End of training
Browse filesThis view is limited to 50 files because it contains too many changes. See raw diff
- .gitattributes +1 -0
- checkpoint-31500/optimizer.bin +3 -0
- checkpoint-31500/querytransformer/config.json +69 -0
- checkpoint-31500/querytransformer/model.safetensors +3 -0
- checkpoint-31500/random_states_0.pkl +3 -0
- checkpoint-31500/scheduler.bin +3 -0
- checkpoint-32000/optimizer.bin +3 -0
- checkpoint-32000/querytransformer/config.json +69 -0
- checkpoint-32000/querytransformer/model.safetensors +3 -0
- checkpoint-32000/random_states_0.pkl +3 -0
- checkpoint-32000/scheduler.bin +3 -0
- config.json +40 -6
- model.safetensors +2 -2
- validation_images/step_10000_val_cross_0.mp4 +0 -0
- validation_images/step_10000_val_recon_0.mp4 +0 -0
- validation_images/step_1000_val_cross_0.mp4 +0 -0
- validation_images/step_1000_val_recon_0.mp4 +0 -0
- validation_images/step_10500_val_cross_0.mp4 +0 -0
- validation_images/step_10500_val_recon_0.mp4 +0 -0
- validation_images/step_11000_val_cross_0.mp4 +0 -0
- validation_images/step_11000_val_recon_0.mp4 +0 -0
- validation_images/step_11500_val_cross_0.mp4 +0 -0
- validation_images/step_11500_val_recon_0.mp4 +0 -0
- validation_images/step_12000_val_cross_0.mp4 +0 -0
- validation_images/step_12000_val_recon_0.mp4 +0 -0
- validation_images/step_12500_val_cross_0.mp4 +0 -0
- validation_images/step_12500_val_recon_0.mp4 +0 -0
- validation_images/step_13000_val_cross_0.mp4 +0 -0
- validation_images/step_13000_val_recon_0.mp4 +0 -0
- validation_images/step_13500_val_cross_0.mp4 +0 -0
- validation_images/step_13500_val_recon_0.mp4 +0 -0
- validation_images/step_14000_val_cross_0.mp4 +0 -0
- validation_images/step_14000_val_recon_0.mp4 +0 -0
- validation_images/step_14500_val_cross_0.mp4 +0 -0
- validation_images/step_14500_val_recon_0.mp4 +0 -0
- validation_images/step_15000_val_cross_0.mp4 +0 -0
- validation_images/step_15000_val_recon_0.mp4 +0 -0
- validation_images/step_1500_val_cross_0.mp4 +0 -0
- validation_images/step_1500_val_recon_0.mp4 +0 -0
- validation_images/step_15500_val_cross_0.mp4 +0 -0
- validation_images/step_15500_val_recon_0.mp4 +0 -0
- validation_images/step_16000_val_cross_0.mp4 +0 -0
- validation_images/step_16000_val_recon_0.mp4 +0 -0
- validation_images/step_16500_val_cross_0.mp4 +0 -0
- validation_images/step_16500_val_recon_0.mp4 +0 -0
- validation_images/step_17000_val_cross_0.mp4 +0 -0
- validation_images/step_17000_val_recon_0.mp4 +0 -0
- validation_images/step_17500_val_cross_0.mp4 +0 -0
- validation_images/step_17500_val_recon_0.mp4 +0 -0
- validation_images/step_18000_val_cross_0.mp4 +0 -0
.gitattributes
CHANGED
|
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
+
validation_images/step_1_val_cross_0.mp4 filter=lfs diff=lfs merge=lfs -text
|
checkpoint-31500/optimizer.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:117aee311bcb13a8651a64f65418f968c5ee0200c51f066a58382fff82309d24
