lykeven commited on
Commit
da4f6a0
1 Parent(s): 2a6fa8f

add cogagent

Browse files
.gitignore ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ en_core_web_sm-3.6.0/*
2
+ daily_receive.*
3
+ weekly_train_generate.*
4
+ app[2-3].py
5
+ test_gradio_client.py
6
+ boxes.py
7
+ *.tar.gz
8
+ *.pyi
9
+ __pycache__/
10
+ examples/*grounding*
app.py CHANGED
@@ -2,77 +2,38 @@
2
 
3
  import gradio as gr
4
  import os
5
- import re
6
- from PIL import Image
7
- import base64
8
- import time
 
9
 
10
- DESCRIPTION = '''# <a href="https://github.com/THUDM/CogVLM">VisualGLM</a>'''
11
 
12
  MAINTENANCE_NOTICE1 = 'Hint 1: If the app report "Something went wrong, connection error out", please turn off your proxy and retry.<br>Hint 2: If you upload a large size of image like 10MB, it may take some time to upload and process. Please be patient and wait.'
13
 
14
- GROUNDING_NOTICE = 'Hint: When you check "Grounding", please use the <a href="https://github.com/THUDM/CogVLM/blob/main/utils/template.py#L344">corresponding prompt</a> or the examples below.'
15
 
 
16
 
17
- NOTES = 'This app is adapted from <a href="https://github.com/THUDM/CogVLM">https://github.com/THUDM/CogVLM</a>. It would be recommended to check out the repo if you want to see the detail of our model.'
18
-
19
- import json
20
- import requests
21
- import base64
22
- import hashlib
23
- from utils import parse_response
24
 
25
  default_chatbox = [("", "Hi, What do you want to know about this image?")]
26
 
27
  URL = os.environ.get("URL")
28
 
29
- def process_image(image_prompt):
30
- image = Image.open(image_prompt)
31
- print(f"height:{image.height}, width:{image.width}")
32
- resized_image = image.resize((224, 224), )
33
- timestamp = int(time.time())
34
- file_ext = os.path.splitext(image_prompt)[1]
35
- filename = f"examples/{timestamp}{file_ext}"
36
- resized_image.save(filename)
37
- print(f"temporal filename {filename}")
38
- with open(filename, "rb") as image_file:
39
- bytes = base64.b64encode(image_file.read())
40
- encoded_img = str(bytes, encoding='utf-8')
41
- image_hash = hashlib.sha256(bytes).hexdigest()
42
- os.remove(filename)
43
- return encoded_img, image_hash
44
-
45
-
46
- def process_image_without_resize(image_prompt):
47
- image = Image.open(image_prompt)
48
- print(f"height:{image.height}, width:{image.width}")
49
- timestamp = int(time.time())
50
- file_ext = os.path.splitext(image_prompt)[1]
51
- filename = f"examples/{timestamp}{file_ext}"
52
- filename_grounding = f"examples/{timestamp}_grounding{file_ext}"
53
- image.save(filename)
54
- print(f"temporal filename {filename}")
55
- with open(filename, "rb") as image_file:
56
- bytes = base64.b64encode(image_file.read())
57
- encoded_img = str(bytes, encoding='utf-8')
58
- image_hash = hashlib.sha256(bytes).hexdigest()
59
- os.remove(filename)
60
- return image, encoded_img, image_hash, filename_grounding
61
-
62
-
63
- def is_chinese(text):
64
- zh_pattern = re.compile(u'[\u4e00-\u9fa5]+')
65
- return zh_pattern.search(text)
66
-
67
 
68
  def post(
69
  input_text,
70
  temperature,
71
  top_p,
 
72
  image_prompt,
73
  result_previous,
74
  hidden_image,
75
- grounding
 
 
 
76
  ):
77
  result_text = [(ele[0], ele[1]) for ele in result_previous]
78
  for i in range(len(result_text)-1, -1, -1):
@@ -107,16 +68,35 @@ def post(
107
  result_text = []
108
  hidden_image = image_hash
109
  else:
110
- encoded_img = None
 
 
 
 
 
 
 
 
 
 
 
 
111
 
112
- print('request chat model...' if not grounding else 'request grounding model...')
 
