Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,119 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_10200_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_10400_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_11000_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_11200_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_1200_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_12200_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_13200_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_1400_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_14800_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_15800_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_16800_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_1800_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_18600_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_20000_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_2000_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_2200_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_2400_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_2600_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_2800_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_3200_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_3400_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_3600_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_3800_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_4000_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_400_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_4200_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_4600_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_4800_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_5000_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_5200_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/.ipynb_checkpoints/step_800_val_img_0-checkpoint.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_10000_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_1000_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_10200_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_10400_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_10800_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_11000_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_11200_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_11400_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_11600_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_11800_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_1200_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_12200_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_12600_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_13200_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_13400_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_13600_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_13800_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_14000_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_1400_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_14200_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_14400_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_14600_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_14800_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_15000_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_15200_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_15400_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_15600_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_15800_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_16000_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_1600_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_16200_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_16400_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_16800_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_17200_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_17400_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_17600_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_17800_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_18000_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_1800_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_18200_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_18400_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_18600_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_19800_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_1_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_20000_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_2000_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_200_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_2200_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_2400_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_2600_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_2800_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_3200_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_3400_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_3600_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_3800_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_4000_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_400_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_4200_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_4400_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_4600_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_4800_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_5000_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_5200_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_5400_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_5600_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_5800_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_6000_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_600_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_6200_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_6400_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_6600_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_6800_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_7000_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_7200_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_7400_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_7800_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_800_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_8200_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_8400_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_8600_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_8800_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_9000_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_9400_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_9600_val_img_0.gif filter=lfs diff=lfs merge=lfs -text
+validation_images/step_9800_val_img_0.gif filter=lfs diff=lfs merge=lfs -text

checkpoint-20000/optimizer.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8dbb8bdc6984612c013386c74a21a7fe8b82b95333abeaf9f194e9e4d9887e26
+size 3181312562

checkpoint-20000/random_states_0.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dff3014ed80498952cf8a2401034c0c9ad35e3f247941180256b772a3d81b4ce
+size 14408

checkpoint-20000/scaler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2330f31807d6efda321f3a9cc30fdde8953b88badace54b329f0507349d6d950
+size 988

checkpoint-20000/scheduler.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1d23edd1999591b9c0d5ea54521ae0a56e61ea0c3a1640c60297d52604f4a7a
+size 1000

checkpoint-20000/unet/config.json

@@ -0,0 +1,38 @@
+{
+  "_class_name": "UNetSpatioTemporalConditionModel",
+  "_diffusers_version": "0.32.1",
+  "_name_or_path": "stabilityai/stable-video-diffusion-img2vid-xt",
+  "addition_time_embed_dim": 256,
+  "block_out_channels": [
+    320,
+    640,
+    1280,
+    1280
+  ],
+  "cross_attention_dim": 1024,
+  "down_block_types": [
+    "CrossAttnDownBlockSpatioTemporal",
+    "CrossAttnDownBlockSpatioTemporal",
+    "CrossAttnDownBlockSpatioTemporal",
+    "DownBlockSpatioTemporal"
+  ],
+  "in_channels": 8,
+  "layers_per_block": 2,
+  "num_attention_heads": [
+    5,
+    10,
+    20,
+    20
+  ],
+  "num_frames": 25,
+  "out_channels": 4,
+  "projection_class_embeddings_input_dim": 768,
+  "sample_size": 96,
+  "transformer_layers_per_block": 1,
+  "up_block_types": [
+    "UpBlockSpatioTemporal",
+    "CrossAttnUpBlockSpatioTemporal",
+    "CrossAttnUpBlockSpatioTemporal",
+    "CrossAttnUpBlockSpatioTemporal"
+  ]
+}

checkpoint-20000/unet/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d33f4bf3d3fe30db4ee1b502e96954245142b2e15f512cc99c1a8fffbd1d3e6
+size 6098682464

feature_extractor/preprocessor_config.json

@@ -0,0 +1,27 @@
+{
+  "crop_size": {
+    "height": 224,
+    "width": 224
+  },
+  "do_center_crop": true,
+  "do_convert_rgb": true,
+  "do_normalize": true,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.48145466,
+    0.4578275,
+    0.40821073
+  ],
+  "image_processor_type": "CLIPImageProcessor",
+  "image_std": [
+    0.26862954,
+    0.26130258,
+    0.27577711
+  ],
+  "resample": 3,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "shortest_edge": 224
+  }
+}

