MRaCL / CGFormer /exp /refcoco /cgformer /train.log
dianecy's picture
Upload folder using huggingface_hub
ea1014e verified
2023-07-23 21:05:48.257 | INFO | model:build_model:29 - Window size 12!
2023-07-23 21:05:49.774 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_window12_384_22k.pth
2023-07-23 21:06:44.926 | INFO | model:build_model:29 - Window size 12!
2023-07-23 21:06:45.837 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-23 21:06:49.628 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-23 21:06:53.778 | INFO | __mp_main__:main_worker:95 - amsgrad: True
batch_size: 64
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://localhost:12345
dropout: 0.0
epochs: 50
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 647631818
mask_root: path/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
rank: 0
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: path/refcoco/train.lmdb
train_split: train
val_lmdb: path/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1
2023-07-23 21:11:10.978 | INFO | model:build_model:29 - Window size 12!
2023-07-23 21:11:11.873 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-23 21:11:13.445 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-23 21:11:17.331 | INFO | __mp_main__:main_worker:95 - amsgrad: True
batch_size: 64
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://localhost:12345
dropout: 0.0
epochs: 50
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 994652464
mask_root: data/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
rank: 0
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: data/lmdb/refcoco/train.lmdb
train_split: train
val_lmdb: data/lmdb/refcoco/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1
2023-07-23 21:19:21.659 | INFO | model:build_model:29 - Window size 12!
2023-07-23 21:19:22.565 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-23 21:19:24.200 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-23 21:19:28.186 | INFO | __mp_main__:main_worker:96 - amsgrad: True
batch_size: 64
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://localhost:12345
dropout: 0.0
epochs: 50
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 12776955
mask_root: data/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
rank: 0
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: data/lmdb/refcoco/train.lmdb
train_split: train
val_lmdb: data/lmdb/refcoco/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1
2023-07-23 21:24:05.865 | INFO | model:build_model:29 - Window size 12!
2023-07-23 21:24:06.668 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-23 21:24:08.095 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-23 21:24:11.919 | INFO | __mp_main__:main_worker:96 - amsgrad: True
batch_size: 32
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://localhost:12345
dropout: 0.0
epochs: 50
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 1029525181
mask_root: data/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
rank: 0
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: data/lmdb/refcoco/train.lmdb
train_split: train
val_lmdb: data/lmdb/refcoco/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1
2023-07-23 21:27:30.199 | INFO | model:build_model:29 - Window size 12!
2023-07-23 21:27:30.993 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-23 21:27:32.462 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-23 21:27:36.506 | INFO | __mp_main__:main_worker:96 - amsgrad: True
batch_size: 16
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://localhost:12345
dropout: 0.0
epochs: 50
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 847009764
mask_root: data/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
rank: 0
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: data/lmdb/refcoco/train.lmdb
train_split: train
val_lmdb: data/lmdb/refcoco/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1
2023-07-23 21:31:04.012 | INFO | model:build_model:29 - Window size 12!
2023-07-23 21:31:04.829 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-23 21:31:06.470 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-23 21:31:10.411 | INFO | __mp_main__:main_worker:96 - amsgrad: True
batch_size: 8
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://localhost:12345
dropout: 0.0
epochs: 50
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 2081107713
mask_root: data/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
rank: 0
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: data/lmdb/refcoco/train.lmdb
train_split: train
