45875200 padded size):
793
+ module.decoder.layers.9.mlp.experts.linear_fc1.weight0
794
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight6
795
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight7
796
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight5
797
+ Params for bucket 6 (45875200 elements, 45875200 padded size):
798
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight4
799
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight2
800
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight0
801
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight3
802
+ module.decoder.layers.8.mlp.experts.linear_fc2.weight1
803
+ Params for bucket 7 (55050240 elements, 55050240 padded size):
804
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight7
805
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight6
806
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight5
807
+ Params for bucket 8 (55050240 elements, 55050240 padded size):
808
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight4
809
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight3
810
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight2
811
+ Params for bucket 9 (45875200 elements, 45875200 padded size):
812
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight1
813
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight7
814
+ module.decoder.layers.8.mlp.experts.linear_fc1.weight0
815
+ Params for bucket 10 (45875200 elements, 45875200 padded size):
816
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight6
817
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight4
818
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight2
819
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight5
820
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight3
821
+ Params for bucket 11 (55050240 elements, 55050240 padded size):
822
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight1
823
+ module.decoder.layers.7.mlp.experts.linear_fc2.weight0
824
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight6
825
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight7
826
+ Params for bucket 12 (55050240 elements, 55050240 padded size):
827
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight5
828
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight3
829
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight4
830
+ Params for bucket 13 (55050240 elements, 55050240 padded size):
831
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight2
832
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight1
833
+ module.decoder.layers.7.mlp.experts.linear_fc1.weight0
834
+ Params for bucket 14 (45875200 elements, 45875200 padded size):
835
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight6
836
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight4
837
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight7
838
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight3
839
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight5
840
+ Params for bucket 15 (45875200 elements, 45875200 padded size):
841
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight2
842
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight0
843
+ module.decoder.layers.6.mlp.experts.linear_fc2.weight1
844
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight7
845
+ Params for bucket 16 (55050240 elements, 55050240 padded size):
846
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight6
847
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight5
848
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight4
849
+ Params for bucket 17 (55050240 elements, 55050240 padded size):
850
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight3
851
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight2
852
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight1
853
+ Params for bucket 18 (45875200 elements, 45875200 padded size):
854
+ module.decoder.layers.6.mlp.experts.linear_fc1.weight0
855
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight5
856
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight6
857
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight7
858
+ Params for bucket 19 (45875200 elements, 45875200 padded size):
859
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight4
860
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight2
861
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight0
862
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight3
863
+ module.decoder.layers.5.mlp.experts.linear_fc2.weight1
864
+ Params for bucket 20 (55050240 elements, 55050240 padded size):
865
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight7
866
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight5
867
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight6
868
+ Params for bucket 21 (55050240 elements, 55050240 padded size):
869
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight4
870
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight2
871
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight3
872
+ Params for bucket 22 (45875200 elements, 45875200 padded size):
873
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight1
874
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight7
875
+ module.decoder.layers.5.mlp.experts.linear_fc1.weight0
876
+ Params for bucket 23 (45875200 elements, 45875200 padded size):
877
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight6
878
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight2
879
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight4
880
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight3
881
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight5
882
+ Params for bucket 24 (55050240 elements, 55050240 padded size):
883
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight1
884
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight6
885
+ module.decoder.layers.4.mlp.experts.linear_fc2.weight0
886
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight7
887
+ Params for bucket 25 (55050240 elements, 55050240 padded size):
888
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight5
889
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight3
890
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight4
891
+ Params for bucket 26 (55050240 elements, 55050240 padded size):
892
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight2
893
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight0
894
+ module.decoder.layers.4.mlp.experts.linear_fc1.weight1
895
+ Params for bucket 27 (45875200 elements, 45875200 padded size):
896
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight7
897
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight3
898
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight5
899
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight6
900
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight4