 
 
113
  data = json.dumps({
114
- 'text': input_text,
 
 
 
115
  'image': encoded_img,
116
  'temperature': temperature,
117
  'top_p': top_p,
118
- 'history': result_text,
119
- 'is_grounding': grounding
 
120
  })
121
  try:
122
  response = requests.request("POST", URL, headers=headers, data=data, timeout=(60, 100)).json()
@@ -161,6 +141,9 @@ def main():
161
 
162
  with gr.Blocks(css='style.css') as demo:
163
 
 
 
 
164
  with gr.Row():
165
  with gr.Column(scale=4.5):
166
  with gr.Group():
@@ -172,28 +155,33 @@ def main():
172
  image_prompt = gr.Image(type="filepath", label="Image Prompt", value=None)
173
  with gr.Row():
174
  grounding = gr.Checkbox(label="Grounding")
 
175
  with gr.Row():
176
- grounding_notice = gr.Markdown(GROUNDING_NOTICE)
 
 
 
177
 
178
  with gr.Row():
179
- temperature = gr.Slider(maximum=1, value=0.8, minimum=0, label='Temperature')
180
- top_p = gr.Slider(maximum=1, value=0.4, minimum=0, label='Top P')
 
 
181
  with gr.Column(scale=5.5):
182
- result_text = gr.components.Chatbot(label='Multi-round conversation History', value=[("", "Hi, What do you want to know about this image?")]).style(height=550)
183
  hidden_image_hash = gr.Textbox(visible=False)
184
 
185
- gr_examples = gr.Examples(examples=[[example["text"], example["image"]] for example in examples],
186
- inputs=[input_text, image_prompt],
187
  label="Example Inputs (Click to insert an examplet into the input box)",
188
  examples_per_page=6)
189
 
190
  gr.Markdown(MAINTENANCE_NOTICE1)
191
- gr.Markdown(NOTES)
192
 
193
  print(gr.__version__)
194
- run_button.click(fn=post,inputs=[input_text, temperature, top_p, image_prompt, result_text, hidden_image_hash, grounding],
195
  outputs=[input_text, result_text, hidden_image_hash])
196
- input_text.submit(fn=post,inputs=[input_text, temperature, top_p, image_prompt, result_text, hidden_image_hash, grounding],
197
  outputs=[input_text, result_text, hidden_image_hash])
198
  clear_button.click(fn=clear_fn, inputs=clear_button, outputs=[input_text, result_text, image_prompt])
199
  image_prompt.upload(fn=clear_fn2, inputs=clear_button, outputs=[result_text])
@@ -202,7 +190,7 @@ def main():
202
  print(gr.__version__)
203
 
204
  demo.queue(concurrency_count=10)
205
- demo.launch()
206
 
207
  if __name__ == '__main__':
208
  main()
 
2
 
3
  import gradio as gr
4
  import os
5
+ import json
6
+ import requests
7
+ from utils import is_chinese, process_image_without_resize, parse_response, templates_agent_cogagent, template_grounding_cogvlm, postprocess_text
8
+
9
+ DESCRIPTION = '''<h2 style='text-align: center'> <a href="https://github.com/THUDM/CogVLM"> CogVLM & CogAgent Chat Demo</a> </h2>'''
10
 
11
+ NOTES = 'This app is adapted from <a href="https://github.com/THUDM/CogVLM">https://github.com/THUDM/CogVLM</a>. It would be recommended to check out the repo if you want to see the detail of our model.\n\n该demo仅作为测试使用,不支持批量请求。如有大批量需求,欢迎联系[智谱AI](mailto:[email protected])。\n\n请注意CoogVLM-17B目前仅支持英文。'
12
 
13
  MAINTENANCE_NOTICE1 = 'Hint 1: If the app report "Something went wrong, connection error out", please turn off your proxy and retry.<br>Hint 2: If you upload a large size of image like 10MB, it may take some time to upload and process. Please be patient and wait.'
14
 
15
+ GROUNDING_NOTICE = 'Hint: When you check "Grounding", please use the <a href="https://github.com/THUDM/CogVLM/blob/main/utils/utils/template.py#L344">corresponding prompt</a> or the examples below.'
16
 
17
+ AGENT_NOTICE = 'Hint: When you check "CogAgent", please use the <a href="https://github.com/THUDM/CogVLM/blob/main/utils/utils/template.py#L761C1-L761C17">corresponding prompt</a> or the examples below.'
18
 
 
 
 
 
 
 
 
19
 
20
  default_chatbox = [("", "Hi, What do you want to know about this image?")]
21
 
22
  URL = os.environ.get("URL")
23
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
24
 
25
  def post(
26
  input_text,
27
  temperature,
28
  top_p,
29
+ top_k,
30
  image_prompt,
31
  result_previous,
32
  hidden_image,
33
+ grounding,
34
+ cogagent,
35
+ grounding_template,
36
+ agent_template
37
  ):
38
  result_text = [(ele[0], ele[1]) for ele in result_previous]
39
  for i in range(len(result_text)-1, -1, -1):
 
