add model_index.json; update configs
Browse files- model_index.json +24 -0
- transformer/config.json +5 -1
- vae/config.json +29 -3
model_index.json
ADDED
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_class_name": "AllegroPipeline",
|
3 |
+
"_diffusers_version": "0.31.0.dev0",
|
4 |
+
"scheduler": [
|
5 |
+
"diffusers",
|
6 |
+
"EulerAncestralDiscreteScheduler"
|
7 |
+
],
|
8 |
+
"text_encoder": [
|
9 |
+
"transformers",
|
10 |
+
"T5EncoderModel"
|
11 |
+
],
|
12 |
+
"tokenizer": [
|
13 |
+
"transformers",
|
14 |
+
"T5Tokenizer"
|
15 |
+
],
|
16 |
+
"transformer": [
|
17 |
+
"diffusers",
|
18 |
+
"AllegroTransformer3DModel"
|
19 |
+
],
|
20 |
+
"vae": [
|
21 |
+
"diffusers",
|
22 |
+
"AutoencoderKLAllegro"
|
23 |
+
]
|
24 |
+
}
|
transformer/config.json
CHANGED
@@ -1,6 +1,6 @@
|
|
1 |
{
|
2 |
"_class_name": "AllegroTransformer3DModel",
|
3 |
-
"_diffusers_version": "0.
|
4 |
"activation_fn": "gelu-approximate",
|
5 |
"attention_bias": true,
|
6 |
"attention_head_dim": 96,
|
@@ -25,12 +25,16 @@
|
|
25 |
"out_channels": 4,
|
26 |
"patch_size": 2,
|
27 |
"patch_size_t": 1,
|
|
|
28 |
"sa_attention_mode": "flash",
|
|
|
|
|
29 |
"sample_size": [
|
30 |
90,
|
31 |
160
|
32 |
],
|
33 |
"sample_size_t": 22,
|
|
|
34 |
"upcast_attention": false,
|
35 |
"use_additional_conditions": null,
|
36 |
"use_linear_projection": false,
|
|
|
1 |
{
|
2 |
"_class_name": "AllegroTransformer3DModel",
|
3 |
+
"_diffusers_version": "0.31.0.dev0",
|
4 |
"activation_fn": "gelu-approximate",
|
5 |
"attention_bias": true,
|
6 |
"attention_head_dim": 96,
|
|
|
25 |
"out_channels": 4,
|
26 |
"patch_size": 2,
|
27 |
"patch_size_t": 1,
|
28 |
+
"patch_size_temporal": 1,
|
29 |
"sa_attention_mode": "flash",
|
30 |
+
"sample_frames": 22,
|
31 |
+
"sample_height": 90,
|
32 |
"sample_size": [
|
33 |
90,
|
34 |
160
|
35 |
],
|
36 |
"sample_size_t": 22,
|
37 |
+
"sample_width": 160,
|
38 |
"upcast_attention": false,
|
39 |
"use_additional_conditions": null,
|
40 |
"use_linear_projection": false,
|
vae/config.json
CHANGED
@@ -1,6 +1,6 @@
|
|
1 |
{
|
2 |
-
"_class_name": "
|
3 |
-
"_diffusers_version": "0.
|
4 |
"act_fn": "silu",
|
5 |
"block_out_channels": [
|
6 |
128,
|
@@ -22,6 +22,12 @@
|
|
22 |
],
|
23 |
"chunk_len": 24,
|
24 |
"down_block_num": 4,
|
|
|
|
|
|
|
|
|
|
|
|
|
25 |
"force_upcast": true,
|
26 |
"in_channels": 3,
|
27 |
"latent_channels": 4,
|
@@ -31,10 +37,30 @@
|
|
31 |
"out_channels": 3,
|
32 |
"sample_size": 320,
|
33 |
"scale_factor": 0.13,
|
|
|
34 |
"t_over": 8,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
35 |
"tile_overlap": [
|
36 |
120,
|
37 |
80
|
38 |
],
|
39 |
-
"up_block_num": 4
|
|
|
|
|
|
|
|
|
|
|
|
|
40 |
}
|
|
|
1 |
{
|
2 |
+
"_class_name": "AutoencoderKLAllegro",
|
3 |
+
"_diffusers_version": "0.31.0.dev0",
|
4 |
"act_fn": "silu",
|
5 |
"block_out_channels": [
|
6 |
128,
|
|
|
22 |
],
|
23 |
"chunk_len": 24,
|
24 |
"down_block_num": 4,
|
25 |
+
"down_block_types": [
|
26 |
+
"AllegroDownBlock3D",
|
27 |
+
"AllegroDownBlock3D",
|
28 |
+
"AllegroDownBlock3D",
|
29 |
+
"AllegroDownBlock3D"
|
30 |
+
],
|
31 |
"force_upcast": true,
|
32 |
"in_channels": 3,
|
33 |
"latent_channels": 4,
|
|
|
37 |
"out_channels": 3,
|
38 |
"sample_size": 320,
|
39 |
"scale_factor": 0.13,
|
40 |
+
"scaling_factor": 0.13235,
|
41 |
"t_over": 8,
|
42 |
+
"temporal_compression_ratio": 4,
|
43 |
+
"temporal_downsample_blocks": [
|
44 |
+
true,
|
45 |
+
true,
|
46 |
+
false,
|
47 |
+
false
|
48 |
+
],
|
49 |
+
"temporal_upsample_blocks": [
|
50 |
+
false,
|
51 |
+
true,
|
52 |
+
true,
|
53 |
+
false
|
54 |
+
],
|
55 |
"tile_overlap": [
|
56 |
120,
|
57 |
80
|
58 |
],
|
59 |
+
"up_block_num": 4,
|
60 |
+
"up_block_types": [
|
61 |
+
"AllegroUpBlock3D",
|
62 |
+
"AllegroUpBlock3D",
|
63 |
+
"AllegroUpBlock3D",
|
64 |
+
"AllegroUpBlock3D"
|
65 |
+
]
|
66 |
}
|