val_lmdb: data/lmdb/refcoco/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1
2023-07-23 21:34:51.445 | INFO | utils.misc:display:108 - Training: Epoch=[1/50] [ 100/5300] Batch=1.03 (2.19) Data=0.00 (0.72) Lr=0.000100 Loss=1.2246 (1.2555) IoU=18.07 (16.25) Prec@50=12.50 (3.62)
2023-07-23 21:36:34.966 | INFO | utils.misc:display:108 - Training: Epoch=[1/50] [ 200/5300] Batch=1.04 (1.61) Data=0.00 (0.36) Lr=0.000100 Loss=1.1269 (1.1840) IoU=27.32 (20.41) Prec@50=12.50 (6.00)
2023-07-23 21:38:18.614 | INFO | utils.misc:display:108 - Training: Epoch=[1/50] [ 300/5300] Batch=1.04 (1.42) Data=0.00 (0.24) Lr=0.000100 Loss=1.1624 (1.1640) IoU=26.60 (21.62) Prec@50=12.50 (7.33)
2023-07-23 21:40:02.132 | INFO | utils.misc:display:108 - Training: Epoch=[1/50] [ 400/5300] Batch=1.03 (1.32) Data=0.00 (0.18) Lr=0.000100 Loss=1.0729 (1.1448) IoU=26.17 (22.45) Prec@50=12.50 (8.47)
2023-07-23 21:41:45.237 | INFO | utils.misc:display:108 - Training: Epoch=[1/50] [ 500/5300] Batch=1.02 (1.27) Data=0.00 (0.14) Lr=0.000100 Loss=1.0978 (1.1276) IoU=19.13 (23.53) Prec@50=12.50 (9.88)
2023-07-23 21:43:27.794 | INFO | utils.misc:display:108 - Training: Epoch=[1/50] [ 600/5300] Batch=1.03 (1.23) Data=0.00 (0.12) Lr=0.000100 Loss=1.1317 (1.1059) IoU=26.98 (24.96) Prec@50=12.50 (12.10)
2023-07-23 21:45:10.395 | INFO | utils.misc:display:108 - Training: Epoch=[1/50] [ 700/5300] Batch=1.03 (1.20) Data=0.00 (0.10) Lr=0.000100 Loss=1.1210 (1.0860) IoU=22.44 (26.26) Prec@50=12.50 (14.07)
2023-07-23 21:47:52.502 | INFO | model:build_model:29 - Window size 12!
2023-07-23 21:47:53.334 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-23 21:47:54.773 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-23 21:47:58.875 | INFO | __mp_main__:main_worker:96 - amsgrad: True
batch_size: 8
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://localhost:12345
dropout: 0.0
epochs: 50
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 2035714905
mask_root: data/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
rank: 0
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: data/lmdb/refcoco/train.lmdb
train_split: train
val_lmdb: data/lmdb/refcoco/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1
2023-07-23 21:51:12.135 | INFO | model:build_model:29 - Window size 12!
2023-07-23 21:51:12.964 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-23 21:51:14.485 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-23 21:51:18.365 | INFO | __mp_main__:main_worker:96 - amsgrad: True
batch_size: 8
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://localhost:12345
dropout: 0.0
epochs: 50
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 1926948477
mask_root: data/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
rank: 0
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: data/lmdb/refcoco/train.lmdb
train_split: train
val_lmdb: data/lmdb/refcoco/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1
2023-07-23 21:54:59.248 | INFO | model:build_model:29 - Window size 12!
2023-07-23 21:55:00.083 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-23 21:55:01.671 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-23 21:55:05.843 | INFO | __mp_main__:main_worker:96 - amsgrad: True
batch_size: 8
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://localhost:12345
dropout: 0.0
epochs: 50
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 1910753006
mask_root: data/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
rank: 0
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: data/lmdb/refcoco/train.lmdb
train_split: train
val_lmdb: data/lmdb/refcoco/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1
2023-07-23 21:58:47.847 | INFO | model:build_model:29 - Window size 12!
2023-07-23 21:58:48.766 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-23 21:58:50.270 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-23 21:58:54.166 | INFO | __mp_main__:main_worker:96 - amsgrad: True
batch_size: 8
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://localhost:12345
dropout: 0.0
epochs: 50
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 912278313
mask_root: data/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
rank: 0
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: data/lmdb/refcoco/train.lmdb
train_split: train
val_lmdb: data/lmdb/refcoco/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1
2023-07-23 22:01:38.585 | INFO | engine.engine:validate:153 - Evaluation: Epoch=[1/50] mIoU=0.00 oIoU=0.00 Pr@50: 0.00 Pr@60: 0.00 Pr@70: 0.00 Pr@80: 0.00 Pr@90: 0.00
2023-07-23 22:05:00.549 | INFO | model:build_model:29 - Window size 12!