68
  result_text = []
69
  hidden_image = image_hash
70
  else:
71
+ encoded_img = None
72
+
73
+ model_use = "vlm_chat"
74
+ if not cogagent and grounding:
75
+ model_use = "vlm_grounding"
76
+ if grounding_template:
77
+ input_text = postprocess_text(grounding_template, input_text)
78
+ elif cogagent:
79
+ model_use = "agent_chat"
80
+ if agent_template:
81
+ input_text = postprocess_text(agent_template, input_text)
82
+
83
+ prompt = input_text
84
 
85
+ if grounding:
86
+ prompt += "(with grounding)"
87
+
88
+ print(f'request {model_use} model... with prompt {prompt}, grounding_template {grounding_template}, agent_template {agent_template}')
89
  data = json.dumps({
90
+ 'model_use': model_use,
91
+ 'is_grounding': grounding,
92
+ 'text': prompt,
93
+ 'history': result_text,
94
  'image': encoded_img,
95
  'temperature': temperature,
96
  'top_p': top_p,
97
+ 'top_k': top_k,
98
+ 'do_sample': True,
99
+ 'max_new_tokens': 2048
100
  })
101
  try:
102
  response = requests.request("POST", URL, headers=headers, data=data, timeout=(60, 100)).json()
 
141
 
142
  with gr.Blocks(css='style.css') as demo:
143
 
144
+ gr.Markdown(DESCRIPTION)
145
+ gr.Markdown(NOTES)
146
+
147
  with gr.Row():
148
  with gr.Column(scale=4.5):
149
  with gr.Group():
 
155
  image_prompt = gr.Image(type="filepath", label="Image Prompt", value=None)
156
  with gr.Row():
157
  grounding = gr.Checkbox(label="Grounding")
158
+ cogagent = gr.Checkbox(label="CogAgent")
159
  with gr.Row():
160
+ # grounding_notice = gr.Markdown(GROUNDING_NOTICE)
161
+ grounding_template = gr.Dropdown(choices=template_grounding_cogvlm, label="Grounding Template", value=template_grounding_cogvlm[0])
162
+ # agent_notice = gr.Markdown(AGENT_NOTICE)
163
+ agent_template = gr.Dropdown(choices=templates_agent_cogagent, label="Agent Template", value=templates_agent_cogagent[0])
164
 
165
  with gr.Row():
166
+ temperature = gr.Slider(maximum=1, value=0.9, minimum=0, label='Temperature')
167
+ top_p = gr.Slider(maximum=1, value=0.8, minimum=0, label='Top P')
168
+ top_k = gr.Slider(maximum=50, value=5, minimum=1, step=1, label='Top K')
169
+
170
  with gr.Column(scale=5.5):
171
+ result_text = gr.components.Chatbot(label='Multi-round conversation History', value=[("", "Hi, What do you want to know about this image?")], height=550)
172
  hidden_image_hash = gr.Textbox(visible=False)
173
 
174
+ gr_examples = gr.Examples(examples=[[example["text"], example["image"], example["grounding"], example["cogagent"]] for example in examples],
175
+ inputs=[input_text, image_prompt, grounding, cogagent],
176
  label="Example Inputs (Click to insert an examplet into the input box)",
177
  examples_per_page=6)
178
 
179
  gr.Markdown(MAINTENANCE_NOTICE1)
 
180
 
181
  print(gr.__version__)
182
+ run_button.click(fn=post,inputs=[input_text, temperature, top_p, top_k, image_prompt, result_text, hidden_image_hash, grounding, cogagent, grounding_template, agent_template],
183
  outputs=[input_text, result_text, hidden_image_hash])
184
+ input_text.submit(fn=post,inputs=[input_text, temperature, top_p, top_k, image_prompt, result_text, hidden_image_hash, grounding, cogagent, grounding_template, agent_template],
185
  outputs=[input_text, result_text, hidden_image_hash])
186
  clear_button.click(fn=clear_fn, inputs=clear_button, outputs=[input_text, result_text, image_prompt])
187
  image_prompt.upload(fn=clear_fn2, inputs=clear_button, outputs=[result_text])
 
