Add files using large-upload tool
Browse files- params/2017.yaml +99 -0
- params/2018.yaml +99 -0
- params/2019.yaml +99 -0
- params/2020.yaml +99 -0
- params/2021.yaml +99 -0
- params/2022.yaml +99 -0
params/2017.yaml
ADDED
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
accum_freq: 1
|
2 |
+
aug_cfg: {}
|
3 |
+
batch_size: 2816
|
4 |
+
beta1: 0.9
|
5 |
+
beta2: 0.98
|
6 |
+
checkpoint_path: /tmp/restart_2017/datacomp_xlarge-basic_cumulative_seq_2017/checkpoints
|
7 |
+
coca_caption_loss_weight: 2.0
|
8 |
+
coca_contrastive_loss_weight: 1.0
|
9 |
+
copy_codebase: False
|
10 |
+
csv_caption_key: title
|
11 |
+
csv_img_key: filepath
|
12 |
+
csv_separator:
|
13 |
+
dataset_resampled: True
|
14 |
+
dataset_type: webdataset
|
15 |
+
ddp_static_graph: True
|
16 |
+
debug: False
|
17 |
+
decay_fraction: 0.2
|
18 |
+
delete_previous_checkpoint: False
|
19 |
+
device: cuda:0
|
20 |
+
dist_backend: nccl
|
21 |
+
dist_url: env://
|
22 |
+
distill: False
|
23 |
+
distill_model: None
|
24 |
+
distill_pretrained: None
|
25 |
+
distributed: True
|
26 |
+
epochs: 32
|
27 |
+
epochs_cooldown: None
|
28 |
+
eps: 1e-06
|
29 |
+
force_custom_text: False
|
30 |
+
force_image_size: None
|
31 |
+
force_patch_dropout: None
|
32 |
+
force_qk_norm: False
|
33 |
+
force_qk_norm_eps: 1e-05
|
34 |
+
force_quick_gelu: False
|
35 |
+
gather_with_grad: True
|
36 |
+
grad_checkpointing: True
|
37 |
+
grad_clip_norm: None
|
38 |
+
horovod: False
|
39 |
+
image_mean: None
|
40 |
+
image_std: None
|
41 |
+
imagenet_v2: None
|
42 |
+
imagenet_val: ../imagenet_validation
|
43 |
+
is_iteration_based: True
|
44 |
+
local_loss: True
|
45 |
+
local_rank: 0
|
46 |
+
lock_image: False
|
47 |
+
lock_image_freeze_bn_stats: False
|
48 |
+
lock_image_unlocked_groups: 0
|
49 |
+
lock_text: False
|
50 |
+
lock_text_freeze_layer_norm: False
|
51 |
+
lock_text_unlocked_layers: 0
|
52 |
+
log_every_n_steps: 100
|
53 |
+
log_level: 20
|
54 |
+
log_local: False
|
55 |
+
log_path: /tmp/restart_2017/datacomp_xlarge-basic_cumulative_seq_2017/out.log
|
56 |
+
logs: /tmp/restart_2017
|
57 |
+
lr: 0.001
|
58 |
+
lr_cooldown_end: 0.0
|
59 |
+
lr_cooldown_power: 1.0
|
60 |
+
lr_scheduler: cosine
|
61 |
+
max_iterations: 20500
|
62 |
+
model: ViT-B-16
|
63 |
+
name: datacomp_xlarge-basic_cumulative_seq_2017
|
64 |
+
new_run: True
|
65 |
+
no_set_device_rank: False
|
66 |
+
precision: amp
|
67 |
+
pretrained:
|
68 |
+
pretrained_image: False
|
69 |
+
rank: 0
|
70 |
+
remote_sync: xlarge_CL_basic_filter/restart_2017/
|
71 |
+
remote_sync_frequency: 300
|
72 |
+
remote_sync_protocol: s3
|
73 |
+
report_to: wandb
|
74 |
+
resume: xlarge_CL_basic_filter/cumulative_sequential_2016/datacomp_xlarge-basic-2023-09-12_05-47-41_cumulative_seq_2016/checkpoints/epoch_8.pt
|
75 |
+
save_frequency: 1
|
76 |
+
save_most_recent: True
|
77 |
+
seed: 0
|
78 |
+
skip_scheduler: False
|
79 |
+
tensorboard: False
|
80 |
+
tensorboard_path:
|
81 |
+
torchcompile: False
|
82 |
+
torchscript: False
|
83 |
+
trace: False
|
84 |
+
train_data: xlarge_basic_filter/2017/1/{00000000..00000681}.tar::xlarge_basic_filter/2017/10/{00000000..00000639}.tar::xlarge_basic_filter/2017/11/{00000000..00000703}.tar::xlarge_basic_filter/2017/12/{00000000..00000703}.tar::xlarge_basic_filter/2017/13/{00000000..00000730}.tar::xlarge_basic_filter/2017/14/{00000000..00000703}.tar::xlarge_basic_filter/2017/15/{00000000..00000672}.tar::xlarge_basic_filter/2017/16/{00000000..00000639}.tar::xlarge_basic_filter/2017/17/{00000000..00000630}.tar::xlarge_basic_filter/2017/18/{00000000..00000723}.tar::xlarge_basic_filter/2017/19/{00000000..00000703}.tar::xlarge_basic_filter/2017/2/{00000000..00000703}.tar::xlarge_basic_filter/2017/20/{00000000..00000665}.tar::xlarge_basic_filter/2017/21/{00000000..00000639}.tar::xlarge_basic_filter/2017/22/{00000000..00000639}.tar::xlarge_basic_filter/2017/23/{00000000..00000767}.tar::xlarge_basic_filter/2017/24/{00000000..00000511}.tar::xlarge_basic_filter/2017/25/{00000000..00000511}.tar::xlarge_basic_filter/2017/26/{00000000..00000063}.tar::xlarge_basic_filter/2017/27/{00000000..00000139}.tar::xlarge_basic_filter/2017/28/{00000000..00000831}.tar::xlarge_basic_filter/2017/29/{00000000..00000831}.tar::xlarge_basic_filter/2017/3/{00000000..00000703}.tar::xlarge_basic_filter/2017/30/{00000000..00000770}.tar::xlarge_basic_filter/2017/31/{00000000..00000772}.tar::xlarge_basic_filter/2017/32/{00000000..00000778}.tar::xlarge_basic_filter/2017/33/{00000000..00000831}.tar::xlarge_basic_filter/2017/34/{00000000..00000767}.tar::xlarge_basic_filter/2017/35/{00000000..00000767}.tar::xlarge_basic_filter/2017/36/{00000000..00000831}.tar::xlarge_basic_filter/2017/37/{00000000..00000799}.tar::xlarge_basic_filter/2017/38/{00000000..00000719}.tar::xlarge_basic_filter/2017/39/{00000000..00000703}.tar::xlarge_basic_filter/