Spaces:
Running
on
Zero
Running
on
Zero
Commit
·
4d0481d
1
Parent(s):
46c8f02
Refactor message handling in conversation and prediction functions to improve clarity and functionality
Browse files- app.py +26 -14
- conversation.py +15 -6
- models.py +2 -9
app.py
CHANGED
@@ -15,7 +15,7 @@ from filelock import FileLock
|
|
15 |
from io import BytesIO
|
16 |
from PIL import Image, ImageDraw, ImageFont
|
17 |
from models import load_image
|
18 |
-
from constants import LOGDIR
|
19 |
from utils import (
|
20 |
build_logger,
|
21 |
server_error_msg,
|
@@ -164,6 +164,10 @@ def add_text(state, message, system_prompt, request: gr.Request):
|
|
164 |
|
165 |
if len(images) > 0 and len(state.get_images(source=state.USER)) > 0:
|
166 |
state = init_state(state)
|
|
|
|
|
|
|
|
|
167 |
state.set_system_message(system_prompt)
|
168 |
state.append_message(Conversation.USER, text, images)
|
169 |
state.skip_next = False
|
@@ -183,19 +187,29 @@ tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True, us
|
|
183 |
@spaces.GPU
|
184 |
def predict(message,
|
185 |
image_path,
|
186 |
-
|
187 |
max_input_tiles=6,
|
188 |
temperature=1.0,
|
189 |
max_output_tokens=700,
|
190 |
top_p=0.7,
|
191 |
repetition_penalty=2.5):
|
192 |
-
|
|
|
|
|
|
|
193 |
generation_config = dict(temperature=temperature, max_new_tokens= max_output_tokens, top_p=top_p, do_sample=False, num_beams = 3, repetition_penalty=repetition_penalty)
|
194 |
-
|
195 |
-
|
196 |
-
|
197 |
-
|
198 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
199 |
response, conv_history = model.chat(tokenizer, pixel_values, question, generation_config, history=history, return_history=True)
|
200 |
return response, conv_history
|
201 |
|
@@ -246,21 +260,19 @@ def http_bot(
|
|
246 |
|
247 |
try:
|
248 |
# Stream output
|
249 |
-
message = state.
|
250 |
logger.info(f"==== User message ====\n{message}")
|
251 |
logger.info(f"==== Image paths ====\n{all_image_paths}")
|
252 |
|
253 |
-
|
254 |
-
logger.info(f"==== History ====\n{history}")
|
255 |
-
response, conv_history = predict(message,
|
256 |
all_image_paths[0],
|
257 |
-
|
258 |
max_input_tiles,
|
259 |
temperature,
|
260 |
max_new_tokens,
|
261 |
top_p,
|
262 |
repetition_penalty)
|
263 |
-
logger.info(f"==== AI history ====\n{conv_history}")
|
264 |
|
265 |
|
266 |
# response = "This is a test response"
|
|
|
15 |
from io import BytesIO
|
16 |
from PIL import Image, ImageDraw, ImageFont
|
17 |
from models import load_image
|
18 |
+
from constants import LOGDIR, DEFAULT_IMAGE_TOKEN
|
19 |
from utils import (
|
20 |
build_logger,
|
21 |
server_error_msg,
|
|
|
164 |
|
165 |
if len(images) > 0 and len(state.get_images(source=state.USER)) > 0:
|
166 |
state = init_state(state)
|
167 |
+
|
168 |
+
if len(images) > 0 and len(state.get_images(source=state.USER)) == 0:
|
169 |
+
text = DEFAULT_IMAGE_TOKEN + "\n" + text
|
170 |
+
|
171 |
state.set_system_message(system_prompt)
|
172 |
state.append_message(Conversation.USER, text, images)
|
173 |
state.skip_next = False
|
|
|
187 |
@spaces.GPU
|
188 |
def predict(message,
|
189 |
image_path,
|
190 |
+
state,
|
191 |
max_input_tiles=6,
|
192 |
temperature=1.0,
|
193 |
max_output_tokens=700,
|
194 |
top_p=0.7,
|
195 |
repetition_penalty=2.5):
|
196 |
+
|
197 |
+
history = state.get_prompt()
|
198 |
+
logger.info(f"==== History ====\n{history}")
|
199 |
+
|
200 |
generation_config = dict(temperature=temperature, max_new_tokens= max_output_tokens, top_p=top_p, do_sample=False, num_beams = 3, repetition_penalty=repetition_penalty)
|
201 |
+
|
202 |
+
question = message
|
203 |
+
pixel_values = None
|
204 |
+
if image_path is not None:
|
205 |
+
pixel_values = load_image(image_path, max_num=max_input_tiles).to(torch.bfloat16).cuda()
|
206 |
+
if pixel_values is not None:
|
207 |
+
# Check the first user message to see if it is an image
|
208 |
+
index, first_user_message = state.get_user_message(source=state.USER, position='first')
|
209 |
+
if first_user_message is not None and \
|
210 |
+
DEFAULT_IMAGE_TOKEN not in first_user_message:
|
211 |
+
state.messages[index]['content'] = DEFAULT_IMAGE_TOKEN + "\n" + first_user_message
|
212 |
+
|
213 |
response, conv_history = model.chat(tokenizer, pixel_values, question, generation_config, history=history, return_history=True)
|
214 |
return response, conv_history
|
215 |
|
|
|
260 |
|
261 |
try:
|
262 |
# Stream output
|
263 |
+
message = state.get_user_message(source=state.USER, position='last')
|
264 |
logger.info(f"==== User message ====\n{message}")
|
265 |
logger.info(f"==== Image paths ====\n{all_image_paths}")
|
266 |
|
267 |
+
response, _ = predict(message,
|
|
|
|
|
268 |
all_image_paths[0],
|
269 |
+
state,
|
270 |
max_input_tiles,
|
271 |
temperature,
|
272 |
max_new_tokens,
|
273 |
top_p,
|
274 |
repetition_penalty)
|
275 |
+
# logger.info(f"==== AI history ====\n{conv_history}")
|
276 |
|
277 |
|
278 |
# response = "This is a test response"
|
conversation.py
CHANGED
@@ -174,14 +174,23 @@ class Conversation:
|
|
174 |
|
175 |
return images
|
176 |
|
177 |
-
def
|
178 |
assert len(self.messages) > 0, "No message in the conversation."
