[03/10 23:18:00 libai]: Rank of current process: 0. World size: 1
[03/10 23:18:00 libai]: Command line arguments: Namespace(config_file='configs/swin_imagenet.py', eval_only=False, fast_dev_run=False, opts=['model.cfg.hidden_dropout_prob=0.1', 'model.cfg.attention_probs_dropout_prob=0.1', 'model.cfg.bias_dropout_fusion=true', 'model.cfg.hidden_layers=12', 'model.cfg.hidden_size=768', 'model.cfg.num_attention_heads=12', 'model.cfg.intermediate_size=3072', 'model.cfg.ffn_hidden_size=3072', 'model.cfg.head_size=64', 'graph.enabled=true', 'train.dist.pipeline_num_layers=12', '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.num_accumulation_steps=8', 'train.evaluation.enabled=false', 'train.train_iter=220', 'train.train_epoch=0', 'train.log_period=100', 'train.zero_optimization.enabled=true', 'train.zero_optimization.stage=2', 'train.load_weight=', 'train.output_dir=test_logs/oneflow-28/NVIDIA_GeForce_RTX_3080_Ti/1ea2bb7/LibAI_swin_imagenet_graph_nl12_nah12_hs768_FP16_actrue_DP1_MP1_PP1_zerotrue_stage2_mbs128_gbs1024_acc8_1n1g'], resume=False)
[03/10 23:18:00 libai]: Contents of args.config_file=configs/swin_imagenet.py:
from libai.config import LazyCall
from .common.models.swin.swin_tiny_patch4_window7_224 import model
from .common.models.graph import graph
from .common.train import train
from .common.optim import optim
from .common.data.imagenet import dataloader

from flowvision.data import Mixup
from flowvision.loss.cross_entropy import SoftTargetCrossEntropy

# Refine data path to imagenet
dataloader.train.dataset[0].root = "/ssd/dataset/ImageNet/extract"
dataloader.test[0].dataset.root = "/ssd/dataset/ImageNet/extract"

# Add Mixup Func
dataloader.train.mixup_func = LazyCall(Mixup)(
    mixup_alpha=0.8,
    cutmix_alpha=1.0,
    prob=1.0,
    switch_prob=0.5,
    mode="batch",
    num_classes=1000,
)

# Refine model cfg for vit training on imagenet
model.cfg.num_classes = 1000
model.cfg.loss_func = SoftTargetCrossEntropy()
# Refine optimizer cfg for vit model
optim.lr = 1e-3
optim.eps = 1e-8
optim.weight_decay = 0.05
optim.params.clip_grad_max_norm = None
optim.params.clip_grad_norm_type = None

# Refine train cfg for vit model
train.train_micro_batch_size = 128
train.test_micro_batch_size = 128
train.train_epoch = 300
train.warmup_ratio = 20 / 300
train.eval_period = 1562
train.log_period = 100