|
| 3 |
+
size 126654
|
checkpoint-31500/querytransformer/config.json
ADDED
|
@@ -0,0 +1,69 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"QueryTransformerEmbedding"
|
| 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": 8,
|
| 18 |
+
"num_channels": 8,
|
| 19 |
+
"num_frames": 24,
|
| 20 |
+
"num_hidden_layers": 6,
|
| 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 |
+
"unet_attention_names": {
|
| 35 |
+
"down_blocks_0_attentions_0_temporal_transformer_blocks_0_attn1": 320,
|
| 36 |
+
"down_blocks_0_attentions_0_transformer_blocks_0_attn1": 320,
|
| 37 |
+
"down_blocks_0_attentions_1_temporal_transformer_blocks_0_attn1": 320,
|
| 38 |
+
"down_blocks_0_attentions_1_transformer_blocks_0_attn1": 320,
|
| 39 |
+
"down_blocks_1_attentions_0_temporal_transformer_blocks_0_attn1": 640,
|
| 40 |
+
"down_blocks_1_attentions_0_transformer_blocks_0_attn1": 640,
|
| 41 |
+
"down_blocks_1_attentions_1_temporal_transformer_blocks_0_attn1": 640,
|
| 42 |
+
"down_blocks_1_attentions_1_transformer_blocks_0_attn1": 640,
|
| 43 |
+
"down_blocks_2_attentions_0_temporal_transformer_blocks_0_attn1": 1280,
|
| 44 |
+
"down_blocks_2_attentions_0_transformer_blocks_0_attn1": 1280,
|
| 45 |
+
"down_blocks_2_attentions_1_temporal_transformer_blocks_0_attn1": 1280,
|
| 46 |
+
"down_blocks_2_attentions_1_transformer_blocks_0_attn1": 1280,
|
| 47 |
+
"mid_block_attentions_0_temporal_transformer_blocks_0_attn1": 1280,
|
| 48 |
+
"mid_block_attentions_0_transformer_blocks_0_attn1": 1280,
|
| 49 |
+
"up_blocks_1_attentions_0_temporal_transformer_blocks_0_attn1": 1280,
|
| 50 |
+
"up_blocks_1_attentions_0_transformer_blocks_0_attn1": 1280,
|
| 51 |
+
"up_blocks_1_attentions_1_temporal_transformer_blocks_0_attn1": 1280,
|
| 52 |
+
"up_blocks_1_attentions_1_transformer_blocks_0_attn1": 1280,
|
| 53 |
+
"up_blocks_1_attentions_2_temporal_transformer_blocks_0_attn1": 1280,
|
| 54 |
+
"up_blocks_1_attentions_2_transformer_blocks_0_attn1": 1280,
|
| 55 |
+
"up_blocks_2_attentions_0_temporal_transformer_blocks_0_attn1": 640,
|
| 56 |
+
"up_blocks_2_attentions_0_transformer_blocks_0_attn1": 640,
|
| 57 |
+
"up_blocks_2_attentions_1_temporal_transformer_blocks_0_attn1": 640,
|
| 58 |
+
"up_blocks_2_attentions_1_transformer_blocks_0_attn1": 640,
|
| 59 |
+
"up_blocks_2_attentions_2_temporal_transformer_blocks_0_attn1": 640,
|
| 60 |
+
"up_blocks_2_attentions_2_transformer_blocks_0_attn1": 640,
|
| 61 |
+
"up_blocks_3_attentions_0_temporal_transformer_blocks_0_attn1": 320,
|
| 62 |
+
"up_blocks_3_attentions_0_transformer_blocks_0_attn1": 320,
|
| 63 |
+
"up_blocks_3_attentions_1_temporal_transformer_blocks_0_attn1": 320,
|
| 64 |
+
"up_blocks_3_attentions_1_transformer_blocks_0_attn1": 320,
|
| 65 |
+
"up_blocks_3_attentions_2_temporal_transformer_blocks_0_attn1": 320,
|
| 66 |
+
"up_blocks_3_attentions_2_transformer_blocks_0_attn1": 320
|
| 67 |
+
},
|
| 68 |
+
"zero_init_theta": false
