Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -208,155 +208,148 @@ with gr.Blocks(theme="Nymbo/Nymbo_Theme") as demo:
|
|
208 |
# Send button for messages
|
209 |
submit_btn = gr.Button("Send", variant="primary")
|
210 |
|
211 |
-
# Create
|
212 |
with gr.Accordion("Settings", open=False):
|
213 |
-
#
|
214 |
-
|
215 |
-
|
216 |
-
|
217 |
-
|
218 |
-
|
219 |
-
label="System Prompt"
|
220 |
-
)
|
221 |
-
|
222 |
-
# Generation parameters
|
223 |
-
with gr.Row():
|
224 |
-
with gr.Column():
|
225 |
-
max_tokens_slider = gr.Slider(
|
226 |
-
minimum=1,
|
227 |
-
maximum=4096,
|
228 |
-
value=512,
|
229 |
-
step=1,
|
230 |
-
label="Max tokens"
|
231 |
-
)
|
232 |
-
|
233 |
-
temperature_slider = gr.Slider(
|
234 |
-
minimum=0.1,
|
235 |
-
maximum=4.0,
|
236 |
-
value=0.7,
|
237 |
-
step=0.1,
|
238 |
-
label="Temperature"
|
239 |
-
)
|
240 |
-
|
241 |
-
with gr.Column():
|
242 |
-
top_p_slider = gr.Slider(
|
243 |
-
minimum=0.1,
|
244 |
-
maximum=1.0,
|
245 |
-
value=0.95,
|
246 |
-
step=0.05,
|
247 |
-
label="Top-P"
|
248 |
-
)
|
249 |
-
|
250 |
-
frequency_penalty_slider = gr.Slider(
|
251 |
-
minimum=-2.0,
|
252 |
-
maximum=2.0,
|
253 |
-
value=0.0,
|
254 |
-
step=0.1,
|
255 |
-
label="Frequency Penalty"
|
256 |
-
)
|
257 |
-
|
258 |
-
seed_slider = gr.Slider(
|
259 |
-
minimum=-1,
|
260 |
-
maximum=65535,
|
261 |
-
value=-1,
|
262 |
-
step=1,
|
263 |
-
label="Seed (-1 for random)"
|
264 |
-
)
|
265 |
|
266 |
-
#
|
267 |
-
with gr.
|
268 |
-
with gr.
|
269 |
-
|
270 |
-
|
271 |
-
|
272 |
-
|
273 |
-
|
274 |
-
|
275 |
-
|
276 |
-
"novita", # Novita AI
|
277 |
-
"cohere", # Cohere
|
278 |
-
"fireworks-ai", # Fireworks AI
|
279 |
-
"hyperbolic", # Hyperbolic
|
280 |
-
"nebius", # Nebius
|
281 |
-
]
|
282 |
-
|
283 |
-
provider_radio = gr.Radio(
|
284 |
-
choices=providers_list,
|
285 |
-
value="hf-inference",
|
286 |
-
label="Inference Provider",
|
287 |
-
info="[View all models here](https://huggingface.co/models?inference_provider=all&sort=trending)"
|
288 |
-
)
|
289 |
-
|
290 |
-
# New BYOK textbox
|
291 |
-
byok_textbox = gr.Textbox(
|
292 |
-
value="",
|
293 |
-
label="BYOK (Bring Your Own Key)",
|
294 |
-
info="Enter a custom Hugging Face API key here. When empty, only 'hf-inference' provider can be used.",
|
295 |
-
placeholder="Enter your Hugging Face API token",
|
296 |
-
type="password" # Hide the API key for security
|
297 |
-
)
|
298 |
-
|
299 |
-
with gr.Column():
|
300 |
-
# Custom model box
|
301 |
-
custom_model_box = gr.Textbox(
|
302 |
-
value="",
|
303 |
-
label="Custom Model",
|
304 |
-
info="(Optional) Provide a custom Hugging Face model path. Overrides any selected featured model.",
|
305 |
-
placeholder="meta-llama/Llama-3.3-70B-Instruct"
|
306 |
-
)
|
307 |
-
|
308 |
-
# Model search
|
309 |
-
model_search_box = gr.Textbox(
|
310 |
-
label="Filter Models",
|
311 |
-
placeholder="Search for a featured model...",
|
312 |
-
lines=1
|
313 |
-
)
|
314 |
-
|
315 |
-
# Featured models list
|
316 |
-
