Update app.py
Browse files
app.py
CHANGED
@@ -128,47 +128,48 @@ def flag_last_response(state, model_selector, request: gr.Request):
|
|
128 |
return ("",) + (disable_btn,) * 3
|
129 |
|
130 |
|
131 |
-
def regenerate(state,
|
132 |
logger.info(f"regenerate. ip: {request.client.host}")
|
133 |
state.messages[-1][-1] = None
|
134 |
prev_human_msg = state.messages[-2]
|
135 |
if type(prev_human_msg[1]) in (tuple, list):
|
136 |
-
prev_human_msg[1] = (*prev_human_msg[1][:2],
|
137 |
state.skip_next = False
|
138 |
-
return (state, state.to_gradio_chatbot(), "", None
|
139 |
|
140 |
|
141 |
def clear_history(request: gr.Request):
|
142 |
logger.info(f"clear_history. ip: {request.client.host}")
|
143 |
state = default_conversation.copy()
|
144 |
-
return (state, state.to_gradio_chatbot(), "", None
|
145 |
|
146 |
|
147 |
-
def add_text(state, text,
|
148 |
logger.info(f"add_text. ip: {request.client.host}. len: {len(text)}")
|
149 |
-
if len(text) <= 0 and
|
150 |
state.skip_next = True
|
151 |
-
return (state, state.to_gradio_chatbot(), "", None
|
152 |
if args.moderate:
|
153 |
flagged = violates_moderation(text)
|
154 |
if flagged:
|
155 |
state.skip_next = True
|
156 |
-
return (state, state.to_gradio_chatbot(), moderation_msg, None
|
157 |
no_change_btn,
|
158 |
) * 5
|
159 |
|
160 |
text = text[:1536] # Hard cut-off
|
161 |
-
if
|
162 |
text = text[:1200] # Hard cut-off for images
|
163 |
if "<image>" not in text:
|
|
|
164 |
text = text + "\n<image>"
|
165 |
-
text = (text,
|
166 |
if len(state.get_images(return_pil=True)) > 0:
|
167 |
state = default_conversation.copy()
|
168 |
state.append_message(state.roles[0], text)
|
169 |
state.append_message(state.roles[1], None)
|
170 |
state.skip_next = False
|
171 |
-
return (state, state.to_gradio_chatbot(), "", None
|
172 |
|
173 |
|
174 |
def http_bot(
|
@@ -179,10 +180,12 @@ def http_bot(
|
|
179 |
model_name = model_selector
|
180 |
|
181 |
if state.skip_next:
|
|
|
182 |
yield (state, state.to_gradio_chatbot()) + (no_change_btn,) * 5
|
183 |
return
|
184 |
|
185 |
if len(state.messages) == state.offset + 2:
|
|
|
186 |
if "llava" in model_name.lower():
|
187 |
if "llama-2" in model_name.lower():
|
188 |
template_name = "llava_llama_2"
|
@@ -219,6 +222,7 @@ def http_bot(
|
|
219 |
new_state.append_message(new_state.roles[1], None)
|
220 |
state = new_state
|
221 |
|
|
|
222 |
controller_url = args.controller_url
|
223 |
ret = requests.post(
|
224 |
controller_url + "/get_worker_address", json={"model": model_name}
|
@@ -226,6 +230,7 @@ def http_bot(
|
|
226 |
worker_addr = ret.json()["address"]
|
227 |
logger.info(f"model_name: {model_name}, worker_addr: {worker_addr}")
|
228 |
|
|
|
229 |
if worker_addr == "":
|
230 |
state.messages[-1][-1] = server_error_msg
|
231 |
yield (
|
@@ -239,6 +244,7 @@ def http_bot(
|
|
239 |
)
|
240 |
return
|
241 |
|
|
|
242 |
prompt = state.get_prompt()
|
243 |
|
244 |
all_images = state.get_images(return_pil=True)
|
@@ -252,6 +258,7 @@ def http_bot(
|
|
252 |
os.makedirs(os.path.dirname(filename), exist_ok=True)
|
253 |
image.save(filename)
|
254 |
|
|
|
255 |
pload = {
|
256 |
"model": model_name,
|
257 |
"prompt": prompt,
|
@@ -271,6 +278,7 @@ def http_bot(
|
|
271 |
yield (state, state.to_gradio_chatbot()) + (disable_btn,) * 5
|
272 |
|
273 |
try:
|
|
|
274 |
response = requests.post(
|
275 |
worker_addr + "/worker_generate_stream",
|
276 |
headers=headers,
|
@@ -331,13 +339,17 @@ def http_bot(
|
|
331 |
title_markdown = """
|
332 |
# 🌋 AI Tutor Vision: Large Language and Vision Assistant
|
333 |
[[website]](https://myapps.ai) [[Paper]](https://arxiv.org/abs/2304.08485) [[Model]](https://github.com/haotian-liu/LLaVA/blob/main/docs/MODEL_ZOO.md)
|
|
|
334 |
ONLY WORKS WITH GPU!
