Delete decoder/1.5B_laion2B_ddim/decoder_config.json
Browse files
decoder/1.5B_laion2B_ddim/decoder_config.json
DELETED
@@ -1,117 +0,0 @@
|
|
1 |
-
{
|
2 |
-
"seed": 1,
|
3 |
-
|
4 |
-
"decoder": {
|
5 |
-
"unets": [
|
6 |
-
{
|
7 |
-
"dim": 320,
|
8 |
-
"cond_dim": 512,
|
9 |
-
"image_embed_dim": 768,
|
10 |
-
"text_embed_dim": 768,
|
11 |
-
"cond_on_text_encodings": true,
|
12 |
-
"channels": 3,
|
13 |
-
"dim_mults": [1, 2, 3, 4],
|
14 |
-
"num_resnet_blocks": 4,
|
15 |
-
"attn_heads": 8,
|
16 |
-
"attn_dim_head": 64,
|
17 |
-
"sparse_attn": true,
|
18 |
-
"memory_efficient": true,
|
19 |
-
"self_attn": [false, true, true, true]
|
20 |
-
}
|
21 |
-
],
|
22 |
-
"clip": {
|
23 |
-
"make": "openai",
|
24 |
-
"model": "ViT-L/14"
|
25 |
-
},
|
26 |
-
"image_sizes": [64],
|
27 |
-
"channels": 3,
|
28 |
-
"timesteps": 1000,
|
29 |
-
"sample_timesteps": 250,
|
30 |
-
"loss_type": "l2",
|
31 |
-
"beta_schedule": ["cosine"],
|
32 |
-
"learned_variance": true
|
33 |
-
},
|
34 |
-
"data": {
|
35 |
-
"webdataset_base_url": "pipe:aws s3 cp --quiet s3://s-datasets/laion5b/laion2B-data/{}.tar -",
|
36 |
-
"num_workers": 12,
|
37 |
-
"batch_size": 26,
|
38 |
-
"start_shard": 0,
|
39 |
-
"end_shard": 231349,
|
40 |
-
"shard_width": 6,
|
41 |
-
"index_width": 4,
|
42 |
-
"splits": {
|
43 |
-
"train": 0.75,
|
44 |
-
"val": 0.15,
|
45 |
-
"test": 0.1
|
46 |
-
},
|
47 |
-
"shuffle_train": false,
|
48 |
-
"resample_train": true,
|
49 |
-
"preprocessing": {
|
50 |
-
"RandomResizedCrop": {
|
51 |
-
"size": [224, 224],
|
52 |
-
"scale": [0.75, 1.0],
|
53 |
-
"ratio": [1.0, 1.0]
|
54 |
-
},
|
55 |
-
"ToTensor": true
|
56 |
-
}
|
57 |
-
},
|
58 |
-
"train": {
|
59 |
-
"epochs": 1000,
|
60 |
-
"lr": 1e-4,
|
61 |
-
"wd": 0.01,
|
62 |
-
"max_grad_norm": 0.5,
|
63 |
-
"save_every_n_samples": 5000000,
|
64 |
-
"n_sample_images": 10,
|
65 |
-
"device": "cuda:0",
|
66 |
-
"epoch_samples": 10000000,
|
67 |
-
"validation_samples": 100000,
|
68 |
-
"use_ema": true,
|
69 |
-
"ema_beta": 0.99,
|
70 |
-
"save_all": false,
|
71 |
-
"save_latest": true,
|
72 |
-
"save_best": true,
|
73 |
-
"unet_training_mask": [true]
|
74 |
-
},
|
75 |
-
"evaluate": {
|
76 |
-
"n_evaluation_samples": 30,
|
77 |
-
"FID": {
|
78 |
-
"feature": 64
|
79 |
-
},
|
80 |
-
"LPIPS": {
|
81 |
-
"net_type": "vgg",
|
82 |
-
"reduction": "mean"
|
83 |
-
}
|
84 |
-
},
|
85 |
-
"tracker": {
|
86 |
-
"data_path": "/fsx/dalle2/.tracker-laion2B",
|
87 |
-
"overwrite_data_path": true,
|
88 |
-
|
89 |
-
"log": {
|
90 |
-
"log_type": "wandb",
|
91 |
-
|
92 |
-
"wandb_entity": "rom1504",
|
93 |
-
"wandb_project": "dalle2_train_decoder",
|
94 |
-
"wandb_resume": false,
|
95 |
-
|
96 |
-
"verbose": true
|
97 |
-
},
|
98 |
-
|
99 |
-
"load": {
|
100 |
-
"load_from": null
|
101 |
-
},
|
102 |
-
|
103 |
-
"save": [{
|
104 |
-
"save_to": "wandb"
|
105 |
-
},
|
106 |
-
{
|
107 |
-
"save_to": "huggingface",
|
108 |
-
"huggingface_repo": "laion/DALLE2-PyTorch",
|
109 |
-
"save_meta_to": "decoder/1.5B_laion2B_ddim/",
|
110 |
-
|
111 |
-
"save_latest_to": "decoder/1.5B_laion2B_ddim/latest.pth",
|
112 |
-
|
113 |
-
"save_type": "model"
|
114 |
-
}]
|
115 |
-
}
|
116 |
-
}
|
117 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|