190
  print(gr.__version__)
191
 
192
  demo.queue(concurrency_count=10)
193
+ demo.launch(server_port=7862)
194
 
195
  if __name__ == '__main__':
196
  main()
examples/1.jpeg DELETED
Binary file (236 kB)
 
examples/2.jpeg DELETED
Binary file (6.7 kB)
 
examples/3.jpeg DELETED
Binary file (53.9 kB)
 
examples/3.jpg CHANGED
examples/{6.jpg → 4.jpg} RENAMED
File without changes
examples/4.png DELETED
Binary file (196 kB)
 
examples/5.jpeg ADDED
examples/5.jpg DELETED
Binary file (343 kB)
 
examples/6.jpeg ADDED
examples/example_inputs.jsonl CHANGED
@@ -1,6 +1,6 @@
1
- {"id":1, "text": "Describe this image", "image": "examples/1.png"}
2
- {"id":2, "text": "What is written in the image?", "image": "examples/2.jpg"}
3
- {"id":3, "text": "How many houses are there in this cartoon?", "image": "examples/3.jpg"}
4
- {"id":4, "text": "Can you provide a description of the image and include the coordinates [[x0,y0,x1,y1]] for each mentioned object?", "image": "examples/4.png"}
5
- {"id":5, "text": "Where is the tree closer to the sun?", "image": "examples/5.jpg"}
6
- {"id":6, "text": "What color are the clothes of the girl whose hands are holding flowers? Let's think step by step", "image": "examples/6.jpg"}
 
1
+ {"id":1, "text": "Describe this image", "image": "examples/1.png", "grounding": false, "cogagent": false}
2
+ {"id":2, "text": "What is written in the image?", "image": "examples/2.jpg", "grounding": false, "cogagent": false}
3
+ {"id":3, "text": "the tree closer to the sun", "image": "examples/3.jpg", "grounding": true, "cogagent": false}
4
+ {"id":4, "text": "What color are the clothes of the girl whose hands are holding flowers? Let's think step by step", "image": "examples/4.jpg", "grounding": true, "cogagent": false}
5
+ {"id":5, "text": "search CogVLM", "image": "examples/5.jpeg", "grounding": true, "cogagent": true}
6
+ {"id":6, "text": "Insert a new slide named 'In-context learning: Details' with a Two Content layout after the current slide.", "image": "examples/6.jpeg", "grounding": false, "cogagent": true}
utils.py CHANGED
@@ -3,9 +3,38 @@ from PIL import Image, ImageDraw, ImageFont
3
  import matplotlib.font_manager
4
  import spacy
5
  import re
 
 
 
 
 
 
 
6
 
7
  nlp = spacy.load("en_core_web_sm-3.6.0")
8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9
  def draw_boxes(image, boxes, texts, output_fn='output.png'):
10
  box_width = 5
11
  color_palette = sns.color_palette("husl", len(boxes))
@@ -83,4 +112,312 @@ def parse_response(img, response, output_fn='output.png'):
83
  boxes = []
84
  else:
85
  texts, boxes = zip(*dic.items())
86
- draw_boxes(new_img, boxes, texts, output_fn=output_fn)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  import matplotlib.font_manager
4
  import spacy
5
  import re
6
+ import base64
7
+ import time
8
+ import re
9
+ from PIL import Image
10
+ import base64
11
+ import hashlib
12
+ import os
13
 
14
  nlp = spacy.load("en_core_web_sm-3.6.0")
15
 
16
+ def process_image_without_resize(image_prompt):
17
+ image = Image.open(image_prompt)
18
+ print(f"height:{image.height}, width:{image.width}")
19
+ timestamp = time.time()
20
+ file_ext = os.path.splitext(image_prompt)[1]
21
+ filename = f"examples/{timestamp}{file_ext}"
22
+ filename_grounding = f"examples/{timestamp}_grounding{file_ext}"
23
+ image.save(filename)
24
+ print(f"temporal filename {filename}")
25
+ with open(filename, "rb") as image_file:
26
+ bytes = base64.b64encode(image_file.read())
27
+ encoded_img = str(bytes, encoding='utf-8')
28
+ image_hash = hashlib.sha256(bytes).hexdigest()
29
+ os.remove(filename)
30
+ return image, encoded_img, image_hash, filename_grounding
31
+
32
+
33
+ def is_chinese(text):
34
+ zh_pattern = re.compile(u'[\u4e00-\u9fa5]+')
35
+ return zh_pattern.search(text)
36
+
37
+
38
  def draw_boxes(image, boxes, texts, output_fn='output.png'):
39
  box_width = 5
40
  color_palette = sns.color_palette("husl", len(boxes))
 
