change-weights-path

VADER-VideoCrafter/scripts/main/train_t2v_lora.py

@@ -580,12 +580,14 @@ def run_training(args, model, **kwargs):
 
     # load the pretrained LoRA model
     if args.lora_ckpt_path != "Base Model":
-        if args.lora_ckpt_path == "huggingface-hps-aesthetic":  # download the pretrained LoRA model from huggingface
-            snapshot_download(repo_id='zheyangqin/VADER', local_dir ='VADER-VideoCrafter/checkpoints/pretrained_lora')
-            args.lora_ckpt_path = 'VADER-VideoCrafter/checkpoints/pretrained_lora/vader_videocrafter_hps_aesthetic.pt'
-        elif args.lora_ckpt_path == "huggingface-pickscore":    # download the pretrained LoRA model from huggingface
-            snapshot_download(repo_id='zheyangqin/VADER', local_dir ='VADER-VideoCrafter/checkpoints/pretrained_lora')
-            args.lora_ckpt_path = 'VADER-VideoCrafter/checkpoints/pretrained_lora/vader_videocrafter_pickscore.pt'
+        if args.lora_ckpt_path == "huggingface-pickscore":  # download the pretrained LoRA model from huggingface
+            os.makedirs('VADER-VideoCrafter/checkpoints/pretrained_lora_pickScore', exist_ok=True)
+            snapshot_download(repo_id='zheyangqin/VADER_VideoCrafter_PickScore', local_dir ='VADER-VideoCrafter/checkpoints/pretrained_lora_pickScore')
+            args.lora_ckpt_path = 'VADER-VideoCrafter/checkpoints/pretrained_lora_pickScore/vader_videocrafter_pickscore.pt'
+        elif args.lora_ckpt_path == "huggingface-hps-aesthetic":    # download the pretrained LoRA model from huggingface
+            os.makedirs('VADER-VideoCrafter/checkpoints/pretrained_lora_hps_aesthetic', exist_ok=True)
+            snapshot_download(repo_id='zheyangqin/VADER_VideoCrafter_HPS_Aesthetic', local_dir ='VADER-VideoCrafter/checkpoints/pretrained_lora_hps_aesthetic')
+            args.lora_ckpt_path = 'VADER-VideoCrafter/checkpoints/pretrained_lora_hps_aesthetic/vader_videocrafter_hps_aesthetic.pt'
         # load the pretrained LoRA model
         peft.set_peft_model_state_dict(peft_model, torch.load(args.lora_ckpt_path))
 

app.py

@@ -51,7 +51,7 @@ def gradio_main_fn(prompt, lora_model, lora_rank, seed, height, width, unconditi
     return video_path
 
 def reset_fn():
-    return ("A mermaid with flowing hair and a shimmering tail discovers a hidden underwater kingdom adorned with coral palaces, glowing pearls, and schools of colorful fish, encountering both wonders and dangers along the way.", 
+    return ("A brown dog eagerly eats from a bowl in a kitchen.", 
             200, 384, 512, 12.0, 25, 1.0, 24, 16, 10, "huggingface-pickscore")
 
 def update_lora_rank(lora_model):
@@ -198,7 +198,7 @@ with gr.Blocks(css=custom_css) as demo:
             )
             lora_rank = gr.Slider(minimum=8, maximum=16, label="LoRA Rank", step = 8, value=16)
             prompt = gr.Textbox(placeholder="Enter prompt text here", lines=4, label="Text Prompt",
-                                value="A mermaid with flowing hair and a shimmering tail discovers a hidden underwater kingdom adorned with coral palaces, glowing pearls, and schools of colorful fish, encountering both wonders and dangers along the way.")
+                                value="A brown dog eagerly eats from a bowl in a kitchen.")
             run_btn = gr.Button("Run Inference")
 
         with gr.Column():