Upload OpenSora-STDiT-v4-i2v model
Browse files- config.json +43 -0
- model.safetensors +3 -0
config.json
ADDED
|
@@ -0,0 +1,43 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"STDiT3"
|
| 4 |
+
],
|
| 5 |
+
"caption_channels": 4096,
|
| 6 |
+
"class_dropout_prob": 0.0,
|
| 7 |
+
"depth": 28,
|
| 8 |
+
"drop_path": 0.0,
|
| 9 |
+
"enable_flash_attn": true,
|
| 10 |
+
"enable_layernorm_kernel": true,
|
| 11 |
+
"enable_sequence_parallelism": false,
|
| 12 |
+
"freeze_y_embedder": false,
|
| 13 |
+
"hidden_size": 1152,
|
| 14 |
+
"in_channels": 16,
|
| 15 |
+
"input_size": [
|
| 16 |
+
null,
|
| 17 |
+
null,
|
| 18 |
+
null
|
| 19 |
+
],
|
| 20 |
+
"input_sq_size": 512,
|
| 21 |
+
"kernel_size": [
|
| 22 |
+
8,
|
| 23 |
+
8,
|
| 24 |
+
-1
|
| 25 |
+
],
|
| 26 |
+
"mlp_ratio": 4.0,
|
| 27 |
+
"model_max_length": 300,
|
| 28 |
+
"model_type": "STDiT3",
|
| 29 |
+
"num_heads": 16,
|
| 30 |
+
"only_train_temporal": false,
|
| 31 |
+
"patch_size": [
|
| 32 |
+
1,
|
| 33 |
+
2,
|
| 34 |
+
2
|
| 35 |
+
],
|
| 36 |
+
"pred_sigma": true,
|
| 37 |
+
"qk_norm": true,
|
| 38 |
+
"skip_temporal": false,
|
| 39 |
+
"skip_y_embedder": false,
|
| 40 |
+
"torch_dtype": "float32",
|
| 41 |
+
"transformers_version": "4.39.3",
|
| 42 |
+
"use_spatial_rope": true
|
| 43 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:18cdce9fa99ebcf8d5a31e6c9dac1e9f14f701f5b18dfce0829405b5f64d7dde
|
| 3 |
+
size 4836861776
|