112
  boxes = []
113
  else:
114
  texts, boxes = zip(*dic.items())
115
+ draw_boxes(new_img, boxes, texts, output_fn=output_fn)
116
+
117
+ def postprocess_text(template, text):
118
+ quoted_text = f'"{text.strip()}"'
119
+ return template.replace("<TASK>", quoted_text).strip() if template != "" else text.strip()
120
+
121
+
122
+ # The templates is for CogAgent_Agent Template
123
+ templates_agent_cogagent = [
124
+ "Can you advise me on how to <TASK>?",
125
+ "I'm looking for guidance on how to <TASK>.",
126
+ "What steps do I need to take to <TASK>?",
127
+ "Could you provide instructions for <TASK>?",
128
+ "I'm wondering what the process is for <TASK>.",
129
+ "How can I go about <TASK>?",
130
+ "I need assistance with planning to <TASK>.",
131
+ "Do you have any recommendations for <TASK>?",
132
+ "Please share some tips for <TASK>.",
133
+ "I'd like to know the best way to <TASK>.",
134
+ "What's the most effective way to <TASK>?",
135
+ "I'm seeking advice on accomplishing <TASK>.",
136
+ "Could you guide me through the steps to <TASK>?",
137
+ "I'm unsure how to start with <TASK>.",
138
+ "Is there a strategy for successfully <TASK>?",
139
+ "What's the proper procedure for <TASK>?",
140
+ "How should I prepare for <TASK>?",
141
+ "I'm not sure where to begin with <TASK>.",
142
+ "I need some insights on <TASK>.",
143
+ "Can you explain how to tackle <TASK>?",
144
+ "I'm interested in the process of <TASK>.",
145
+ "Could you enlighten me on <TASK>?",
146
+ "What are the recommended steps for <TASK>?",
147
+ "Is there a preferred method for <TASK>?",
148
+ "I'd appreciate your advice on <TASK>.",
149
+ "Can you shed light on <TASK>?",
150
+ "What would be the best approach to <TASK>?",
151
+ "How do I get started with <TASK>?",
152
+ "I'm inquiring about the procedure for <TASK>.",
153
+ "Could you share your expertise on <TASK>?",
154
+ "I'd like some guidance on <TASK>.",
155
+ "What's your recommendation for <TASK>?",
156
+ "I'm seeking your input on how to <TASK>.",
157
+ "Can you provide some insights into <TASK>?",
158
+ "How can I successfully accomplish <TASK>?",
159
+ "What steps are involved in <TASK>?",
160
+ "I'm curious about the best way to <TASK>.",
161
+ "Could you show me the ropes for <TASK>?",
162
+ "I need to know how to go about <TASK>.",
163
+ "What are the essential steps for <TASK>?",
164
+ "Is there a specific method for <TASK>?",
165
+ "I'd like to get some advice on <TASK>.",
166
+ "Can you explain the process of <TASK>?",
167
+ "I'm looking for guidance on how to approach <TASK>.",
168
+ "What's the proper way to handle <TASK>?",
169
+ "How should I proceed with <TASK>?",
170
+ "I'm interested in your expertise on <TASK>.",
171
+ "Could you walk me through the steps for <TASK>?",
172
+ "I'm not sure where to begin when it comes to <TASK>.",
173
+ "What should I prioritize when doing <TASK>?",
174
+ "How can I ensure success with <TASK>?",
175
+ "I'd appreciate some tips on <TASK>.",
176
+ "Can you provide a roadmap for <TASK>?",
177
+ "What's the recommended course of action for <TASK>?",
178
+ "I'm seeking your guidance on <TASK>.",
179
+ "Could you offer some suggestions for <TASK>?",
180
+ "I'd like to know the steps to take for <TASK>.",
181
+ "What's the most effective way to achieve <TASK>?",
182
+ "How can I make the most of <TASK>?",
183
+ "I'm wondering about the best approach to <TASK>.",
184
+ "Can you share your insights on <TASK>?",
185
+ "What steps should I follow to complete <TASK>?",
186
+ "I'm looking for advice on <TASK>.",
187
+ "What's the strategy for successfully completing <TASK>?",
188
+ "How should I prepare myself for <TASK>?",
189
+ "I'm not sure where to start with <TASK>.",
190
+ "What's the procedure for <TASK>?",
191
+ "Could you provide some guidance on <TASK>?",
192
+ "I'd like to get some tips on how to <TASK>.",
193
+ "Can you explain how to tackle <TASK> step by step?",
194
+ "I'm interested in understanding the process of <TASK>.",
195
+ "What are the key steps to <TASK>?",
196
+ "Is there a specific method that works for <TASK>?",
197
+ "I'd appreciate your advice on successfully completing <TASK>.",
198
+ "Can you shed light on the best way to <TASK>?",
199
+ "What would you recommend as the first step to <TASK>?",
200
+ "How do I initiate <TASK>?",
201
+ "I'm inquiring about the recommended steps for <TASK>.",
202
+ "Could you share some insights into <TASK>?",
203
+ "I'm seeking your expertise on <TASK>.",
204
+ "What's your recommended approach for <TASK>?",
205
+ "I'd like some guidance on where to start with <TASK>.",
206
+ "Can you provide recommendations for <TASK>?",