# Updated to include multimodal models
|
317 |
-
models_list = [
|
318 |
-
# Multimodal models
|
319 |
-
"meta-llama/Llama-3.3-70B-Vision",
|
320 |
-
"Alibaba-NLP/NephilaV-16B-Chat",
|
321 |
-
"mistralai/Mistral-Large-Vision-2407",
|
322 |
-
"OpenGVLab/InternVL-Chat-V1-5",
|
323 |
-
"microsoft/Phi-3.5-vision-instruct",
|
324 |
-
"Qwen/Qwen2.5-VL-7B-Instruct",
|
325 |
-
"liuhaotian/llava-v1.6-mistral-7b",
|
326 |
|
327 |
-
|
328 |
-
|
329 |
-
|
330 |
-
|
331 |
-
|
332 |
-
|
333 |
-
|
334 |
-
|
335 |
-
|
336 |
-
|
337 |
-
|
338 |
-
|
339 |
-
|
340 |
-
|
341 |
-
|
342 |
-
|
343 |
-
|
344 |
-
|
345 |
-
|
346 |
-
|
347 |
-
|
348 |
-
|
349 |
-
|
350 |
-
|
351 |
-
|
352 |
-
|
353 |
-
|
354 |
-
|
355 |
-
|
356 |
-
|
357 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
358 |
|
359 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
360 |
|
361 |
# Chat history state
|
362 |
chat_history = gr.State([])
|
|
|
208 |
# Send button for messages
|
209 |
submit_btn = gr.Button("Send", variant="primary")
|
210 |
|
211 |
+
# Create accordion for settings
|
212 |
with gr.Accordion("Settings", open=False):
|
213 |
+
# System message
|
214 |
+
system_message_box = gr.Textbox(
|
215 |
+
value="You are a helpful AI assistant that can understand images and text.",
|
216 |
+
placeholder="You are a helpful assistant.",
|
217 |
+
label="System Prompt"
|
218 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
219 |
|
220 |
+
# Generation parameters
|
221 |
+
with gr.Row():
|
222 |
+
with gr.Column():
|
223 |
+
max_tokens_slider = gr.Slider(
|
224 |
+
minimum=1,
|
225 |
+
maximum=4096,
|
226 |
+
value=512,
|
227 |
+
step=1,
|
228 |
+
label="Max tokens"
|
229 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
230 |
|
231 |
+
temperature_slider = gr.Slider(
|
232 |
+
minimum=0.1,
|
233 |
+
maximum=4.0,
|
234 |
+
value=0.7,
|
235 |
+
step=0.1,
|
236 |
+
label="Temperature"
|
237 |
+
)
|
238 |
+
|
239 |
+
top_p_slider = gr.Slider(
|
240 |
+
minimum=0.1,
|
241 |
+
maximum=1.0,
|
242 |
+
value=0.95,
|
243 |
+
step=0.05,
|
244 |
+
label="Top-P"
|
245 |
+
)
|
246 |
+
|
247 |
+
with gr.Column():
|
248 |
+
frequency_penalty_slider = gr.Slider(
|
249 |
+
minimum=-2.0,
|
250 |
+
maximum=2.0,
|
251 |
+
value=0.0,
|
252 |
+
step=0.1,
|
253 |
+
label="Frequency Penalty"
|
254 |
+
)
|
255 |
+
|
256 |
+
seed_slider = gr.Slider(
|
257 |
+
minimum=-1,
|
258 |
+
maximum=65535,
|
259 |
+
value=-1,
|
260 |
+
step=1,
|
261 |
+
label="Seed (-1 for random)"
|
262 |
+
)
|
263 |
+
|
264 |
+
# Provider selection
|
265 |
+
providers_list = [
|
266 |
+
"hf-inference", # Default Hugging Face Inference
|
267 |
+
"cerebras", # Cerebras provider
|
268 |
+
"together", # Together AI
|
269 |
+
"sambanova", # SambaNova
|
270 |
+
"novita", # Novita AI
|
271 |
+
"cohere", # Cohere
|
272 |
+
"fireworks-ai", # Fireworks AI
|
273 |
+
"hyperbolic", # Hyperbolic
|
274 |
+
"nebius", # Nebius
|
275 |
+
]
|
276 |
+
|
277 |
+
provider_radio = gr.Radio(
|
278 |
+
choices=providers_list,
|
279 |
+
value="hf-inference",
|
280 |