901
+ Params for bucket 28 (45875200 elements, 45875200 padded size):
902
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight2
903
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight0
904
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight7
905
+ module.decoder.layers.3.mlp.experts.linear_fc2.weight1
906
+ Params for bucket 29 (55050240 elements, 55050240 padded size):
907
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight6
908
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight4
909
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight5
910
+ Params for bucket 30 (55050240 elements, 55050240 padded size):
911
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight3
912
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight1
913
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight2
914
+ Params for bucket 31 (45875200 elements, 45875200 padded size):
915
+ module.decoder.layers.3.mlp.experts.linear_fc1.weight0
916
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight7
917
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight5
918
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight6
919
+ Params for bucket 32 (45875200 elements, 45875200 padded size):
920
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight4
921
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight0
922
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight2
923
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight1
924
+ module.decoder.layers.2.mlp.experts.linear_fc2.weight3
925
+ Params for bucket 33 (55050240 elements, 55050240 padded size):
926
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight7
927
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight6
928
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight5
929
+ Params for bucket 34 (55050240 elements, 55050240 padded size):
930
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight4
931
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight3
932
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight2
933
+ Params for bucket 35 (45875200 elements, 45875200 padded size):
934
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight1
935
+ module.decoder.layers.2.mlp.experts.linear_fc1.weight0
936
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight7
937
+ Params for bucket 36 (45875200 elements, 45875200 padded size):
938
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight6
939
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight4
940
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight2
941
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight3
942
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight5
943
+ Params for bucket 37 (55050240 elements, 55050240 padded size):
944
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight1
945
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight6
946
+ module.decoder.layers.1.mlp.experts.linear_fc2.weight0
947
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight7
948
+ Params for bucket 38 (55050240 elements, 55050240 padded size):
949
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight5
950
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight3
951
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight4
952
+ Params for bucket 39 (55050240 elements, 55050240 padded size):
953
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight2
954
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight0
955
+ module.decoder.layers.1.mlp.experts.linear_fc1.weight1
956
+ Params for bucket 40 (45875200 elements, 45875200 padded size):
957
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight6
958
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight4
959
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight7
960
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight3
961
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight5
962
+ Params for bucket 41 (45875200 elements, 45875200 padded size):
963
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight2
964
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight7
965
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight0
966
+ module.decoder.layers.0.mlp.experts.linear_fc2.weight1
967
+ Params for bucket 42 (55050240 elements, 55050240 padded size):
968
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight6
969
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight5
970
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight4
971
+ Params for bucket 43 (55050240 elements, 55050240 padded size):
972
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight3
973
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight1
974
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight2
975
+ Params for bucket 44 (18350080 elements, 18350080 padded size):
976
+ module.decoder.layers.0.mlp.experts.linear_fc1.weight0
977
+ 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 0x7e8368cfcc80>, config_logger_dir='')
978
+ INFO:megatron.core.optimizer_param_scheduler:> learning rate decay style: cosine
979
+ [after model, optimizer, and learning rate scheduler are built] datetime: 2026-08-30 00:57:38
980
+ > building train, validation, and test datasets ...
981
+ > datasets target sizes (minimum size):
982
+ train: 80
983
+ validation: 0
984
+ test: 0
985
+ INFO:megatron.core.datasets.blended_megatron_dataset_config:Let split_matrix = [(0, 1.0), None, None]
986
+ > building train, validation, and test datasets for GPT ...
987
+ INFO:megatron.core.datasets.blended_megatron_dataset_builder:Building GPTDataset splits with sizes=(80, 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 0x7e8373b75250>, 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)
988
+ INFO:megatron.core.datasets.indexed_dataset:Load the _IndexReader from /workspace/data/wikitext/wikitext103_gpt2_text_document.idx
989
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence lengths
990
+ INFO:megatron.core.datasets.indexed_dataset: Extract the sequence pointers
991
+ INFO:megatron.core.datasets.indexed_dataset: Extract the document indices
992
+ INFO:megatron.core.datasets.indexed_dataset:> total number of sequences: 1165029
993
+ INFO:megatron.core.datasets.indexed_dataset:> total number of documents: 1801350
994
+ INFO:megatron.core.datasets.gpt_dataset:Load the GPTDataset train indices
995
+ INFO:megatron.core.datasets.gpt_dataset: Load the document index from 71e99bc7c46bc3412b339a272e0f9cf3-GPTDataset-train-document_index.npy
996
+ INFO:megatron.core.datasets.gpt_dataset: Load the sample index from 71e99bc7c46bc3412b339a272e0f9cf3-GPTDataset-train-sample_index.npy
997
+ INFO:megatron.core.datasets.gpt_dataset: Load the shuffle index from 71e99bc7c46bc3412b339a272e0f9cf3-GPTDataset-train-shuffle_index.npy
998
+ INFO:megatron.core.datasets.gpt_dataset:> total number of samples: 29073
999
+ > finished creating GPT datasets ...
