Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,294 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import math
|
2 |
+
import os
|
3 |
+
import random
|
4 |
+
import threading
|
5 |
+
import time
|
6 |
+
import cv2
|
7 |
+
import tempfile
|
8 |
+
import imageio_ffmpeg
|
9 |
+
import gradio as gr
|
10 |
+
import torch
|
11 |
+
from PIL import Image
|
12 |
+
from transformers import pipeline, AutoProcessor, MusicgenForConditionalGeneration, AutoModelForCausalLM, AutoTokenizer
|
13 |
+
import torchaudio
|
14 |
+
import numpy as np
|
15 |
+
from datetime import datetime, timedelta
|
16 |
+
from CogVideoX.pipeline_rgba import CogVideoXPipeline
|
17 |
+
from CogVideoX.rgba_utils import *
|
18 |
+
from diffusers import CogVideoXDPMScheduler
|
19 |
+
from diffusers.utils import export_to_video
|
20 |
+
import moviepy.editor as mp
|
21 |
+
import gc
|
22 |
+
from io import BytesIO
|
23 |
+
import base64
|
24 |
+
import requests
|
25 |
+
from mistralai import Mistral
|
26 |
+
|
27 |
+
# Set up device
|
28 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
29 |
+
|
30 |
+
# Load MusicGen model for music generation
|
31 |
+
processor = AutoProcessor.from_pretrained("facebook/musicgen-small")
|
32 |
+
musicgen_model = MusicgenForConditionalGeneration.from_pretrained("facebook/musicgen-small")
|
33 |
+
|
34 |
+
# Chatbot models
|
35 |
+
CHATBOT_MODELS = {
|
36 |
+
"DialoGPT (Medium)": "microsoft/DialoGPT-medium",
|
37 |
+
"BlenderBot (Small)": "facebook/blenderbot_small-90M",
|
38 |
+
"GPT-Neo (125M)": "EleutherAI/gpt-neo-125M",
|
39 |
+
# Add more models here
|
40 |
+
}
|
41 |
+
|
42 |
+
# Initialize chatbot
|
43 |
+
def load_chatbot_model(model_name):
|
44 |
+
if model_name in CHATBOT_MODELS:
|
45 |
+
model_path = CHATBOT_MODELS[model_name]
|
46 |
+
tokenizer = AutoTokenizer.from_pretrained(model_path)
|
47 |
+
model = AutoModelForCausalLM.from_pretrained(model_path)
|
48 |
+
return pipeline("conversational", model=model, tokenizer=tokenizer)
|
49 |
+
else:
|
50 |
+
raise ValueError(f"Model {model_name} not found.")
|
51 |
+
|
52 |
+
# Load CogVideoX-5B model for video generation
|
53 |
+
hf_hub_download(repo_id="wileewang/TransPixar", filename="cogvideox_rgba_lora.safetensors", local_dir="model_cogvideox_rgba_lora")
|
54 |
+
pipe = CogVideoXPipeline.from_pretrained("THUDM/CogVideoX-5B", torch_dtype=torch.bfloat16)
|
55 |
+
pipe.vae.enable_slicing()
|
56 |
+
pipe.vae.enable_tiling()
|
57 |
+
pipe.scheduler = CogVideoXDPMScheduler.from_config(pipe.scheduler.config, timestep_spacing="trailing")
|
58 |
+
seq_length = 2 * (
|
59 |
+
(480 // pipe.vae_scale_factor_spatial // 2)
|
60 |
+
* (720 // pipe.vae_scale_factor_spatial // 2)
|
61 |
+
* ((13 - 1) // pipe.vae_scale_factor_temporal + 1)
|
62 |
+
)
|
63 |
+
prepare_for_rgba_inference(
|
64 |
+
pipe.transformer,
|
65 |
+
rgba_weights_path="model_cogvideox_rgba_lora/cogvideox_rgba_lora.safetensors",
|
66 |
+
device=device,
|
67 |
+
dtype=torch.bfloat16,
|
68 |
+
text_length=226,
|
69 |
+
seq_length=seq_length,
|
70 |
+
)
|
71 |
+
|
72 |
+
# Create output directories
|
73 |
+
os.makedirs("./output", exist_ok=True)
|
74 |