2023-07-23 22:05:01.472 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-23 22:05:02.944 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-23 22:05:06.910 | INFO | __mp_main__:main_worker:96 - amsgrad: True
batch_size: 8
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://localhost:12345
dropout: 0.0
epochs: 3
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 336376068
mask_root: data/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
rank: 0
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: data/lmdb/refcoco/train.lmdb
train_split: train
val_lmdb: data/lmdb/refcoco/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1
2023-07-23 22:07:52.969 | INFO | engine.engine:validate:153 - Evaluation: Epoch=[1/3] mIoU=0.00 oIoU=0.00 Pr@50: 0.00 Pr@60: 0.00 Pr@70: 0.00 Pr@80: 0.00 Pr@90: 0.00
2023-07-23 22:10:58.358 | INFO | engine.engine:validate:153 - Evaluation: Epoch=[2/3] mIoU=5.43 oIoU=6.35 Pr@50: 1.00 Pr@60: 1.00 Pr@70: 0.00 Pr@80: 0.00 Pr@90: 0.00
2023-07-23 22:14:02.747 | INFO | engine.engine:validate:153 - Evaluation: Epoch=[3/3] mIoU=13.40 oIoU=16.25 Pr@50: 5.00 Pr@60: 3.00 Pr@70: 0.00 Pr@80: 0.00 Pr@90: 0.00
2023-07-23 22:14:31.253 | INFO | __mp_main__:main_worker:216 - * Best IoU=0.16250132024288177 *
2023-07-23 22:14:31.254 | INFO | __mp_main__:main_worker:219 - * Training time 0:09:24 *
2023-07-24 23:21:38.905 | INFO | model:build_model:29 - Window size 12!
2023-07-24 23:21:40.004 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-24 23:21:43.309 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-24 23:21:47.412 | INFO | __mp_main__:main_worker:133 - amsgrad: True
batch_size: 8
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://localhost:12345
dropout: 0.0
epochs: 3
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 636138922
mask_root: data/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
rank: 0
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: data/lmdb/refcoco/train.lmdb
train_split: train
val_lmdb: data/lmdb/refcoco/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1
2023-07-24 23:25:39.582 | INFO | model:build_model:29 - Window size 12!
2023-07-24 23:25:40.367 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-24 23:25:41.788 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-24 23:25:45.956 | INFO | __mp_main__:main_worker:136 - amsgrad: True
batch_size: 8
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://localhost:12345
dropout: 0.0
epochs: 3
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 1399806621
mask_root: data/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
rank: 0
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: data/lmdb/refcoco/train.lmdb
train_split: train
val_lmdb: data/lmdb/refcoco/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1
2023-07-24 23:30:31.733 | INFO | model:build_model:29 - Window size 12!
2023-07-24 23:30:32.534 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-24 23:30:34.030 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-24 23:30:38.036 | INFO | __mp_main__:main_worker:136 - amsgrad: True
batch_size: 8
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://localhost:12345
dropout: 0.0
epochs: 3
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
local_rank: 0
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 31329146
mask_root: data/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: data/lmdb/refcoco/train.lmdb
train_split: train
val_lmdb: data/lmdb/refcoco/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1
2023-07-25 00:42:09.924 | INFO | model:build_model:29 - Window size 12!
2023-07-25 00:42:10.757 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-25 00:42:12.256 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-25 00:42:16.565 | INFO | __mp_main__:main_worker:139 - amsgrad: True
batch_size: 8
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://127.0.0.1:14325
dropout: 0.0
epochs: 3
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 193784491
mask_root: data/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
rank: 0
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: data/lmdb/refcoco/train.lmdb
train_split: train
val_lmdb: data/lmdb/refcoco/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1
2023-07-25 00:45:54.939 | INFO | model:build_model:29 - Window size 12!
2023-07-25 00:45:55.714 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-25 00:45:57.056 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-25 00:46:01.058 | INFO | __mp_main__:main_worker:137 - amsgrad: True
batch_size: 8
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://127.0.0.1:14325
dropout: 0.0
epochs: 3
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 551183181
mask_root: data/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
rank: 0
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: data/lmdb/refcoco/train.lmdb
train_split: train
val_lmdb: data/lmdb/refcoco/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1
2023-07-25 00:50:33.797 | INFO | model:build_model:29 - Window size 12!
2023-07-25 00:50:34.577 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-25 00:50:35.966 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-25 00:50:39.766 | INFO | __mp_main__:main_worker:136 - amsgrad: True
batch_size: 8
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://127.0.0.1:14325
dropout: 0.0
epochs: 3
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 653388287
mask_root: data/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
rank: 0
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: data/lmdb/refcoco/train.lmdb
train_split: train
val_lmdb: data/lmdb/refcoco/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1
2023-07-25 00:51:38.538 | INFO | model:build_model:29 - Window size 12!
2023-07-25 00:51:39.332 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-25 00:51:40.766 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-25 00:51:44.682 | INFO | __mp_main__:main_worker:136 - amsgrad: True
batch_size: 8
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://127.0.0.1:14325
dropout: 0.0
epochs: 3
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 1276931077
mask_root: data/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
rank: 0
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: data/lmdb/refcoco/train.lmdb
train_split: train
val_lmdb: data/lmdb/refcoco/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1
2023-07-25 00:59:14.853 | INFO | model:build_model:29 - Window size 12!