2017/4/{00000000..00000679}.tar::xlarge_basic_filter/2017/40/{00000000..00000703}.tar::xlarge_basic_filter/2017/41/{00000000..00000639}.tar::xlarge_basic_filter/2017/42/{00000000..00000703}.tar::xlarge_basic_filter/2017/43/{00000000..00000767}.tar::xlarge_basic_filter/2017/44/{00000000..00000767}.tar::xlarge_basic_filter/2017/45/{00000000..00000767}.tar::xlarge_basic_filter/2017/46/{00000000..00000755}.tar::xlarge_basic_filter/2017/47/{00000000..00000767}.tar::xlarge_basic_filter/2017/48/{00000000..00000831}.tar::xlarge_basic_filter/2017/49/{00000000..00000831}.tar::xlarge_basic_filter/2017/5/{00000000..00000682}.tar::xlarge_basic_filter/2017/50/{00000000..00000831}.tar::xlarge_basic_filter/2017/51/{00000000..00000831}.tar::xlarge_basic_filter/2017/52/{00000000..00000831}.tar::xlarge_basic_filter/2017/53/{00000000..00001279}.tar::xlarge_basic_filter/2017/54/{00000000..00001227}.tar::xlarge_basic_filter/2017/55/{00000000..00000435}.tar::xlarge_basic_filter/2017/56/{00000000..00000433}.tar::xlarge_basic_filter/2017/57/{00000000..00000383}.tar::xlarge_basic_filter/2017/58/{00000000..00000383}.tar::xlarge_basic_filter/2017/59/{00000000..00000383}.tar::xlarge_basic_filter/2017/6/{00000000..00000679}.tar::xlarge_basic_filter/2017/7/{00000000..00000767}.tar::xlarge_basic_filter/2017/8/{00000000..00000719}.tar::xlarge_basic_filter/2017/9/{00000000..00000639}.tar
|
85 |
+
train_data_upsampling_factors: None
|
86 |
+
train_num_samples: None
|
87 |
+
use_bn_sync: False
|
88 |
+
use_bnb_linear: None
|
89 |
+
val_data: None
|
90 |
+
val_frequency: 1
|
91 |
+
val_num_samples: None
|
92 |
+
wandb: True
|
93 |
+
wandb_notes:
|
94 |
+
wandb_project_name: datacomp-xlarge-CL
|
95 |
+
warmup: 2000
|
96 |
+
wd: 0.2
|
97 |
+
workers: 4
|
98 |
+
world_size: 32
|
99 |
+
zeroshot_frequency: 1
|
params/2018.yaml
ADDED
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
accum_freq: 1
|
2 |
+
aug_cfg: {}
|
3 |
+
batch_size: 1408
|
4 |
+
beta1: 0.9
|
5 |
+
beta2: 0.98
|
6 |
+
checkpoint_path: /tmp/restart_2018/datacomp_xlarge-basic_cumulative_seq_2018/checkpoints
|
7 |
+
coca_caption_loss_weight: 2.0
|
8 |
+
coca_contrastive_loss_weight: 1.0
|
9 |
+
copy_codebase: False
|
10 |
+
csv_caption_key: title
|
11 |
+
csv_img_key: filepath
|
12 |
+
csv_separator:
|
13 |
+
dataset_resampled: True
|
14 |
+
dataset_type: webdataset
|
15 |
+
ddp_static_graph: True
|
16 |
+
debug: False
|
17 |
+
decay_fraction: 0.2
|
18 |
+
delete_previous_checkpoint: False
|
19 |
+
device: cuda:0
|
20 |
+
dist_backend: nccl
|
21 |
+
dist_url: env://
|
22 |
+
distill: False
|
23 |
+
distill_model: None
|
24 |
+
distill_pretrained: None
|
25 |
+
distributed: True
|
26 |
+
epochs: 32
|
27 |
+
epochs_cooldown: None
|
28 |
+
eps: 1e-06
|
29 |
+
force_custom_text: False
|
30 |
+
force_image_size: None
|
31 |
+
force_patch_dropout: None
|
32 |
+
force_qk_norm: False
|
33 |
+
force_qk_norm_eps: 1e-05
|
34 |
+
force_quick_gelu: False
|
35 |
+
gather_with_grad: True
|
36 |
+
grad_checkpointing: True
|
37 |
+
grad_clip_norm: None
|
38 |
+
horovod: False
|
39 |
+
image_mean: None
|
40 |
+
image_std: None
|
41 |
+
imagenet_v2: None
|
42 |
+
imagenet_val: ../imagenet_validation
|
43 |
+
is_iteration_based: True
|
44 |
+
local_loss: True
|
45 |
+
local_rank: 0
|
46 |
+
lock_image: False
|
47 |
+
lock_image_freeze_bn_stats: False
|
48 |
+
lock_image_unlocked_groups: 0
|
49 |
+
lock_text: False
|
50 |
+
lock_text_freeze_layer_norm: False
|
51 |
+
lock_text_unlocked_layers: 0
|
52 |
+
log_every_n_steps: 100
|
53 |
+
log_level: 20
|
54 |
+
log_local: False
|
55 |
+
log_path: /tmp/restart_2018/datacomp_xlarge-basic_cumulative_seq_2018/out.log
|
56 |
+
logs: /tmp/restart_2018
|
57 |
+
lr: 0.001
|
58 |
+
lr_cooldown_end: 0.0
|
59 |
+
lr_cooldown_power: 1.0
|
60 |
+
lr_scheduler: cosine
|
61 |
+
max_iterations: 20500
|
62 |
+
model: ViT-B-16
|
63 |
+
name: datacomp_xlarge-basic_cumulative_seq_2018
|
64 |
+
new_run: True
|
65 |
+
no_set_device_rank: False
|
66 |
+
precision: amp
|
67 |
+
pretrained:
|
68 |
+
pretrained_image: False
|
69 |
+
rank: 0
|
70 |
+
remote_sync: xlarge_CL_basic_filter/restart_2018/
|
71 |
+
remote_sync_frequency: 300
|
72 |
+
remote_sync_protocol: s3
|
73 |
+
report_to: wandb
|
74 |
+
resume: xlarge_CL_basic_filter/restart_2017/datacomp_xlarge-basic_cumulative_seq_2017/checkpoints/epoch_5.pt
|
75 |
+
save_frequency: 1
|
76 |
+
save_most_recent: True
|
77 |
+
seed: 0
|
78 |
+
skip_scheduler: False
|
79 |
+
tensorboard: False
|
80 |
+
tensorboard_path:
|
81 |
+
torchcompile: False
|
82 |
+
torchscript: False
|
83 |
+
trace: False
|
84 |
+