# Scheduler
train.scheduler.warmup_factor = 0.001
train.scheduler.alpha = 0.01
train.scheduler.warmup_method = "linear"

# Set fp16 ON
train.amp.enabled = True

[03/10 23:18:00 libai]: Full config saved to test_logs/oneflow-28/NVIDIA_GeForce_RTX_3080_Ti/1ea2bb7/LibAI_swin_imagenet_graph_nl12_nah12_hs768_FP16_actrue_DP1_MP1_PP1_zerotrue_stage2_mbs128_gbs1024_acc8_1n1g/config.yaml
[03/10 23:18:00 lb.engine.default]: > compiling dataset index builder ...
make: Entering directory '/ssd/home/ouyangyu/libai_week_test/libai/libai/data/data_utils'
make: Nothing to be done for 'default'.
make: Leaving directory '/ssd/home/ouyangyu/libai_week_test/libai/libai/data/data_utils'
[03/10 23:18:00 lb.engine.default]: >>> done with dataset index builder. Compilation time: 0.050 seconds
[03/10 23:18:00 lb.engine.default]: >>> done with compiling. Compilation time: 0.051 seconds
[03/10 23:18:00 lb.engine.default]: Prepare training, validating, testing set
[03/10 23:18:04 lb.engine.default]: Prepare testing set
[03/10 23:18:04 lb.engine.default]: Auto-scaling the config to train.train_iter=220, train.warmup_iter=15
[03/10 23:18:04 libai]: > Start building model...
[03/10 23:18:05 lb.engine.default]: Model:
SwinTransformer(
  (patch_embed): PatchEmbed(
    (proj): Conv2d(3, 96, kernel_size=(4, 4), stride=(4, 4))
    (norm): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
  )
  (pos_drop): Dropout(p=0.0, inplace=False)
  (layers): ModuleList(
    (0): BasicLayer(
      (blocks): ModuleList(
        (0): SwinTransformerBlock(
          (norm1): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
          (attn): WindowAttention(
            (qkv): Linear1D(in_features=96, out_features=288, bias=True, parallel=data)
            (attn_drop): Dropout(p=0.0, inplace=False)
            (proj): Linear1D(in_features=96, out_features=96, bias=True, parallel=data)
            (proj_drop): Dropout(p=0.0, inplace=False)
            (softmax): Softmax(dim=-1)
          )
          (drop_path): Identity()
          (norm2): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
          (mlp): MLP(
            bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.0
            (dense_h_to_4h): Linear1D(in_features=96, out_features=384, bias=True, parallel=col)
            (dense_4h_to_h): Linear1D(in_features=384, out_features=96, bias=True, parallel=row)
          )
        )
        (1): SwinTransformerBlock(
          (norm1): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
          (attn): WindowAttention(
            (qkv): Linear1D(in_features=96, out_features=288, bias=True, parallel=data)
            (attn_drop): Dropout(p=0.0, inplace=False)
            (proj): Linear1D(in_features=96, out_features=96, bias=True, parallel=data)
            (proj_drop): Dropout(p=0.0, inplace=False)
            (softmax): Softmax(dim=-1)
          )
          (drop_path): DropPath()
          (norm2): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
          (mlp): MLP(
            bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.0
            (dense_h_to_4h): Linear1D(in_features=96, out_features=384, bias=True, parallel=col)
            (dense_4h_to_h): Linear1D(in_features=384, out_features=96, bias=True, parallel=row)
          )
        )
      )
      (downsample): PatchMerging(
        (reduction): Linear1D(in_features=384, out_features=192, bias=False, parallel=data)
        (norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
      )
    )
    (1): BasicLayer(
      (blocks): ModuleList(
        (0): SwinTransformerBlock(
          (norm1): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (attn): WindowAttention(
            (qkv): Linear1D(in_features=192, out_features=576, bias=True, parallel=data)
            (attn_drop): Dropout(p=0.0, inplace=False)
            (proj): Linear1D(in_features=192, out_features=192, bias=True, parallel=data)
            (proj_drop): Dropout(p=0.0, inplace=False)
            (softmax): Softmax(dim=-1)
          )
          (drop_path): DropPath()
          (norm2): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (mlp): MLP(
            bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.0
            (dense_h_to_4h): Linear1D(in_features=192, out_features=768, bias=True, parallel=col)
            (dense_4h_to_h): Linear1D(in_features=768, out_features=192, bias=True, parallel=row)
          )
        )
        (1): SwinTransformerBlock(
          (norm1): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (attn): WindowAttention(
            (qkv): Linear1D(in_features=192, out_features=576, bias=True, parallel=data)
            (attn_drop): Dropout(p=0.0, inplace=False)
            (proj): Linear1D(in_features=192, out_features=192, bias=True, parallel=data)
            (proj_drop): Dropout(p=0.0, inplace=False)
            (softmax): Softmax(dim=-1)
          )
          (drop_path): DropPath()
          (norm2): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (mlp): MLP(
            bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.0
