Spaces:
Sleeping
Sleeping
Update src/app.py
Browse files- src/app.py +189 -150
src/app.py
CHANGED
@@ -1,29 +1,51 @@
|
|
1 |
"""Developed by Ruslan Magana Vsevolodovna"""
|
|
|
2 |
from collections.abc import Iterator
|
3 |
from datetime import datetime
|
4 |
from pathlib import Path
|
5 |
from threading import Thread
|
|
|
|
|
|
|
|
|
6 |
import gradio as gr
|
7 |
import spaces
|
8 |
import torch
|
9 |
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
|
10 |
from transformers import LlavaNextProcessor, LlavaNextForConditionalGeneration
|
11 |
-
|
12 |
from themes.research_monochrome import theme
|
13 |
|
14 |
# =============================================================================
|
15 |
# Constants & Prompts
|
16 |
# =============================================================================
|
17 |
-
today_date = datetime.today().strftime("%B %-d, %Y")
|
18 |
-
SYS_PROMPT =
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
19 |
TITLE = "IBM Granite 3.1 8b Reasoning & Vision Preview"
|
20 |
-
DESCRIPTION = """
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
MAX_INPUT_TOKEN_LENGTH = 128_000
|
22 |
MAX_NEW_TOKENS = 1024
|
23 |
-
TEMPERATURE = 0.
|
24 |
TOP_P = 0.85
|
25 |
TOP_K = 50
|
26 |
REPETITION_PENALTY = 1.05
|
|
|
27 |
# Vision defaults (advanced settings)
|
28 |
VISION_TEMPERATURE = 0.2
|
29 |
VISION_TOP_P = 0.95
|
@@ -32,13 +54,13 @@ VISION_MAX_TOKENS = 128
|
|
32 |
|
33 |
if not torch.cuda.is_available():
|
34 |
print("This demo may not work on CPU.")
|
|
|
35 |
# =============================================================================
|
36 |
# Text Model Loading
|
37 |
# =============================================================================
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
granite_text_model="ruslanmv/granite-3.1-8b-Reasoning"
|
42 |
text_model = AutoModelForCausalLM.from_pretrained(
|
43 |
granite_text_model,
|
44 |
torch_dtype=torch.float16,
|
@@ -46,6 +68,7 @@ text_model = AutoModelForCausalLM.from_pretrained(
|
|
46 |
)
|
47 |
tokenizer = AutoTokenizer.from_pretrained(granite_text_model)
|
48 |
tokenizer.use_default_system_prompt = False
|
|
|
49 |
# =============================================================================
|
50 |
# Vision Model Loading
|
51 |
# =============================================================================
|
@@ -55,8 +78,63 @@ vision_model = LlavaNextForConditionalGeneration.from_pretrained(
|
|
55 |
vision_model_path,
|
56 |
torch_dtype=torch.float16,
|
57 |
device_map="auto",
|
58 |
-
trust_remote_code=True # Ensure the custom code is used so that weight shapes match.
|
59 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
60 |
# =============================================================================
|
61 |
# Text Generation Function (for text-only chat)
|
62 |
# =============================================================================
|
@@ -70,7 +148,10 @@ def generate(
|
|
70 |
top_k: float = TOP_K,
|
71 |
max_new_tokens: int = MAX_NEW_TOKENS,
|
72 |
) -> Iterator[str]:
|
73 |
-
"""
|
|
|
|
|
|
|
74 |
conversation = []
|
75 |
conversation.append({"role": "system", "content": SYS_PROMPT})
|
76 |
conversation.extend(chat_history)
|
@@ -84,17 +165,17 @@ def generate(
|
|
84 |
)
|
85 |
input_ids = input_ids.to(text_model.device)
|
86 |
streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
|
87 |
-
generate_kwargs =
|
88 |
-
|
89 |
-
streamer
|
90 |
-
max_new_tokens
|
91 |
-
do_sample
|
92 |
-
top_p
|
93 |
-
top_k
|
94 |
-
temperature
|
95 |
-
num_beams
|
96 |
-
repetition_penalty
|
97 |
-
|
98 |
t = Thread(target=text_model.generate, kwargs=generate_kwargs)
|
99 |
t.start()
|
100 |
|
@@ -112,43 +193,36 @@ def generate(
|
|
112 |
reasoning_started = True
|
113 |
reasoning_start_index = current_output.find("<reasoning>") + len("<reasoning>")
|
114 |