|
|
|
335 |
You can load the model with 4-bit or 8-bit quantization to make it fit in smaller hardwares. Setting the environment variable `bits` to control the quantization.
|
336 |
*Note: 8-bit seems to be slower than both 4-bit/16-bit. Although it has enough VRAM to support 8-bit, until we figure out the inference speed issue, we recommend 4-bit for A10G for the best efficiency.*
|
|
|
337 |
Recommended configurations:
|
338 |
| Hardware | T4-Small (16G) | A10G-Small (24G) | A100-Large (40G) |
|
339 |
|-------------------|-----------------|------------------|------------------|
|
340 |
| **Bits** | 4 (default) | 4 | 16 |
|
|
|
341 |
"""
|
342 |
|
343 |
tos_markdown = """
|
@@ -355,9 +367,11 @@ The service is a research preview intended for non-commercial use only, subject
|
|
355 |
"""
|
356 |
|
357 |
block_css = """
|
|
|
358 |
#buttons button {
|
359 |
min-width: min(120px,100%);
|
360 |
}
|
|
|
361 |
"""
|
362 |
|
363 |
|
@@ -384,15 +398,8 @@ def build_demo(embed_mode):
|
|
384 |
container=False,
|
385 |
)
|
386 |
|
387 |
-
|
388 |
-
|
389 |
-
image_process_mode1 = gr.Radio(
|
390 |
-
["Crop", "Resize", "Pad", "Default"],
|
391 |
-
value="Default",
|
392 |
-
label="Preprocess for non-square image",
|
393 |
-
visible=False,
|
394 |
-
)
|
395 |
-
image_process_mode2 = gr.Radio(
|
396 |
["Crop", "Resize", "Pad", "Default"],
|
397 |
value="Default",
|
398 |
label="Preprocess for non-square image",
|
@@ -411,7 +418,7 @@ def build_demo(embed_mode):
|
|
411 |
"What are the things I should be cautious about when I visit here?",
|
412 |
],
|
413 |
],
|
414 |
-
inputs=[
|
415 |
)
|
416 |
|
417 |
with gr.Accordion("Parameters", open=False) as parameter_row:
|
@@ -442,18 +449,20 @@ def build_demo(embed_mode):
|
|
442 |
|
443 |
with gr.Column(scale=8):
|
444 |
chatbot = gr.Chatbot(
|
445 |
-
elem_id="chatbot", label="AI Tutor Vision
|
446 |
)
|
447 |
with gr.Row():
|
448 |
with gr.Column(scale=8):
|
449 |
textbox.render()
|
450 |
with gr.Column(scale=1, min_width=50):
|
451 |
submit_btn = gr.Button(
|
452 |
-
value="Send", variant="primary", interactive=False
|
|
|
453 |
with gr.Row(elem_id="buttons") as button_row:
|
454 |
upvote_btn = gr.Button(value="👍 Upvote", interactive=False)
|
455 |
downvote_btn = gr.Button(value="👎 Downvote", interactive=False)
|
456 |
flag_btn = gr.Button(value="⚠️ Flag", interactive=False)
|
|
|
457 |
regenerate_btn = gr.Button(value="🔄 Regenerate", interactive=False)
|
458 |
clear_btn = gr.Button(value="🗑️ Clear history", interactive=False)
|
459 |
|
@@ -462,6 +471,7 @@ def build_demo(embed_mode):
|
|
462 |
gr.Markdown(learn_more_markdown)
|
463 |
url_params = gr.JSON(visible=False)
|
464 |
|
|
|
465 |
btn_list = [upvote_btn, downvote_btn, flag_btn, regenerate_btn, clear_btn]