image_encoder/config.json

@@ -0,0 +1,23 @@
+{
+  "_name_or_path": "stabilityai/stable-video-diffusion-img2vid-xt",
+  "architectures": [
+    "CLIPVisionModelWithProjection"
+  ],
+  "attention_dropout": 0.0,
+  "dropout": 0.0,
+  "hidden_act": "gelu",
+  "hidden_size": 1280,
+  "image_size": 224,
+  "initializer_factor": 1.0,
+  "initializer_range": 0.02,
+  "intermediate_size": 5120,
+  "layer_norm_eps": 1e-05,
+  "model_type": "clip_vision_model",
+  "num_attention_heads": 16,
+  "num_channels": 3,
+  "num_hidden_layers": 32,
+  "patch_size": 14,
+  "projection_dim": 1024,
+  "torch_dtype": "float16",
+  "transformers_version": "4.47.1"
+}

image_encoder/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ae616c24393dd1854372b0639e5541666f7521cbe219669255e865cb7f89466a
+size 1264217240

model_index.json

@@ -0,0 +1,25 @@
+{
+  "_class_name": "StableVideoDiffusionPipeline",
+  "_diffusers_version": "0.32.1",
+  "_name_or_path": "stabilityai/stable-video-diffusion-img2vid-xt",
+  "feature_extractor": [
+    "transformers",
+    "CLIPImageProcessor"
+  ],
+  "image_encoder": [
+    "transformers",
+    "CLIPVisionModelWithProjection"
+  ],
+  "scheduler": [
+    "diffusers",
+    "EulerDiscreteScheduler"
+  ],
+  "unet": [
+    "diffusers",
+    "UNetSpatioTemporalConditionModel"
+  ],
+  "vae": [
+    "diffusers",
+    "AutoencoderKLTemporalDecoder"
+  ]
+}

scheduler/scheduler_config.json

@@ -0,0 +1,24 @@
+{
+  "_class_name": "EulerDiscreteScheduler",
+  "_diffusers_version": "0.32.1",
+  "beta_end": 0.012,
+  "beta_schedule": "scaled_linear",
+  "beta_start": 0.00085,
+  "clip_sample": false,
+  "final_sigmas_type": "zero",
+  "interpolation_type": "linear",
+  "num_train_timesteps": 1000,
+  "prediction_type": "v_prediction",
+  "rescale_betas_zero_snr": false,
+  "set_alpha_to_one": false,
+  "sigma_max": 700.0,
+  "sigma_min": 0.002,
+  "skip_prk_steps": true,
+  "steps_offset": 1,
+  "timestep_spacing": "leading",
+  "timestep_type": "continuous",
+  "trained_betas": null,
+  "use_beta_sigmas": false,
+  "use_exponential_sigmas": false,
+  "use_karras_sigmas": true
+}

unet/config.json

@@ -0,0 +1,38 @@
+{
+  "_class_name": "UNetSpatioTemporalConditionModel",
+  "_diffusers_version": "0.32.1",
+  "_name_or_path": "stabilityai/stable-video-diffusion-img2vid-xt",
+  "addition_time_embed_dim": 256,
+  "block_out_channels": [
+    320,
+    640,
+    1280,
+    1280
+  ],
+  "cross_attention_dim": 1024,
+  "down_block_types": [
+    "CrossAttnDownBlockSpatioTemporal",
+    "CrossAttnDownBlockSpatioTemporal",
+    "CrossAttnDownBlockSpatioTemporal",
+    "DownBlockSpatioTemporal"
+  ],
+  "in_channels": 8,
+  "layers_per_block": 2,
+  "num_attention_heads": [
+    5,
+    10,
+    20,
+    20
+  ],
+  "num_frames": 25,
+  "out_channels": 4,
+  "projection_class_embeddings_input_dim": 768,
+  "sample_size": 96,
+  "transformer_layers_per_block": 1,
+  "up_block_types": [
+    "UpBlockSpatioTemporal",
+    "CrossAttnUpBlockSpatioTemporal",
+    "CrossAttnUpBlockSpatioTemporal",
+    "CrossAttnUpBlockSpatioTemporal"
+  ]
+}

unet/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d33f4bf3d3fe30db4ee1b502e96954245142b2e15f512cc99c1a8fffbd1d3e6
+size 6098682464

vae/config.json

@@ -0,0 +1,24 @@
+{
+  "_class_name": "AutoencoderKLTemporalDecoder",
+  "_diffusers_version": "0.32.1",
+  "_name_or_path": "stabilityai/stable-video-diffusion-img2vid-xt",
+  "block_out_channels": [
+    128,
+    256,
+    512,
+    512
+  ],
+  "down_block_types": [
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D"
+  ],
+  "force_upcast": true,
+  "in_channels": 3,
+  "latent_channels": 4,
+  "layers_per_block": 2,
+  "out_channels": 3,
+  "sample_size": 768,
+  "scaling_factor": 0.18215
+}

vae/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af602cd0eb4ad6086ec94fbf1438dfb1be5ec9ac03fd0215640854e90d6463a3
+size 195531910