train_data: xlarge_basic_filter/2018/1/{00000000..00000832}.tar::xlarge_basic_filter/2018/10/{00000000..00000768}.tar::xlarge_basic_filter/2018/11/{00000000..00000831}.tar::xlarge_basic_filter/2018/12/{00000000..00000833}.tar::xlarge_basic_filter/2018/13/{00000000..00000895}.tar::xlarge_basic_filter/2018/14/{00000000..00000895}.tar::xlarge_basic_filter/2018/15/{00000000..00000831}.tar::xlarge_basic_filter/2018/16/{00000000..00000799}.tar::xlarge_basic_filter/2018/17/{00000000..00000767}.tar::xlarge_basic_filter/2018/18/{00000000..00000895}.tar::xlarge_basic_filter/2018/19/{00000000..00000879}.tar::xlarge_basic_filter/2018/2/{00000000..00000879}.tar::xlarge_basic_filter/2018/20/{00000000..00000831}.tar::xlarge_basic_filter/2018/21/{00000000..00000767}.tar::xlarge_basic_filter/2018/22/{00000000..00000759}.tar::xlarge_basic_filter/2018/23/{00000000..00000911}.tar::xlarge_basic_filter/2018/24/{00000000..00000629}.tar::xlarge_basic_filter/2018/25/{00000000..00000599}.tar::xlarge_basic_filter/2018/26/{00000000..00000063}.tar::xlarge_basic_filter/2018/27/{00000000..00000191}.tar::xlarge_basic_filter/2018/28/{00000000..00001023}.tar::xlarge_basic_filter/2018/29/{00000000..00000999}.tar::xlarge_basic_filter/2018/3/{00000000..00000895}.tar::xlarge_basic_filter/2018/30/{00000000..00000972}.tar::xlarge_basic_filter/2018/31/{00000000..00000973}.tar::xlarge_basic_filter/2018/32/{00000000..00000981}.tar::xlarge_basic_filter/2018/33/{00000000..00000999}.tar::xlarge_basic_filter/2018/34/{00000000..00000959}.tar::xlarge_basic_filter/2018/35/{00000000..00000959}.tar::xlarge_basic_filter/2018/36/{00000000..00001023}.tar::xlarge_basic_filter/2018/37/{00000000..00000999}.tar::xlarge_basic_filter/2018/38/{00000000..00000895}.tar::xlarge_basic_filter/2018/39/{00000000..00000895}.tar::xlarge_basic_filter/2018/4/{00000000..00000895}.tar::xlarge_basic_filter/2018/40/{00000000..00000877}.tar::xlarge_basic_filter/2018/41/{00000000..00000767}.tar::xlarge_basic_filter/2018/42/{00000000..00000895}.tar::xlarge_basic_filter/2018/43/{00000000..00000959}.tar::xlarge_basic_filter/2018/44/{00000000..00000959}.tar::xlarge_basic_filter/2018/45/{00000000..00000901}.tar::xlarge_basic_filter/2018/46/{00000000..00000896}.tar::xlarge_basic_filter/2018/47/{00000000..00000959}.tar::xlarge_basic_filter/2018/48/{00000000..00001023}.tar::xlarge_basic_filter/2018/49/{00000000..00001023}.tar::xlarge_basic_filter/2018/5/{00000000..00000895}.tar::xlarge_basic_filter/2018/50/{00000000..00001023}.tar::xlarge_basic_filter/2018/51/{00000000..00001023}.tar::xlarge_basic_filter/2018/52/{00000000..00001023}.tar::xlarge_basic_filter/2018/53/{00000000..00001599}.tar::xlarge_basic_filter/2018/54/{00000000..00001535}.tar::xlarge_basic_filter/2018/55/{00000000..00000511}.tar::xlarge_basic_filter/2018/56/{00000000..00000511}.tar::xlarge_basic_filter/2018/57/{00000000..00000479}.tar::xlarge_basic_filter/2018/58/{00000000..00000510}.tar::xlarge_basic_filter/2018/59/{00000000..00000479}.tar::xlarge_basic_filter/2018/6/{00000000..00000895}.tar::xlarge_basic_filter/2018/7/{00000000..00000919}.tar::xlarge_basic_filter/2018/8/{00000000..00000895}.tar::xlarge_basic_filter/2018/9/{00000000..00000830}.tar
|
85 |
+
train_data_upsampling_factors: None
|
86 |
+
train_num_samples: None
|
87 |
+
use_bn_sync: False
|
88 |
+
use_bnb_linear: None
|
89 |
+
val_data: None
|
90 |
+
val_frequency: 1
|
91 |
+
val_num_samples: None
|
92 |
+
wandb: True
|
93 |
+
wandb_notes:
|
94 |
+
wandb_project_name: datacomp-xlarge-CL
|
95 |
+
warmup: 2000
|
96 |
+
wd: 0.2
|
97 |
+
workers: 4
|
98 |
+
world_size: 64
|
99 |
+
zeroshot_frequency: 1
|
params/2019.yaml
ADDED
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
accum_freq: 1
|
2 |
+
aug_cfg: {}
|
3 |
+
batch_size: 1408
|
4 |
+
beta1: 0.9
|
5 |
+
beta2: 0.98
|
6 |
+
checkpoint_path: /tmp/restart_2019/datacomp_xlarge-basic_cumulative_seq_2019/checkpoints
|
7 |
+
coca_caption_loss_weight: 2.0
|
8 |
+
coca_contrastive_loss_weight: 1.0
|
9 |
+
copy_codebase: False
|
10 |
+
csv_caption_key: title
|
11 |
+
csv_img_key: filepath
|
12 |
+
csv_separator:
|
13 |
+
dataset_resampled: True
|
14 |
+
dataset_type: webdataset
|
15 |
+
ddp_static_graph: True
|
16 |
+
debug: False
|
17 |
+
decay_fraction: 0.2
|
18 |
+
delete_previous_checkpoint: False
|
19 |
+
device: cuda:0
|
20 |
+
dist_backend: nccl
|
21 |
+
dist_url: env://
|
22 |
+
distill: False
|
23 |
+
distill_model: None
|
24 |
+
distill_pretrained: None
|
25 |
+
distributed: True
|
26 |
+
epochs: 32
|
27 |
+
epochs_cooldown: None
|
28 |
+
eps: 1e-06
|
29 |
+
force_custom_text: False
|
30 |
+
force_image_size: None
|
31 |
+
force_patch_dropout: None
|
32 |
+
force_qk_norm: False
|
33 |
+
force_qk_norm_eps: 1e-05
|
34 |
+
force_quick_gelu: False
|
35 |
+
gather_with_grad: True
|
36 |
+
grad_checkpointing: True
|
37 |
+
grad_clip_norm: None
|
38 |
+
horovod: False
|
39 |
+
image_mean: None
|
40 |
+
image_std: None
|
41 |
+
imagenet_v2: None
|
42 |
+
imagenet_val: ../imagenet_validation
|
43 |
+
is_iteration_based: True
|
44 |
+
local_loss: True
|
45 |
+
local_rank: 0
