Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
|
@@ -157,18 +157,11 @@ def process_input(
|
|
| 157 |
def create_interface():
|
| 158 |
"""Create and configure Gradio interface"""
|
| 159 |
|
| 160 |
-
with gr.Blocks(title="MiniCPM-V-4.5 Multimodal Chat"
|
| 161 |
gr.Markdown("""
|
| 162 |
-
#
|
| 163 |
|
| 164 |
A powerful 8B parameter multimodal model that can understand images and videos with GPT-4V level performance.
|
| 165 |
-
|
| 166 |
-
**Features:**
|
| 167 |
-
- πΈ Single/Multi-image understanding
|
| 168 |
-
- π₯ High refresh rate video understanding (up to 10 FPS)
|
| 169 |
-
- π Strong OCR and document parsing
|
| 170 |
-
- π§ Controllable fast/deep thinking mode
|
| 171 |
-
- π Multilingual support (30+ languages)
|
| 172 |
""")
|
| 173 |
|
| 174 |
with gr.Row():
|
|
@@ -185,8 +178,7 @@ def create_interface():
|
|
| 185 |
maximum=30,
|
| 186 |
value=5,
|
| 187 |
step=1,
|
| 188 |
-
label="Video FPS"
|
| 189 |
-
info="Frames per second for video processing (only applies to videos)"
|
| 190 |
)
|
| 191 |
|
| 192 |
# Context size
|
|
@@ -195,8 +187,7 @@ def create_interface():
|
|
| 195 |
maximum=4096,
|
| 196 |
value=2048,
|
| 197 |
step=256,
|
| 198 |
-
label="Max Output Tokens"
|
| 199 |
-
info="Maximum number of tokens to generate"
|
| 200 |
)
|
| 201 |
|
| 202 |
# Temperature
|
|
@@ -205,15 +196,13 @@ def create_interface():
|
|
| 205 |
maximum=2.0,
|
| 206 |
value=0.7,
|
| 207 |
step=0.1,
|
| 208 |
-
label="Temperature"
|
| 209 |
-
info="Controls randomness in generation"
|
| 210 |
)
|
| 211 |
|
| 212 |
# Thinking mode
|
| 213 |
enable_thinking = gr.Checkbox(
|
| 214 |
label="Enable Deep Thinking",
|
| 215 |
-
value=False
|
| 216 |
-
info="Enable deep thinking mode for complex problem solving"
|
| 217 |
)
|
| 218 |
|
| 219 |
with gr.Column(scale=2):
|
|
@@ -221,8 +210,7 @@ def create_interface():
|
|
| 221 |
system_prompt = gr.Textbox(
|
| 222 |
label="System Prompt (Optional)",
|
| 223 |
placeholder="Enter system instructions here...",
|
| 224 |
-
lines=3
|
| 225 |
-
info="Set the behavior and context for the model"
|
| 226 |
)
|
| 227 |
|
| 228 |
# User prompt
|
|
@@ -233,31 +221,13 @@ def create_interface():
|
|
| 233 |
)
|
| 234 |
|
| 235 |
# Submit button
|
| 236 |
-
submit_btn = gr.Button("
|
| 237 |
|
| 238 |
# Output
|
| 239 |
output = gr.Textbox(
|
| 240 |
label="Model Response",
|
| 241 |
-
lines=15
|
| 242 |
-
max_lines=25,
|
| 243 |
-
show_copy_button=True
|
| 244 |
)
|
| 245 |
-
|
| 246 |
-
# Examples
|
| 247 |
-
gr.Markdown("## π‘ Example Prompts")
|
| 248 |
-
gr.Examples(
|
| 249 |
-
examples=[
|
| 250 |
-
["What objects do you see in this image?"],
|
| 251 |
-
["Describe the scene in detail."],
|
| 252 |
-
["What is the main action happening in this video?"],
|
| 253 |
-
["Read and transcribe any text visible in the image."],
|
| 254 |
-
["What emotions or mood does this image convey?"],
|
| 255 |
-
["Analyze the composition and visual elements."],
|
| 256 |
-
["What might happen next in this sequence?"]
