faster inference using new diffusers update. vae tiling and cpu offloading, much faster inference on zero gpu.

app.py

@@ -4,54 +4,47 @@ import threading
 import time
 
 import gradio as gr
-import numpy as np
 import torch
-from diffusers import CogVideoXPipeline
+import gc
+from diffusers import CogVideoXPipeline, CogVideoXDDIMScheduler
+from diffusers.utils import export_to_video
 from datetime import datetime, timedelta
 from openai import OpenAI
 import spaces
-import imageio
 import moviepy.editor as mp
 from typing import List, Union
 import PIL
 
+def reset_memory():
+    gc.collect()
+    torch.cuda.empty_cache()
+    torch.cuda.reset_accumulated_memory_stats()
+    torch.cuda.reset_peak_memory_stats()
+
+def print_memory():
+    memory = round(torch.cuda.memory_allocated() / 1024**3, 2)
+    max_memory = round(torch.cuda.max_memory_allocated() / 1024**3, 2)
+    max_reserved = round(torch.cuda.max_memory_reserved() / 1024**3, 2)
+    print(f"{memory=} GB")
+    print(f"{max_memory=} GB")
+    print(f"{max_reserved=} GB")
+
 dtype = torch.float16
 device = "cuda" if torch.cuda.is_available() else "cpu"
-pipe = CogVideoXPipeline.from_pretrained("THUDM/CogVideoX-2b", torch_dtype=dtype).to(device)
+pipe = CogVideoXPipeline.from_pretrained("THUDM/CogVideoX-2b", torch_dtype=dtype)
+pipe.scheduler = CogVideoXDDIMScheduler.from_config(pipe.scheduler.config, timestep_spacing="trailing")
+pipe.enable_sequential_cpu_offload()
+pipe.vae.enable_tiling()
 
 sys_prompt = """You are part of a team of bots that creates videos. You work with an assistant bot that will draw anything you say in square brackets.
-
 For example , outputting " a beautiful morning in the woods with the sun peaking through the trees " will trigger your partner bot to output an video of a forest morning , as described. You will be prompted by people looking to create detailed , amazing videos. The way to accomplish this is to take their short prompts and make them extremely detailed and descriptive.
 There are a few rules to follow:
-
 You will only ever output a single video description per user request.
-
 When modifications are requested , you should not simply make the description longer . You should refactor the entire description to integrate the suggestions.
 Other times the user will not want modifications , but instead want a new image . In this case , you should ignore your previous conversation with the user.
-
 Video descriptions must have the same num of words as examples below. Extra words will be ignored.
 """
 
-
-def export_to_video_imageio(
-        video_frames: Union[List[np.ndarray], List[PIL.Image.Image]], output_video_path: str = None, fps: int = 8
-) -> str:
-    """
-    Export the video frames to a video file using imageio lib to Avoid "green screen" issue (for example CogVideoX)
-    """
-    if output_video_path is None:
-        output_video_path = tempfile.NamedTemporaryFile(suffix=".mp4").name
-
-    if isinstance(video_frames[0], PIL.Image.Image):
-        video_frames = [np.array(frame) for frame in video_frames]
-
-    with imageio.get_writer(output_video_path, fps=fps) as writer:
-        for frame in video_frames:
-            writer.append_data(frame)
-
-    return output_video_path
-
-
 def convert_prompt(prompt: str, retry_times: int = 3) -> str:
     if not os.environ.get("OPENAI_API_KEY"):
         return prompt
@@ -87,7 +80,6 @@ def convert_prompt(prompt: str, retry_times: int = 3) -> str:
             return response.choices[0].message.content
     return prompt
 
-
 @spaces.GPU(duration=240)
 def infer(
         prompt: str,
@@ -95,34 +87,25 @@ def infer(
         guidance_scale: float,
         progress=gr.Progress(track_tqdm=True)
 ):
-    torch.cuda.empty_cache()
-
-    prompt_embeds, _ = pipe.encode_prompt(
-        prompt=prompt,
-        negative_prompt=None,
-        do_classifier_free_guidance=True,
-        num_videos_per_prompt=1,
-        max_sequence_length=226,
-        device=device,
-        dtype=dtype,
-    )
-
-    video = pipe(
-        num_inference_steps=num_inference_steps,
-        guidance_scale=guidance_scale,
-        prompt_embeds=prompt_embeds,
-        negative_prompt_embeds=torch.zeros_like(prompt_embeds),
-    ).frames[0]
-
-
+    reset_memory()
+
+    with torch.cuda.amp.autocast():
+        video = pipe(
+            prompt=prompt,
+            num_frames=48,
+            guidance_scale=guidance_scale,
+            num_inference_steps=num_inference_steps,
+            generator=torch.Generator().manual_seed(42)
+        ).frames[0]
+
+    print_memory()
     return video
 
-
 def save_video(tensor):
     timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
     video_path = f"./output/{timestamp}.mp4"
     os.makedirs(os.path.dirname(video_path), exist_ok=True)
-    export_to_video_imageio(tensor[1:], video_path)
+    export_to_video(tensor, video_path, fps=8)
     return video_path
 
 def convert_to_gif(video_path):
@@ -133,7 +116,6 @@ def convert_to_gif(video_path):
     clip.write_gif(gif_path, fps=8)
     return gif_path
 
-
 def delete_old_files():
     while True:
         now = datetime.now()
@@ -147,7 +129,6 @@ def delete_old_files():
                     os.remove(file_path)
         time.sleep(600)  # Sleep for 10 minutes
 