|
46 |
+
lock_image: False
|
47 |
+
lock_image_freeze_bn_stats: False
|
48 |
+
lock_image_unlocked_groups: 0
|
49 |
+
lock_text: False
|
50 |
+
lock_text_freeze_layer_norm: False
|
51 |
+
lock_text_unlocked_layers: 0
|
52 |
+
log_every_n_steps: 100
|
53 |
+
log_level: 20
|
54 |
+
log_local: False
|
55 |
+
log_path: /tmp/restart_2019/datacomp_xlarge-basic_cumulative_seq_2019/out.log
|
56 |
+
logs: /tmp/restart_2019
|
57 |
+
lr: 0.001
|
58 |
+
lr_cooldown_end: 0.0
|
59 |
+
lr_cooldown_power: 1.0
|
60 |
+
lr_scheduler: cosine
|
61 |
+
max_iterations: 20500
|
62 |
+
model: ViT-B-16
|
63 |
+
name: datacomp_xlarge-basic_cumulative_seq_2019
|
64 |
+
new_run: True
|
65 |
+
no_set_device_rank: False
|
66 |
+
precision: amp
|
67 |
+
pretrained:
|
68 |
+
pretrained_image: False
|
69 |
+
rank: 0
|
70 |
+
remote_sync: xlarge_CL_basic_filter/restart_2019/
|
71 |
+
remote_sync_frequency: 300
|
72 |
+
remote_sync_protocol: s3
|
73 |
+
report_to: wandb
|
74 |
+
resume: xlarge_CL_basic_filter/restart_2018/datacomp_xlarge-basic_cumulative_seq_2018/checkpoints/epoch_4.pt
|
75 |
+
save_frequency: 1
|
76 |
+
save_most_recent: True
|
77 |
+
seed: 0
|
78 |
+
skip_scheduler: False
|
79 |
+
tensorboard: False
|
80 |
+
tensorboard_path:
|
81 |
+
torchcompile: False
|
82 |
+
torchscript: False
|
83 |
+
trace: False
|
84 |
+
train_data: xlarge_basic_filter/2019/1/{00000000..00000895}.tar::xlarge_basic_filter/2019/10/{00000000..00000831}.tar::xlarge_basic_filter/2019/11/{00000000..00000846}.tar::xlarge_basic_filter/2019/12/{00000000..00000887}.tar::xlarge_basic_filter/2019/13/{00000000..00000940}.tar::xlarge_basic_filter/2019/14/{00000000..00000895}.tar::xlarge_basic_filter/2019/15/{00000000..00000831}.tar::xlarge_basic_filter/2019/16/{00000000..00000831}.tar::xlarge_basic_filter/2019/17/{00000000..00000820}.tar::xlarge_basic_filter/2019/18/{00000000..00000925}.tar::xlarge_basic_filter/2019/19/{00000000..00000895}.tar::xlarge_basic_filter/2019/2/{00000000..00000935}.tar::xlarge_basic_filter/2019/20/{00000000..00000831}.tar::xlarge_basic_filter/2019/21/{00000000..00000831}.tar::xlarge_basic_filter/2019/22/{00000000..00000760}.tar::xlarge_basic_filter/2019/23/{00000000..00000959}.tar::xlarge_basic_filter/2019/24/{00000000..00000639}.tar::xlarge_basic_filter/2019/25/{00000000..00000609}.tar::xlarge_basic_filter/2019/26/{00000000..00000063}.tar::xlarge_basic_filter/2019/27/{00000000..00000191}.tar::xlarge_basic_filter/2019/28/{00000000..00001074}.tar::xlarge_basic_filter/2019/29/{00000000..00001060}.tar::xlarge_basic_filter/2019/3/{00000000..00000910}.tar::xlarge_basic_filter/2019/30/{00000000..00001023}.tar::xlarge_basic_filter/2019/31/{00000000..00001023}.tar::xlarge_basic_filter/2019/32/{00000000..00001023}.tar::xlarge_basic_filter/2019/33/{00000000..00001024}.tar::xlarge_basic_filter/2019/34/{00000000..00001023}.tar::xlarge_basic_filter/2019/35/{00000000..00001023}.tar::xlarge_basic_filter/2019/36/{00000000..00001086}.tar::xlarge_basic_filter/2019/37/{00000000..00001087}.tar::xlarge_basic_filter/2019/38/{00000000..00000959}.tar::xlarge_basic_filter/2019/39/{00000000..00000906}.tar::xlarge_basic_filter/2019/4/{00000000..00000897}.tar::xlarge_basic_filter/2019/40/{00000000..00000897}.tar::xlarge_basic_filter/2019/41/{00000000..00000831}.tar::xlarge_basic_filter/2019/42/{00000000..00000958}.tar::xlarge_basic_filter/2019/43/{00000000..00000966}.tar::xlarge_basic_filter/2019/44/{00000000..00001011}.tar::xlarge_basic_filter/2019/45/{00000000..00000959}.tar::xlarge_basic_filter/2019/46/{00000000..00000959}.tar::xlarge_basic_filter/2019/47/{00000000..00000967}.tar::xlarge_basic_filter/2019/48/{00000000..00001082}.tar::xlarge_basic_filter/2019/49/{00000000..00001066}.tar::xlarge_basic_filter/2019/5/{00000000..00000959}.tar::xlarge_basic_filter/2019/50/{00000000..00001067}.tar::xlarge_basic_filter/2019/51/{00000000..00001075}.tar::xlarge_basic_filter/2019/52/{00000000..00001075}.tar::xlarge_basic_filter/2019/53/{00000000..00001646}.tar::xlarge_basic_filter/2019/54/{00000000..00001599}.tar::xlarge_basic_filter/2019/55/{00000000..00000511}.tar::xlarge_basic_filter/2019/56/{00000000..00000511}.tar::xlarge_basic_filter/2019/57/{00000000..00000511}.tar::xlarge_basic_filter/2019/58/{00000000..00000511}.tar::xlarge_basic_filter/2019/59/{00000000..00000511}.tar::xlarge_basic_filter/2019/6/{00000000..00000959}.tar::xlarge_basic_filter/2019/7/{00000000..00000959}.tar::xlarge_basic_filter/2019/8/{00000000..00000895}.tar::xlarge_basic_filter/2019/9/{00000000..00000831}.tar
|
85 |
+
train_data_upsampling_factors: None
|
86 |
+
train_num_samples: None
|
87 |
+
use_bn_sync: False
|
88 |
+
use_bnb_linear: None
|
89 |
+
val_data: None
|
90 |
+
val_frequency: 1
|
91 |
+
val_num_samples: None
|
92 |
+
wandb: True
|
93 |
+
wandb_notes:
|
94 |
+
wandb_project_name: datacomp-xlarge-CL