|
| 69 |
+
}
|
checkpoint-31500/querytransformer/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6fcfa5f299a42181a50b91adb2c4c91f2861d4b743536edf8195b21abcffb751
|
| 3 |
+
size 53944
|
checkpoint-31500/random_states_0.pkl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b8030c29d1e9d45c4064b2f7cbb60a388973f0f5b8bfcb2db00af710301bcb09
|
| 3 |
+
size 14408
|
checkpoint-31500/scheduler.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:95f142efb6788e98b9e7f658acec6af427fffbb91cc830a18c18f0adc2bd4384
|
| 3 |
+
size 1000
|
checkpoint-32000/optimizer.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0bf73527abd70bbbe3052e09244f00c967c12eba2b5fed072b65177b3f2f5798
|
| 3 |
+
size 126654
|
checkpoint-32000/querytransformer/config.json
ADDED
|
@@ -0,0 +1,69 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"QueryTransformerEmbedding"
|
| 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": 8,
|
| 18 |
+
"num_channels": 8,
|
| 19 |
+
"num_frames": 24,
|
| 20 |
+
"num_hidden_layers": 6,
|
| 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 |
+
"unet_attention_names": {
|
| 35 |
+
"down_blocks_0_attentions_0_temporal_transformer_blocks_0_attn1": 320,
|
| 36 |
+
"down_blocks_0_attentions_0_transformer_blocks_0_attn1": 320,
|
| 37 |
+
"down_blocks_0_attentions_1_temporal_transformer_blocks_0_attn1": 320,
|
| 38 |
+
"down_blocks_0_attentions_1_transformer_blocks_0_attn1": 320,
|
| 39 |
+
"down_blocks_1_attentions_0_temporal_transformer_blocks_0_attn1": 640,
|
| 40 |
+
"down_blocks_1_attentions_0_transformer_blocks_0_attn1": 640,
|
| 41 |
+
"down_blocks_1_attentions_1_temporal_transformer_blocks_0_attn1": 640,
|
| 42 |
+
"down_blocks_1_attentions_1_transformer_blocks_0_attn1": 640,
|
| 43 |
+
"down_blocks_2_attentions_0_temporal_transformer_blocks_0_attn1": 1280,
|
| 44 |
+
"down_blocks_2_attentions_0_transformer_blocks_0_attn1": 1280,
|
| 45 |
+
"down_blocks_2_attentions_1_temporal_transformer_blocks_0_attn1": 1280,
|
| 46 |
+
"down_blocks_2_attentions_1_transformer_blocks_0_attn1": 1280,
|
| 47 |
+
"mid_block_attentions_0_temporal_transformer_blocks_0_attn1": 1280,
|
| 48 |
+
"mid_block_attentions_0_transformer_blocks_0_attn1": 1280,
|
| 49 |
+
"up_blocks_1_attentions_0_temporal_transformer_blocks_0_attn1": 1280,
|
| 50 |
+
"up_blocks_1_attentions_0_transformer_blocks_0_attn1": 1280,
|
| 51 |
+
"up_blocks_1_attentions_1_temporal_transformer_blocks_0_attn1": 1280,
|
| 52 |
+
"up_blocks_1_attentions_1_transformer_blocks_0_attn1": 1280,
|
| 53 |
+
"up_blocks_1_attentions_2_temporal_transformer_blocks_0_attn1": 1280,
|
| 54 |
+
"up_blocks_1_attentions_2_transformer_blocks_0_attn1": 1280,
|
| 55 |
+
"up_blocks_2_attentions_0_temporal_transformer_blocks_0_attn1": 640,
|
| 56 |
+
"up_blocks_2_attentions_0_transformer_blocks_0_attn1": 640,
|
| 57 |
+
"up_blocks_2_attentions_1_temporal_transformer_blocks_0_attn1": 640,
|
| 58 |
+