|
179 |
assert source in [self.USER, self.ASSISTANT, None], f"Invalid source: {source}"
|
180 |
-
|
181 |
-
|
182 |
-
|
183 |
-
|
184 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
185 |
|
186 |
def to_gradio_chatbot(self):
|
187 |
ret = []
|
|
|
174 |
|
175 |
return images
|
176 |
|
177 |
+
def get_user_message(self, source: Union[str, None] = None, position="first"):
|
178 |
assert len(self.messages) > 0, "No message in the conversation."
|
179 |
assert source in [self.USER, self.ASSISTANT, None], f"Invalid source: {source}"
|
180 |
+
|
181 |
+
if position == "first":
|
182 |
+
for i, msg in enumerate(self.messages):
|
183 |
+
if source and msg["role"] != source:
|
184 |
+
continue
|
185 |
+
if msg["role"] == self.USER:
|
186 |
+
return i, msg["content"]
|
187 |
+
|
188 |
+
elif position == "last":
|
189 |
+
for i in range(len(self.messages) - 1, -1, -1):
|
190 |
+
if source and self.messages[i]["role"] != source:
|
191 |
+
continue
|
192 |
+
if self.messages[i]["role"] == self.USER:
|
193 |
+
return i, self.messages[i]["content"]
|
194 |
|
195 |
def to_gradio_chatbot(self):
|
196 |
ret = []
|
models.py
CHANGED
@@ -74,16 +74,12 @@ def dynamic_preprocess(image, min_num=1, max_num=12, image_size=448, use_thumbna
|
|
74 |
return processed_images
|
75 |
|
76 |
def correct_image_orientation(image_path):
|
77 |
-
# Mở ảnh
|
78 |
image = Image.open(image_path)
|
79 |
-
|
80 |
-
# Kiểm tra dữ liệu Exif (nếu có)
|
81 |
try:
|
82 |
exif = image._getexif()
|
83 |
if exif is not None:
|
84 |
for tag, value in exif.items():
|
85 |
if ExifTags.TAGS.get(tag) == "Orientation":
|
86 |
-
# Sửa hướng dựa trên Orientation
|
87 |
if value == 3:
|
88 |
image = image.rotate(180, expand=True)
|
89 |
elif value == 6:
|
@@ -92,7 +88,8 @@ def correct_image_orientation(image_path):
|
|
92 |
image = image.rotate(90, expand=True)
|
93 |
break
|
94 |
except Exception as e:
|
95 |
-
print("
|
|
|
96 |
|
97 |
return image
|
98 |
|
@@ -100,13 +97,9 @@ def load_image(image_file, input_size=448, max_num=12):
|
|
100 |
try:
|
101 |
print("Loading image:", image_file)
|
102 |
image = correct_image_orientation(image_file).convert('RGB')
|
103 |
-
print("Image size:", image.size)
|
104 |
transform = build_transform(input_size=input_size)
|
105 |
-
print("Transform built.")
|
106 |
images = dynamic_preprocess(image, image_size=input_size, use_thumbnail=True, max_num=max_num)
|
107 |
-
print("Number of images:", len(images))
|
108 |
pixel_values = [transform(image) for image in images]
|
109 |
-
print("Images transformed.")
|
110 |
pixel_values = torch.stack(pixel_values)
|
111 |
print("Image loaded successfully.")
|
112 |
except Exception as e:
|
|
|
74 |
return processed_images
|
75 |
|
76 |
def correct_image_orientation(image_path):
|
|
|
77 |
image = Image.open(image_path)
|
|
|
|
|
78 |
try:
|
79 |
exif = image._getexif()
|
80 |
if exif is not None:
|
81 |
for tag, value in exif.items():
|
82 |
if ExifTags.TAGS.get(tag) == "Orientation":
|
|
|
83 |
if value == 3:
|
84 |
image = image.rotate(180, expand=True)
|
85 |
elif value == 6:
|
|
|
88 |
image = image.rotate(90, expand=True)
|
89 |
break
|
90 |
except Exception as e:
|
91 |
+
print("Error reading exif:", e)
|
92 |
+
print(traceback.format_exc())
|
93 |
|
94 |
return image
|
95 |
|
|
|
97 |
try:
|
98 |
print("Loading image:", image_file)
|
99 |
image = correct_image_orientation(image_file).convert('RGB')
|
|
|
100 |
transform = build_transform(input_size=input_size)
|
|
|
101 |
images = dynamic_preprocess(image, image_size=input_size, use_thumbnail=True, max_num=max_num)
|
|
|
102 |
pixel_values = [transform(image) for image in images]
|
|
|
103 |
pixel_values = torch.stack(pixel_values)
|
104 |
print("Image loaded successfully.")
|
105 |
except Exception as e:
|