|
95 |
+
warmup: 2000
|
96 |
+
wd: 0.2
|
97 |
+
workers: 4
|
98 |
+
world_size: 64
|
99 |
+
zeroshot_frequency: 1
|
params/2020.yaml
ADDED
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
accum_freq: 1
|
2 |
+
aug_cfg: {}
|
3 |
+
batch_size: 704
|
4 |
+
beta1: 0.9
|
5 |
+
beta2: 0.98
|
6 |
+
checkpoint_path: /tmp/restart_2020/datacomp_xlarge-basic_cumulative_seq_2020/checkpoints
|
7 |
+
coca_caption_loss_weight: 2.0
|
8 |
+
coca_contrastive_loss_weight: 1.0
|
9 |
+
copy_codebase: False
|
10 |
+
csv_caption_key: title
|
11 |
+
csv_img_key: filepath
|
12 |
+
csv_separator:
|
13 |
+
dataset_resampled: True
|
14 |
+
dataset_type: webdataset
|
15 |
+
ddp_static_graph: True
|
16 |
+
debug: False
|
17 |
+
decay_fraction: 0.2
|
18 |
+
delete_previous_checkpoint: False
|
19 |
+
device: cuda:0
|
20 |
+
dist_backend: nccl
|
21 |
+
dist_url: env://
|
22 |
+
distill: False
|
23 |
+
distill_model: None
|
24 |
+
distill_pretrained: None
|
25 |
+
distributed: True
|
26 |
+
epochs: 32
|
27 |
+
epochs_cooldown: None
|
28 |
+
eps: 1e-06
|
29 |
+
force_custom_text: False
|
30 |
+
force_image_size: None
|
31 |
+
force_patch_dropout: None
|
32 |
+
force_qk_norm: False
|
33 |
+
force_qk_norm_eps: 1e-05
|
34 |
+
force_quick_gelu: False
|
35 |
+
gather_with_grad: True
|
36 |
+
grad_checkpointing: True
|
37 |
+
grad_clip_norm: None
|
38 |
+
horovod: False
|
39 |
+
image_mean: None
|
40 |
+
image_std: None
|
41 |
+
imagenet_v2: None
|
42 |
+
imagenet_val: ../imagenet_validation
|
43 |
+
is_iteration_based: True
|
44 |
+
local_loss: True
|
45 |
+
local_rank: 0
|
46 |
+
lock_image: False
|
47 |
+
lock_image_freeze_bn_stats: False
|
48 |
+
lock_image_unlocked_groups: 0
|
49 |
+
lock_text: False
|
50 |
+
lock_text_freeze_layer_norm: False
|
51 |
+
lock_text_unlocked_layers: 0
|
52 |
+
log_every_n_steps: 100
|
53 |
+
log_level: 20
|
54 |
+
log_local: False
|
55 |
+
log_path: /tmp/restart_2020/datacomp_xlarge-basic_cumulative_seq_2020/out.log
|
56 |
+
logs: /tmp/restart_2020
|
57 |
+
lr: 0.001
|
58 |
+
lr_cooldown_end: 0.0
|
59 |
+
lr_cooldown_power: 1.0
|
60 |
+
lr_scheduler: cosine
|
61 |
+
max_iterations: 20500
|
62 |
+
model: ViT-B-16
|
63 |
+
name: datacomp_xlarge-basic_cumulative_seq_2020
|
64 |
+
new_run: True
|
65 |
+
no_set_device_rank: False
|
66 |
+
precision: amp
|
67 |
+
pretrained:
|
68 |
+
pretrained_image: False
|
69 |
+
rank: 0
|
70 |
+
remote_sync: xlarge_CL_basic_filter/restart_2020/
|
71 |
+
remote_sync_frequency: 300
|
72 |
+
remote_sync_protocol: s3
|
73 |
+
report_to: wandb
|
74 |
+
resume: xlarge_CL_basic_filter/restart_2019/datacomp_xlarge-basic_cumulative_seq_2019/checkpoints/epoch_4.pt
|
75 |
+
save_frequency: 1
|
76 |
+
save_most_recent: True
|
77 |
+
seed: 0
|
78 |
+
skip_scheduler: False
|
79 |
+
tensorboard: False
|
80 |
+
tensorboard_path:
|
81 |
+
torchcompile: False
|
82 |
+
torchscript: False
|
83 |
+
trace: False
|
84 |
+
train_data: xlarge_basic_filter/2020/1/{00000000..00000703}.tar::xlarge_basic_filter/2020/10/{00000000..00000644}.tar::xlarge_basic_filter/2020/11/{00000000..00000703}.tar::xlarge_basic_filter/2020/12/{00000000..00000703}.tar::xlarge_basic_filter/2020/13/{00000000..00000767}.tar::xlarge_basic_filter/2020/14/{00000000..00000727}.tar::xlarge_basic_filter/2020/15/{00000000..00000703}.tar::xlarge_basic_filter/2020/16/{00000000..00000676}.tar::xlarge_basic_filter/2020/17/{00000000..00000639}.tar::xlarge_basic_filter/2020/18/{00000000..00000759}.tar::xlarge_basic_filter/2020/19/{00000000..00000718}.tar::xlarge_basic_filter/2020/2/{00000000..00000767}.tar::xlarge_basic_filter/2020/20/{00000000..00000703}.tar::xlarge_basic_filter/2020/21/{00000000..00000639}.tar::xlarge_basic_filter/2020/22/{00000000..00000639}.tar::xlarge_basic_filter/2020/23/{00000000..00000767}.tar::xlarge_basic_filter/2020/24/{00000000..00000514}.tar::xlarge_basic_filter/2020/25/{00000000..00000511}.tar::xlarge_basic_filter/2020/26/{00000000..00000063}.tar::xlarge_basic_filter/2020/27/{00000000..00000165}.tar::xlarge_basic_filter/2020/28/{00000000..00000831}.tar::xlarge_basic_filter/2020/29/{00000000..00000831}.tar::xlarge_basic_filter/2020/3/{00000000..00000766}.tar::xlarge_basic_filter/2020/30/{00000000..00000831}.tar::xlarge_basic_filter/2020/31/{00000000..00000831}.tar::xlarge_basic_filter/2020/32/{00000000..00000831}.tar::xlarge_basic_filter/2020/33/{00000000..00000831}.tar::xlarge_basic_filter/2020/34/{00000000..00000831}.tar::xlarge_basic_filter/2020/35/{00000000..00000831}.tar::xlarge_basic_filter/2020/36/{00000000..00000831}.tar::xlarge_basic_filter/2020/37/{00000000..00000831}.tar::xlarge_basic_filter/2020/38/{00000000..00000767}.tar::xlarge_basic_filter/2020/39/{00000000..00000760}.tar::xlarge_basic_filter/2020/4/{00000000..00000719}.tar::xlarge_basic_filter/2020/40/{00000000..00000719}.tar::xlarge_basic_filter/2020/41/{00