Upload umt5/checkpoint-309000/model/config.json with huggingface_hub
Browse files
umt5/checkpoint-309000/model/config.json
ADDED
|
@@ -0,0 +1,48 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_class_name": "OpenSoraT2I",
|
| 3 |
+
"_diffusers_version": "0.30.2",
|
| 4 |
+
"_name_or_path": "t2i_ablation_arch_/umt5/checkpoint-270375/model",
|
| 5 |
+
"activation_fn": "gelu-approximate",
|
| 6 |
+
"attention_bias": true,
|
| 7 |
+
"attention_head_dim": 64,
|
| 8 |
+
"caption_channels": 4096,
|
| 9 |
+
"caption_channels_2": null,
|
| 10 |
+
"conv_ffn": false,
|
| 11 |
+
"conv_out": false,
|
| 12 |
+
"deepnorm": false,
|
| 13 |
+
"dropout": 0.0,
|
| 14 |
+
"explicit_uniform_rope": false,
|
| 15 |
+
"in_channels": 32,
|
| 16 |
+
"interpolation_scale_h": 1.0,
|
| 17 |
+
"interpolation_scale_t": 1.0,
|
| 18 |
+
"interpolation_scale_w": 1.0,
|
| 19 |
+
"layerwise_text_mlp": false,
|
| 20 |
+
"light_block": false,
|
| 21 |
+
"norm_cls": "fp32_layer_norm",
|
| 22 |
+
"norm_elementwise_affine": true,
|
| 23 |
+
"norm_eps": 1e-06,
|
| 24 |
+
"norm_skip": false,
|
| 25 |
+
"num_attention_heads": 24,
|
| 26 |
+
"num_layers": [
|
| 27 |
+
12,
|
| 28 |
+
1,
|
| 29 |
+
12
|
| 30 |
+
],
|
| 31 |
+
"out_channels": 32,
|
| 32 |
+
"patch_size": 2,
|
| 33 |
+
"patch_size_t": 1,
|
| 34 |
+
"pooled_projection_dim": 0,
|
| 35 |
+
"prenorm_num": 1000,
|
| 36 |
+
"sample_size_h": 16,
|
| 37 |
+
"sample_size_t": 1,
|
| 38 |
+
"sample_size_w": 16,
|
| 39 |
+
"sandwich_norm": false,
|
| 40 |
+
"shrink": false,
|
| 41 |
+
"skip_connection": false,
|
| 42 |
+
"sparse1d": false,
|
| 43 |
+
"sparse_n": 2,
|
| 44 |
+
"time_as_text_token": false,
|
| 45 |
+
"time_as_x_token": false,
|
| 46 |
+
"timestep_embed_dim": 512,
|
| 47 |
+
"use_text_dim": false
|
| 48 |
+
}
|