1000
+ [after dataloaders are built] datetime: 2026-08-30 00:57:38
1001
+ done with setup ...
1002
+ (min, max) time across ranks (ms):
1003
+ model-and-optimizer-setup ......................: (3437.07, 3443.96)
1004
+ train/valid/test-data-iterators-setup ..........: (99.25, 129.55)
1005
+ training ...
1006
+ Overwriting rerun_state_machine.current_iteration from -1 to 0...
1007
+ [before the start of training step] datetime: 2026-08-30 00:57:39
1008
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1009
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1010
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1011
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1012
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1013
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1014
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1015
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1016
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1017
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1018
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1019
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1020
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1021
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1022
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1023
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1024
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1025
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1026
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1027
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1028
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1029
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1030
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1031
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1032
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1033
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1034
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1035
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1036
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1037
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1038
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1039
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1040
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1041
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1042
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1043
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1044
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1045
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1046
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1047
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1048
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1049
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1050
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1051
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1052
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1053
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1054
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1055
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1056
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1057
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1058
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1059
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1060
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1061
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1062
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1063
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1064
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1065
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1066
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1067
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1068
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1069
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1070
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1071
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1072
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1073
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1074
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1075
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1076
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1077
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1078
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1079
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1080
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1081
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1082
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1083
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1084
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1085
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1086
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:336: UserWarning: The use_host_meta is deprecated, it will be removed in the future.
1087
+ warnings.warn("The use_host_meta is deprecated, it will be removed in the future.")
1088
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1089
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1090
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1091
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1092
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1093
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1094
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1095
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1096
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1097
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1098
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1099
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1100
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1101
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1102
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:482: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1103
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1104
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1105
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1106
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1107
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1108
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1109
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1110
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1111
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1112
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1113
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1114
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1115
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1116
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1117
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
1118
+ /workspace/DeepEP/deep_ep/hybrid_ep_buffer.py:334: UserWarning: The num_dispatched_tokens is deprecated, it will be removed in the future.
1119
+ warnings.warn("The num_dispatched_tokens is deprecated, it will be removed in the future.")
tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/QWEN3_30B__ultraep__matched__seq4096__gbs8__cap1p11328125/nsys/train.log CHANGED
@@ -2837,3 +2837,47 @@ HYBRID-EP ALLGATHER TIMEOUT:SM 1 [0]:expecting 3448 got 3446
2837
  HYBRID-EP ALLGATHER TIMEOUT:SM 7 [0]:expecting 3448 got 3446
2838
  HYBRID-EP ALLGATHER TIMEOUT:SM 1 [0]:expecting 3448 got 3446
2839
  HYBRID-EP ALLGATHER TIMEOUT:SM 12 [0]:expecting 3448 got 3446
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2837
  HYBRID-EP ALLGATHER TIMEOUT:SM 7 [0]:expecting 3448 got 3446
2838
  HYBRID-EP ALLGATHER TIMEOUT:SM 1 [0]:expecting 3448 got 3446
2839
  HYBRID-EP ALLGATHER TIMEOUT:SM 12 [0]:expecting 3448 got 3446
2840
+ Collection stop is not allowed in this state.
