loaded library: /usr/lib/x86_64-linux-gnu/libibverbs.so.1
loaded library: /usr/lib/x86_64-linux-gnu/libibverbs.so.1
[07/05 07:15:34 libai]: Rank of current process: 0. World size: 1
[07/05 07:15:34 libai]: Command line arguments: Namespace(config_file='configs/bert_nl24_nah16_hs1024.py', eval_only=False, fast_dev_run=False, opts=['model.cfg.hidden_layers=24', 'train.dist.pipeline_num_layers=24', 'train.train_micro_batch_size=128', 'train.global_batch_size=1024', 'train.dist.tensor_parallel_size=1', 'train.dist.pipeline_parallel_size=1', 'train.amp.enabled=true', 'train.activation_checkpoint.enabled=true', 'train.train_iter=220', 'train.log_period=100', 'train.output_dir=test_logs/01b1d32/1n1g/LibAI_bert_nl24_nah16_hs1024_FP16_actrue_mp1_pp1_mb128_gb1024_1n1g_20220705_071531848751549'], resume=False)
[07/05 07:15:34 libai]: Contents of args.config_file=configs/bert_nl24_nah16_hs1024.py:
from libai.config import LazyCall
from libai.evaluation import PPLEvaluator
from .common.models.bert import pretrain_model as model
from .common.models.graph import graph
from .common.train import train
from .common.optim import optim
from .common.data.bert_dataset import dataloader, tokenization

#vocab_file = "/workspace/dataset/bert-base-chinese-vocab.txt"
#data_prefix = "/workspace/dataset/loss_compara_content_sentence"

vocab_file = "/dataset/source/dataset/bert-base-chinese-vocab.txt"
data_prefix = "/dataset/source/dataset/loss_compara_content_sentence"

tokenization.tokenizer.vocab_file = vocab_file
dataloader.train.dataset[0].data_prefix = data_prefix
dataloader.train.dataset[0].indexed_dataset.data_prefix = data_prefix
# dataloader.train.num_workers = 4

# Bert-large model config
#model.cfg.hidden_layers = 24
model.cfg.num_attention_heads = 16
model.cfg.hidden_size = 1024

#train.dist.pipeline_num_layers = model.cfg.hidden_layers
train.test_micro_batch_size = 4

train.evaluation.evaluator = LazyCall(PPLEvaluator)()

train.input_placement_device = "cpu"


