Upload config.json
Browse files- config.json +47 -0
config.json
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_class_name": "OpenSoraT2V",
|
| 3 |
+
"_diffusers_version": "0.27.2",
|
| 4 |
+
"activation_fn": "gelu-approximate",
|
| 5 |
+
"attention_bias": true,
|
| 6 |
+
"attention_head_dim": 96,
|
| 7 |
+
"attention_mode": "xformers",
|
| 8 |
+
"attention_type": "default",
|
| 9 |
+
"caption_channels": 4096,
|
| 10 |
+
"cross_attention_dim": 2304,
|
| 11 |
+
"decay": 0.9999,
|
| 12 |
+
"double_self_attention": false,
|
| 13 |
+
"downsampler": null,
|
| 14 |
+
"dropout": 0.0,
|
| 15 |
+
"in_channels": 4,
|
| 16 |
+
"interpolation_scale_h": 1.0,
|
| 17 |
+
"interpolation_scale_t": 1.0,
|
| 18 |
+
"interpolation_scale_w": 1.0,
|
| 19 |
+
"inv_gamma": 1.0,
|
| 20 |
+
"min_decay": 0.0,
|
| 21 |
+
"norm_elementwise_affine": false,
|
| 22 |
+
"norm_eps": 1e-06,
|
| 23 |
+
"norm_num_groups": 32,
|
| 24 |
+
"norm_type": "ada_norm_single",
|
| 25 |
+
"num_attention_heads": 24,
|
| 26 |
+
"num_embeds_ada_norm": 1000,
|
| 27 |
+
"num_layers": 32,
|
| 28 |
+
"num_vector_embeds": null,
|
| 29 |
+
"only_cross_attention": false,
|
| 30 |
+
"optimization_step": 30000,
|
| 31 |
+
"out_channels": 4,
|
| 32 |
+
"patch_size": 2,
|
| 33 |
+
"patch_size_t": 1,
|
| 34 |
+
"power": 0.6666666666666666,
|
| 35 |
+
"sample_size": [
|
| 36 |
+
60,
|
| 37 |
+
80
|
| 38 |
+
],
|
| 39 |
+
"sample_size_t": 8,
|
| 40 |
+
"upcast_attention": false,
|
| 41 |
+
"update_after_step": 0,
|
| 42 |
+
"use_additional_conditions": null,
|
| 43 |
+
"use_ema_warmup": false,
|
| 44 |
+
"use_linear_projection": false,
|
| 45 |
+
"use_rope": true,
|
| 46 |
+
"use_stable_fp32": true
|
| 47 |
+
}
|