collected_reasoning = current_output[reasoning_start_index:]
|
115 |
-
yield "[Reasoning]: "
|
116 |
-
outputs = [collected_reasoning]
|
117 |
|
118 |
elif reasoning_started and "<answer>" in current_output and not answer_started:
|
119 |
answer_started = True
|
120 |
reasoning_end_index = current_output.find("<answer>")
|
121 |
-
collected_reasoning = current_output[len("<reasoning>"):reasoning_end_index]
|
122 |
|
123 |
answer_start_index = current_output.find("<answer>") + len("<answer>")
|
124 |
collected_answer = current_output[answer_start_index:]
|
125 |
-
yield "\n[Answer]: "
|
126 |
-
outputs = [collected_answer]
|
127 |
-
yield collected_answer
|
128 |
|
129 |
elif reasoning_started and not answer_started:
|
130 |
-
|
131 |
-
|
132 |
|
133 |
elif answer_started:
|
134 |
-
collected_answer += text
|
135 |
-
yield text
|
136 |
-
|
137 |
-
|
138 |
|
|
|
|
|
139 |
|
140 |
# =============================================================================
|
141 |
# Vision Chat Inference Function (for image+text chat)
|
142 |
# =============================================================================
|
143 |
-
def get_text_from_content(content):
|
144 |
-
texts = []
|
145 |
-
for item in content:
|
146 |
-
if item["type"] == "text":
|
147 |
-
texts.append(item["text"])
|
148 |
-
elif item["type"] == "image":
|
149 |
-
texts.append("<Image>")
|
150 |
-
return " ".join(texts)
|
151 |
-
|
152 |
@spaces.GPU
|
153 |
def chat_inference(image, text, conversation, temperature=VISION_TEMPERATURE, top_p=VISION_TOP_P, top_k=VISION_TOP_K, max_tokens=VISION_MAX_TOKENS):
|
154 |
if conversation is None:
|
@@ -159,7 +233,7 @@ def chat_inference(image, text, conversation, temperature=VISION_TEMPERATURE, to
|
|
159 |
if text and text.strip():
|
160 |
user_content.append({"type": "text", "text": text.strip()})
|
161 |
if not user_content:
|
162 |
-
return
|
163 |
conversation.append({"role": "user", "content": user_content})
|
164 |
inputs = vision_processor.apply_chat_template(
|
165 |
conversation,
|
@@ -179,131 +253,89 @@ def chat_inference(image, text, conversation, temperature=VISION_TEMPERATURE, to
|
|
179 |
output = vision_model.generate(**inputs, **generation_kwargs)
|
180 |
assistant_response = vision_processor.decode(output[0], skip_special_tokens=True)
|
181 |
|
182 |
-
|
183 |
-
|
184 |
-
|
185 |
-
if vision_reasoning:
|
186 |
-
reasoning = ""
|
187 |
-
answer = ""
|
188 |
-
if "<reasoning>" in assistant_response and "<answer>" in assistant_response:
|
189 |
-
reasoning_start = assistant_response.find("<reasoning>") + len("<reasoning>")
|
190 |
-
reasoning_end = assistant_response.find("</reasoning>")
|
191 |
-
reasoning = assistant_response[reasoning_start:reasoning_end].strip()
|
192 |
-
|
193 |
-
answer_start = assistant_response.find("<answer>") + len("<answer>")
|
194 |
-
answer_end = assistant_response.find("</answer>")
|
195 |
-
|
196 |
-
if answer_end != -1: # Handle cases where answer end tag is present
|
197 |
-
answer = assistant_response[answer_start:answer_end].strip()
|
198 |
-
else: # Fallback if answer end tag is missing (less robust)
|
199 |
-
answer = assistant_response[answer_start:].strip()
|
200 |
-
formatted_response_content = []
|
201 |
-
if reasoning:
|
202 |
-
formatted_response_content.append({"type": "text", "text": f"[Reasoning]: {reasoning}"})
|
203 |
-
formatted_response_content.append({"type": "text", "text": f"[Answer]: {answer}"})
|
204 |
-
conversation.append({"role": "assistant", "content": formatted_response_content})
|
205 |
else:
|
206 |
-
|
207 |
-
|
208 |
-
return display_vision_conversation(conversation), conversation
|
209 |
|
210 |
-
|
211 |
-
|
212 |
-
# =============================================================================
|
213 |
-
def display_text_conversation(conversation):
|
214 |
-
"""Convert a text conversation (list of dicts) into a list of (user, assistant) tuples."""