|
466 |
upvote_btn.click(
|
467 |
upvote_last_response,
|
@@ -480,21 +490,21 @@ def build_demo(embed_mode):
|
|
480 |
)
|
481 |
regenerate_btn.click(
|
482 |
regenerate,
|
483 |
-
[state,
|
484 |
-
[state, chatbot, textbox,
|
485 |
).then(
|
486 |
http_bot,
|
487 |
[state, model_selector, temperature, top_p, max_output_tokens],
|
488 |
[state, chatbot] + btn_list,
|
489 |
)
|
490 |
clear_btn.click(
|
491 |
-
clear_history, None, [state, chatbot, textbox,
|
492 |
)
|
493 |
|
494 |
textbox.submit(
|
495 |
add_text,
|
496 |
-
[state, textbox,
|
497 |
-
[state, chatbot, textbox,
|
498 |
).then(
|
499 |
http_bot,
|
500 |
[state, model_selector, temperature, top_p, max_output_tokens],
|
@@ -502,8 +512,8 @@ def build_demo(embed_mode):
|
|
502 |
)
|
503 |
submit_btn.click(
|
504 |
add_text,
|
505 |
-
[state, textbox,
|
506 |
-
[state, chatbot, textbox,
|
507 |
).then(
|
508 |
http_bot,
|
509 |
[state, model_selector, temperature, top_p, max_output_tokens],
|
@@ -600,6 +610,7 @@ if __name__ == "__main__":
|
|
600 |
controller_proc = start_controller()
|
601 |
worker_proc = start_worker(model_path, bits=bits)
|
602 |
|
|
|
603 |
time.sleep(10)
|
604 |
|
605 |
exit_status = 0
|
@@ -612,4 +623,4 @@ if __name__ == "__main__":
|
|
612 |
worker_proc.kill()
|
613 |
controller_proc.kill()
|
614 |
|
615 |
-
sys.exit(exit_status)
|
|
|
128 |
return ("",) + (disable_btn,) * 3
|
129 |
|
130 |
|
131 |
+
def regenerate(state, image_process_mode, request: gr.Request):
|
132 |
logger.info(f"regenerate. ip: {request.client.host}")
|
133 |
state.messages[-1][-1] = None
|
134 |
prev_human_msg = state.messages[-2]
|
135 |
if type(prev_human_msg[1]) in (tuple, list):
|
136 |
+
prev_human_msg[1] = (*prev_human_msg[1][:2], image_process_mode)
|
137 |
state.skip_next = False
|
138 |
+
return (state, state.to_gradio_chatbot(), "", None) + (disable_btn,) * 5
|
139 |
|
140 |
|
141 |
def clear_history(request: gr.Request):
|
142 |
logger.info(f"clear_history. ip: {request.client.host}")
|
143 |
state = default_conversation.copy()
|
144 |
+
return (state, state.to_gradio_chatbot(), "", None) + (disable_btn,) * 5
|
145 |
|
146 |
|
147 |
+
def add_text(state, text, image, image_process_mode, request: gr.Request):
|
148 |
logger.info(f"add_text. ip: {request.client.host}. len: {len(text)}")
|
149 |
+
if len(text) <= 0 and image is None:
|
150 |
state.skip_next = True
|
151 |
+
return (state, state.to_gradio_chatbot(), "", None) + (no_change_btn,) * 5
|
152 |
if args.moderate:
|
153 |
flagged = violates_moderation(text)
|
154 |
if flagged:
|
155 |
state.skip_next = True
|
156 |
+
return (state, state.to_gradio_chatbot(), moderation_msg, None) + (
|
157 |
no_change_btn,
|
158 |
) * 5
|