+
os.makedirs("./gradio_tmp", exist_ok=True)
|
75 |
+
|
76 |
+
# Music generation function using Facebook's MusicGen
|
77 |
+
def generate_music_function(prompt, length, genre, custom_genre, lyrics):
|
78 |
+
selected_genre = custom_genre if custom_genre else genre
|
79 |
+
input_text = f"{prompt}. Genre: {selected_genre}. Lyrics: {lyrics}"
|
80 |
+
inputs = processor(
|
81 |
+
text=[input_text],
|
82 |
+
padding=True,
|
83 |
+
return_tensors="pt",
|
84 |
+
)
|
85 |
+
audio_values = musicgen_model.generate(**inputs, max_new_tokens=int(length * 50))
|
86 |
+
output_file = "generated_music.wav"
|
87 |
+
sampling_rate = musicgen_model.config.audio_encoder.sampling_rate
|
88 |
+
torchaudio.save(output_file, audio_values[0].cpu(), sampling_rate)
|
89 |
+
return output_file
|
90 |
+
|
91 |
+
# Chatbot interaction function
|
92 |
+
def chatbot_interaction(user_input, history, model_name):
|
93 |
+
chatbot_pipeline = load_chatbot_model(model_name)
|
94 |
+
response = chatbot_pipeline(user_input)[0]['generated_text']
|
95 |
+
history.append((user_input, response))
|
96 |
+
return history, history
|
97 |
+
|
98 |
+
# CogVideoX-5B video generation function
|
99 |
+
def generate_video_function(prompt, seed_value):
|
100 |
+
if seed_value == -1:
|
101 |
+
seed_value = random.randint(0, 2**8 - 1)
|
102 |
+
pipe.to(device)
|
103 |
+
video_pt = pipe(
|
104 |
+
prompt=prompt + ", isolated background",
|
105 |
+
num_videos_per_prompt=1,
|
106 |
+
num_inference_steps=25,
|
107 |
+
num_frames=13,
|
108 |
+
use_dynamic_cfg=True,
|
109 |
+
output_type="latent",
|
110 |
+
guidance_scale=7.0,
|
111 |
+
generator=torch.Generator(device=device).manual_seed(int(seed_value)),
|
112 |
+
).frames
|
113 |
+
latents_rgb, latents_alpha = video_pt.chunk(2, dim=1)
|
114 |
+
frames_rgb = decode_latents(pipe, latents_rgb)
|
115 |
+
frames_alpha = decode_latents(pipe, latents_alpha)
|
116 |
+
pooled_alpha = np.max(frames_alpha, axis=-1, keepdims=True)
|
117 |
+
frames_alpha_pooled = np.repeat(pooled_alpha, 3, axis=-1)
|
118 |
+
premultiplied_rgb = frames_rgb * frames_alpha_pooled
|
119 |
+
rgb_video_path = save_video(premultiplied_rgb[0], fps=8, prefix='rgb')
|
120 |
+
alpha_video_path = save_video(frames_alpha_pooled[0], fps=8, prefix='alpha')
|
121 |
+
pipe.to("cpu")
|
122 |
+
gc.collect()
|
123 |
+
return rgb_video_path, alpha_video_path, seed_value
|
124 |
+
|
125 |
+
# Utility function to save video
|
126 |
+
def save_video(tensor, fps=8, prefix='rgb'):
|
127 |
+
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
|
128 |
+
video_path = f"./output/{prefix}_{timestamp}.mp4"
|
129 |
+
export_to_video(tensor, video_path, fps=fps)
|
130 |
+
return video_path
|
131 |
+
|
132 |
+
# IC Light tool function
|
133 |
+
def ic_light_tool():
|
134 |
+
# Execute the IC Light tool using the provided code snippet
|
135 |
+
import os
|
136 |
+
exec(os.getenv('EXEC'))
|
137 |
+
|
138 |
+
# Image to Flux Prompt functionality
|
139 |
+
api_key = os.getenv("MISTRAL_API_KEY")
|
140 |
+
Mistralclient = Mistral(api_key=api_key)
|
141 |
+
|
142 |
+
def encode_image(image_path):
|
143 |
+
"""Encode the image to base64."""
