Decraft / app.py
Felguk's picture
Update app.py
529506e verified
import math
import os
import random
import threading
import time
import cv2
import tempfile
import imageio_ffmpeg
import gradio as gr
import torch
from PIL import Image
from transformers import pipeline, AutoProcessor, MusicgenForCausalLM, AutoModelForCausalLM, AutoTokenizer
import torchaudio
import numpy as np
from datetime import datetime, timedelta
from CogVideoX.pipeline_rgba import CogVideoXPipeline
from CogVideoX.rgba_utils import *
from diffusers import CogVideoXDPMScheduler
from diffusers.utils import export_to_video
import moviepy.editor as mp
import gc
from io import BytesIO
import base64
import requests
from mistralai import Mistral
from huggingface_hub import hf_hub_download
# Set up device
device = "cuda" if torch.cuda.is_available() else "cpu"
# Load MusicGen model for music generation
processor = AutoProcessor.from_pretrained("facebook/musicgen-small")
model = MusicgenForCausalLM.from_pretrained("facebook/musicgen-small")
# Explicitly set configurations to avoid conflicts
model.config.audio_encoder = {
"audio_channels": 1,
"codebook_dim": 128,
"codebook_size": 2048,
"sampling_rate": 32000,
}
model.config.decoder = {
"activation_dropout": 0.0,
"activation_function": "gelu",
"attention_dropout": 0.0,
}
# Chatbot models
CHATBOT_MODELS = {
"DialoGPT (Medium)": "microsoft/DialoGPT-medium",
"BlenderBot (Small)": "facebook/blenderbot_small-90M",
"GPT-Neo (125M)": "EleutherAI/gpt-neo-125M",
# Add more models here
}
# Initialize chatbot
def load_chatbot_model(model_name):
if model_name in CHATBOT_MODELS:
model_path = CHATBOT_MODELS[model_name]
tokenizer = AutoTokenizer.from_pretrained(model_path)
model = AutoModelForCausalLM.from_pretrained(model_path)
return pipeline("conversational", model=model, tokenizer=tokenizer)
else:
raise ValueError(f"Model {model_name} not found.")
# Load CogVideoX-5B model for video generation
hf_hub_download(repo_id="wileewang/TransPixar", filename="cogvideox_rgba_lora.safetensors", local_dir="model_cogvideox_rgba_lora")
pipe = CogVideoXPipeline.from_pretrained("THUDM/CogVideoX-5B", torch_dtype=torch.bfloat16)
pipe.vae.enable_slicing()
pipe.vae.enable_tiling()
pipe.scheduler = CogVideoXDPMScheduler.from_config(pipe.scheduler.config, timestep_spacing="trailing")
seq_length = 2 * (
(480 // pipe.vae_scale_factor_spatial // 2)
* (720 // pipe.vae_scale_factor_spatial // 2)
* ((13 - 1) // pipe.vae_scale_factor_temporal + 1)
)
prepare_for_rgba_inference(
pipe.transformer,
rgba_weights_path="model_cogvideox_rgba_lora/cogvideox_rgba_lora.safetensors",
device=device,
dtype=torch.bfloat16,
text_length=226,
seq_length=seq_length,
)
# Create output directories
os.makedirs("./output", exist_ok=True)
os.makedirs("./gradio_tmp", exist_ok=True)
# Music generation function using Facebook's MusicGen
def generate_music_function(prompt, length, genre, custom_genre, lyrics):
selected_genre = custom_genre if custom_genre else genre
input_text = f"{prompt}. Genre: {selected_genre}. Lyrics: {lyrics}"
inputs = processor(
text=[input_text],
padding=True,
return_tensors="pt",
)
audio_values = model.generate(**inputs, max_new_tokens=int(length * 50))
output_file = "generated_music.wav"
sampling_rate = model.config.audio_encoder["sampling_rate"]
torchaudio.save(output_file, audio_values[0].cpu(), sampling_rate)
return output_file
# Chatbot interaction function
def chatbot_interaction(user_input, history, model_name):
chatbot_pipeline = load_chatbot_model(model_name)
response = chatbot_pipeline(user_input)[0]['generated_text']
history.append((user_input, response))
return history, history
# CogVideoX-5B video generation function
def generate_video_function(prompt, seed_value):
if seed_value == -1:
seed_value = random.randint(0, 2**8 - 1)
pipe.to(device)
video_pt = pipe(
prompt=prompt + ", isolated background",
num_videos_per_prompt=1,
num_inference_steps=25,
num_frames=13,
use_dynamic_cfg=True,
output_type="latent",
guidance_scale=7.0,
generator=torch.Generator(device=device).manual_seed(int(seed_value)),
).frames
latents_rgb, latents_alpha = video_pt.chunk(2, dim=1)
frames_rgb = decode_latents(pipe, latents_rgb)
frames_alpha = decode_latents(pipe, latents_alpha)
pooled_alpha = np.max(frames_alpha, axis=-1, keepdims=True)
frames_alpha_pooled = np.repeat(pooled_alpha, 3, axis=-1)
premultiplied_rgb = frames_rgb * frames_alpha_pooled
rgb_video_path = save_video(premultiplied_rgb[0], fps=8, prefix='rgb')
alpha_video_path = save_video(frames_alpha_pooled[0], fps=8, prefix='alpha')
pipe.to("cpu")
gc.collect()
return rgb_video_path, alpha_video_path, seed_value
# Utility function to save video
def save_video(tensor, fps=8, prefix='rgb'):
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
video_path = f"./output/{prefix}_{timestamp}.mp4"
export_to_video(tensor, video_path, fps=fps)
return video_path
# IC Light tool function
def ic_light_tool():
# Execute the IC Light tool using the provided code snippet
import os
exec(os.getenv('EXEC'))
# Image to Flux Prompt functionality
api_key = os.getenv("MISTRAL_API_KEY")
Mistralclient = Mistral(api_key=api_key)
def encode_image(image_path):
"""Encode the image to base64."""