|
215 |
-
chat_history = []
|
216 |
-
i = 0
|
217 |
-
while i < len(conversation):
|
218 |
-
if conversation[i]["role"] == "user":
|
219 |
-
user_msg = conversation[i]["content"]
|
220 |
-
assistant_msg = ""
|
221 |
-
if i + 1 < len(conversation) and conversation[i+1]["role"] == "assistant":
|
222 |
-
assistant_msg = conversation[i+1]["content"]
|
223 |
-
i += 2
|
224 |
-
else:
|
225 |
-
i += 1
|
226 |
-
chat_history.append((user_msg, assistant_msg))
|
227 |
-
else:
|
228 |
-
i += 1
|
229 |
-
return chat_history
|
230 |
|
231 |
-
def display_vision_conversation(conversation):
|
232 |
-
"""Convert a vision conversation (with mixed content types) into a list of (user, assistant) tuples."""
|
233 |
-
chat_history = []
|
234 |
-
i = 0
|
235 |
-
while i < len(conversation):
|
236 |
-
if conversation[i]["role"] == "user":
|
237 |
-
user_msg = get_text_from_content(conversation[i]["content"])
|
238 |
-
assistant_msg = ""
|
239 |
-
if i + 1 < len(conversation) and conversation[i+1]["role"] == "assistant":
|
240 |
-
# Extract assistant text; remove any special tokens if present.
|
241 |
-
assistant_content = conversation[i+1]["content"]
|
242 |
-
assistant_text_parts = []
|
243 |
-
for item in assistant_content:
|
244 |
-
if item["type"] == "text":
|
245 |
-
assistant_text_parts.append(item["text"])
|
246 |
-
assistant_msg = "\n".join(assistant_text_parts).strip()
|
247 |
-
i += 2
|
248 |
-
else:
|
249 |
-
i += 1
|
250 |
-
chat_history.append((user_msg, assistant_msg))
|
251 |
-
else:
|
252 |
-
i += 1
|
253 |
-
return chat_history
|
254 |
# =============================================================================
|
255 |
# Unified Send-Message Function
|
|
|
|
|
|
|
|
|
|
|
256 |
# =============================================================================
|
257 |
def send_message(image, text,
|
258 |
text_temperature, text_repetition_penalty, text_top_p, text_top_k, text_max_new_tokens,
|
259 |
vision_temperature, vision_top_p, vision_top_k, vision_max_tokens,
|
260 |
-
|
261 |
-
|
262 |
-
|
263 |
-
|
264 |
-
|
|
|
|
|
265 |
if image is not None:
|
266 |
-
#
|
267 |
-
|
268 |
-
|
269 |
-
|
270 |
-
|
271 |
-
|
272 |
-
|
273 |
-
|
274 |
-
|
275 |
-
|
276 |
-
|
277 |
-
|
|
|
|
|
|
|
|
|
278 |
else:
|
279 |
-
# Text
|
280 |
-
|
281 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
282 |
for chunk in generate(
|
283 |
-
text,
|
284 |
temperature=text_temperature,
|
285 |
repetition_penalty=text_repetition_penalty,
|
286 |
top_p=text_top_p,
|
287 |
top_k=text_top_k,
|
288 |
max_new_tokens=text_max_new_tokens
|
289 |
):
|
290 |
-
|
|
|
|
|
|
|
291 |
|
292 |
-
|
293 |
-
conv.append({"role": "assistant", "content": output_text}) # Store full output with tags
|
294 |
-
text_state = conv
|
295 |
-
chat_history = display_text_conversation(text_state) # Display function handles tag parsing now.