"up_blocks_2_attentions_1_transformer_blocks_0_attn1": 640,
|
| 59 |
+
"up_blocks_2_attentions_2_temporal_transformer_blocks_0_attn1": 640,
|
| 60 |
+
"up_blocks_2_attentions_2_transformer_blocks_0_attn1": 640,
|
| 61 |
+
"up_blocks_3_attentions_0_temporal_transformer_blocks_0_attn1": 320,
|
| 62 |
+
"up_blocks_3_attentions_0_transformer_blocks_0_attn1": 320,
|
| 63 |
+
"up_blocks_3_attentions_1_temporal_transformer_blocks_0_attn1": 320,
|
| 64 |
+
"up_blocks_3_attentions_1_transformer_blocks_0_attn1": 320,
|
| 65 |
+
"up_blocks_3_attentions_2_temporal_transformer_blocks_0_attn1": 320,
|
| 66 |
+
"up_blocks_3_attentions_2_transformer_blocks_0_attn1": 320
|
| 67 |
+
},
|
| 68 |
+
"zero_init_theta": false
|
| 69 |
+
}
|
checkpoint-32000/querytransformer/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1eb5ab499b124bc4f8b5ec476c634fcddab9a577ff5a1254ac2b0efe82d6521e
|
| 3 |
+
size 53944
|
checkpoint-32000/random_states_0.pkl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bd5c8ed682c248a021ac9094cdde93eafe224b01adfe8f750a0c48e25dfdfec5
|
| 3 |
+
size 14408
|
checkpoint-32000/scheduler.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2bbb61109d01da8a9c381c0e2f077875e9118afeba0c4ed8b289ee114f675793
|
| 3 |
+
size 1000
|
config.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
| 1 |
{
|
| 2 |
"architectures": [
|
| 3 |
-
"
|
| 4 |
],
|
| 5 |
"attention_probs_dropout_prob": 0.0,
|
| 6 |
"hidden_act": "gelu_fast",
|
|
@@ -14,22 +14,56 @@
|
|
| 14 |
"intermediate_size": 3072,
|
| 15 |
"layer_norm_eps": 1e-06,
|
| 16 |
"model_type": "vivit",
|
| 17 |
-
"num_attention_heads":
|
| 18 |
-
"num_channels":
|
| 19 |
"num_frames": 24,
|
| 20 |
-
"num_hidden_layers":
|
| 21 |
"qkv_bias": true,
|
| 22 |
"query_dims": [
|
| 23 |
160,
|
| 24 |
320,
|
| 25 |
640
|
| 26 |
],
|
| 27 |
-
"torch_dtype": "
|
| 28 |
"transformers_version": "4.47.1",
|
| 29 |
"tubelet_size": [
|
| 30 |
2,
|
| 31 |
9,
|
| 32 |
16
|
| 33 |
],
|
| 34 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 35 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"architectures": [
|
| 3 |
+
"QueryTransformerEmbedding"
|
| 4 |
],
|
| 5 |
"attention_probs_dropout_prob": 0.0,
|
| 6 |
"hidden_act": "gelu_fast",
|
|
|
|
| 14 |
"intermediate_size": 3072,
|
| 15 |
"layer_norm_eps": 1e-06,
|
| 16 |
"model_type": "vivit",
|
| 17 |
+
"num_attention_heads": 8,
|
| 18 |
+
"num_channels": 8,
|
| 19 |
"num_frames": 24,
|
| 20 |
+
"num_hidden_layers": 6,
|
| 21 |
"qkv_bias": true,
|
| 22 |
"query_dims": [
|
| 23 |
160,
|
| 24 |
320,
|
| 25 |
640
|
| 26 |
],
|
| 27 |
+
"torch_dtype": "bfloat16",
|
| 28 |
"transformers_version": "4.47.1",
|
| 29 |
"tubelet_size": [
|
| 30 |
2,
|
| 31 |
9,
|
| 32 |
16
|
| 33 |
],
|
| 34 |
+
"unet_attention_names": {
|
| 35 |
+
"down_blocks_0_attentions_0_temporal_transformer_blocks_0_attn1": 320,
|
| 36 |
+
"down_blocks_0_attentions_0_transformer_blocks_0_attn1": 320,
|
| 37 |
+
