Upload 2 files

config.yaml

@@ -0,0 +1,46 @@
+pretrained_model_path: ./checkpoints/kyujinpy/KO-stable-diffusion-disney
+output_dir: outputs/bear_guitar
+train_data:
+  video_path: data/bear_guitar.mp4
+  prompt: A bear is playing guitar
+  n_sample_frames: 14
+  width: 512
+  height: 512
+  sample_start_idx: 0
+  sample_frame_rate: 2
+validation_data:
+  prompts:
+  - 잘생긴 왕자가 피아노를 연주하고 있습니다, 모던한 디즈니 스타일
+  - 엘프가 피아노를 연주하고 있습니다, 모던한 디즈니 스타일
+  - 빨간색 머리를 한 공주가 피아노를 연주하고 있습니다, 모던한 디즈니 스타일
+  video_length: 14
+  width: 512
+  height: 512
+  num_inference_steps: 20
+  guidance_scale: 12.5
+  use_inv_latent: true
+  num_inv_steps: 50
+validation_steps: 100
+trainable_modules:
+- attn1.to_q
+- attn2.to_q
+- attn_temp
+train_batch_size: 1
+max_train_steps: 400
+learning_rate: 3.0e-05
+scale_lr: false
+lr_scheduler: constant
+lr_warmup_steps: 0
+adam_beta1: 0.9
+adam_beta2: 0.999
+adam_weight_decay: 0.01
+adam_epsilon: 1.0e-08
+max_grad_norm: 1.0
+gradient_accumulation_steps: 1
+gradient_checkpointing: true
+checkpointing_steps: 1000
+resume_from_checkpoint: null
+mixed_precision: fp16
+use_8bit_adam: false
+enable_xformers_memory_efficient_attention: true
+seed: 33

model_index.json

@@ -1,14 +1,6 @@
 {
-  "_class_name": "StableDiffusionPipeline",
-  "_diffusers_version": "0.6.0",
-  "feature_extractor": [
-    "transformers",
-    "CLIPImageProcessor"
-  ],
-  "safety_checker": [
-    "stable_diffusion",
-    "StableDiffusionSafetyChecker"
-  ],
+  "_class_name": "TuneAVideoPipeline",
+  "_diffusers_version": "0.11.1",
   "scheduler": [
     "diffusers",
     "PNDMScheduler"
@@ -22,8 +14,8 @@
     "CLIPTokenizer"
   ],
   "unet": [
-    "diffusers",
-    "UNet2DConditionModel"
+    "models",
+    "UNet3DConditionModel"
   ],
   "vae": [
     "diffusers",