iskhare commited on
Commit
cb83ea5
·
verified ·
1 Parent(s): 3c44b4f

Upload VAR/tokenizer/checkpoints artifacts

Browse files
VAR/tokenizer/checkpoints/epoch_040_tokenizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e1aefa735b638ba83db20d144c792db308d3a296ca468c0c1b9c9547ea8f9b0
3
+ size 436084970
VAR/tokenizer/checkpoints/resolved_config.json ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "command": [
3
+ "/mnt/research-common-h100-data/home/dylanlim/repos/MsTok/src/image_tokenization/training/train_var_tokenizer.py",
4
+ "--config",
5
+ "configs/baselines/var/tokenizer_imagenet256.yaml"
6
+ ],
7
+ "config": {
8
+ "adam_betas": [
9
+ 0.5,
10
+ 0.9
11
+ ],
12
+ "adaptive_epsilon": 0.0001,
13
+ "adaptive_max": 10000.0,
14
+ "adversarial_weight": 0.5,
15
+ "auto_resume": true,
16
+ "base_channels": 160,
17
+ "beta": 0.25,
18
+ "cache_dir": "/mnt/research-common-h100-data/home/dylanlim/repos/MsTok/data/cache",
19
+ "data_root": "/mnt/research-common-h100-data/home/dylanlim/datasets/imagenet",
20
+ "epochs": 100,
21
+ "eval_batch_size": 16,
22
+ "eval_checkpoint": "",
23
+ "eval_num_workers": 8,
24
+ "eval_output": "",
25
+ "expected_classes": 1000,
26
+ "expected_parameters": 108948355,
27
+ "expected_train_images": 1281167,
28
+ "expected_val_images": 50000,
29
+ "expected_world_size": 16,
30
+ "gan_start_epoch": 3,
31
+ "gradient_clip": 1.0,
32
+ "grid_epochs": [
33
+ 2,
34
+ 5,
35
+ 10,
36
+ 20,
37
+ 30,
38
+ 40,
39
+ 50,
40
+ 60,
41
+ 70,
42
+ 80,
43
+ 90,
44
+ 100
45
+ ],
46
+ "image_size": 256,
47
+ "latent_channels": 32,
48
+ "learning_rate": 0.0001,
49
+ "local_batch_size": 16,
50
+ "log_interval": 50,
51
+ "master_port": 29500,
52
+ "num_workers": 8,
53
+ "output_dir": "/mnt/research-common-h100-data/home/dylanlim/repos/MsTok/outputs",
54
+ "patch_nums": [
55
+ 1,
56
+ 2,
57
+ 3,
58
+ 4,
59
+ 5,
60
+ 6,
61
+ 8,
62
+ 10,
63
+ 13,
64
+ 16
65
+ ],
66
+ "perceptual_weight": 1.0,
67
+ "permanent_epochs": [
68
+ 40,
69
+ 60,
70
+ 80,
71
+ 100
72
+ ],
73
+ "precision": "bf16",
74
+ "primary_epoch": 100,
75
+ "quant_resi": 0.5,
76
+ "real_stats_cache": "/mnt/research-common-h100-data/home/dylanlim/repos/MsTok/data/cache/imagenet256_val_pytorch_fid_stats.npz",
77
+ "reconstruction_weight": 1.0,
78
+ "recovery_interval": 5000,
79
+ "resize_short": 288,
80
+ "resume": "",
81
+ "run_name": "var_imagenet256_v1",
82
+ "seed": 42,
83
+ "share_quant_resi": 4,
84
+ "stop_after_step": 0,
85
+ "usage_interval": 500,
86
+ "vocab_size": 4096,
87
+ "vq_weight": 1.0,
88
+ "wandb_enabled": true,
89
+ "wandb_entity": "mstok",
90
+ "wandb_group": "VAR-tokenizer",
91
+ "wandb_project": "baselines",
92
+ "weight_decay": 0.0
93
+ },
94
+ "environment": {
95
+ "cuda_runtime": "12.1",
96
+ "cudnn": 8902,
97
+ "lpips": "0.1.4",
98
+ "numpy": "1.26.4",
99
+ "pillow": "10.2.0",
100
+ "python": "3.11.15 (main, Jun 11 2026, 15:20:16) [GCC 14.3.0]",
101
+ "pytorch_fid": "0.3.0",
102
+ "scipy": "1.11.4",
103
+ "torch": "2.1.2+cu121",
104
+ "torchvision": "0.16.2",
105
+ "wandb": "0.17.9",
106
+ "world_size": 16
107
+ },
108
+ "git_revision": "961d40ec6dd4023af6605977b76c965a4681d13e"
109
+ }
VAR/tokenizer/checkpoints/tokenizer_imagenet256.yaml ADDED
@@ -0,0 +1,61 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ run_name: var_imagenet256_v1
2
+ output_dir: $MSTOK_OUTPUT_DIR
3
+ data_root: $IMAGENET_ROOT
4
+ cache_dir: $MSTOK_CACHE_DIR
5
+
6
+ image_size: 256
7
+ resize_short: 288
8
+ patch_nums: [1, 2, 3, 4, 5, 6, 8, 10, 13, 16]
9
+ vocab_size: 4096
10
+ latent_channels: 32
11
+ base_channels: 160
12
+ beta: 0.25
13
+ quant_resi: 0.5
14
+ share_quant_resi: 4
15
+ expected_parameters: 108948355
16
+
17
+ epochs: 100
18
+ local_batch_size: 16
19
+ expected_world_size: 16
20
+ num_workers: 8
21
+ seed: 42
22
+
23
+ learning_rate: 0.0001
24
+ adam_betas: [0.5, 0.9]
25
+ weight_decay: 0.0
26
+ gradient_clip: 1.0
27
+
28
+ reconstruction_weight: 1.0
29
+ vq_weight: 1.0
30
+ perceptual_weight: 1.0
31
+ adversarial_weight: 0.5
32
+ gan_start_epoch: 3
33
+ adaptive_epsilon: 0.0001
34
+ adaptive_max: 10000.0
35
+
36
+ precision: bf16
37
+ recovery_interval: 5000
38
+ permanent_epochs: [40, 60, 80, 100]
39
+ primary_epoch: 100
40
+ resume: ""
41
+ auto_resume: true
42
+ stop_after_step: 0
43
+
44
+ log_interval: 50
45
+ usage_interval: 500
46
+ grid_epochs: [2, 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100]
47
+ wandb_enabled: true
48
+ wandb_entity: mstok
49
+ wandb_project: baselines
50
+ wandb_group: VAR-tokenizer
51
+
52
+ eval_batch_size: 16
53
+ eval_num_workers: 8
54
+ eval_checkpoint: ""
55
+ eval_output: ""
56
+ real_stats_cache: $MSTOK_CACHE_DIR/imagenet256_val_pytorch_fid_stats.npz
57
+
58
+ expected_train_images: 1281167
59
+ expected_val_images: 50000
60
+ expected_classes: 1000
61
+ master_port: 29500