159 |
|
160 |
text = text[:1536] # Hard cut-off
|
161 |
+
if image is not None:
|
162 |
text = text[:1200] # Hard cut-off for images
|
163 |
if "<image>" not in text:
|
164 |
+
# text = '<Image><image></Image>' + text
|
165 |
text = text + "\n<image>"
|
166 |
+
text = (text, image, image_process_mode)
|
167 |
if len(state.get_images(return_pil=True)) > 0:
|
168 |
state = default_conversation.copy()
|
169 |
state.append_message(state.roles[0], text)
|
170 |
state.append_message(state.roles[1], None)
|
171 |
state.skip_next = False
|
172 |
+
return (state, state.to_gradio_chatbot(), "", None) + (disable_btn,) * 5
|
173 |
|
174 |
|
175 |
def http_bot(
|
|
|
180 |
model_name = model_selector
|
181 |
|
182 |
if state.skip_next:
|
183 |
+
# This generate call is skipped due to invalid inputs
|
184 |
yield (state, state.to_gradio_chatbot()) + (no_change_btn,) * 5
|
185 |
return
|
186 |
|
187 |
if len(state.messages) == state.offset + 2:
|
188 |
+
# First round of conversation
|
189 |
if "llava" in model_name.lower():
|
190 |
if "llama-2" in model_name.lower():
|
191 |
template_name = "llava_llama_2"
|
|
|
222 |
new_state.append_message(new_state.roles[1], None)
|
223 |
state = new_state
|
224 |
|
225 |
+
# Query worker address
|
226 |
controller_url = args.controller_url
|
227 |
ret = requests.post(
|
228 |
controller_url + "/get_worker_address", json={"model": model_name}
|
|
|
230 |
worker_addr = ret.json()["address"]
|
231 |
logger.info(f"model_name: {model_name}, worker_addr: {worker_addr}")
|
232 |
|
233 |
+
# No available worker
|
234 |
if worker_addr == "":
|
235 |
state.messages[-1][-1] = server_error_msg
|
236 |
yield (
|
|
|
244 |
)
|
245 |
return
|
246 |
|
247 |
+
# Construct prompt
|
248 |
prompt = state.get_prompt()
|
249 |
|
250 |
all_images = state.get_images(return_pil=True)
|
|
|
258 |
os.makedirs(os.path.dirname(filename), exist_ok=True)
|
259 |
image.save(filename)
|
260 |
|
261 |
+
# Make requests
|
262 |
pload = {
|
263 |
"model": model_name,
|
264 |
"prompt": prompt,
|
|
|
278 |
yield (state, state.to_gradio_chatbot()) + (disable_btn,) * 5
|
279 |
|
280 |
try:
|
281 |
+
# Stream output
|
282 |
response = requests.post(
|
283 |
worker_addr + "/worker_generate_stream",
|
284 |
headers=headers,
|
|
|
339 |
title_markdown = """
|
340 |
# 🌋 AI Tutor Vision: Large Language and Vision Assistant
|
341 |
[[website]](https://myapps.ai) [[Paper]](https://arxiv.org/abs/2304.08485) [[Model]](https://github.com/haotian-liu/LLaVA/blob/main/docs/MODEL_ZOO.md)
|
342 |
+
|
343 |
ONLY WORKS WITH GPU!
|
344 |
+
|
345 |
You can load the model with 4-bit or 8-bit quantization to make it fit in smaller hardwares. Setting the environment variable `bits` to control the quantization.