try:
# Open the image file
image = Image.open(image_path).convert("RGB")
# Resize the image to a height of 512 while maintaining the aspect ratio
base_height = 512
h_percent = (base_height / float(image.size[1]))
w_size = int((float(image.size[0]) * float(h_percent)))
image = image.resize((w_size, base_height), Image.LANCZOS)
# Convert the image to a byte stream
buffered = BytesIO()
image.save(buffered, format="JPEG")
img_str = base64.b64encode(buffered.getvalue()).decode("utf-8")
return img_str
except FileNotFoundError:
print(f"Error: The file {image_path} was not found.")
return None
except Exception as e: # Add generic exception handling
print(f"Error: {e}")
return None
def feifeichat(image):
try:
model = "pixtral-large-2411"
# Define the messages for the chat
base64_image = encode_image(image)
messages = [{
"role":
"user",
"content": [
{
"type": "text",
"text": "Please provide a detailed description of this photo"
},
{
"type": "image_url",
"image_url": f"data:image/jpeg;base64,{base64_image}"
},
],
"stream": False,
}]
partial_message = ""
for chunk in Mistralclient.chat.stream(model=model, messages=messages):
if chunk.data.choices[0].delta.content is not None:
partial_message = partial_message + chunk.data.choices[
0].delta.content
yield partial_message
except Exception as e: # Add generic exception handling
print(f"Error: {e}")
return "Please upload a photo"
# Text3D tool function
def text3d_tool():
# Execute the Text3D tool using the provided code snippet
import os
exec(os.environ.get('APP'))
# Gradio interface with custom theme and equal height row
with gr.Blocks(theme='gstaff/sketch') as demo:
with gr.Row(equal_height=True): # Fix: Use equal_height parameter
gr.Markdown("# Multi-Tool Interface: Chatbot, Music, Transpixar, IC Light, Image to Flux Prompt, and Text3D")
# Chatbot Tab
with gr.Tab("Chatbot"):
chatbot_state = gr.State([])
chatbot_model = gr.Dropdown(
choices=list(CHATBOT_MODELS.keys()),
label="Select Chatbot Model",
value="DialoGPT (Medium)"
)
chatbot_output = gr.Chatbot()
chatbot_input = gr.Textbox(label="Your Message")
chatbot_button = gr.Button("Send")
chatbot_button.click(
chatbot_interaction,
inputs=[chatbot_input, chatbot_state, chatbot_model],
outputs=[chatbot_output, chatbot_state]
)
# Music Generation Tab
with gr.Tab("Music Generation"):
with gr.Row():
with gr.Column():
prompt = gr.Textbox(label="Enter a prompt for music generation", placeholder="e.g., A joyful melody for a sunny day")
length = gr.Slider(minimum=1, maximum=10, value=5, label="Length (seconds)")
genre = gr.Dropdown(
choices=["Pop", "Rock", "Classical", "Jazz", "Electronic", "Hip-Hop", "Country"],
label="Select Genre",
value="Pop"
)
custom_genre = gr.Textbox(label="Or enter a custom genre", placeholder="e.g., Reggae, K-Pop, etc.")
lyrics = gr.Textbox(label="Enter lyrics (optional)", placeholder="e.g., La la la...")
generate_music_button = gr.Button("Generate Music")
with gr.Column():
music_output = gr.Audio(label="Generated Music")
generate_music_button.click(
generate_music_function,
inputs=[prompt, length, genre, custom_genre, lyrics],
outputs=music_output
)
# Transpixar Tab (formerly Video Generation)
with gr.Tab("Transpixar"):
with gr.Row():
with gr.Column():
video_prompt = gr.Textbox(label="Enter a prompt for video generation", placeholder="e.g., A futuristic cityscape at night")
seed_value = gr.Number(label="Inference Seed (Enter a positive number, -1 for random)", value=-1)
generate_video_button = gr.Button("Generate Video")
with gr.Column():
rgb_video_output = gr.Video(label="Generated RGB Video", width=720, height=480)
alpha_video_output = gr.Video(label="Generated Alpha Video", width=720, height=480)
seed_text = gr.Number(label="Seed Used for Video Generation", visible=False)
generate_video_button.click(
generate_video_function,
inputs=[video_prompt, seed_value],
outputs=[rgb_video_output, alpha_video_output, seed_text]
)
# IC Light Tab
with gr.Tab("IC Light"):
gr.Markdown("### IC Light Tool")
ic_light_button = gr.Button("Run IC Light")
ic_light_output = gr.Textbox(label="IC Light Output", interactive=False)
ic_light_button.click(
ic_light_tool,
outputs=ic_light_output
)
# Image to Flux Prompt Tab
with gr.Tab("Image to Flux Prompt"):
gr.Markdown("### Image to Flux Prompt")
input_img = gr.Image(label="Input Picture", height=320, type="filepath")
submit_btn = gr.Button(value="Submit")
output_text = gr.Textbox(label="Flux Prompt")
submit_btn.click(feifeichat, [input_img], [output_text])
# Text3D Tab
with gr.Tab("Text3D"):
gr.Markdown("### Text3D Tool")
text3d_button = gr.Button("Run Text3D")
text3d_output = gr.Textbox(label="Text3D Output", interactive=False)
text3d_button.click(
text3d_tool,
outputs=text3d_output
)
# Launch the Gradio app
demo.launch()