207
+ "What's your advice for someone looking to <TASK>?",
208
+ "I'm seeking your input on the process of <TASK>.",
209
+ "How can I achieve success with <TASK>?",
210
+ "What's the best way to navigate <TASK>?",
211
+ "I'm curious about the steps required for <TASK>.",
212
+ "Could you show me the proper way to <TASK>?",
213
+ "I need to know the necessary steps for <TASK>.",
214
+ "What's the most efficient method for <TASK>?",
215
+ "I'd appreciate your guidance on <TASK>.",
216
+ "Can you explain the steps involved in <TASK>?",
217
+ "I'm looking for recommendations on how to approach <TASK>.",
218
+ "What's the right way to handle <TASK>?",
219
+ "How should I manage <TASK>?",
220
+ "I'm interested in your insights on <TASK>.",
221
+ "Could you provide a step-by-step guide for <TASK>?",
222
+ "I'm not sure how to start when it comes to <TASK>.",
223
+ "What are the key factors to consider for <TASK>?",
224
+ "How can I ensure a successful outcome with <TASK>?",
225
+ "I'd like some tips and tricks for <TASK>.",
226
+ "Can you offer a roadmap for accomplishing <TASK>?",
227
+ "What's the preferred course of action for <TASK>?",
228
+ "I'm seeking your expert advice on <TASK>.",
229
+ "Could you suggest some best practices for <TASK>?",
230
+ "I'd like to understand the necessary steps to complete <TASK>.",
231
+ "What's the most effective strategy for <TASK>?",
232
+ ]
233
+
234
+ template_grounding_cogvlm = [
235
+ "Where is <TASK>?",
236
+ "Where is <TASK> in the image?",
237
+ "Where is <TASK>? answer in [[x0,y0,x1,y1]] format.",
238
+ "Can you point out <TASK> in the image and provide the bounding boxes of its location?",
239
+ "Help me to locate <TASK> in and give me its bounding boxes, please.",
240
+ "In the given, could you find and tell me the bounding boxes of <TASK>?",
241
+ "Guide me to the location of <TASK> within the image by providing its bounding boxes.",
242
+ "I'd like to know the exact bounding boxes of <TASK> in the photo.",
243
+ "Would you kindly provide the bounding boxes of <TASK> located in the picture?",
244
+ "Can you find <TASK> in and give me the bounding boxes of where it is located?",
245
+ "I'm trying to locate <TASK> in. Can you determine its bounding boxes for me?",
246
+ "What are the bounding boxes of <TASK> in the image?",
247
+ "Can you disclose the position of <TASK> in the photograph by stating its bounding boxes?",
248
+ "In, could you let me know the location of <TASK> in the form of bounding boxes?",
249
+ "I need the bounding boxes of <TASK> in, can you please assist me with that?",
250
+ "Where in is <TASK> located? Provide me with its bounding boxes, please.",
251
+ "May I have the bounding boxes of <TASK>?",
252
+ "In the photograph, could you pinpoint the location of <TASK> and tell me its bounding boxes?",
253
+ "Can you please search and find <TASK> in, then let me know its bounding boxes?",
254
+ "Please, point out the position of <TASK> in the image by giving its bounding boxes.",
255
+ "What are the exact bounding boxes of <TASK> in the provided picture?",
256
+ "Detect the location of <TASK> in and share the bounding boxes with me, please.",
257
+ "In the picture, I'd like you to locate <TASK> and provide its coordinates.",
258
+ "Please indicate the location of <TASK> in the photo by giving bounding boxes.",
259
+ "Find <TASK> in and share its coordinates with me.",
260
+ "Could you please help me find the bounding boxes of <TASK> in the image?",
261
+ "I am looking for the position of <TASK> in. Can you provide its bounding boxes?",
262
+ "In the image, can you locate <TASK> and let me know its coordinates?",
263
+ "I'd appreciate if you could find and tell me the bounding boxes of <TASK>.",
264
+ "In, I need the bounding box bounding boxes of <TASK>.",
265
+ "Point me to the location of <TASK> in the picture by providing its bounding boxes.",
266
+ "Could you trace <TASK> in and tell me its bounding boxes?",
267
+ "Can you assist me in locating <TASK> in, and then provide its bounding boxes?",
268
+ "I'm curious, what are the bounding boxes of <TASK> in the photo?",
269
+ "Kindly share the bounding boxes of <TASK> located in the image.",
270
+ "I would like to find <TASK> in. Can you give me its bounding boxes?",
271
+ "Can you spot <TASK> in and disclose its bounding boxes to me?",
272
+ "Please, reveal the location of <TASK> in the provided photograph as coordinates.",
273
+ "Help me locate and determine the bounding boxes of <TASK>.",
274
+ "I request the bounding boxes of <TASK> in the image.",
275
+ "In the given, can you find <TASK> and tell me its bounding boxes?",