-
 threading.Thread(target=delete_old_files, daemon=True).start()
 
 with gr.Blocks() as demo:
@@ -160,7 +141,6 @@ with gr.Blocks() as demo:
                <a href="https://github.com/THUDM/CogVideo">🌐 Github</a> |
                <a href="https://arxiv.org/pdf/2408.06072">📜 arxiv </a>
            </div>
-
            <div style="text-align: center; font-size: 15px; font-weight: bold; color: red; margin-bottom: 20px;">
             ⚠️ This demo is for academic research and experiential use only. 
             Users should strictly adhere to local laws and ethics.
@@ -177,9 +157,9 @@ with gr.Blocks() as demo:
             with gr.Column():
                 gr.Markdown("**Optional Parameters** (default values are recommended)<br>"
                             "Turn Inference Steps larger if you want more detailed video, but it will be slower.<br>"
-                            "50 steps are recommended for most cases. will cause 120 seconds for inference.<br>")
+                            "30 steps are recommended for most cases. will cause about 60 seconds for inference.<br>")
                 with gr.Row():
-                    num_inference_steps = gr.Number(label="Inference Steps", value=50)
+                    num_inference_steps = gr.Number(label="Inference Steps", value=30)
                     guidance_scale = gr.Number(label="Guidance Scale", value=6.0)
                 generate_button = gr.Button("🎬 Generate Video")
 
@@ -200,31 +180,30 @@ with gr.Blocks() as demo:
             <tr>
                 <td>A detailed wooden toy ship with intricately carved masts and sails is seen gliding smoothly over a plush, blue carpet that mimics the waves of the sea. The ship's hull is painted a rich brown, with tiny windows. The carpet, soft and textured, provides a perfect backdrop, resembling an oceanic expanse. Surrounding the ship are various other toys and children's items, hinting at a playful environment. The scene captures the innocence and imagination of childhood, with the toy ship's journey symbolizing endless adventures in a whimsical, indoor setting.</td>
                 <td><a href="https://github.com/THUDM/CogVideo/raw/main/resources/videos/1.mp4">Video 1</a></td>
-                <td>50</td>
+                <td>30</td>
                 <td>6</td>
             </tr>
             <tr>
-                <td>The camera follows behind a white vintage SUV with a black roof rack as it speeds up a steep dirt road surrounded by pine trees on a steep mountain slope, dust kicks up from it’s tires, the sunlight shines on the SUV as it speeds along the dirt road, casting a warm glow over the scene. The dirt road curves gently into the distance, with no other cars or vehicles in sight. The trees on either side of the road are redwoods, with patches of greenery scattered throughout. The car is seen from the rear following the curve with ease, making it seem as if it is on a rugged drive through the rugged terrain. The dirt road itself is surrounded by steep hills and mountains, with a clear blue sky above with wispy clouds.</td>
+                <td>The camera follows behind a white vintage SUV with a black roof rack as it speeds up a steep dirt road surrounded by pine trees on a steep mountain slope, dust kicks up from it's tires, the sunlight shines on the SUV as it speeds along the dirt road, casting a warm glow over the scene. The dirt road curves gently into the distance, with no other cars or vehicles in sight. The trees on either side of the road are redwoods, with patches of greenery scattered throughout. The car is seen from the rear following the curve with ease, making it seem as if it is on a rugged drive through the rugged terrain. The dirt road itself is surrounded by steep hills and mountains, with a clear blue sky above with wispy clouds.</td>
                 <td><a href="https://github.com/THUDM/CogVideo/raw/main/resources/videos/2.mp4">Video 2</a></td>
-                <td>50</td>
+                <td>30</td>
                 <td>6</td>
             </tr>
             <tr>
                 <td>A street artist, clad in a worn-out denim jacket and a colorful bandana, stands before a vast concrete wall in the heart, holding a can of spray paint, spray-painting a colorful bird on a mottled wall.</td>
                 <td><a href="https://github.com/THUDM/CogVideo/raw/main/resources/videos/3.mp4">Video 3</a></td>
-                <td>50</td>
+                <td>30</td>
                 <td>6</td>
             </tr>
             <tr>
                 <td>In the haunting backdrop of a war-torn city, where ruins and crumbled walls tell a story of devastation, a poignant close-up frames a young girl. Her face is smudged with ash, a silent testament to the chaos around her. Her eyes glistening with a mix of sorrow and resilience, capturing the raw emotion of a world that has lost its innocence to the ravages of conflict.</td>
                 <td><a href="https://github.com/THUDM/CogVideo/raw/main/resources/videos/4.mp4">Video 4</a></td>
-                <td>50</td>
+                <td>30</td>
                 <td>6</td>
             </tr>
         </table>
     """)
 
-
     def generate(prompt, num_inference_steps, guidance_scale, progress=gr.Progress(track_tqdm=True)):
         tensor = infer(prompt, num_inference_steps, guidance_scale, progress=progress)
         video_path = save_video(tensor)
@@ -234,11 +213,9 @@ with gr.Blocks() as demo:
 
         return video_path, video_update, gif_update
 
-
     def enhance_prompt_func(prompt):
         return convert_prompt(prompt, retry_times=1)
 
-
     generate_button.click(
         generate,
         inputs=[prompt, num_inference_steps, guidance_scale],
@@ -252,4 +229,4 @@ with gr.Blocks() as demo:
     )
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()