Spaces:
Sleeping
Sleeping
space update
Browse files
app.py
CHANGED
@@ -189,6 +189,11 @@ def respond(
|
|
189 |
max_tokens,
|
190 |
temperature,
|
191 |
top_p,
|
|
|
|
|
|
|
|
|
|
|
192 |
):
|
193 |
"""
|
194 |
Generate a response using the UstaModel
|
@@ -234,44 +239,88 @@ def respond(
|
|
234 |
except Exception as e:
|
235 |
yield f"Sorry, I encountered an error: {str(e)}"
|
236 |
|
237 |
-
# Create
|
238 |
-
|
239 |
-
|
240 |
-
|
241 |
-
|
242 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
243 |
label="System message"
|
244 |
-
)
|
245 |
-
gr.Slider(minimum=1, maximum=30, value=20, step=1, label="Max new tokens")
|
246 |
-
gr.Slider(minimum=0.1, maximum=2.0, value=1.0, step=0.1, label="Temperature")
|
247 |
-
gr.Slider(
|
248 |
-
|
249 |
-
|
250 |
-
|
251 |
-
|
252 |
-
|
253 |
-
|
254 |
-
|
255 |
-
|
256 |
-
|
257 |
-
|
258 |
-
|
259 |
-
|
260 |
-
|
261 |
-
|
262 |
-
|
263 |
-
|
264 |
-
|
265 |
-
|
266 |
-
|
267 |
-
|
268 |
-
|
269 |
-
|
270 |
-
|
271 |
-
|
272 |
-
|
|
|
|
|
|
|
|
|
|
|
273 |
|
274 |
# Set up event handlers
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
275 |
load_file_btn.click(
|
276 |
load_model_from_file,
|
277 |
inputs=[model_file],
|
@@ -284,8 +333,5 @@ def setup_events():
|
|
284 |
outputs=[status_display]
|
285 |
)
|
286 |
|
287 |
-
# Set up events after interface creation
|
288 |
-
demo.load(setup_events)
|
289 |
-
|
290 |
if __name__ == "__main__":
|
291 |
demo.launch()
|
|
|
189 |
max_tokens,
|
190 |
temperature,
|
191 |
top_p,
|
192 |
+
model_file,
|
193 |
+
model_url,
|
194 |
+
load_file_btn,
|
195 |
+
load_url_btn,
|
196 |
+
status_display
|
197 |
):
|
198 |
"""
|
199 |
Generate a response using the UstaModel
|
|
|
239 |
except Exception as e:
|
240 |
yield f"Sorry, I encountered an error: {str(e)}"
|
241 |
|
242 |
+
# Create a Blocks interface to properly handle events
|
243 |
+
with gr.Blocks(title="π€ Usta Model Chat", theme=gr.themes.Soft()) as demo:
|
244 |
+
gr.Markdown("# π€ Usta Model Chat")
|
245 |
+
gr.Markdown("Chat with a custom transformer language model built from scratch! Upload your own model file or provide a URL to load a different model.")
|
246 |
+
|
247 |
+
# Model loading section
|
248 |
+
with gr.Accordion("π§ Model Loading Options", open=False):
|
249 |
+
with gr.Row():
|
250 |
+
with gr.Column():
|
251 |
+
gr.Markdown("### π Upload Model File")
|
252 |
+
model_file = gr.File(label="Upload Model File (.pth)", file_types=[".pth", ".pt"])
|
253 |
+
load_file_btn = gr.Button("Load from File", variant="primary")
|
254 |
+
|
255 |
+
with gr.Column():
|
256 |
+
gr.Markdown("### π Load from URL")
|
257 |
+
model_url = gr.Textbox(label="Model URL", placeholder="https://github.com/malibayram/llm-from-scratch/raw/main/u_model_4000.pth")
|
258 |
+
load_url_btn = gr.Button("Load from URL", variant="primary")
|
259 |
+
|
260 |
+
status_display = gr.Textbox(label="Model Status", value=model_status, interactive=False)
|
261 |
+
|
262 |
+
# Chat interface (simpler version)
|
263 |
+
chatbot = gr.Chatbot(label="Chat", type="messages")
|
264 |
+
msg = gr.Textbox(label="Message", placeholder="Type your message here...")
|
265 |
+
|
266 |
+
# Generation settings
|
267 |
+
with gr.Accordion("βοΈ Generation Settings", open=False):
|
268 |
+
system_msg = gr.Textbox(
|
269 |
+
value="You are Usta, a geographical knowledge assistant trained from scratch.",
|
270 |
label="System message"
|
271 |
+
)
|
272 |
+
max_tokens = gr.Slider(minimum=1, maximum=30, value=20, step=1, label="Max new tokens")
|
273 |
+
temperature = gr.Slider(minimum=0.1, maximum=2.0, value=1.0, step=0.1, label="Temperature")
|
274 |
+
top_p = gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p")
|
275 |
+
|
276 |
+
# Button controls
|
277 |
+
with gr.Row():
|
278 |
+
submit_btn = gr.Button("Send", variant="primary")
|
279 |
+
clear_btn = gr.Button("Clear Chat", variant="secondary")
|
280 |
+
|
281 |
+
# Event handlers
|
282 |
+
def chat_respond(message, history, sys_msg, max_tok, temp, top_p_val):
|
283 |
+
if not message.strip():
|
284 |
+
return history, ""
|
285 |
+
|
286 |
+
# Convert messages format for our respond function
|
287 |
+
tuple_history = [(h["role"], h["content"]) for h in history if h["role"] != "system"]
|
288 |
+
|
289 |
+
# Generate response using our existing function
|
290 |
+
response_gen = respond(
|
291 |
+
message, tuple_history, sys_msg, max_tok, temp, top_p_val,
|
292 |
+
None, None, None, None, None # Dummy values for unused params
|
293 |
+
)
|
294 |
+
|
295 |
+
# Get the response
|
296 |
+
response = ""
|
297 |
+
for r in response_gen:
|
298 |
+
response = r
|
299 |
+
|
300 |
+
# Add to history in messages format
|
301 |
+
history.append({"role": "user", "content": message})
|
302 |
+
history.append({"role": "assistant", "content": response})
|
303 |
+
|
304 |
+
return history, ""
|
305 |
|
306 |
# Set up event handlers
|
307 |
+
submit_btn.click(
|
308 |
+
chat_respond,
|
309 |
+
inputs=[msg, chatbot, system_msg, max_tokens, temperature, top_p],
|
310 |
+
outputs=[chatbot, msg]
|
311 |
+
)
|
312 |
+
|
313 |
+
msg.submit(
|
314 |
+
chat_respond,
|
315 |
+
inputs=[msg, chatbot, system_msg, max_tokens, temperature, top_p],
|
316 |
+
outputs=[chatbot, msg]
|
317 |
+
)
|
318 |
+
|
319 |
+
clear_btn.click(
|
320 |
+
lambda: ([], ""),
|
321 |
+
outputs=[chatbot, msg]
|
322 |
+
)
|
323 |
+
|
324 |
load_file_btn.click(
|
325 |
load_model_from_file,
|
326 |
inputs=[model_file],
|
|
|
333 |
outputs=[status_display]
|
334 |
)
|
335 |
|
|
|
|
|
|
|
336 |
if __name__ == "__main__":
|
337 |
demo.launch()
|