276
+ "I need to know the position of <TASK> in as bounding boxes.",
277
+ "Locate <TASK> in and provide its bounding boxes, please.",
278
+ "Assist me in finding <TASK> in the photo and provide the bounding box bounding boxes.",
279
+ "In, can you guide me to the location of <TASK> by providing bounding boxes?",
280
+ "I'd like the bounding boxes of <TASK> as it appears in the image.",
281
+ "What location does <TASK> hold in the picture? Inform me of its bounding boxes.",
282
+ "Identify the position of <TASK> in and share its bounding boxes.",
283
+ "I'd like to request the bounding boxes of <TASK> within the photo.",
284
+ "How can I locate <TASK> in the image? Please provide the bounding boxes.",
285
+ "I am interested in knowing the bounding boxes of <TASK> in the picture.",
286
+ "Assist me in locating the position of <TASK> in the photograph and its bounding box bounding boxes.",
287
+ "In the image, I need to find <TASK> and know its bounding boxes. Can you please help?"
288
+ "Can you give me a description of the region <TASK> in image?",
289
+ "In the provided image, would you mind describing the selected area <TASK>?",
290
+ "I need details about the area <TASK> located within image.",
291
+ "Could you please share some information on the region <TASK> in this photograph?",
292
+ "Describe what's happening within the coordinates <TASK> of the given image.",
293
+ "What can you tell me about the selected region <TASK> in the photo?",
294
+ "Please, can you help me understand what's inside the region <TASK> in image?",
295
+ "Give me a comprehensive description of the specified area <TASK> in the picture.",
296
+ "I'm curious about the area <TASK> in the following image. Can you describe it?",
297
+ "Please elaborate on the area with the coordinates <TASK> in the visual.",
298
+ "In the displayed image, help me understand the region defined by <TASK>.",
299
+ "Regarding the image, what's going on in the section <TASK>?",
300
+ "In the given photograph, can you explain the area with coordinates <TASK>?",
301
+ "Kindly describe what I should be seeing in the area <TASK> of image.",
302
+ "Within the input image, what can be found in the region defined by <TASK>?",
303
+ "Tell me what you see within the designated area <TASK> in the picture.",
304
+ "Please detail the contents of the chosen region <TASK> in the visual input.",
305
+ "What's inside the area <TASK> of the provided graphic?",
306
+ "I'd like some information about the specific region <TASK> in the image.",
307
+ "Help me understand the details within the area <TASK> in photograph.",
308
+ "Can you break down the region <TASK> in the image for me?",
309
+ "What is taking place within the specified area <TASK> in this capture?",
310
+ "Care to elaborate on the targeted area <TASK> in the visual illustration?",
311
+ "What insights can you provide about the area <TASK> in the selected picture?",
312
+ "What does the area <TASK> within the given visual contain?",
313
+ "Analyze and describe the region <TASK> in the included photo.",
314
+ "Please provide details for the area marked as <TASK> in this photographic.",
315
+ "For the image, can you assess and describe what's happening at <TASK>?",
316
+ "Fill me in about the selected portion <TASK> within the presented image.",
317
+ "In the image, elaborate on the details found within the section <TASK>.",
318
+ "Please interpret and describe the area <TASK> inside the given picture.",
319
+ "What information can you give me about the coordinates <TASK> in image?",
320
+ "Regarding the coordinates <TASK> in image, can you provide a description?",
321
+ "In the photo, can you delve into the details of the region <TASK>?",
322
+ "Please provide insights on the specified area <TASK> within the graphic.",
323
+ "Detail the chosen region <TASK> in the depicted scene.",
324
+ "Can you discuss the entities within the region <TASK> of image?",
325
+ "I'd appreciate a breakdown of the area <TASK> in the displayed image.",
326
+ "What's the story in the section <TASK> of the included visual?",
327
+ "Please enlighten me about the region <TASK> in the given photo.",
328
+ "Offer a thorough description of the area <TASK> within the illustration.",
329
+ "What can you share about the area <TASK> in the presented image?",
330
+ "Help me grasp the context of the region <TASK> within image.",
331
+ "Kindly give an overview of the section <TASK> in photo.",
332
+ "What details can you provide about the region <TASK> in the snapshot?",
333
+ "Can you divulge the contents of the area <TASK> within the given image?",