|
346 |
*Note: 8-bit seems to be slower than both 4-bit/16-bit. Although it has enough VRAM to support 8-bit, until we figure out the inference speed issue, we recommend 4-bit for A10G for the best efficiency.*
|
347 |
+
|
348 |
Recommended configurations:
|
349 |
| Hardware | T4-Small (16G) | A10G-Small (24G) | A100-Large (40G) |
|
350 |
|-------------------|-----------------|------------------|------------------|
|
351 |
| **Bits** | 4 (default) | 4 | 16 |
|
352 |
+
|
353 |
"""
|
354 |
|
355 |
tos_markdown = """
|
|
|
367 |
"""
|
368 |
|
369 |
block_css = """
|
370 |
+
|
371 |
#buttons button {
|
372 |
min-width: min(120px,100%);
|
373 |
}
|
374 |
+
|
375 |
"""
|
376 |
|
377 |
|
|
|
398 |
container=False,
|
399 |
)
|
400 |
|
401 |
+
imagebox = gr.Image(type="pil")
|
402 |
+
image_process_mode = gr.Radio(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
403 |
["Crop", "Resize", "Pad", "Default"],
|
404 |
value="Default",
|
405 |
label="Preprocess for non-square image",
|
|
|
418 |
"What are the things I should be cautious about when I visit here?",
|
419 |
],
|
420 |
],
|
421 |
+
inputs=[imagebox, textbox],
|
422 |
)
|
423 |
|
424 |
with gr.Accordion("Parameters", open=False) as parameter_row:
|
|
|
449 |
|
450 |
with gr.Column(scale=8):
|
451 |
chatbot = gr.Chatbot(
|
452 |
+
elem_id="chatbot", label="AI Tutor Vision", height=550
|
453 |
)
|
454 |
with gr.Row():
|
455 |
with gr.Column(scale=8):
|
456 |
textbox.render()
|
457 |
with gr.Column(scale=1, min_width=50):
|
458 |
submit_btn = gr.Button(
|
459 |
+
value="Send", variant="primary", interactive=False
|
460 |
+
)
|
461 |
with gr.Row(elem_id="buttons") as button_row:
|
462 |
upvote_btn = gr.Button(value="👍 Upvote", interactive=False)
|
463 |
downvote_btn = gr.Button(value="👎 Downvote", interactive=False)
|
464 |
flag_btn = gr.Button(value="⚠️ Flag", interactive=False)
|
465 |
+
# stop_btn = gr.Button(value="⏹️ Stop Generation", interactive=False)
|
466 |
regenerate_btn = gr.Button(value="🔄 Regenerate", interactive=False)
|
467 |
clear_btn = gr.Button(value="🗑️ Clear history", interactive=False)
|
468 |
|
|
|
471 |
gr.Markdown(learn_more_markdown)
|
472 |
url_params = gr.JSON(visible=False)
|
473 |
|
474 |
+
# Register listeners
|
475 |
btn_list = [upvote_btn, downvote_btn, flag_btn, regenerate_btn, clear_btn]
|
476 |
upvote_btn.click(
|
477 |
upvote_last_response,
|
|
|
490 |
)
|
491 |
regenerate_btn.click(
|
492 |
regenerate,
|
493 |
+
[state, image_process_mode],
|
494 |
+
[state, chatbot, textbox, imagebox] + btn_list,
|
495 |
).then(
|
496 |
http_bot,
|
497 |
[state, model_selector, temperature, top_p, max_output_tokens],
|
498 |
[state, chatbot] + btn_list,
|
499 |
)
|
500 |
clear_btn.click(
|
501 |
+
clear_history, None, [state, chatbot, textbox, imagebox] + btn_list
|
502 |
)
|
503 |
|
504 |
textbox.submit(
|
505 |
add_text,
|
506 |
+
[state, textbox, imagebox, image_process_mode],
|
507 |
+
[state, chatbot, textbox, imagebox] + btn_list,
|
508 |
).then(
|
509 |
http_bot,
|
510 |
[state, model_selector, temperature, top_p, max_output_tokens],
|
|
|
512 |
)
|
513 |
submit_btn.click(
|
514 |
add_text,
|
515 |
+
[state, textbox, imagebox, image_process_mode],
|
516 |
+
[state, chatbot, textbox, imagebox] + btn_list,
|
517 |
).then(
|
518 |
http_bot,
|
519 |
[state, model_selector, temperature, top_p, max_output_tokens],
|
|
|
610 |
controller_proc = start_controller()
|
611 |
worker_proc = start_worker(model_path, bits=bits)
|
612 |
|
613 |
+
# Wait for worker and controller to start
|
614 |
time.sleep(10)
|
615 |
|
616 |
exit_status = 0
|
|
|
623 |
worker_proc.kill()
|
624 |
controller_proc.kill()
|
625 |
|
626 |
+
sys.exit(exit_status)
|