|
296 |
-
return chat_history, text_state, vision_state
|
297 |
|
|
|
|
|
|
|
298 |
def clear_chat():
|
299 |
-
# Clear
|
300 |
-
return [], [], [],
|
|
|
301 |
# =============================================================================
|
302 |
# UI Layout with Gradio
|
303 |
# =============================================================================
|
304 |
css_file_path = Path(Path(__file__).parent / "app.css")
|
305 |
head_file_path = Path(Path(__file__).parent / "app_head.html")
|
306 |
-
|
|
|
307 |
gr.HTML(f"<h1>{TITLE}</h1>", elem_classes=["gr_title"])
|
308 |
gr.HTML(DESCRIPTION)
|
309 |
|
@@ -325,12 +357,17 @@ with gr.Blocks(fill_height=True, css_paths=css_file_path, head_paths=head_file_p
|
|
325 |
vision_top_p_slider = gr.Slider(minimum=0.0, maximum=1.0, value=VISION_TOP_P, step=0.01, label="Vision Top p", elem_classes=["gr_accordion_element"])
|
326 |
vision_top_k_slider = gr.Slider(minimum=0, maximum=100, value=VISION_TOP_K, step=1, label="Vision Top k", elem_classes=["gr_accordion_element"])
|
327 |
vision_max_tokens_slider = gr.Slider(minimum=10, maximum=300, value=VISION_MAX_TOKENS, step=1, label="Vision Max Tokens", elem_classes=["gr_accordion_element"])
|
328 |
-
|
|
|
329 |
clear_button = gr.Button("Clear Chat")
|
330 |
|
331 |
-
# Conversation state variables
|
332 |
-
|
|
|
|
|
|
|
333 |
vision_state = gr.State([])
|
|
|
334 |
|
335 |
send_button.click(
|
336 |
send_message,
|
@@ -338,20 +375,21 @@ with gr.Blocks(fill_height=True, css_paths=css_file_path, head_paths=head_file_p
|
|
338 |
image_input, text_input,
|
339 |
text_temperature_slider, repetition_penalty_slider, top_p_slider, top_k_slider, max_new_tokens_slider,
|
340 |
vision_temperature_slider, vision_top_p_slider, vision_top_k_slider, vision_max_tokens_slider,
|
341 |
-
|
342 |
],
|
343 |
-
outputs=[chatbot,
|
344 |
)
|
345 |
|
346 |
clear_button.click(
|
347 |
clear_chat,
|
348 |
inputs=None,
|
349 |
-
outputs=[chatbot,
|
350 |
)
|
351 |
|
352 |
gr.Examples(
|
353 |
examples=[
|
354 |
["https://raw.githubusercontent.com/gradio-app/gradio/main/test/test_files/cheetah1.jpg", "What is in this image?"],
|
|
|
355 |
[None, "Explain quantum computing to a beginner."],
|
356 |
[None, "What is OpenShift?"],
|
357 |
[None, "Importance of low latency inference"],
|
@@ -362,6 +400,7 @@ with gr.Blocks(fill_height=True, css_paths=css_file_path, head_paths=head_file_p
|
|
362 |
inputs=[image_input, text_input],
|
363 |
example_labels=[
|
364 |
"Vision Example: What is in this image?",
|
|
|
365 |
"Explain quantum computing",
|
366 |
"What is OpenShift?",
|
367 |
"Importance of low latency inference",
|
@@ -373,4 +412,4 @@ with gr.Blocks(fill_height=True, css_paths=css_file_path, head_paths=head_file_p
|
|
373 |
)
|
374 |
|
375 |
if __name__ == "__main__":
|
376 |
-
demo.queue().launch()
|
|
|
1 |
"""Developed by Ruslan Magana Vsevolodovna"""
|
2 |
+
|
3 |
from collections.abc import Iterator
|
4 |
from datetime import datetime
|
5 |
from pathlib import Path
|
6 |
from threading import Thread
|
7 |
+
import io
|
8 |
+
import base64
|
9 |
+
import random
|
10 |
+
|
11 |
import gradio as gr
|
12 |
import spaces
|
13 |
import torch
|
14 |
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
|
15 |
from transformers import LlavaNextProcessor, LlavaNextForConditionalGeneration
|
16 |
+
|
17 |
from themes.research_monochrome import theme
|
18 |
|
19 |
# =============================================================================
|
20 |
# Constants & Prompts
|
21 |
# =============================================================================
|
22 |
+
today_date = datetime.today().strftime("%B %-d, %Y")
|
23 |
+
SYS_PROMPT = """
|
24 |
+
Respond in the following format:
|
25 |
+
<reasoning>
|
26 |
+
...