            (dense_h_to_4h): Linear1D(in_features=192, out_features=768, bias=True, parallel=col)
            (dense_4h_to_h): Linear1D(in_features=768, out_features=192, bias=True, parallel=row)
          )
        )
      )
      (downsample): PatchMerging(
        (reduction): Linear1D(in_features=768, out_features=384, bias=False, parallel=data)
        (norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
      )
    )
    (2): BasicLayer(
      (blocks): ModuleList(
        (0): SwinTransformerBlock(
          (norm1): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (attn): WindowAttention(
            (qkv): Linear1D(in_features=384, out_features=1152, bias=True, parallel=data)
            (attn_drop): Dropout(p=0.0, inplace=False)
            (proj): Linear1D(in_features=384, out_features=384, bias=True, parallel=data)
            (proj_drop): Dropout(p=0.0, inplace=False)
            (softmax): Softmax(dim=-1)
          )
          (drop_path): DropPath()
          (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (mlp): MLP(
            bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.0
            (dense_h_to_4h): Linear1D(in_features=384, out_features=1536, bias=True, parallel=col)
            (dense_4h_to_h): Linear1D(in_features=1536, out_features=384, bias=True, parallel=row)
          )
        )
        (1): SwinTransformerBlock(
          (norm1): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (attn): WindowAttention(
            (qkv): Linear1D(in_features=384, out_features=1152, bias=True, parallel=data)
            (attn_drop): Dropout(p=0.0, inplace=False)
            (proj): Linear1D(in_features=384, out_features=384, bias=True, parallel=data)
            (proj_drop): Dropout(p=0.0, inplace=False)
            (softmax): Softmax(dim=-1)
          )
          (drop_path): DropPath()
          (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (mlp): MLP(
            bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.0
            (dense_h_to_4h): Linear1D(in_features=384, out_features=1536, bias=True, parallel=col)
            (dense_4h_to_h): Linear1D(in_features=1536, out_features=384, bias=True, parallel=row)
          )
        )
        (2): SwinTransformerBlock(
          (norm1): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (attn): WindowAttention(
            (qkv): Linear1D(in_features=384, out_features=1152, bias=True, parallel=data)
            (attn_drop): Dropout(p=0.0, inplace=False)
            (proj): Linear1D(in_features=384, out_features=384, bias=True, parallel=data)
            (proj_drop): Dropout(p=0.0, inplace=False)
            (softmax): Softmax(dim=-1)
          )
          (drop_path): DropPath()
          (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (mlp): MLP(
            bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.0
            (dense_h_to_4h): Linear1D(in_features=384, out_features=1536, bias=True, parallel=col)
            (dense_4h_to_h): Linear1D(in_features=1536, out_features=384, bias=True, parallel=row)
          )
        )
        (3): SwinTransformerBlock(
          (norm1): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (attn): WindowAttention(
            (qkv): Linear1D(in_features=384, out_features=1152, bias=True, parallel=data)
            (attn_drop): Dropout(p=0.0, inplace=False)
            (proj): Linear1D(in_features=384, out_features=384, bias=True, parallel=data)
            (proj_drop): Dropout(p=0.0, inplace=False)
            (softmax): Softmax(dim=-1)
          )
          (drop_path): DropPath()
          (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (mlp): MLP(
            bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.0
            (dense_h_to_4h): Linear1D(in_features=384, out_features=1536, bias=True, parallel=col)
            (dense_4h_to_h): Linear1D(in_features=1536, out_features=384, bias=True, parallel=row)
          )
        )
        (4): SwinTransformerBlock(
          (norm1): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (attn): WindowAttention(
            (qkv): Linear1D(in_features=384, out_features=1152, bias=True, parallel=data)
            (attn_drop): Dropout(p=0.0, inplace=False)
            (proj): Linear1D(in_features=384, out_features=384, bias=True, parallel=data)
            (proj_drop): Dropout(p=0.0, inplace=False)
            (softmax): Softmax(dim=-1)
          )
          (drop_path): DropPath()
          (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (mlp): MLP(
            bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.0
            (dense_h_to_4h): Linear1D(in_features=384, out_features=1536, bias=True, parallel=col)
            (dense_4h_to_h): Linear1D(in_features=1536, out_features=384, bias=True, parallel=row)
          )
        )
        (5): SwinTransformerBlock(
          (norm1): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (attn): WindowAttention(
            (qkv): Linear1D(in_features=384, out_features=1152, bias=True, parallel=data)
