remove video

app.py

@@ -4,14 +4,9 @@ import torch.nn as nn
 import torch.nn.functional as F
 from transformers import AutoTokenizer, SiglipTextModel
 from peft import LoraConfig, get_peft_model
-import random
 from torchvision.transforms import ToPILImage
-import numpy as np
-from moviepy.editor import ImageSequenceClip
 import random
 import gradio as gr
-import tempfile
-import os
 
 class SimpleAdapter(nn.Module):
     def __init__(self, input_dim=512, hidden_dim=1024, out_dim=1024):
@@ -117,35 +112,6 @@ class InrenceTextVAR(nn.Module):
         image = ToPILImage()(generated_images[0].cpu())
         return image
 
-    @torch.no_grad()
-    def generate_video(self, text, start_beta, target_beta, fps, length, top_k=0, top_p=0.9, seed=None,
-                       more_smooth=False,
-                       output_filename='output_video.mp4'):
-
-        if seed is None:
-            seed = random.randint(0, 2 ** 32 - 1)
-
-        num_frames = int(fps * length)
-        images = []
-
-        # Define an easing function for smoother interpolation
-        def ease_in_out(t):
-            return t * t * (3 - 2 * t)
-
-        # Generate t values between 0 and 1
-        t_values = np.linspace(0, 1, num_frames)
-        # Apply the easing function
-        eased_t_values = ease_in_out(t_values)
-        # Interpolate beta values using the eased t values
-        beta_values = start_beta + (target_beta - start_beta) * eased_t_values
-
-        for beta in beta_values:
-            image = self.generate_image(text, beta=beta, seed=seed, more_smooth=more_smooth, top_k=top_k, top_p=top_p)
-            images.append(np.array(image))
-
-        # Create a video from images
-        clip = ImageSequenceClip(images, fps=fps)
-        clip.write_videofile(output_filename, codec='libx264')
 
 if __name__ == '__main__':
 
@@ -167,37 +133,6 @@ if __name__ == '__main__':
         image = model.generate_image(text, beta=beta, seed=seed, more_smooth=more_smooth, top_k=int(top_k), top_p=top_p)
         return image
 
-    def generate_video_gradio(text, start_beta=1.0, target_beta=1.0, fps=10, length=5.0, top_k=0, top_p=0.9, seed=None, more_smooth=False, progress=gr.Progress()):
-        print(f"Generating video for text: {text}\n"
-              f"start_beta: {start_beta}\n"
-              f"target_beta: {target_beta}\n"
-              f"seed: {seed}\n"
-              f"more_smooth: {more_smooth}\n"
-              f"top_k: {top_k}\n"
-              f"top_p: {top_p}"
-              f"fps: {fps}\n"
-              f"length: {length}\n")
-        with tempfile.NamedTemporaryFile(suffix='.mp4', delete=False) as tmpfile:
-            output_filename = tmpfile.name
-        num_frames = int(fps * length)
-        beta_values = np.linspace(start_beta, target_beta, num_frames)
-        images = []
-
-        for i, beta in enumerate(beta_values):
-            image = model.generate_image(text, beta=beta, seed=seed, more_smooth=more_smooth, top_k=top_k, top_p=top_p)
-            images.append(np.array(image))
-            # Update progress
-            progress((i + 1) / num_frames)
-            # Yield the frame image to update the GUI
-            yield image, gr.update()
-
-        # After generating all frames, create the video
-        clip = ImageSequenceClip(images, fps=fps)
-        clip.write_videofile(output_filename, codec='libx264')
-
-        # Yield the final video output
-        yield gr.update(), output_filename
-
     with gr.Blocks() as demo:
         gr.Markdown("# Text to Image/Video Generator")
         with gr.Tab("Generate Image"):
@@ -215,25 +150,4 @@ if __name__ == '__main__':
                 outputs=image_output
             )
 
-        with gr.Tab("Generate Video"):
-            text_input_video = gr.Textbox(label="Input Text")
-            start_beta_input = gr.Slider(label="Start Beta", minimum=0.0, maximum=2.5, step=0.05, value=0)
-            target_beta_input = gr.Slider(label="Target Beta",minimum=0.0, maximum=2.5, step=0.05, value=1.0)
-            fps_input = gr.Number(label="FPS", value=10)
-            length_input = gr.Number(label="Length (seconds)", value=5.0)
-            seed_input_video = gr.Number(label="Seed", value=None)
-            more_smooth_input_video = gr.Checkbox(label="More Smooth", value=False)
-            top_k_input_video = gr.Number(label="Top K", value=0)
-            top_p_input_video = gr.Slider(label="Top P", minimum=0.0, maximum=1.0, step=0.01, value=0.9)
-            generate_video_button = gr.Button("Generate Video")
-            frame_output = gr.Image(label="Current Frame")
-            video_output = gr.Video(label="Generated Video")
-
-            generate_video_button.click(
-                generate_video_gradio,
-                inputs=[text_input_video, start_beta_input, target_beta_input, fps_input, length_input, top_k_input_video, top_p_input_video, seed_input_video, more_smooth_input_video],
-                outputs=[frame_output, video_output],
-                queue=True  # Enable queuing to allow for progress updates
-            )
-
     demo.launch()