|
27 |
+
</reasoning>
|
28 |
+
<answer>
|
29 |
+
...
|
30 |
+
</answer>
|
31 |
+
"""
|
32 |
+
|
33 |
TITLE = "IBM Granite 3.1 8b Reasoning & Vision Preview"
|
34 |
+
DESCRIPTION = """
|
35 |
+
<p>Granite 3.1 8b Reasoning is an open‐source LLM supporting a 128k context window and Granite Vision 3.1 2B Preview for vision‐language capabilities. Start with one of the sample prompts
|
36 |
+
or enter your own. Keep in mind that AI can occasionally make mistakes.
|
37 |
+
<span class="gr_docs_link">
|
38 |
+
<a href="https://www.ibm.com/granite/docs/">View Documentation <i class="fa fa-external-link"></i></a>
|
39 |
+
</span>
|
40 |
+
</p>
|
41 |
+
"""
|
42 |
MAX_INPUT_TOKEN_LENGTH = 128_000
|
43 |
MAX_NEW_TOKENS = 1024
|
44 |
+
TEMPERATURE = 0.5
|
45 |
TOP_P = 0.85
|
46 |
TOP_K = 50
|
47 |
REPETITION_PENALTY = 1.05
|
48 |
+
|
49 |
# Vision defaults (advanced settings)
|
50 |
VISION_TEMPERATURE = 0.2
|
51 |
VISION_TOP_P = 0.95
|
|
|
54 |
|
55 |
if not torch.cuda.is_available():
|
56 |
print("This demo may not work on CPU.")
|
57 |
+
|
58 |
# =============================================================================
|
59 |
# Text Model Loading
|
60 |
# =============================================================================
|
61 |
+
|
62 |
+
granite_text_model = "ruslanmv/granite-3.1-8b-Reasoning"
|
63 |
+
|
|
|
64 |
text_model = AutoModelForCausalLM.from_pretrained(
|
65 |
granite_text_model,
|
66 |
torch_dtype=torch.float16,
|
|
|
68 |
)
|
69 |
tokenizer = AutoTokenizer.from_pretrained(granite_text_model)
|
70 |
tokenizer.use_default_system_prompt = False
|
71 |
+
|
72 |
# =============================================================================
|
73 |
# Vision Model Loading
|
74 |
# =============================================================================
|
|
|
78 |
vision_model_path,
|
79 |
torch_dtype=torch.float16,
|
80 |
device_map="auto",
|
81 |
+
trust_remote_code=True # Ensure the custom code is used so that weight shapes match.
|
82 |
)
|
83 |
+
|
84 |
+
# =============================================================================
|
85 |
+
# Unified Display Function
|
86 |
+
# =============================================================================
|
87 |
+
def get_text_from_content(content):
|
88 |
+
"""Helper to extract text from a list of content items."""
|
89 |
+
texts = []
|
90 |
+
for item in content:
|
91 |
+
if isinstance(item, dict):
|
92 |
+
if item.get("type") == "text":
|
93 |
+
texts.append(item.get("text", ""))
|
94 |
+
elif item.get("type") == "image":
|
95 |
+
image = item.get("image")
|
96 |
+
if image is not None:
|
97 |
+
buffered = io.BytesIO()
|
98 |
+
image.save(buffered, format="JPEG")
|
99 |
+
img_str = base64.b64encode(buffered.getvalue()).decode()
|
100 |
+
texts.append(f'<img src="data:image/jpeg;base64,{img_str}" style="max-width: 200px; max-height: 200px;">')
|
101 |
+
else:
|
102 |
+
texts.append("<image>")
|
103 |
+
else:
|
104 |
+
texts.append(str(item))
|
105 |
+
return " ".join(texts)
|
106 |
+
|
107 |
+
def display_unified_conversation(conversation):
|
108 |
+
"""
|
109 |
+
Combine both text-only and vision messages.