            (attn_drop): Dropout(p=0.0, inplace=False)
            (proj): Linear1D(in_features=384, out_features=384, bias=True, parallel=data)
            (proj_drop): Dropout(p=0.0, inplace=False)
            (softmax): Softmax(dim=-1)
          )
          (drop_path): DropPath()
          (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (mlp): MLP(
            bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.0
            (dense_h_to_4h): Linear1D(in_features=384, out_features=1536, bias=True, parallel=col)
            (dense_4h_to_h): Linear1D(in_features=1536, out_features=384, bias=True, parallel=row)
          )
        )
      )
      (downsample): PatchMerging(
        (reduction): Linear1D(in_features=1536, out_features=768, bias=False, parallel=data)
        (norm): LayerNorm((1536,), eps=1e-05, elementwise_affine=True)
      )
    )
    (3): BasicLayer(
      (blocks): ModuleList(
        (0): SwinTransformerBlock(
          (norm1): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
          (attn): WindowAttention(
            (qkv): Linear1D(in_features=768, out_features=2304, bias=True, parallel=data)
            (attn_drop): Dropout(p=0.0, inplace=False)
            (proj): Linear1D(in_features=768, out_features=768, bias=True, parallel=data)
            (proj_drop): Dropout(p=0.0, inplace=False)
            (softmax): Softmax(dim=-1)
          )
          (drop_path): DropPath()
          (norm2): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
          (mlp): MLP(
            bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.0
            (dense_h_to_4h): Linear1D(in_features=768, out_features=3072, bias=True, parallel=col)
            (dense_4h_to_h): Linear1D(in_features=3072, out_features=768, bias=True, parallel=row)
          )
        )
        (1): SwinTransformerBlock(
          (norm1): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
          (attn): WindowAttention(
            (qkv): Linear1D(in_features=768, out_features=2304, bias=True, parallel=data)
            (attn_drop): Dropout(p=0.0, inplace=False)
            (proj): Linear1D(in_features=768, out_features=768, bias=True, parallel=data)
            (proj_drop): Dropout(p=0.0, inplace=False)
            (softmax): Softmax(dim=-1)
          )
          (drop_path): DropPath()
          (norm2): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
          (mlp): MLP(
            bias_gelu_fusion=True, bias_dropout_fusion=True, dropout=0.0
            (dense_h_to_4h): Linear1D(in_features=768, out_features=3072, bias=True, parallel=col)
            (dense_4h_to_h): Linear1D(in_features=3072, out_features=768, bias=True, parallel=row)
          )
        )
      )
    )
  )
  (norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
  (avgpool): AdaptiveAvgPool1d()
  (head): Linear1D(in_features=768, out_features=1000, bias=True, parallel=data)
  (loss_func): SoftTargetCrossEntropy()
)
[03/10 23:18:05 libai]: >>> done with building model. Building time: 1.264 seconds
[03/10 23:18:05 lb.engine.trainer]: Starting training from iteration 0
[03/10 23:18:08 lb.models.utils.graph_base]: Start compiling 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]
2023/03/10 23:23:17.112, NVIDIA GeForce RTX 3080 Ti, 515.65.01, 100 %, 82 %, 12288 MiB, 6061 MiB, 5992 MiB
2023/03/10 23:23:17.113, NVIDIA GeForce RTX 3080 Ti, 515.65.01, 0 %, 0 %, 12288 MiB, 12051 MiB, 2 MiB
2023/03/10 23:23:17.114, NVIDIA GeForce RTX 3080 Ti, 515.65.01, 0 %, 0 %, 12288 MiB, 12051 MiB, 2 MiB
2023/03/10 23:23:17.115, NVIDIA GeForce RTX 3080 Ti, 515.65.01, 0 %, 0 %, 12288 MiB, 12051 MiB, 2 MiB
2023/03/10 23:23:17.116, NVIDIA GeForce RTX 3080 Ti, 515.65.01, 0 %, 0 %, 12288 MiB, 12051 MiB, 2 MiB
2023/03/10 23:23:17.117, NVIDIA GeForce RTX 3080 Ti, 515.65.01, 0 %, 0 %, 12288 MiB, 12051 MiB, 2 MiB
2023/03/10 23:23:17.118, NVIDIA GeForce RTX 3080 Ti, 515.65.01, 0 %, 0 %, 12288 MiB, 12051 MiB, 2 MiB
2023/03/10 23:23:17.119, NVIDIA GeForce RTX 3080 Ti, 515.65.01, 0 %, 0 %, 12288 MiB, 12051 MiB, 2 MiB
[03/10 23:23:19 lb.utils.events]:  eta: 0:05:49  iteration: 99/220  consumed_samples: 102400  total_loss: 6.927  time: 2.9761 s/iter  data_time: 0.8175 s/iter total_throughput: 344.08 samples/s lr: 5.82e-04  
[03/10 23:28:18 lb.utils.events]:  eta: 0:00:57  iteration: 199/220  consumed_samples: 204800  total_loss: 6.916  time: 2.9793 s/iter  data_time: 0.8006 s/iter total_throughput: 343.71 samples/s lr: 3.21e-05  
[03/10 23:29:17 lb.utils.events]:  eta: 0:00:00  iteration: 219/220  consumed_samples: 225280  total_loss: 6.911  time: 2.9766 s/iter  data_time: 0.8165 s/iter total_throughput: 344.02 samples/s lr: 1.01e-05  
[03/10 23:29:17 lb.engine.hooks]: Overall training speed: 218 iterations in 0:10:48 (2.9766 s / it)
[03/10 23:29:17 lb.engine.hooks]: Total training time: 0:10:49 (0:00:00 on hooks)
oneflow-version(git_commit)=0.9.1.dev20230309+cu117
oneflow-commit(git_commit)=1ea2bb7
oneflow-libai(git_commit)=50a973dc5de635b8613ad7666c073c763e238850