000000..00000639}.tar::xlarge_basic_filter/2020/42/{00000000..00000726}.tar::xlarge_basic_filter/2020/43/{00000000..00000767}.tar::xlarge_basic_filter/2020/44/{00000000..00000795}.tar::xlarge_basic_filter/2020/45/{00000000..00000759}.tar::xlarge_basic_filter/2020/46/{00000000..00000759}.tar::xlarge_basic_filter/2020/47/{00000000..00000759}.tar::xlarge_basic_filter/2020/48/{00000000..00000839}.tar::xlarge_basic_filter/2020/49/{00000000..00000831}.tar::xlarge_basic_filter/2020/5/{00000000..00000767}.tar::xlarge_basic_filter/2020/50/{00000000..00000839}.tar::xlarge_basic_filter/2020/51/{00000000..00000831}.tar::xlarge_basic_filter/2020/52/{00000000..00000839}.tar::xlarge_basic_filter/2020/53/{00000000..00001323}.tar::xlarge_basic_filter/2020/54/{00000000..00001279}.tar::xlarge_basic_filter/2020/55/{00000000..00000403}.tar::xlarge_basic_filter/2020/56/{00000000..00000401}.tar::xlarge_basic_filter/2020/57/{00000000..00000430}.tar::xlarge_basic_filter/2020/58/{00000000..00000383}.tar::xlarge_basic_filter/2020/59/{00000000..00000399}.tar::xlarge_basic_filter/2020/6/{00000000..00000767}.tar::xlarge_basic_filter/2020/7/{00000000..00000769}.tar::xlarge_basic_filter/2020/8/{00000000..00000767}.tar::xlarge_basic_filter/2020/9/{00000000..00000703}.tar
|
85 |
+
train_data_upsampling_factors: None
|
86 |
+
train_num_samples: None
|
87 |
+
use_bn_sync: False
|
88 |
+
use_bnb_linear: None
|
89 |
+
val_data: None
|
90 |
+
val_frequency: 1
|
91 |
+
val_num_samples: None
|
92 |
+
wandb: True
|
93 |
+
wandb_notes:
|
94 |
+
wandb_project_name: datacomp-xlarge-CL
|
95 |
+
warmup: 2000
|
96 |
+
wd: 0.2
|
97 |
+
workers: 4
|
98 |
+
world_size: 128
|
99 |
+
zeroshot_frequency: 1
|
params/2021.yaml
ADDED
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
accum_freq: 1
|
2 |
+
aug_cfg: {}
|
3 |
+
batch_size: 704
|
4 |
+
beta1: 0.9
|
5 |
+
beta2: 0.98
|
6 |
+
checkpoint_path: /tmp/restart_2021/datacomp_xlarge-basic_cumulative_seq_2021/checkpoints
|
7 |
+
coca_caption_loss_weight: 2.0
|
8 |
+
coca_contrastive_loss_weight: 1.0
|
9 |
+
copy_codebase: False
|
10 |
+
csv_caption_key: title
|
11 |
+
csv_img_key: filepath
|
12 |
+
csv_separator:
|
13 |
+
dataset_resampled: True
|
14 |
+
dataset_type: webdataset
|
15 |
+
ddp_static_graph: True
|
16 |
+
debug: False
|
17 |
+
decay_fraction: 0.2
|
18 |
+
delete_previous_checkpoint: False
|
19 |
+
device: cuda:0
|
20 |
+
dist_backend: nccl
|
21 |
+
dist_url: env://
|
22 |
+
distill: False
|
23 |
+
distill_model: None
|
24 |
+
distill_pretrained: None
|
25 |
+
distributed: True
|
26 |
+
epochs: 32
|
27 |
+
epochs_cooldown: None
|
28 |
+
eps: 1e-06
|
29 |
+
force_custom_text: False
|
30 |
+
force_image_size: None
|
31 |
+
force_patch_dropout: None
|
32 |
+
force_qk_norm: False
|
33 |
+
force_qk_norm_eps: 1e-05
|
34 |
+
force_quick_gelu: False
|
35 |
+
gather_with_grad: True
|
36 |
+
grad_checkpointing: True
|
37 |
+
grad_clip_norm: None
|
38 |
+
horovod: False
|
39 |
+
image_mean: None
|
40 |
+
image_std: None
|
41 |
+
imagenet_v2: None
|
42 |
+
imagenet_val: ../imagenet_validation
|
43 |
+
is_iteration_based: True
|
44 |
+
local_loss: True
|
45 |
+
local_rank: 0
|
46 |
+
lock_image: False
|
47 |
+
lock_image_freeze_bn_stats: False
|
48 |
+
lock_image_unlocked_groups: 0
|
49 |
+
lock_text: False
|
50 |
+
lock_text_freeze_layer_norm: False
|
51 |
+
lock_text_unlocked_layers: 0
|
52 |
+
log_every_n_steps: 100
|
53 |
+
log_level: 20
|
54 |
+
log_local: False
|
55 |
+
log_path: /tmp/restart_2021/datacomp_xlarge-basic_cumulative_seq_2021/out.log
|
56 |
+
logs: /tmp/restart_2021
|
57 |
+
lr: 0.001
|
58 |
+
lr_cooldown_end: 0.0
|
59 |
+
lr_cooldown_power: 1.0
|
60 |
+
lr_scheduler: cosine
|
61 |
+
max_iterations: 20500
|
62 |
+
model: ViT-B-16
|
63 |
+
name: datacomp_xlarge-basic_cumulative_seq_2021
|
64 |
+
new_run: True
|
65 |
+
no_set_device_rank: False
|
66 |
+
precision: amp
|
67 |
+
pretrained:
|
68 |
+
pretrained_image: False
|
69 |
+
rank: 0
|
70 |
+
remote_sync: xlarge_CL_basic_filter/restart_2021/
|
71 |
+
remote_sync_frequency: 300
|
72 |
+
remote_sync_protocol: s3
|
73 |
+
report_to: wandb
|
74 |
+
resume: xlarge_CL_basic_filter/restart_2020/datacomp_xlarge-basic_cumulative_seq_2020/checkpoints/epoch_5.pt
|
75 |
+
save_frequency: 1
|
76 |
+
save_most_recent: True
|
77 |
+
seed: 0
|
78 |
+
skip_scheduler: False
|
79 |
+
tensorboard: False
|
80 |
+
tensorboard_path:
|
81 |
+
torchcompile: False
|
82 |
+
torchscript: False
|
83 |
+
trace: False
|
84 |
+