|
110 |
+
Each conversation entry is expected to be a dict with keys:
|
111 |
+
- role: "user" or "assistant"
|
112 |
+
- content: either a string (for text) or a list of content items (for vision)
|
113 |
+
"""
|
114 |
+
chat_history = []
|
115 |
+
i = 0
|
116 |
+
while i < len(conversation):
|
117 |
+
if conversation[i]["role"] == "user":
|
118 |
+
user_content = conversation[i]["content"]
|
119 |
+
if isinstance(user_content, list):
|
120 |
+
user_msg = get_text_from_content(user_content)
|
121 |
+
else:
|
122 |
+
user_msg = user_content
|
123 |
+
assistant_msg = ""
|
124 |
+
if i + 1 < len(conversation) and conversation[i+1]["role"] == "assistant":
|
125 |
+
asst_content = conversation[i+1]["content"]
|
126 |
+
if isinstance(asst_content, list):
|
127 |
+
assistant_msg = get_text_from_content(asst_content)
|
128 |
+
else:
|
129 |
+
assistant_msg = asst_content
|
130 |
+
i += 2
|
131 |
+
else:
|
132 |
+
i += 1
|
133 |
+
chat_history.append((user_msg, assistant_msg))
|
134 |
+
else:
|
135 |
+
i += 1
|
136 |
+
return chat_history
|
137 |
+
|
138 |
# =============================================================================
|
139 |
# Text Generation Function (for text-only chat)
|
140 |
# =============================================================================
|
|
|
148 |
top_k: float = TOP_K,
|
149 |
max_new_tokens: int = MAX_NEW_TOKENS,
|
150 |
) -> Iterator[str]:
|
151 |
+
"""
|
152 |
+
Generate function for text chat. It streams tokens and stops once the generated answer
|
153 |
+
contains the closing </answer> tag.
|
154 |
+
"""
|
155 |
conversation = []
|
156 |
conversation.append({"role": "system", "content": SYS_PROMPT})
|
157 |
conversation.extend(chat_history)
|
|
|
165 |
)
|
166 |
input_ids = input_ids.to(text_model.device)
|
167 |
streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
|
168 |
+
generate_kwargs = {
|
169 |
+
"input_ids": input_ids,
|
170 |
+
"streamer": streamer,
|
171 |
+
"max_new_tokens": max_new_tokens,
|
172 |
+
"do_sample": True,
|
173 |
+
"top_p": top_p,
|
174 |
+
"top_k": top_k,
|
175 |
+
"temperature": temperature,
|
176 |
+
"num_beams": 1,
|
177 |
+
"repetition_penalty": repetition_penalty,
|
178 |
+
}
|
179 |
t = Thread(target=text_model.generate, kwargs=generate_kwargs)
|
180 |
t.start()
|
181 |
|
|
|
193 |
reasoning_started = True
|
194 |
reasoning_start_index = current_output.find("<reasoning>") + len("<reasoning>")
|
195 |
collected_reasoning = current_output[reasoning_start_index:]
|
196 |
+
yield "[Reasoning]: "
|
197 |
+
outputs = [collected_reasoning]
|
198 |
|
199 |
elif reasoning_started and "<answer>" in current_output and not answer_started:
|
200 |
answer_started = True
|
201 |
reasoning_end_index = current_output.find("<answer>")
|
202 |
+
collected_reasoning = current_output[len("<reasoning>"):reasoning_end_index]
|
203 |
|
204 |
answer_start_index = current_output.find("<answer>") + len("<answer>")
|
205 |
collected_answer = current_output[answer_start_index:]
|
206 |
+
yield "\n[Answer]: "
|
207 |
+
outputs = [collected_answer]
|
208 |
+
yield collected_answer
|
209 |
|
210 |
elif reasoning_started and not answer_started:
|
211 |
+
collected_reasoning += text
|
212 |
+
yield text
|
213 |
|
214 |
elif answer_started:
|
215 |
+
collected_answer += text
|
216 |
+
yield text
|
217 |
+
if "</answer>" in collected_answer:
|
218 |
+
break
|
219 |
|
220 |
+
else:
|
221 |
+
yield text
|
222 |
|
223 |
# =============================================================================
|
224 |
# Vision Chat Inference Function (for image+text chat)
|
225 |
# =============================================================================
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
226 |
@spaces.GPU
|
227 |
def chat_inference(image, text, conversation, temperature=VISION_TEMPERATURE, top_p=VISION_TOP_P, top_k=VISION_TOP_K, max_tokens=VISION_MAX_TOKENS):
|
228 |
if conversation is None:
|
|
|
233 |
if text and text.strip():
|
234 |
user_content.append({"type": "text", "text": text.strip()})
|
235 |
if not user_content:
|
236 |
+
return display_unified_conversation(conversation), conversation
|
237 |
conversation.append({"role": "user", "content": user_content})