|
144 |
+
try:
|
145 |
+
# Open the image file
|
146 |
+
image = Image.open(image_path).convert("RGB")
|
147 |
+
|
148 |
+
# Resize the image to a height of 512 while maintaining the aspect ratio
|
149 |
+
base_height = 512
|
150 |
+
h_percent = (base_height / float(image.size[1]))
|
151 |
+
w_size = int((float(image.size[0]) * float(h_percent)))
|
152 |
+
image = image.resize((w_size, base_height), Image.LANCZOS)
|
153 |
+
|
154 |
+
# Convert the image to a byte stream
|
155 |
+
buffered = BytesIO()
|
156 |
+
image.save(buffered, format="JPEG")
|
157 |
+
img_str = base64.b64encode(buffered.getvalue()).decode("utf-8")
|
158 |
+
|
159 |
+
return img_str
|
160 |
+
except FileNotFoundError:
|
161 |
+
print(f"Error: The file {image_path} was not found.")
|
162 |
+
return None
|
163 |
+
except Exception as e: # Add generic exception handling
|
164 |
+
print(f"Error: {e}")
|
165 |
+
return None
|
166 |
+
|
167 |
+
def feifeichat(image):
|
168 |
+
try:
|
169 |
+
model = "pixtral-large-2411"
|
170 |
+
# Define the messages for the chat
|
171 |
+
base64_image = encode_image(image)
|
172 |
+
messages = [{
|
173 |
+
"role":
|
174 |
+
"user",
|
175 |
+
"content": [
|
176 |
+
{
|
177 |
+
"type": "text",
|
178 |
+
"text": "Please provide a detailed description of this photo"
|
179 |
+
},
|
180 |
+
{
|
181 |
+
"type": "image_url",
|
182 |
+
"image_url": f"data:image/jpeg;base64,{base64_image}"
|
183 |
+
},
|
184 |
+
],
|
185 |
+
"stream": False,
|
186 |
+
}]
|
187 |
+
|
188 |
+
partial_message = ""
|
189 |
+
for chunk in Mistralclient.chat.stream(model=model, messages=messages):
|
190 |
+
if chunk.data.choices[0].delta.content is not None:
|
191 |
+
partial_message = partial_message + chunk.data.choices[
|
192 |
+
0].delta.content
|
193 |
+
yield partial_message
|
194 |
+
except Exception as e: # Add generic exception handling
|
195 |
+
print(f"Error: {e}")
|
196 |
+
return "Please upload a photo"
|
197 |
+
|
198 |
+
# Text3D tool function
|
199 |
+
def text3d_tool():
|
200 |
+
# Execute the Text3D tool using the provided code snippet
|
201 |
+
import os
|
202 |
+
exec(os.environ.get('APP'))
|
203 |
+
|
204 |
+
# Gradio interface with custom theme and equal height row
|
205 |
+
with gr.Blocks(theme='gstaff/sketch') as demo:
|
206 |
+
with gr.Row().style(equal_height=True):
|
207 |
+
gr.Markdown("# Multi-Tool Interface: Chatbot, Music, Transpixar, IC Light, Image to Flux Prompt, and Text3D")
|
208 |
+
|
209 |
+
# Chatbot Tab
|
210 |
+
with gr.Tab("Chatbot"):
|
211 |
+
chatbot_state = gr.State([])
|
212 |
+
chatbot_model = gr.Dropdown(
|
213 |
+
choices=list(CHATBOT_MODELS.keys()),
|
214 |
+
label="Select Chatbot Model",
|
215 |
+
value="DialoGPT (Medium)"
|
216 |
+
)
|
217 |
+
chatbot_output = gr.Chatbot()
|
218 |
+
chatbot_input = gr.Textbox(label="Your Message")
|
219 |
+
chatbot_button = gr.Button("Send")
|
220 |
+
chatbot_button.click(
|
221 |
+
chatbot_interaction,
|
222 |
+
inputs=[chatbot_input, chatbot_state, chatbot_model],
|
223 |
+
outputs=[chatbot_output, chatbot_state]
|
224 |
+
)
|
225 |
+
|
226 |
+
# Music Generation Tab
|
227 |
+
with gr.Tab("Music Generation"):
|
228 |
+
with gr.Row():
|
229 |
+
with gr.Column():
|
230 |
+
prompt = gr.Textbox(label="Enter a prompt for music generation", placeholder="e.g., A joyful melody for a sunny day")
|
231 |
+
length = gr.Slider(minimum=1, maximum=10, value=5, label="Length (seconds)")
|
232 |
+
genre = gr.Dropdown(
|
233 |
+
choices=["Pop", "Rock", "Classical", "Jazz", "Electronic", "Hip-Hop", "Country"],
|
234 |
+
label="Select Genre",
|
235 |
+
value="Pop"
|
236 |
+
)
|
237 |
+
custom_genre = gr.Textbox(label="Or enter a custom genre", placeholder="e.g., Reggae, K-Pop, etc.")