train_data: xlarge_basic_filter/2021/1/{00000000..00000831}.tar::xlarge_basic_filter/2021/10/{00000000..00000767}.tar::xlarge_basic_filter/2021/11/{00000000..00000839}.tar::xlarge_basic_filter/2021/12/{00000000..00000831}.tar::xlarge_basic_filter/2021/13/{00000000..00000895}.tar::xlarge_basic_filter/2021/14/{00000000..00000877}.tar::xlarge_basic_filter/2021/15/{00000000..00000799}.tar::xlarge_basic_filter/2021/16/{00000000..00000770}.tar::xlarge_basic_filter/2021/17/{00000000..00000767}.tar::xlarge_basic_filter/2021/18/{00000000..00000895}.tar::xlarge_basic_filter/2021/19/{00000000..00000839}.tar::xlarge_basic_filter/2021/2/{00000000..00000895}.tar::xlarge_basic_filter/2021/20/{00000000..00000799}.tar::xlarge_basic_filter/2021/21/{00000000..00000759}.tar::xlarge_basic_filter/2021/22/{00000000..00000757}.tar::xlarge_basic_filter/2021/23/{00000000..00000919}.tar::xlarge_basic_filter/2021/24/{00000000..00000599}.tar::xlarge_basic_filter/2021/25/{00000000..00000599}.tar::xlarge_basic_filter/2021/26/{00000000..00000079}.tar::xlarge_basic_filter/2021/27/{00000000..00000191}.tar::xlarge_basic_filter/2021/28/{00000000..00000999}.tar::xlarge_basic_filter/2021/29/{00000000..00000999}.tar::xlarge_basic_filter/2021/3/{00000000..00000894}.tar::xlarge_basic_filter/2021/30/{00000000..00000959}.tar::xlarge_basic_filter/2021/31/{00000000..00000959}.tar::xlarge_basic_filter/2021/32/{00000000..00000959}.tar::xlarge_basic_filter/2021/33/{00000000..00000959}.tar::xlarge_basic_filter/2021/34/{00000000..00000959}.tar::xlarge_basic_filter/2021/35/{00000000..00000959}.tar::xlarge_basic_filter/2021/36/{00000000..00000999}.tar::xlarge_basic_filter/2021/37/{00000000..00001020}.tar::xlarge_basic_filter/2021/38/{00000000..00000879}.tar::xlarge_basic_filter/2021/39/{00000000..00000841}.tar::xlarge_basic_filter/2021/4/{00000000..00000842}.tar::xlarge_basic_filter/2021/40/{00000000..00000862}.tar::xlarge_basic_filter/2021/41/{00000000..00000767}.tar::xlarge_basic_filter/2021/42/{00000000..00000895}.tar::xlarge_basic_filter/2021/43/{00000000..00000950}.tar::xlarge_basic_filter/2021/44/{00000000..00000919}.tar::xlarge_basic_filter/2021/45/{00000000..00000893}.tar::xlarge_basic_filter/2021/46/{00000000..00000879}.tar::xlarge_basic_filter/2021/47/{00000000..00000902}.tar::xlarge_basic_filter/2021/48/{00000000..00000999}.tar::xlarge_basic_filter/2021/49/{00000000..00000996}.tar::xlarge_basic_filter/2021/5/{00000000..00000895}.tar::xlarge_basic_filter/2021/50/{00000000..00001014}.tar::xlarge_basic_filter/2021/51/{00000000..00001021}.tar::xlarge_basic_filter/2021/52/{00000000..00000989}.tar::xlarge_basic_filter/2021/53/{00000000..00001559}.tar::xlarge_basic_filter/2021/54/{00000000..00001511}.tar::xlarge_basic_filter/2021/55/{00000000..00000511}.tar::xlarge_basic_filter/2021/56/{00000000..00000511}.tar::xlarge_basic_filter/2021/57/{00000000..00000511}.tar::xlarge_basic_filter/2021/58/{00000000..00000486}.tar::xlarge_basic_filter/2021/59/{00000000..00000510}.tar::xlarge_basic_filter/2021/6/{00000000..00000879}.tar::xlarge_basic_filter/2021/7/{00000000..00000919}.tar::xlarge_basic_filter/2021/8/{00000000..00000879}.tar::xlarge_basic_filter/2021/9/{00000000..00000775}.tar
|
85 |
+
train_data_upsampling_factors: None
|
86 |
+
train_num_samples: None
|
87 |
+
use_bn_sync: False
|
88 |
+
use_bnb_linear: None
|
89 |
+
val_data: None
|
90 |
+
val_frequency: 1
|
91 |
+
val_num_samples: None
|
92 |
+
wandb: True
|
93 |
+
wandb_notes:
|
94 |
+
wandb_project_name: datacomp-xlarge-CL
|
95 |
+
warmup: 2000
|
96 |
+
wd: 0.2
|
97 |
+
workers: 4
|
98 |
+
world_size: 128
|
99 |
+
zeroshot_frequency: 1
|
params/2022.yaml
ADDED
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
accum_freq: 1
|
2 |
+
aug_cfg: {}
|
3 |
+
batch_size: 704
|
4 |
+
beta1: 0.9
|
5 |
+
beta2: 0.98
|
6 |
+
checkpoint_path: /tmp/restart_2022/datacomp_xlarge-basic_cumulative_seq_2022/checkpoints
|
7 |
+
coca_caption_loss_weight: 2.0
|
8 |
+
coca_contrastive_loss_weight: 1.0
|
9 |
+
copy_codebase: False
|
10 |
+
csv_caption_key: title
|
11 |
+
csv_img_key: filepath
|
12 |
+
csv_separator:
|
13 |
+
dataset_resampled: True
|
14 |
+
dataset_type: webdataset
|
15 |
+
ddp_static_graph: True
|
16 |
+
debug: False
|
17 |
+
decay_fraction: 0.2
|
18 |
+
delete_previous_checkpoint: False
|
19 |
+
device: cuda:0
|
20 |
+
dist_backend: nccl
|
21 |
+
dist_url: env://
|
22 |
+
distill: False
|
23 |
+
distill_model: None
|
24 |
+
distill_pretrained: None
|
25 |
+
distributed: True
|
26 |
+
epochs: 32
|
27 |
+
epochs_cooldown: None
|
28 |
+
eps: 1e-06
|
29 |
+
force_custom_text: False
|
30 |
+
force_image_size: None
|
31 |
+
force_patch_dropout: None
|
32 |
+
force_qk_norm: False
|
33 |
+
force_qk_norm_eps: 1e-05
|
34 |
+
force_quick_gelu: False
|
35 |
+
gather_with_grad: True
|
36 |
+
grad_checkpointing: True
|
37 |
+
grad_clip_norm: None
|
38 |
+
horovod: False
|
39 |
+
image_mean: None
|
40 |
+
image_std: None
|
41 |
+
imagenet_v2: None
|
42 |
+
imagenet_val: ../imagenet_validation
|
43 |
+
is_iteration_based: True
|
44 |
+
local_loss: True
|
45 |
+
local_rank: 0
|
46 |
+
lock_image: False
|
47 |
+
lock_image_freeze_bn_stats: False
|
48 |
+
lock_image_unlocked_groups: 0
|
49 |