2841
+ W0830 00:54:15.857000 339798 torch/distributed/elastic/agent/server/api.py:719] Received 15 death signal, shutting down workers
2842
+ W0830 00:54:15.858000 339798 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 339915 closing signal SIGTERM
2843
+ Generated:
2844
+ /workspace/tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/QWEN3_30B__ultraep__matched__seq4096__gbs8__cap1p11328125/nsys/QWEN3_30B__ultraep__matched__seq4096__gbs8__cap1p11328125.nsys-rep
2845
+ W0830 00:54:15.861000 339798 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 339916 closing signal SIGTERM
2846
+ W0830 00:54:15.863000 339798 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 339917 closing signal SIGTERM
2847
+ W0830 00:54:15.869000 339798 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 339918 closing signal SIGTERM
2848
+ W0830 00:54:15.870000 339798 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 339919 closing signal SIGTERM
2849
+ W0830 00:54:15.872000 339798 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 339920 closing signal SIGTERM
2850
+ W0830 00:54:15.879000 339798 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 339921 closing signal SIGTERM
2851
+ W0830 00:54:15.881000 339798 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 339922 closing signal SIGTERM
2852
+ Traceback (most recent call last):
2853
+ File "/usr/local/bin/torchrun", line 8, in <module>
2854
+ sys.exit(main())
2855
+ ^^^^^^
2856
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper
2857
+ return f(*args, **kwargs)
2858
+ ^^^^^^^^^^^^^^^^^^
2859
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 892, in main
2860
+ run(args)
2861
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/run.py", line 883, in run
2862
+ elastic_launch(
2863
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 139, in __call__
2864
+ return launch_agent(self._config, self._entrypoint, list(args))
2865
+ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2866
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/launcher/api.py", line 261, in launch_agent
2867
+ result = agent.run()
2868
+ ^^^^^^^^^^^
2869
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/metrics/api.py", line 138, in wrapper
2870
+ result = f(*args, **kwargs)
2871
+ ^^^^^^^^^^^^^^^^^^
2872
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 711, in run
2873
+ result = self._invoke_run(role)
2874
+ ^^^^^^^^^^^^^^^^^^^^^^
2875
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/agent/server/api.py", line 870, in _invoke_run
2876
+ time.sleep(monitor_interval)
2877
+ File "/usr/local/lib/python3.12/dist-packages/torch/distributed/elastic/multiprocessing/api.py", line 84, in _terminate_process_handler
2878
+ raise SignalException(f"Process {os.getpid()} got signal: {sigval}", sigval=sigval)
2879
+ torch.distributed.elastic.multiprocessing.api.SignalException: Process 339798 got signal: 15
2880
+ The target application terminated. One or more process it created re-parented.
2881
+ Waiting for termination of re-parented processes.
2882
+ Use the `--wait` option to modify this behavior.
2883
+ There are no active sessions.
tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/QWEN3_30B__ultraep__matched__seq4096__gbs8__cap1p11328125/nsys/verify_trace_ranks.log ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ report : /workspace/tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/QWEN3_30B__ultraep__matched__seq4096__gbs8__cap1p11328125/nsys/QWEN3_30B__ultraep__matched__seq4096__gbs8__cap1p11328125.nsys-rep
2
+ kernels : 294800
3
+ devices : 8 of 8 -> [0, 1, 2, 3, 4, 5, 6, 7]
4
+ processes: 1
5
+ markers : 39936 ultraep_trace__ kernels on devices [0, 1, 2, 3, 4, 5, 6, 7]
6
+ PASS: all ranks present and backend markers emitted
tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/trace_results.tsv CHANGED
@@ -1,2 +1,4 @@
1
  model backend capacity status report async_overlap_json
2
  QWEN3_235B ultraep 1.083984375 OK /workspace/tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__ultraep__matched__seq8192__gbs8__cap1p083984375/nsys/QWEN3_235B__ultraep__matched__seq8192__gbs8__cap1p083984375.nsys-rep
 
 
 
1
  model backend capacity status report async_overlap_json
2
  QWEN3_235B ultraep 1.083984375 OK /workspace/tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/QWEN3_235B__ultraep__matched__seq8192__gbs8__cap1p083984375/nsys/QWEN3_235B__ultraep__matched__seq8192__gbs8__cap1p083984375.nsys-rep
3
+ QWEN3_30B ultraep 1.11328125 OK /workspace/tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/QWEN3_30B__ultraep__matched__seq4096__gbs8__cap1p11328125/nsys/QWEN3_30B__ultraep__matched__seq4096__gbs8__cap1p11328125.nsys-rep
4
+ DSV2_LITE_L24 ultraep 1.0755208333333333 OK /workspace/tierb-fixed-moonep-full/ultraep/tierbtrace/tierbtrace_gbs8/DSV2_LITE_L24__ultraep__matched__seq8192__gbs8__cap1p0755208333333333/nsys/DSV2_LITE_L24__ultraep__matched__seq8192__gbs8__cap1p0755208333333333.nsys-rep