Veldrovive
commited on
Commit
•
f47bf78
1
Parent(s):
1c75b29
Upload text_conditioned_config.json
Browse files- text_conditioned_config.json +99 -0
text_conditioned_config.json
ADDED
@@ -0,0 +1,99 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"seed": 1,
|
3 |
+
|
4 |
+
"decoder": {
|
5 |
+
"unets": [
|
6 |
+
{
|
7 |
+
"dim": 416,
|
8 |
+
"cond_dim": 512,
|
9 |
+
"image_embed_dim": 768,
|
10 |
+
"channels": 3,
|
11 |
+
"dim_mults": [1, 2, 3, 4],
|
12 |
+
"num_resnet_blocks": 4,
|
13 |
+
"attn_heads": 8,
|
14 |
+
"attn_dim_head": 64,
|
15 |
+
"sparse_attn": true,
|
16 |
+
"memory_efficient": true
|
17 |
+
}
|
18 |
+
],
|
19 |
+
"clip": {
|
20 |
+
"make": "openai",
|
21 |
+
"model": "ViT-L/14"
|
22 |
+
},
|
23 |
+
"image_sizes": [64],
|
24 |
+
"channels": 3,
|
25 |
+
"timesteps": 1000,
|
26 |
+
"loss_type": "l2",
|
27 |
+
"beta_schedule": ["cosine"],
|
28 |
+
"condition_on_text_encodings": true,
|
29 |
+
"learned_variance": true
|
30 |
+
},
|
31 |
+
"data": {
|
32 |
+
"webdataset_base_url": "pipe:aws s3 cp --quiet s3://s-datasets/laion-aesthetic/data/laion2B-en-aesthetic/{}.tar -",
|
33 |
+
"img_embeddings_url": "s3://s-datasets/laion-aesthetic/ordered_embeddings/",
|
34 |
+
"num_workers": 4,
|
35 |
+
"batch_size": 22,
|
36 |
+
"start_shard": 0,
|
37 |
+
"end_shard": 5247,
|
38 |
+
"shard_width": 5,
|
39 |
+
"index_width": 4,
|
40 |
+
"splits": {
|
41 |
+
"train": 0.75,
|
42 |
+
"val": 0.15,
|
43 |
+
"test": 0.1
|
44 |
+
},
|
45 |
+
"shuffle_train": false,
|
46 |
+
"resample_train": true,
|
47 |
+
"preprocessing": {
|
48 |
+
"RandomResizedCrop": {
|
49 |
+
"size": [64, 64],
|
50 |
+
"scale": [0.75, 1.0],
|
51 |
+
"ratio": [1.0, 1.0]
|
52 |
+
},
|
53 |
+
"ToTensor": true
|
54 |
+
}
|
55 |
+
},
|
56 |
+
"train": {
|
57 |
+
"epochs": 40,
|
58 |
+
"lr":1e-4,
|
59 |
+
"wd": 0.01,
|
60 |
+
"max_grad_norm": 0.5,
|
61 |
+
"save_every_n_samples": 200000,
|
62 |
+
"n_sample_images": 10,
|
63 |
+
"device": "cuda:0",
|
64 |
+
"epoch_samples": 2000000,
|
65 |
+
"validation_samples": 100000,
|
66 |
+
"use_ema": true,
|
67 |
+
"ema_beta": 0.99,
|
68 |
+
"save_all": false,
|
69 |
+
"save_latest": true,
|
70 |
+
"save_best": true,
|
71 |
+
"unet_training_mask": [true]
|
72 |
+
},
|
73 |
+
"evaluate": {
|
74 |
+
"n_evaluation_samples": 30,
|
75 |
+
"FID": {
|
76 |
+
"feature": 64
|
77 |
+
},
|
78 |
+
"LPIPS": {
|
79 |
+
"net_type": "vgg",
|
80 |
+
"reduction": "mean"
|
81 |
+
}
|
82 |
+
},
|
83 |
+
"tracker": {
|
84 |
+
"tracker_type": "wandb",
|
85 |
+
"data_path": ".tracker-data",
|
86 |
+
"verbose": true,
|
87 |
+
|
88 |
+
"wandb_entity": "nousr_laion",
|
89 |
+
"wandb_project": "dalle2_train_decoder"
|
90 |
+
},
|
91 |
+
"load": {
|
92 |
+
"source": null,
|
93 |
+
|
94 |
+
"run_path": "",
|
95 |
+
"file_path": "",
|
96 |
+
|
97 |
+
"resume": false
|
98 |
+
}
|
99 |
+
}
|