+
lock_text: False
|
50 |
+
lock_text_freeze_layer_norm: False
|
51 |
+
lock_text_unlocked_layers: 0
|
52 |
+
log_every_n_steps: 100
|
53 |
+
log_level: 20
|
54 |
+
log_local: False
|
55 |
+
log_path: /tmp/restart_2022/datacomp_xlarge-basic_cumulative_seq_2022/out.log
|
56 |
+
logs: /tmp/restart_2022
|
57 |
+
lr: 0.001
|
58 |
+
lr_cooldown_end: 0.0
|
59 |
+
lr_cooldown_power: 1.0
|
60 |
+
lr_scheduler: cosine
|
61 |
+
max_iterations: 20500
|
62 |
+
model: ViT-B-16
|
63 |
+
name: datacomp_xlarge-basic_cumulative_seq_2022
|
64 |
+
new_run: True
|
65 |
+
no_set_device_rank: False
|
66 |
+
precision: amp
|
67 |
+
pretrained:
|
68 |
+
pretrained_image: False
|
69 |
+
rank: 0
|
70 |
+
remote_sync: xlarge_CL_basic_filter/restart_2022/
|
71 |
+
remote_sync_frequency: 300
|
72 |
+
remote_sync_protocol: s3
|
73 |
+
report_to: wandb
|
74 |
+
resume: xlarge_CL_basic_filter/restart_2021/datacomp_xlarge-basic_cumulative_seq_2021/checkpoints/epoch_4.pt
|
75 |
+
save_frequency: 1
|
76 |
+
save_most_recent: True
|
77 |
+
seed: 0
|
78 |
+
skip_scheduler: False
|
79 |
+
tensorboard: False
|
80 |
+
tensorboard_path:
|
81 |
+
torchcompile: False
|
82 |
+
torchscript: False
|
83 |
+
trace: False
|
84 |
+
train_data: xlarge_basic_filter/2022/1/{00000000..00000639}.tar::xlarge_basic_filter/2022/10/{00000000..00000575}.tar::xlarge_basic_filter/2022/11/{00000000..00000639}.tar::xlarge_basic_filter/2022/12/{00000000..00000639}.tar::xlarge_basic_filter/2022/13/{00000000..00000703}.tar::xlarge_basic_filter/2022/14/{00000000..00000639}.tar::xlarge_basic_filter/2022/15/{00000000..00000639}.tar::xlarge_basic_filter/2022/16/{00000000..00000599}.tar::xlarge_basic_filter/2022/17/{00000000..00000571}.tar::xlarge_basic_filter/2022/18/{00000000..00000703}.tar::xlarge_basic_filter/2022/19/{00000000..00000639}.tar::xlarge_basic_filter/2022/2/{00000000..00000698}.tar::xlarge_basic_filter/2022/20/{00000000..00000599}.tar::xlarge_basic_filter/2022/21/{00000000..00000599}.tar::xlarge_basic_filter/2022/22/{00000000..00000559}.tar::xlarge_basic_filter/2022/23/{00000000..00000703}.tar::xlarge_basic_filter/2022/24/{00000000..00000447}.tar::xlarge_basic_filter/2022/25/{00000000..00000447}.tar::xlarge_basic_filter/2022/26/{00000000..00000063}.tar::xlarge_basic_filter/2022/27/{00000000..00000127}.tar::xlarge_basic_filter/2022/28/{00000000..00000767}.tar::xlarge_basic_filter/2022/29/{00000000..00000767}.tar::xlarge_basic_filter/2022/3/{00000000..00000698}.tar::xlarge_basic_filter/2022/30/{00000000..00000740}.tar::xlarge_basic_filter/2022/31/{00000000..00000703}.tar::xlarge_basic_filter/2022/32/{00000000..00000703}.tar::xlarge_basic_filter/2022/33/{00000000..00000732}.tar::xlarge_basic_filter/2022/34/{00000000..00000703}.tar::xlarge_basic_filter/2022/35/{00000000..00000713}.tar::xlarge_basic_filter/2022/36/{00000000..00000767}.tar::xlarge_basic_filter/2022/37/{00000000..00000767}.tar::xlarge_basic_filter/2022/38/{00000000..00000703}.tar::xlarge_basic_filter/2022/39/{00000000..00000640}.tar::xlarge_basic_filter/2022/4/{00000000..00000640}.tar::xlarge_basic_filter/2022/40/{00000000..00000639}.tar::xlarge_basic_filter/2022/41/{00000000..00000575}.tar::xlarge_basic_filter/2022/42/{00000000..00000649}.tar::xlarge_basic_filter/2022/43/{00000000..00000689}.tar::xlarge_basic_filter/2022/44/{00000000..00000703}.tar::xlarge_basic_filter/2022/45/{00000000..00000703}.tar::xlarge_basic_filter/2022/46/{00000000..00000679}.tar::xlarge_basic_filter/2022/47/{00000000..00000703}.tar::xlarge_basic_filter/2022/48/{00000000..00000767}.tar::xlarge_basic_filter/2022/49/{00000000..00000767}.tar::xlarge_basic_filter/2022/5/{00000000..00000703}.tar::xlarge_basic_filter/2022/50/{00000000..00000767}.tar::xlarge_basic_filter/2022/51/{00000000..00000767}.tar::xlarge_basic_filter/2022/52/{00000000..00000756}.tar::xlarge_basic_filter/2022/53/{00000000..00001159}.tar::xlarge_basic_filter/2022/54/{00000000..00001151}.tar::xlarge_basic_filter/2022/55/{00000000..00000359}.tar::xlarge_basic_filter/2022/56/{00000000..00000359}.tar::xlarge_basic_filter/2022/57/{00000000..00000383}.tar::xlarge_basic_filter/2022/58/{00000000..00000383}.tar::xlarge_basic_filter/2022/59/{00000000..00000359}.tar::xlarge_basic_filter/2022/6/{00000000..00000702}.tar::xlarge_basic_filter/2022/7/{00000000..00000703}.tar::xlarge_basic_filter/2022/8/{00000000..00000699}.tar::xlarge_basic_filter/2022/9/{00000000..00000617}.tar
|
85 |
+
train_data_upsampling_factors: None
|
86 |
+
train_num_samples: None
|
87 |
+
use_bn_sync: False
|
88 |
+
use_bnb_linear: None
|
89 |
+
val_data: None
|
90 |
+
val_frequency: 1
|
91 |
+
val_num_samples: None
|
92 |
+
wandb: True
|
93 |
+
wandb_notes:
|
94 |
+
wandb_project_name: datacomp-xlarge-CL
|
95 |
+
warmup: 2000
|
96 |
+
wd: 0.2
|
97 |
+
workers: 4
|
98 |
+
world_size: 128
|
99 |
+
zeroshot_frequency: 1
|