|
| 257 |
-
],
|
| 258 |
-
inputs=[user_prompt],
|
| 259 |
-
label="Click any example to use it"
|
| 260 |
-
)
|
| 261 |
|
| 262 |
# Event handlers
|
| 263 |
submit_btn.click(
|
|
@@ -271,11 +241,9 @@ def create_interface():
|
|
| 271 |
temperature,
|
| 272 |
enable_thinking
|
| 273 |
],
|
| 274 |
-
outputs=output
|
| 275 |
-
show_progress=True
|
| 276 |
)
|
| 277 |
|
| 278 |
-
# Also allow Enter key submission
|
| 279 |
user_prompt.submit(
|
| 280 |
fn=process_input,
|
| 281 |
inputs=[
|
|
@@ -287,37 +255,12 @@ def create_interface():
|
|
| 287 |
temperature,
|
| 288 |
enable_thinking
|
| 289 |
],
|
| 290 |
-
outputs=output
|
| 291 |
-
show_progress=True
|
| 292 |
)
|
| 293 |
-
|
| 294 |
-
# Information section
|
| 295 |
-
with gr.Accordion("π Model Information", open=False):
|
| 296 |
-
gr.Markdown("""
|
| 297 |
-
### MiniCPM-V-4.5 Specifications
|
| 298 |
-
|
| 299 |
-
- **Parameters**: 8B (Qwen3-8B + SigLIP2-400M)
|
| 300 |
-
- **Video Compression**: 96x compression rate (6 frames β 64 tokens)
|
| 301 |
-
- **Max Resolution**: Up to 1.8M pixels (1344x1344)
|
| 302 |
-
- **Languages**: 30+ languages supported
|
| 303 |
-
- **Performance**: Surpasses GPT-4o-latest on multiple benchmarks
|
| 304 |
-
|
| 305 |
-
### Usage Tips
|
| 306 |
-
|
| 307 |
-
1. **For Images**: Upload any image format and ask questions about content, objects, text, or analysis
|
| 308 |
-
2. **For Videos**: Adjust FPS based on video content (higher FPS for action, lower for static scenes)
|
| 309 |
-
3. **System Prompt**: Use to set specific roles like "You are an expert art critic" or "Analyze this from a medical perspective"
|
| 310 |
-
4. **Deep Thinking**: Enable for complex reasoning tasks, analysis, or problem-solving
|
| 311 |
-
5. **Temperature**: Lower (0.1-0.3) for factual responses, higher (0.7-1.0) for creative outputs
|
| 312 |
-
|
| 313 |
-
### Supported Formats
|
| 314 |
-
- **Images**: JPG, PNG, JPEG, BMP, GIF, WEBP
|
| 315 |
-
- **Videos**: MP4, AVI, MOV, MKV, WEBM, M4V
|
| 316 |
-
""")
|
| 317 |
|
| 318 |
return iface
|
| 319 |
|
| 320 |
if __name__ == "__main__":
|
| 321 |
# Create and launch interface
|
| 322 |
demo = create_interface()
|
| 323 |
-
demo.launch()
|
|
|
|
| 157 |
def create_interface():
|
| 158 |
"""Create and configure Gradio interface"""
|
| 159 |
|
| 160 |
+
with gr.Blocks(title="MiniCPM-V-4.5 Multimodal Chat") as iface:
|
| 161 |
gr.Markdown("""
|
| 162 |
+
# MiniCPM-V-4.5 Multimodal Chat
|
| 163 |
|
| 164 |
A powerful 8B parameter multimodal model that can understand images and videos with GPT-4V level performance.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 165 |
""")
|
| 166 |
|
| 167 |
with gr.Row():
|
|
|
|
| 178 |
maximum=30,
|
| 179 |
value=5,
|
| 180 |
step=1,
|
| 181 |
+
label="Video FPS"
|
|
|
|
| 182 |
)
|
| 183 |
|
| 184 |
# Context size
|
|
|
|
| 187 |
maximum=4096,
|
| 188 |
value=2048,
|
| 189 |
step=256,
|
| 190 |
+
label="Max Output Tokens"
|
|
|
|
| 191 |
)
|
| 192 |
|
| 193 |
# Temperature
|
|
|
|
| 196 |
maximum=2.0,
|
| 197 |
value=0.7,
|
| 198 |
step=0.1,
|
| 199 |
+
label="Temperature"
|
|
|
|
| 200 |
)
|
| 201 |
|
| 202 |
# Thinking mode
|
| 203 |
enable_thinking = gr.Checkbox(
|
| 204 |
label="Enable Deep Thinking",
|
| 205 |
+
value=False
|
|
|
|
| 206 |
)
|
| 207 |
|
| 208 |
with gr.Column(scale=2):
|
|
|
|
| 210 |
system_prompt = gr.Textbox(
|
| 211 |
label="System Prompt (Optional)",
|
| 212 |
placeholder="Enter system instructions here...",
|
| 213 |
+
lines=3
|
|
|
|
| 214 |
)
|
| 215 |
|
| 216 |
# User prompt
|
|
|
|
| 221 |
)
|
| 222 |
|
| 223 |
# Submit button
|
| 224 |
+
submit_btn = gr.Button("Generate Response", variant="primary")
|
| 225 |
|
| 226 |
# Output
|
| 227 |
output = gr.Textbox(
|
| 228 |
label="Model Response",
|
| 229 |
+
lines=15
|
|
|
|
|
|
|
| 230 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 231 |
|
| 232 |
# Event handlers
|
| 233 |
submit_btn.click(
|
|
|
|
| 241 |
temperature,
|
| 242 |
enable_thinking
|
| 243 |
],
|
| 244 |
+
outputs=output
|
|
|
|
| 245 |
)
|
| 246 |
|
|
|
|
| 247 |
user_prompt.submit(
|
| 248 |
fn=process_input,
|
| 249 |
inputs=[
|
|
|
|
| 255 |
temperature,
|
| 256 |
enable_thinking
|
| 257 |
],
|
| 258 |
+
outputs=output
|
|
|
|
| 259 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 260 |
|
| 261 |
return iface
|
| 262 |
|
| 263 |
if __name__ == "__main__":
|
| 264 |
# Create and launch interface
|
| 265 |
demo = create_interface()
|
| 266 |
+
demo.launch(share=True)
|