Update app.py

app.py

@@ -13,7 +13,7 @@ from diffusers.utils import load_image
 from PIL import Image
 
 import torch
-
+'''
 torch.backends.cuda.matmul.allow_tf32 = False
 torch.backends.cuda.matmul.allow_bf16_reduced_precision_reduction = False
 torch.backends.cuda.matmul.allow_fp16_reduced_precision_reduction = False
@@ -23,7 +23,7 @@ torch.backends.cudnn.benchmark = False
 torch.backends.cuda.preferred_blas_library="cublas"
 torch.backends.cuda.preferred_linalg_library="cusolver"
 torch.set_float32_matmul_precision("highest")
-
+'''
 os.putenv("HF_HUB_ENABLE_HF_TRANSFER","1")
 os.environ["SAFETENSORS_FAST_GPU"] = "1"
 os.putenv("TOKENIZERS_PARALLELISM","False")
@@ -42,12 +42,11 @@ def init_predictor():
         )
     )
     
-@spaces.GPU(duration=120)
-def generate_video(prompt, seed, image, size, steps, frames, progress=gr.Progress(track_tqdm=True) ):
+@spaces.GPU(duration=100)
+def generate_video(prompt, image, size, steps, frames, guidance_scale, progress=gr.Progress(track_tqdm=True) ):
     print(f"image:{type(image)}")
-    if seed == -1:
-        random.seed(time.time())
-        seed = int(random.randrange(4294967294))
+    random.seed(time.time())
+    seed = int(random.randrange(4294967294))
     kwargs = {
         "prompt": prompt,
         "height": size,
@@ -55,7 +54,7 @@ def generate_video(prompt, seed, image, size, steps, frames, progress=gr.Progres
         "num_frames": frames,
         "num_inference_steps": steps,
         "seed": seed,
-        "guidance_scale": 6.0,
+        "guidance_scale": guidance_scale,
         "embedded_guidance_scale": 1.0,
         "negative_prompt": "Aerial view, aerial view, overexposed, low quality, deformation, a poor composition, bad hands, bad teeth, bad eyes, bad limbs, distortion",
         "cfg_for": False,
@@ -76,20 +75,19 @@ def create_gradio_interface():
             with gr.Row():
                 image = gr.Image(label="Upload Image", type="filepath")
                 prompt = gr.Textbox(label="Input Prompt")
-                seed = gr.Number(label="Random Seed", value=-1)
                 size = gr.Slider(
                     label="Size",
                     minimum=256,
                     maximum=1024,
                     step=16,
-                    value=448,
+                    value=368,
                 )
                 frames = gr.Slider(
                     label="Number of Frames",
                     minimum=16,
                     maximum=256,
-                    step=8,
-                    value=64,
+                    step=12,
+                    value=48,
                 )
                 steps = gr.Slider(
                     label="Number of Steps",
@@ -98,11 +96,18 @@ def create_gradio_interface():
                     step=1,
                     value=20,
                 )
+                guidance_scale = gr.Slider(
+                    label="Guidance Scale",
+                    minimum=1.0,
+                    maximum=16.0,
+                    step=.1,
+                    value=6.0,
+                )
             submit_button = gr.Button("Generate Video")
             output_video = gr.Video(label="Generated Video")
             submit_button.click(
                 fn=generate_video,
-                inputs=[prompt, seed, image, size, steps, frames],
+                inputs=[prompt, image, size, steps, frames, guidance_scale],
                 outputs=[output_video],
             )
         return demo