train.evaluation.enabled = False
train.evaluation.eval_iter = 30

[07/05 07:15:34 libai]: Full config saved to test_logs/01b1d32/1n1g/LibAI_bert_nl24_nah16_hs1024_FP16_actrue_mp1_pp1_mb128_gb1024_1n1g_20220705_071531848751549/config.yaml
[07/05 07:15:34 lb.engine.default]: > compiling dataset index builder ...
make: Entering directory '/dataset/xyn/libai_bench/libai/libai/data/data_utils'
make: Nothing to be done for 'default'.
make: Leaving directory '/dataset/xyn/libai_bench/libai/libai/data/data_utils'
[07/05 07:15:34 lb.engine.default]: >>> done with dataset index builder. Compilation time: 0.041 seconds
[07/05 07:15:34 lb.engine.default]: >>> done with compiling. Compilation time: 0.041 seconds
[07/05 07:15:34 lb.engine.default]: Prepare training, validating, testing set
[07/05 07:15:34 lb.data.data_utils.indexed_dataset]: building dataset index ...
[07/05 07:15:34 lb.data.data_utils.indexed_dataset]: warming up index mmap file...
[07/05 07:15:34 lb.data.data_utils.indexed_dataset]: reading sizes...
[07/05 07:15:34 lb.data.data_utils.indexed_dataset]: reading pointers...
[07/05 07:15:34 lb.data.data_utils.indexed_dataset]: reading document index...
[07/05 07:15:34 lb.data.data_utils.indexed_dataset]: warming up data mmap file...
[07/05 07:15:34 lb.data.data_utils.indexed_dataset]: creating numpy buffer of mmap...
[07/05 07:15:34 lb.data.data_utils.indexed_dataset]: creating memory view of numpy buffer...
[07/05 07:15:34 lb.data.data_utils.indexed_dataset]: Finished creating indexed dataset in 0.094710 seconds
[07/05 07:15:34 lb.data.data_utils.indexed_dataset]: indexed dataset stats:
[07/05 07:15:34 lb.data.data_utils.indexed_dataset]: number of documents: 50000
[07/05 07:15:34 lb.data.data_utils.indexed_dataset]: number of sentences: 1249934
[07/05 07:15:34 lb.data.data_utils.dataset_utils]:  > loading indexed mapping from /dataset/source/dataset/loss_compara_content_sentence_bert_indexmap_225280mns_509msl_0.10ssp_1234s.npy
[07/05 07:15:34 lb.data.data_utils.dataset_utils]:     loaded indexed file in 0.005 seconds
[07/05 07:15:34 lb.data.data_utils.dataset_utils]:     total number of samples: 226136
[07/05 07:15:34 lb.data.data_utils.dataset_utils]:  > loading indexed mapping from /dataset/source/dataset/loss_compara_content_sentence_bert_indexmap_4mns_509msl_0.10ssp_1234s.npy
[07/05 07:15:34 lb.data.data_utils.dataset_utils]:     loaded indexed file in 0.001 seconds
[07/05 07:15:34 lb.data.data_utils.dataset_utils]:     total number of samples: 5884
[07/05 07:15:34 lb.data.data_utils.dataset_utils]:  > loading indexed mapping from /dataset/source/dataset/loss_compara_content_sentence_bert_indexmap_4mns_509msl_0.10ssp_1234s.npy
[07/05 07:15:34 lb.data.data_utils.dataset_utils]:     loaded indexed file in 0.000 seconds
[07/05 07:15:34 lb.data.data_utils.dataset_utils]:     total number of samples: 5884
W20220705 07:15:34.657053  4530 env_global_objects_scope.cpp:273] Skip init RDMA because only one process in this group!
[07/05 07:15:34 lb.engine.default]: Auto-scaling the config to train.train_iter=220, train.warmup_iter=0
[07/05 07:15:37 lb.engine.default]: Model:
BertForPreTraining(
  (bert): BertModel(
    (embeddings): BertEmbeddings(
      (vocab_embeddings): VocabEmbedding(num_embeddings=21248, embedding_dim=1024)
      (position_embeddings): Embedding(num_embeddings=512, embedding_dim=1024)
      (tokentype_embeddings): Embedding(num_embeddings=2, embedding_dim=1024)
      (embedding_dropout): Dropout(p=0.1, inplace=False)
    )
    (extended_attn_mask): BertExtendedAttnMask()
    (encoders): ModuleList(
      (0): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (1): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (2): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (3): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (4): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (5): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (6): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (7): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (8): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (9): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (10): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (11): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (12): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (13): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (14): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (15): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (16): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (17): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (18): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (19): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (20): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (21): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (22): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
      (23): TransformerLayer(
        (drop_path): Identity()
        (input_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (self_attention): MultiheadAttention(
          hidden_size=1024, num_heads=16, is_cross_attention=False
          (dropout): Dropout(p=0.1, inplace=False)
          (query_key_value): Linear1D(in_features=1024, out_features=3072, bias=True, parallel=col)
          (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=row)
        )
        (post_attention_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (mlp): MLP(
          bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.1
          (dense_h_to_4h): Linear1D(in_features=1024, out_features=4096, bias=True, parallel=col)
          (dense_4h_to_h): Linear1D(in_features=4096, out_features=1024, bias=True, parallel=row)
        )
      )
    )
    (final_layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
    (pooler): BertPooler(
      (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=col)
      (activation_func): Tanh()
    )
  )
  (cls_head): BertPreTrainingHeads(
    (predictions): BertLMPredictionHead(
      (dense): Linear1D(in_features=1024, out_features=1024, bias=True, parallel=data)
      (activation_func): GELU()
      (layernorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
    )
    (seq_relationship): Linear1D(in_features=1024, out_features=2, bias=True, parallel=data)
    (lm_logits): LMLogits()
    (loss_func): BertLoss(
      (lm_loss): ParallelCrossEntropyLoss()
    )
  )
)
WARNING [07/05 07:15:37 lb.scheduler.lr_scheduler]: warmup iters equals to zero, return CosineLR
[07/05 07:15:46 lb.engine.trainer]: Starting training from iteration 0
[07/05 07:15:49 lb.models.utils.graph_base]: Start compling the train graph which may take some time. Please wait for a moment ...
timestamp, name, driver_version, utilization.gpu [%], utilization.memory [%], memory.total [MiB], memory.free [MiB], memory.used [MiB]
2022/07/05 08:03:21.735, Tesla V100-SXM2-32GB, 470.57.02, 100 %, 52 %, 32510 MiB, 7348 MiB, 25162 MiB
2022/07/05 08:03:21.736, Tesla V100-SXM2-32GB, 470.57.02, 0 %, 0 %, 32510 MiB, 32507 MiB, 3 MiB
2022/07/05 08:03:21.736, Tesla V100-SXM2-32GB, 470.57.02, 0 %, 0 %, 32510 MiB, 32507 MiB, 3 MiB
2022/07/05 08:03:21.737, Tesla V100-SXM2-32GB, 470.57.02, 0 %, 0 %, 32510 MiB, 32507 MiB, 3 MiB
2022/07/05 08:03:21.737, Tesla V100-SXM2-32GB, 470.57.02, 0 %, 0 %, 32510 MiB, 32507 MiB, 3 MiB
2022/07/05 08:03:21.738, Tesla V100-SXM2-32GB, 470.57.02, 0 %, 0 %, 32510 MiB, 32507 MiB, 3 MiB
2022/07/05 08:03:21.739, Tesla V100-SXM2-32GB, 470.57.02, 0 %, 0 %, 32510 MiB, 32507 MiB, 3 MiB
2022/07/05 08:03:21.739, Tesla V100-SXM2-32GB, 470.57.02, 0 %, 0 %, 32510 MiB, 32507 MiB, 3 MiB
[07/05 08:03:50 lb.utils.events]:  eta: 0:57:19  iteration: 99/220  consumed_samples: 102400  total_loss: 7.979  lm_loss: 7.284  sop_loss: 0.6965  time: 28.6675 s/iter  data_time: 0.0051 s/iter total_throughput: 35.72 samples/s lr: 5.82e-05  
[07/05 08:51:34 lb.utils.events]:  eta: 0:09:32  iteration: 199/220  consumed_samples: 204800  total_loss: 7.899  lm_loss: 7.204  sop_loss: 0.6942  time: 28.6562 s/iter  data_time: 0.0049 s/iter total_throughput: 35.73 samples/s lr: 3.21e-06  
[07/05 09:01:07 lb.utils.events]:  eta: 0:00:00  iteration: 219/220  consumed_samples: 225280  total_loss: 7.895  lm_loss: 7.201  sop_loss: 0.6939  time: 28.6553 s/iter  data_time: 0.1348 s/iter total_throughput: 35.74 samples/s lr: 1.01e-06  
[07/05 09:01:07 lb.engine.hooks]: Overall training speed: 218 iterations in 1:44:06 (28.6553 s / it)
[07/05 09:01:07 lb.engine.hooks]: Total training time: 1:44:06 (0:00:00 on hooks)