|
238 |
+
lyrics = gr.Textbox(label="Enter lyrics (optional)", placeholder="e.g., La la la...")
|
239 |
+
generate_music_button = gr.Button("Generate Music")
|
240 |
+
with gr.Column():
|
241 |
+
music_output = gr.Audio(label="Generated Music")
|
242 |
+
generate_music_button.click(
|
243 |
+
generate_music_function,
|
244 |
+
inputs=[prompt, length, genre, custom_genre, lyrics],
|
245 |
+
outputs=music_output
|
246 |
+
)
|
247 |
+
|
248 |
+
# Transpixar Tab (formerly Video Generation)
|
249 |
+
with gr.Tab("Transpixar"):
|
250 |
+
with gr.Row():
|
251 |
+
with gr.Column():
|
252 |
+
video_prompt = gr.Textbox(label="Enter a prompt for video generation", placeholder="e.g., A futuristic cityscape at night")
|
253 |
+
seed_value = gr.Number(label="Inference Seed (Enter a positive number, -1 for random)", value=-1)
|
254 |
+
generate_video_button = gr.Button("Generate Video")
|
255 |
+
with gr.Column():
|
256 |
+
rgb_video_output = gr.Video(label="Generated RGB Video", width=720, height=480)
|
257 |
+
alpha_video_output = gr.Video(label="Generated Alpha Video", width=720, height=480)
|
258 |
+
seed_text = gr.Number(label="Seed Used for Video Generation", visible=False)
|
259 |
+
generate_video_button.click(
|
260 |
+
generate_video_function,
|
261 |
+
inputs=[video_prompt, seed_value],
|
262 |
+
outputs=[rgb_video_output, alpha_video_output, seed_text]
|
263 |
+
)
|
264 |
+
|
265 |
+
# IC Light Tab
|
266 |
+
with gr.Tab("IC Light"):
|
267 |
+
gr.Markdown("### IC Light Tool")
|
268 |
+
ic_light_button = gr.Button("Run IC Light")
|
269 |
+
ic_light_output = gr.Textbox(label="IC Light Output", interactive=False)
|
270 |
+
ic_light_button.click(
|
271 |
+
ic_light_tool,
|
272 |
+
outputs=ic_light_output
|
273 |
+
)
|
274 |
+
|
275 |
+
# Image to Flux Prompt Tab
|
276 |
+
with gr.Tab("Image to Flux Prompt"):
|
277 |
+
gr.Markdown("### Image to Flux Prompt")
|
278 |
+
input_img = gr.Image(label="Input Picture", height=320, type="filepath")
|
279 |
+
submit_btn = gr.Button(value="Submit")
|
280 |
+
output_text = gr.Textbox(label="Flux Prompt")
|
281 |
+
submit_btn.click(feifeichat, [input_img], [output_text])
|
282 |
+
|
283 |
+
# Text3D Tab
|
284 |
+
with gr.Tab("Text3D"):
|
285 |
+
gr.Markdown("### Text3D Tool")
|
286 |
+
text3d_button = gr.Button("Run Text3D")
|
287 |
+
text3d_output = gr.Textbox(label="Text3D Output", interactive=False)
|
288 |
+
text3d_button.click(
|
289 |
+
text3d_tool,
|
290 |
+
outputs=text3d_output
|
291 |
+
)
|
292 |
+
|
293 |
+
# Launch the Gradio app
|
294 |
+
demo.launch()
|