Update config.json
Browse files- config.json +77 -2
config.json
CHANGED
@@ -1,3 +1,4 @@
|
|
|
|
1 |
{
|
2 |
"architectures": [
|
3 |
"Qwen2VLVAEForConditionalGeneration"
|
@@ -33,7 +34,80 @@
|
|
33 |
"patch_size": 2,
|
34 |
"embed_dim": 1280,
|
35 |
"vae_path": "genmo/mochi-1-preview",
|
36 |
-
"vae_subfolder": "vae"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
37 |
},
|
38 |
"rope_scaling": {
|
39 |
"type": "mrope",
|
@@ -44,4 +118,5 @@
|
|
44 |
]
|
45 |
},
|
46 |
"vocab_size": 152064
|
47 |
-
}
|
|
|
|
1 |
+
|
2 |
{
|
3 |
"architectures": [
|
4 |
"Qwen2VLVAEForConditionalGeneration"
|
|
|
34 |
"patch_size": 2,
|
35 |
"embed_dim": 1280,
|
36 |
"vae_path": "genmo/mochi-1-preview",
|
37 |
+
"vae_subfolder": "vae",
|
38 |
+
"vae_config": {
|
39 |
+
"_class_name": "AutoencoderKLMochi",
|
40 |
+
"_diffusers_version": "0.32.0.dev0",
|
41 |
+
"act_fn": "silu",
|
42 |
+
"add_attention_block": [
|
43 |
+
false,
|
44 |
+
true,
|
45 |
+
true,
|
46 |
+
true,
|
47 |
+
true
|
48 |
+
],
|
49 |
+
"decoder_block_out_channels": [
|
50 |
+
128,
|
51 |
+
256,
|
52 |
+
512,
|
53 |
+
768
|
54 |
+
],
|
55 |
+
"encoder_block_out_channels": [
|
56 |
+
64,
|
57 |
+
128,
|
58 |
+
256,
|
59 |
+
384
|
60 |
+
],
|
61 |
+
"in_channels": 15,
|
62 |
+
"latent_channels": 12,
|
63 |
+
"latents_mean": [
|
64 |
+
-0.06730895953510081,
|
65 |
+
-0.038011381506090416,
|
66 |
+
-0.07477820912866141,
|
67 |
+
-0.05565264470995561,
|
68 |
+
0.012767231469026969,
|
69 |
+
-0.04703542746246419,
|
70 |
+
0.043896967884726704,
|
71 |
+
-0.09346305707025976,
|
72 |
+
-0.09918314763016893,
|
73 |
+
-0.008729793427399178,
|
74 |
+
-0.011931556316503654,
|
75 |
+
-0.0321993391887285
|
76 |
+
],
|
77 |
+
"latents_std": [
|
78 |
+
0.9263795028493863,
|
79 |
+
0.9248894543193766,
|
80 |
+
0.9393059390890617,
|
81 |
+
0.959253732819592,
|
82 |
+
0.8244560132752793,
|
83 |
+
0.917259975397747,
|
84 |
+
0.9294154431013696,
|
85 |
+
1.3720942357788521,
|
86 |
+
0.881393668867029,
|
87 |
+
0.9168315692124348,
|
88 |
+
0.9185249279345552,
|
89 |
+
0.9274757570805041
|
90 |
+
],
|
91 |
+
"layers_per_block": [
|
92 |
+
3,
|
93 |
+
3,
|
94 |
+
4,
|
95 |
+
6,
|
96 |
+
3
|
97 |
+
],
|
98 |
+
"out_channels": 3,
|
99 |
+
"scaling_factor": 1.0,
|
100 |
+
"spatial_expansions": [
|
101 |
+
2,
|
102 |
+
2,
|
103 |
+
2
|
104 |
+
],
|
105 |
+
"temporal_expansions": [
|
106 |
+
1,
|
107 |
+
2,
|
108 |
+
3
|
109 |
+
]
|
110 |
+
}
|
111 |
},
|
112 |
"rope_scaling": {
|
113 |
"type": "mrope",
|
|
|
118 |
]
|
119 |
},
|
120 |
"vocab_size": 152064
|
121 |
+
}
|
122 |
+
|