|
238 |
inputs = vision_processor.apply_chat_template(
|
239 |
conversation,
|
|
|
253 |
output = vision_model.generate(**inputs, **generation_kwargs)
|
254 |
assistant_response = vision_processor.decode(output[0], skip_special_tokens=True)
|
255 |
|
256 |
+
if "<|assistant|>" in assistant_response:
|
257 |
+
assistant_response_parts = assistant_response.split("<|assistant|>")
|
258 |
+
assistant_response_text = assistant_response_parts[-1].strip()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
259 |
else:
|
260 |
+
assistant_response_text = assistant_response.strip()
|
|
|
|
|
261 |
|
262 |
+
conversation.append({"role": "assistant", "content": [{"type": "text", "text": assistant_response_text.strip()}]})
|
263 |
+
return display_unified_conversation(conversation), conversation
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
264 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
265 |
# =============================================================================
|
266 |
# Unified Send-Message Function
|
267 |
+
#
|
268 |
+
# We now maintain two histories:
|
269 |
+
# - unified_state: complete conversation (for display)
|
270 |
+
# - internal_text_state: only text turns (for text generation)
|
271 |
+
# Vision turns update only unified_state.
|
272 |
# =============================================================================
|
273 |
def send_message(image, text,
|
274 |
text_temperature, text_repetition_penalty, text_top_p, text_top_k, text_max_new_tokens,
|
275 |
vision_temperature, vision_top_p, vision_top_k, vision_max_tokens,
|
276 |
+
unified_state, vision_state, internal_text_state):
|
277 |
+
# Initialize states if empty
|
278 |
+
if unified_state is None:
|
279 |
+
unified_state = []
|
280 |
+
if internal_text_state is None:
|
281 |
+
internal_text_state = []
|
282 |
+
|
283 |
if image is not None:
|
284 |
+
# Use vision inference.
|
285 |
+
user_msg = []
|
286 |
+
user_msg.append({"type": "image", "image": image})
|
287 |
+
if text and text.strip():
|
288 |
+
user_msg.append({"type": "text", "text": text.strip()})
|
289 |
+
unified_state.append({"role": "user", "content": user_msg})
|
290 |
+
chat_history, updated_vision_conv = chat_inference(image, text, vision_state,
|
291 |
+
temperature=vision_temperature,
|
292 |
+
top_p=vision_top_p,
|
293 |
+
top_k=vision_top_k,
|
294 |
+
max_tokens=vision_max_tokens)
|
295 |
+
vision_state = updated_vision_conv
|
296 |
+
if updated_vision_conv and updated_vision_conv[-1]["role"] == "assistant":
|
297 |
+
unified_state.append(updated_vision_conv[-1])
|
298 |
+
yield display_unified_conversation(unified_state), unified_state, vision_state, internal_text_state
|
299 |
+
|
300 |
else:
|
301 |
+
# Text-only mode: update both unified and internal text states.
|
302 |
+
unified_state.append({"role": "user", "content": text})
|
303 |
+
internal_text_state.append({"role": "user", "content": text})
|
304 |
+
unified_state.append({"role": "assistant", "content": ""})
|
305 |
+
internal_text_state.append({"role": "assistant", "content": ""})
|
306 |
+
yield display_unified_conversation(unified_state), unified_state, vision_state, internal_text_state
|
307 |
+
|
308 |
+
base_conv = internal_text_state[:-1]
|
309 |
+
assistant_text = ""
|
310 |
for chunk in generate(
|
311 |
+
text, base_conv,
|
312 |
temperature=text_temperature,
|
313 |
repetition_penalty=text_repetition_penalty,
|
314 |
top_p=text_top_p,
|
315 |
top_k=text_top_k,
|
316 |
max_new_tokens=text_max_new_tokens
|
317 |
):
|
318 |
+
assistant_text += chunk
|
319 |
+
unified_state[-1]["content"] = assistant_text
|
320 |
+
internal_text_state[-1]["content"] = assistant_text
|
321 |
+
yield display_unified_conversation(unified_state), unified_state, vision_state, internal_text_state
|
322 |
|
323 |
+
yield display_unified_conversation(unified_state), unified_state, vision_state, internal_text_state
|
|
|
|
|
|
|
|
|
324 |
|
325 |
+
# =============================================================================
|
326 |
+
# Clear Chat Function
|
327 |
+
# =============================================================================
|
328 |
def clear_chat():