+
label="Inference Provider",
|
281 |
+
info="[View all models here](https://huggingface.co/models?inference_provider=all&sort=trending)"
|
282 |
+
)
|
283 |
+
|
284 |
+
# New BYOK textbox
|
285 |
+
byok_textbox = gr.Textbox(
|
286 |
+
value="",
|
287 |
+
label="BYOK (Bring Your Own Key)",
|
288 |
+
info="Enter a custom Hugging Face API key here. When empty, only 'hf-inference' provider can be used.",
|
289 |
+
placeholder="Enter your Hugging Face API token",
|
290 |
+
type="password" # Hide the API key for security
|
291 |
+
)
|
292 |
+
|
293 |
+
# Custom model box
|
294 |
+
custom_model_box = gr.Textbox(
|
295 |
+
value="",
|
296 |
+
label="Custom Model",
|
297 |
+
info="(Optional) Provide a custom Hugging Face model path. Overrides any selected featured model.",
|
298 |
+
placeholder="meta-llama/Llama-3.3-70B-Instruct"
|
299 |
+
)
|
300 |
+
|
301 |
+
# Model search
|
302 |
+
model_search_box = gr.Textbox(
|
303 |
+
label="Filter Models",
|
304 |
+
placeholder="Search for a featured model...",
|
305 |
+
lines=1
|
306 |
+
)
|
307 |
+
|
308 |
+
# Featured models list
|
309 |
+
# Updated to include multimodal models
|
310 |
+
models_list = [
|
311 |
+
# Multimodal models
|
312 |
+
"meta-llama/Llama-3.3-70B-Vision",
|
313 |
+
"Alibaba-NLP/NephilaV-16B-Chat",
|
314 |
+
"mistralai/Mistral-Large-Vision-2407",
|
315 |
+
"OpenGVLab/InternVL-Chat-V1-5",
|
316 |
+
"microsoft/Phi-3.5-vision-instruct",
|
317 |
+
"Qwen/Qwen2.5-VL-7B-Instruct",
|
318 |
+
"liuhaotian/llava-v1.6-mistral-7b",
|
319 |
|
320 |
+
# Standard text models
|
321 |
+
"meta-llama/Llama-3.3-70B-Instruct",
|
322 |
+
"meta-llama/Llama-3.1-70B-Instruct",
|
323 |
+
"meta-llama/Llama-3.0-70B-Instruct",
|
324 |
+
"meta-llama/Llama-3.2-3B-Instruct",
|
325 |
+
"meta-llama/Llama-3.2-1B-Instruct",
|
326 |
+
"meta-llama/Llama-3.1-8B-Instruct",
|
327 |
+
"NousResearch/Hermes-3-Llama-3.1-8B",
|
328 |
+
"NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO",
|
329 |
+
"mistralai/Mistral-Nemo-Instruct-2407",
|
330 |
+
"mistralai/Mixtral-8x7B-Instruct-v0.1",
|
331 |
+
"mistralai/Mistral-7B-Instruct-v0.3",
|
332 |
+
"mistralai/Mistral-7B-Instruct-v0.2",
|
333 |
+
"Qwen/Qwen3-235B-A22B",
|
334 |
+
"Qwen/Qwen3-32B",
|
335 |
+
"Qwen/Qwen2.5-72B-Instruct",
|
336 |
+
"Qwen/Qwen2.5-3B-Instruct",
|
337 |
+
"Qwen/Qwen2.5-0.5B-Instruct",
|
338 |
+
"Qwen/QwQ-32B",
|
339 |
+
"Qwen/Qwen2.5-Coder-32B-Instruct",
|
340 |
+
"microsoft/Phi-3.5-mini-instruct",
|
341 |
+
"microsoft/Phi-3-mini-128k-instruct",
|
342 |
+
"microsoft/Phi-3-mini-4k-instruct",
|
343 |
+
]
|
344 |
+
|
345 |
+
featured_model_radio = gr.Radio(
|
346 |
+
label="Select a model below",
|
347 |
+
choices=models_list,
|
348 |
+
value="meta-llama/Llama-3.3-70B-Vision", # Default to a multimodal model
|
349 |
+
interactive=True
|
350 |
+
)
|
351 |
+
|
352 |
+
gr.Markdown("[View all multimodal models](https://huggingface.co/models?pipeline_tag=image-to-text&sort=trending)")
|
353 |
|
354 |
# Chat history state
|
355 |
chat_history = gr.State([])
|