Upload config.json with huggingface_hub
Browse files- config.json +12 -17
config.json
CHANGED
@@ -1,5 +1,5 @@
|
|
1 |
{
|
2 |
-
"save_path": "
|
3 |
"model": {
|
4 |
"base_learning_rate": 8e-05,
|
5 |
"target": "latent_diffusion.ldm.models.diffusion.ddpm.LatentDiffusion",
|
@@ -18,7 +18,7 @@
|
|
18 |
48
|
19 |
],
|
20 |
"channels": 3,
|
21 |
-
"cond_stage_trainable":
|
22 |
"conditioning_key": "hybrid",
|
23 |
"monitor": "val/loss_simple_ema",
|
24 |
"unet_config": {
|
@@ -28,7 +28,7 @@
|
|
28 |
64,
|
29 |
48
|
30 |
],
|
31 |
-
"in_channels":
|
32 |
"out_channels": 4,
|
33 |
"model_channels": 192,
|
34 |
"attention_resolutions": [
|
@@ -44,9 +44,8 @@
|
|
44 |
5
|
45 |
],
|
46 |
"num_head_channels": 32,
|
47 |
-
"use_spatial_transformer":
|
48 |
-
"transformer_depth": 1
|
49 |
-
"context_dim": 768
|
50 |
}
|
51 |
},
|
52 |
"first_stage_config": {
|
@@ -76,20 +75,14 @@
|
|
76 |
}
|
77 |
}
|
78 |
},
|
79 |
-
"cond_stage_config":
|
80 |
-
"target": "latent_diffusion.ldm.modules.encoders.modules.GPTEmbedder",
|
81 |
-
"params": {
|
82 |
-
"n_embed": 768,
|
83 |
-
"n_layer": 12
|
84 |
-
}
|
85 |
-
}
|
86 |
}
|
87 |
},
|
88 |
"data": {
|
89 |
"target": "data.data_processing.datasets.DataModule",
|
90 |
"params": {
|
91 |
-
"batch_size":
|
92 |
-
"num_workers":
|
93 |
"wrap": false,
|
94 |
"shuffle": true,
|
95 |
"drop_last": true,
|
@@ -99,7 +92,9 @@
|
|
99 |
"train": {
|
100 |
"target": "data.data_processing.datasets.ActionsData",
|
101 |
"params": {
|
102 |
-
"data_csv_path": "train_dataset/
|
|
|
|
|
103 |
}
|
104 |
}
|
105 |
}
|
@@ -107,7 +102,7 @@
|
|
107 |
"lightning": {
|
108 |
"trainer": {
|
109 |
"benchmark": false,
|
110 |
-
"max_epochs":
|
111 |
"limit_val_batches": 0,
|
112 |
"accelerator": "gpu",
|
113 |
"gpus": 1,
|
|
|
1 |
{
|
2 |
+
"save_path": "saved_standard_challenging_context32_nocond_cont_cont_all_cont",
|
3 |
"model": {
|
4 |
"base_learning_rate": 8e-05,
|
5 |
"target": "latent_diffusion.ldm.models.diffusion.ddpm.LatentDiffusion",
|
|
|
18 |
48
|
19 |
],
|
20 |
"channels": 3,
|
21 |
+
"cond_stage_trainable": false,
|
22 |
"conditioning_key": "hybrid",
|
23 |
"monitor": "val/loss_simple_ema",
|
24 |
"unet_config": {
|
|
|
28 |
64,
|
29 |
48
|
30 |
],
|
31 |
+
"in_channels": 166,
|
32 |
"out_channels": 4,
|
33 |
"model_channels": 192,
|
34 |
"attention_resolutions": [
|
|
|
44 |
5
|
45 |
],
|
46 |
"num_head_channels": 32,
|
47 |
+
"use_spatial_transformer": false,
|
48 |
+
"transformer_depth": 1
|
|
|
49 |
}
|
50 |
},
|
51 |
"first_stage_config": {
|
|
|
75 |
}
|
76 |
}
|
77 |
},
|
78 |
+
"cond_stage_config": "__is_unconditional__"
|
|
|
|
|
|
|
|
|
|
|
|
|
79 |
}
|
80 |
},
|
81 |
"data": {
|
82 |
"target": "data.data_processing.datasets.DataModule",
|
83 |
"params": {
|
84 |
+
"batch_size": 256,
|
85 |
+
"num_workers": 16,
|
86 |
"wrap": false,
|
87 |
"shuffle": true,
|
88 |
"drop_last": true,
|
|
|
92 |
"train": {
|
93 |
"target": "data.data_processing.datasets.ActionsData",
|
94 |
"params": {
|
95 |
+
"data_csv_path": "train_dataset/train_dataset.target_frames.csv",
|
96 |
+
"normalization": "standard",
|
97 |
+
"context_length": 32
|
98 |
}
|
99 |
}
|
100 |
}
|
|
|
102 |
"lightning": {
|
103 |
"trainer": {
|
104 |
"benchmark": false,
|
105 |
+
"max_epochs": 6400,
|
106 |
"limit_val_batches": 0,
|
107 |
"accelerator": "gpu",
|
108 |
"gpus": 1,
|