"down_blocks_0_attentions_1_temporal_transformer_blocks_0_attn1": 320,
|
| 38 |
+
"down_blocks_0_attentions_1_transformer_blocks_0_attn1": 320,
|
| 39 |
+
"down_blocks_1_attentions_0_temporal_transformer_blocks_0_attn1": 640,
|
| 40 |
+
"down_blocks_1_attentions_0_transformer_blocks_0_attn1": 640,
|
| 41 |
+
"down_blocks_1_attentions_1_temporal_transformer_blocks_0_attn1": 640,
|
| 42 |
+
"down_blocks_1_attentions_1_transformer_blocks_0_attn1": 640,
|
| 43 |
+
"down_blocks_2_attentions_0_temporal_transformer_blocks_0_attn1": 1280,
|
| 44 |
+
"down_blocks_2_attentions_0_transformer_blocks_0_attn1": 1280,
|
| 45 |
+
"down_blocks_2_attentions_1_temporal_transformer_blocks_0_attn1": 1280,
|
| 46 |
+
"down_blocks_2_attentions_1_transformer_blocks_0_attn1": 1280,
|
| 47 |
+
"mid_block_attentions_0_temporal_transformer_blocks_0_attn1": 1280,
|
| 48 |
+
"mid_block_attentions_0_transformer_blocks_0_attn1": 1280,
|
| 49 |
+
"up_blocks_1_attentions_0_temporal_transformer_blocks_0_attn1": 1280,
|
| 50 |
+
"up_blocks_1_attentions_0_transformer_blocks_0_attn1": 1280,
|
| 51 |
+
"up_blocks_1_attentions_1_temporal_transformer_blocks_0_attn1": 1280,
|
| 52 |
+
"up_blocks_1_attentions_1_transformer_blocks_0_attn1": 1280,
|
| 53 |
+
"up_blocks_1_attentions_2_temporal_transformer_blocks_0_attn1": 1280,
|
| 54 |
+
"up_blocks_1_attentions_2_transformer_blocks_0_attn1": 1280,
|
| 55 |
+
"up_blocks_2_attentions_0_temporal_transformer_blocks_0_attn1": 640,
|
| 56 |
+
"up_blocks_2_attentions_0_transformer_blocks_0_attn1": 640,
|
| 57 |
+
"up_blocks_2_attentions_1_temporal_transformer_blocks_0_attn1": 640,
|
| 58 |
+
"up_blocks_2_attentions_1_transformer_blocks_0_attn1": 640,
|
| 59 |
+
"up_blocks_2_attentions_2_temporal_transformer_blocks_0_attn1": 640,
|
| 60 |
+
"up_blocks_2_attentions_2_transformer_blocks_0_attn1": 640,
|
| 61 |
+
"up_blocks_3_attentions_0_temporal_transformer_blocks_0_attn1": 320,
|
| 62 |
+
"up_blocks_3_attentions_0_transformer_blocks_0_attn1": 320,
|
| 63 |
+
"up_blocks_3_attentions_1_temporal_transformer_blocks_0_attn1": 320,
|
| 64 |
+
"up_blocks_3_attentions_1_transformer_blocks_0_attn1": 320,
|
| 65 |
+
"up_blocks_3_attentions_2_temporal_transformer_blocks_0_attn1": 320,
|
| 66 |
+
"up_blocks_3_attentions_2_transformer_blocks_0_attn1": 320
|
| 67 |
+
},
|
| 68 |
+
"zero_init_theta": false
|
| 69 |
}
|
model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5ce6e732b58004553623d901af2efbce705669863fae78a330504b7c8072abbc
|
| 3 |
+
size 29008
|
validation_images/step_10000_val_cross_0.mp4
ADDED
|
Binary file (895 kB). View file
|
|
|
validation_images/step_10000_val_recon_0.mp4
ADDED
|
Binary file (473 kB). View file
|
|
|
validation_images/step_1000_val_cross_0.mp4
ADDED
|
Binary file (910 kB). View file
|
|
|
validation_images/step_1000_val_recon_0.mp4
ADDED
|
Binary file (549 kB). View file
|
|
|
validation_images/step_10500_val_cross_0.mp4
ADDED
|
Binary file (850 kB). View file
|
