End of training
Browse files- config.json +35 -0
- model.safetensors +3 -0
- validation_images/step_1_val_cross_0.mp4 +0 -0
- validation_images/step_1_val_recon_0.mp4 +0 -0
config.json
ADDED
|
@@ -0,0 +1,35 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"QueryTransformerEncoder"
|
| 4 |
+
],
|
| 5 |
+
"attention_probs_dropout_prob": 0.0,
|
| 6 |
+
"hidden_act": "gelu_fast",
|
| 7 |
+
"hidden_dropout_prob": 0.0,
|
| 8 |
+
"hidden_size": 768,
|
| 9 |
+
"image_size": [
|
| 10 |
+
72,
|
| 11 |
+
128
|
| 12 |
+
],
|
| 13 |
+
"initializer_range": 0.02,
|
| 14 |
+
"intermediate_size": 3072,
|
| 15 |
+
"layer_norm_eps": 1e-06,
|
| 16 |
+
"model_type": "vivit",
|
| 17 |
+
"num_attention_heads": 16,
|
| 18 |
+
"num_channels": 4,
|
| 19 |
+
"num_frames": 24,
|
| 20 |
+
"num_hidden_layers": 12,
|
| 21 |
+
"qkv_bias": true,
|
| 22 |
+
"query_dims": [
|
| 23 |
+
160,
|
| 24 |
+
320,
|
| 25 |
+
640
|
| 26 |
+
],
|
| 27 |
+
"torch_dtype": "float32",
|
| 28 |
+
"transformers_version": "4.47.1",
|
| 29 |
+
"tubelet_size": [
|
| 30 |
+
2,
|
| 31 |
+
9,
|
| 32 |
+
16
|
| 33 |
+
],
|
| 34 |
+
"zero_init_theta": true
|
| 35 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fa570c60b29edfb504dbd670f92b327e0edbf54d59821779553446f6e2b35d7b
|
| 3 |
+
size 349599384
|
validation_images/step_1_val_cross_0.mp4
ADDED
|
Binary file (815 kB). View file
|
|
|
validation_images/step_1_val_recon_0.mp4
ADDED
|
Binary file (544 kB). View file
|
|
|