2023-07-25 00:59:15.634 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-25 00:59:17.018 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-25 00:59:20.911 | INFO | __mp_main__:main_worker:136 - amsgrad: True
batch_size: 8
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://127.0.0.1:14325
dropout: 0.0
epochs: 3
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 1356008501
mask_root: data/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
rank: 0
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: data/lmdb/refcoco/train.lmdb
train_split: train
val_lmdb: data/lmdb/refcoco/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1
2023-07-25 01:04:31.476 | INFO | model:build_model:29 - Window size 12!
2023-07-25 01:04:32.252 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-25 01:04:33.639 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-25 01:04:37.629 | INFO | __mp_main__:main_worker:136 - amsgrad: True
batch_size: 2
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://127.0.0.1:14325
dropout: 0.0
epochs: 3
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 231773943
mask_root: data/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
rank: 0
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: data/lmdb/refcoco/train.lmdb
train_split: train
val_lmdb: data/lmdb/refcoco/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1
2023-07-25 01:12:13.956 | INFO | model:build_model:29 - Window size 12!
2023-07-25 01:12:14.747 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-25 01:12:16.132 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-25 01:12:20.148 | INFO | __mp_main__:main_worker:136 - amsgrad: True
batch_size: 1
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://127.0.0.1:14325
dropout: 0.0
epochs: 3
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 1469706493
mask_root: data/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
rank: 0
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: data/lmdb/refcoco/train.lmdb
train_split: train
val_lmdb: data/lmdb/refcoco/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1
2023-07-25 01:17:10.358 | INFO | model:build_model:29 - Window size 12!
2023-07-25 01:17:11.143 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-25 01:17:12.552 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-25 01:17:16.372 | INFO | __mp_main__:main_worker:136 - amsgrad: True
batch_size: 1
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://127.0.0.1:14325
dropout: 0.0
epochs: 3
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 2130061625
mask_root: data/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
rank: 0
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: data/lmdb/refcoco/train.lmdb
train_split: train
val_lmdb: data/lmdb/refcoco/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1
2023-07-25 01:24:52.013 | INFO | utils.misc:display:108 - Training: Epoch=[1/3] [100/100] Batch=2.70 (4.47) Data=0.00 (1.62) Lr=0.000667 Loss=1.5772 (1.6640) IoU=7.44 (10.28) Prec@50=0.00 (0.00)
2023-07-25 21:57:03.906 | INFO | model:build_model:29 - Window size 12!
2023-07-25 21:57:04.713 | INFO | model:build_model:49 - Initializing Multi-modal Swin Transformer weights from ckpts/swin_base_patch4_window12_384_22k.pth
2023-07-25 21:57:06.063 | INFO | model.backbone:init_weights:459 - loading swin success !!!
2023-07-25 21:57:10.219 | INFO | __mp_main__:main_worker:136 - amsgrad: True
batch_size: 1
batch_size_val: 16
bert: bert-base-uncased
dataset: refcoco
dist_backend: nccl
dist_url: tcp://127.0.0.1:14325
dropout: 0.0
epochs: 3
evaluate: True
exp_name: cgformer
fusion_drop: 0.0
gpu: 0
input_size: 480
lr: 0.0001
lr_backbone: 5e-05
lr_text_encoder: 5e-05
manual_seed: 931331093
mask_root: data/masks/refcoco
mha: 8-8-8-8
multiprocessing_distributed: True
ngpus_per_node: 1
num_token: 2
output_dir: exp/refcoco/cgformer
output_folder: exp/refcoco/
print_freq: 100
rank: 0
resume: None
save_freq: 1
start_epoch: 0
swin_pretrain: ckpts/swin_base_patch4_window12_384_22k.pth
swin_type: base
sync_bn: True
test_lmdb: data/lmdb/refcoco/val.lmdb
test_split: val
token_dim: 512
train_lmdb: data/lmdb/refcoco/train.lmdb
train_split: train
val_lmdb: data/lmdb/refcoco/val.lmdb
val_split: val
vis_dim: 512
visualize: False
weight: None
weight_decay: 0.0001
window12: True
word_dim: 768
word_len: 20
workers: 32
workers_val: 8
world_size: 1