|
329 |
+
# Clear unified conversation, vision state, and internal text state.
|
330 |
+
return [], [], [], "", None
|
331 |
+
|
332 |
# =============================================================================
|
333 |
# UI Layout with Gradio
|
334 |
# =============================================================================
|
335 |
css_file_path = Path(Path(__file__).parent / "app.css")
|
336 |
head_file_path = Path(Path(__file__).parent / "app_head.html")
|
337 |
+
|
338 |
+
with gr.Blocks(fill_height=True, css_paths=[str(css_file_path)], head_paths=[str(head_file_path)], theme=theme, title=TITLE) as demo:
|
339 |
gr.HTML(f"<h1>{TITLE}</h1>", elem_classes=["gr_title"])
|
340 |
gr.HTML(DESCRIPTION)
|
341 |
|
|
|
357 |
vision_top_p_slider = gr.Slider(minimum=0.0, maximum=1.0, value=VISION_TOP_P, step=0.01, label="Vision Top p", elem_classes=["gr_accordion_element"])
|
358 |
vision_top_k_slider = gr.Slider(minimum=0, maximum=100, value=VISION_TOP_K, step=1, label="Vision Top k", elem_classes=["gr_accordion_element"])
|
359 |
vision_max_tokens_slider = gr.Slider(minimum=10, maximum=300, value=VISION_MAX_TOKENS, step=1, label="Vision Max Tokens", elem_classes=["gr_accordion_element"])
|
360 |
+
|
361 |
+
send_button = gr.Button("Send Message")
|
362 |
clear_button = gr.Button("Clear Chat")
|
363 |
|
364 |
+
# Conversation state variables:
|
365 |
+
# - unified_state: complete conversation for display (text and vision)
|
366 |
+
# - vision_state: state for vision turns
|
367 |
+
# - internal_text_state: only text turns (for text-generation)
|
368 |
+
unified_state = gr.State([])
|
369 |
vision_state = gr.State([])
|
370 |
+
internal_text_state = gr.State([])
|
371 |
|
372 |
send_button.click(
|
373 |
send_message,
|
|
|
375 |
image_input, text_input,
|
376 |
text_temperature_slider, repetition_penalty_slider, top_p_slider, top_k_slider, max_new_tokens_slider,
|
377 |
vision_temperature_slider, vision_top_p_slider, vision_top_k_slider, vision_max_tokens_slider,
|
378 |
+
unified_state, vision_state, internal_text_state
|
379 |
],
|
380 |
+
outputs=[chatbot, unified_state, vision_state, internal_text_state],
|
381 |
)
|
382 |
|
383 |
clear_button.click(
|
384 |
clear_chat,
|
385 |
inputs=None,
|
386 |
+
outputs=[chatbot, unified_state, vision_state, internal_text_state, text_input, image_input]
|
387 |
)
|
388 |
|
389 |
gr.Examples(
|
390 |
examples=[
|
391 |
["https://raw.githubusercontent.com/gradio-app/gradio/main/test/test_files/cheetah1.jpg", "What is in this image?"],
|
392 |
+
[None, "Compute Pi."],
|
393 |
[None, "Explain quantum computing to a beginner."],
|
394 |
[None, "What is OpenShift?"],
|
395 |
[None, "Importance of low latency inference"],
|
|
|
400 |
inputs=[image_input, text_input],
|
401 |
example_labels=[
|
402 |
"Vision Example: What is in this image?",
|
403 |
+
"Compute Pi.",
|
404 |
"Explain quantum computing",
|
405 |
"What is OpenShift?",
|
406 |
"Importance of low latency inference",
|
|
|
412 |
)
|
413 |
|
414 |
if __name__ == "__main__":
|
415 |
+
demo.queue().launch(debug=True, share=False)
|