Upload decoder_config.json with huggingface_hub
Browse files- decoder_config.json +14 -14
decoder_config.json
CHANGED
@@ -4,7 +4,7 @@
|
|
4 |
"decoder": {
|
5 |
"unets": [
|
6 |
{
|
7 |
-
"dim":
|
8 |
"cond_dim": 512,
|
9 |
"image_embed_dim": 768,
|
10 |
"text_embed_dim": 768,
|
@@ -15,7 +15,8 @@
|
|
15 |
"attn_heads": 8,
|
16 |
"attn_dim_head": 64,
|
17 |
"sparse_attn": true,
|
18 |
-
"memory_efficient": true
|
|
|
19 |
}
|
20 |
],
|
21 |
"clip": {
|
@@ -31,9 +32,9 @@
|
|
31 |
},
|
32 |
"data": {
|
33 |
"webdataset_base_url": "pipe:aws s3 cp --quiet s3://s-datasets/laion-aesthetic/data/laion2B-en-aesthetic/{}.tar -",
|
34 |
-
"
|
35 |
"num_workers": 12,
|
36 |
-
"batch_size":
|
37 |
"start_shard": 0,
|
38 |
"end_shard": 5247,
|
39 |
"shard_width": 5,
|
@@ -59,10 +60,10 @@
|
|
59 |
"lr":1e-4,
|
60 |
"wd": 0.01,
|
61 |
"max_grad_norm": 0.5,
|
62 |
-
"save_every_n_samples":
|
63 |
"n_sample_images": 10,
|
64 |
"device": "cuda:0",
|
65 |
-
"epoch_samples":
|
66 |
"validation_samples": 100000,
|
67 |
"use_ema": true,
|
68 |
"ema_beta": 0.99,
|
@@ -88,19 +89,17 @@
|
|
88 |
"log": {
|
89 |
"log_type": "wandb",
|
90 |
|
91 |
-
"wandb_entity": "
|
92 |
"wandb_project": "dalle2_train_decoder",
|
93 |
-
|
94 |
-
"
|
95 |
-
|
96 |
-
"wandb_resume": true,
|
97 |
|
98 |
"verbose": true
|
99 |
},
|
100 |
|
101 |
"load": {
|
102 |
-
"
|
103 |
-
"
|
104 |
},
|
105 |
|
106 |
"save": [{
|
@@ -110,9 +109,10 @@
|
|
110 |
"save_to": "huggingface",
|
111 |
"huggingface_repo": "laion/DALLE2-PyTorch",
|
112 |
|
113 |
-
"save_latest_to": "
|
114 |
|
115 |
"save_type": "model"
|
116 |
}]
|
117 |
}
|
118 |
}
|
|
|
|
4 |
"decoder": {
|
5 |
"unets": [
|
6 |
{
|
7 |
+
"dim": 256,
|
8 |
"cond_dim": 512,
|
9 |
"image_embed_dim": 768,
|
10 |
"text_embed_dim": 768,
|
|
|
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": {
|
|
|
32 |
},
|
33 |
"data": {
|
34 |
"webdataset_base_url": "pipe:aws s3 cp --quiet s3://s-datasets/laion-aesthetic/data/laion2B-en-aesthetic/{}.tar -",
|
35 |
+
"img_embeddings_url": "s3://s-datasets/laion-aesthetic/ordered_embeddings/",
|
36 |
"num_workers": 12,
|
37 |
+
"batch_size": 60,
|
38 |
"start_shard": 0,
|
39 |
"end_shard": 5247,
|
40 |
"shard_width": 5,
|
|
|
60 |
"lr":1e-4,
|
61 |
"wd": 0.01,
|
62 |
"max_grad_norm": 0.5,
|
63 |
+
"save_every_n_samples": 2000000,
|
64 |
"n_sample_images": 10,
|
65 |
"device": "cuda:0",
|
66 |
+
"epoch_samples": 40000000,
|
67 |
"validation_samples": 100000,
|
68 |
"use_ema": true,
|
69 |
"ema_beta": 0.99,
|
|
|
89 |
"log": {
|
90 |
"log_type": "wandb",
|
91 |
|
92 |
+
"wandb_entity": "rom1504",
|
93 |
"wandb_project": "dalle2_train_decoder",
|
94 |
+
"wandb_run_id": "3tmnv289",
|
95 |
+
"wandb_resume": true,
|
|
|
|
|
96 |
|
97 |
"verbose": true
|
98 |
},
|
99 |
|
100 |
"load": {
|
101 |
+
"source": "url",
|
102 |
+
"wandb_file_path": "https://api.wandb.ai/files/rom1504/dalle2_train_decoder/3tmnv289/latest.pth"
|
103 |
},
|
104 |
|
105 |
"save": [{
|
|
|
109 |
"save_to": "huggingface",
|
110 |
"huggingface_repo": "laion/DALLE2-PyTorch",
|
111 |
|
112 |
+
"save_latest_to": "decoder/small_32gpus/latest.pth",
|
113 |
|
114 |
"save_type": "model"
|
115 |
}]
|
116 |
}
|
117 |
}
|
118 |
+
|