|
|
validation_images/step_10500_val_recon_0.mp4
ADDED
|
Binary file (453 kB). View file
|
|
|
validation_images/step_11000_val_cross_0.mp4
ADDED
|
Binary file (818 kB). View file
|
|
|
validation_images/step_11000_val_recon_0.mp4
ADDED
|
Binary file (610 kB). View file
|
|
|
validation_images/step_11500_val_cross_0.mp4
ADDED
|
Binary file (890 kB). View file
|
|
|
validation_images/step_11500_val_recon_0.mp4
ADDED
|
Binary file (458 kB). View file
|
|
|
validation_images/step_12000_val_cross_0.mp4
ADDED
|
Binary file (812 kB). View file
|
|
|
validation_images/step_12000_val_recon_0.mp4
ADDED
|
Binary file (541 kB). View file
|
|
|
validation_images/step_12500_val_cross_0.mp4
ADDED
|
Binary file (893 kB). View file
|
|
|
validation_images/step_12500_val_recon_0.mp4
ADDED
|
Binary file (581 kB). View file
|
|
|
validation_images/step_13000_val_cross_0.mp4
ADDED
|
Binary file (704 kB). View file
|
|
|
validation_images/step_13000_val_recon_0.mp4
ADDED
|
Binary file (299 kB). View file
|
|
|
validation_images/step_13500_val_cross_0.mp4
ADDED
|
Binary file (951 kB). View file
|
|
|
validation_images/step_13500_val_recon_0.mp4
ADDED
|
Binary file (523 kB). View file
|
|
|
validation_images/step_14000_val_cross_0.mp4
ADDED
|
Binary file (929 kB). View file
|
|
|
validation_images/step_14000_val_recon_0.mp4
ADDED
|
Binary file (542 kB). View file
|
|
|
validation_images/step_14500_val_cross_0.mp4
ADDED
|
Binary file (867 kB). View file
|
|
|
validation_images/step_14500_val_recon_0.mp4
ADDED
|
Binary file (571 kB). View file
|
|
|
validation_images/step_15000_val_cross_0.mp4
ADDED
|
Binary file (706 kB). View file
|
|
|
validation_images/step_15000_val_recon_0.mp4
ADDED
|
Binary file (518 kB). View file
|
|
|
validation_images/step_1500_val_cross_0.mp4
ADDED
|
Binary file (870 kB). View file
|
|
|
validation_images/step_1500_val_recon_0.mp4
ADDED
|
Binary file (552 kB). View file
|
|
|
validation_images/step_15500_val_cross_0.mp4
ADDED
|
Binary file (870 kB). View file
|
|
|
validation_images/step_15500_val_recon_0.mp4
ADDED
|
Binary file (378 kB). View file
|
|
|
validation_images/step_16000_val_cross_0.mp4
ADDED
|
Binary file (926 kB). View file
|
|
|
validation_images/step_16000_val_recon_0.mp4
ADDED
|
Binary file (454 kB). View file
|
|
|
validation_images/step_16500_val_cross_0.mp4
ADDED
|
Binary file (860 kB). View file
|
|
|
validation_images/step_16500_val_recon_0.mp4
ADDED
|
Binary file (464 kB). View file
|
|
|
validation_images/step_17000_val_cross_0.mp4
ADDED
|
Binary file (900 kB). View file
|
|
|
validation_images/step_17000_val_recon_0.mp4
ADDED
|
Binary file (496 kB). View file
|
|
|
validation_images/step_17500_val_cross_0.mp4
ADDED
|
Binary file (926 kB). View file
|
|
|
validation_images/step_17500_val_recon_0.mp4
ADDED
|
Binary file (258 kB). View file
|
|
|
validation_images/step_18000_val_cross_0.mp4
ADDED
|
Binary file (863 kB). View file
|
|
|