334
+ "In the submitted image, please give a synopsis of the area <TASK>.",
335
+ "In the image, please describe the bounding box <TASK>.",
336
+ "Please describe the region <TASK> in the picture.",
337
+ "Describe the bbox <TASK> in the provided photo.",
338
+ "What can you tell me about the area <TASK> within the image?",
339
+ "Could you give me a description of the rectangular region <TASK> found in?",
340
+ "In, what elements can be found within the coordinates <TASK>?",
341
+ "Please provide details for the area within the bounding box <TASK> in.",
342
+ "Can you generate a description for the selected region <TASK> in the image?",
343
+ "Kindly describe the objects or scenery in the bounding box <TASK> within.",
344
+ "What details can you provide for the rectangle defined by the coordinates <TASK> in?",
345
+ "In relation to the picture, please describe the content of the area marked by <TASK>.",
346
+ "I'd like to know more about the area <TASK> in the given image. Can you describe it?",
347
+ "Can you help me by describing the part of that lies within the bounding box <TASK>?",
348
+ "What's happening in the section of the photo enclosed by the coordinates <TASK>?",
349
+ "Describe the image content present in the specified rectangular area <TASK> of.",
350
+ "Please provide information about the area within the bounding box <TASK> in the picture.",
351
+ "Could you offer a description of the contents in the selected area <TASK> of the image?",
352
+ "I'm curious about the area <TASK> in. Can you provide a description of it?",
353
+ "What can be observed in the rectangular region <TASK> in the photograph?",
354
+ "Please explain what is contained in the portion of defined by the box <TASK>.",
355
+ "In the photograph, can you describe the objects or scenery enclosed by <TASK>?",
356
+ "Can you give a brief explanation of the specified area <TASK> in the image?",
357
+ "What does the area <TASK> look like in the context of the image?",
358
+ "Could you please describe the contents of the bounding box <TASK> in the given image?",
359
+ "I would like to know more about the rectangular region <TASK> within the picture. Can you describe it?",
360
+ "Please tell me about the area <TASK> in the image. What does it contain?",
361
+ "Help me understand what's happening in the selected bounding box <TASK> within.",
362
+ "Can you provide a description of the area <TASK> in the image?",
363
+ "What sort of things can be seen in the region <TASK> of the photo?",
364
+ "Describe what can be found within the bounds of <TASK> in the image.",
365
+ "In, can you paint a picture of the area enclosed by coordinates <TASK>?",
366
+ "Please provide a detailed account of the area covered by the bounding box <TASK> in.",
367
+ "Give me a vivid description of what's happening in the area <TASK> within the snapshot.",
368
+ "In the image, what do you observe within the rectangular box defined by the coordinates <TASK>?",
369
+ "Could you give me a breakdown of the content in the specified area <TASK> of the picture?",
370
+ "Please elucidate the area<TASK> of the image.",
371
+ "I'd appreciate it if you could describe the portion of that lies within the rectangle <TASK>.",
372
+ "Can you share some insights about the rectangular region <TASK> in the image?",
373
+ "Help me visualize the section of the photo enclosed by the bounding box <TASK>.",
374
+ "Would you kindly provide a description for the content within the rectangular area <TASK> of?",
375
+ "In, can you tell me more about the area specified by the bounding box <TASK>?",
376
+ "Please describe what can be seen in the rectangular region <TASK> of the image.",
377
+ "Can you analyze the content of the area <TASK> within the photograph?",
378
+ "In the provided image, please explain the content within the region <TASK>.",
379
+ "I'm interested in the selected rectangle <TASK> in. Can you tell me more about it?",
380
+ "Explain what can be found in the bounding box <TASK> in the context of the image.",
381
+ "Kindly share your observations about the rectangular region <TASK> within.",
382
+ "I'd like a thorough description of the area <TASK> in the image.",
383
+ "Could you please provide a description of the rectangular area <TASK> in?",
384
+ "Please describe the section of the picture defined by the bbox <TASK>.",
385
+ "Tell me more about the scenery or objects within the rectangular region <TASK> in.",
386
+ "Would you kindly describe the content of the area enclosed by <TASK> in the image?",
387
+ "Help me understand the objects or scenery within the bounding box <TASK> in the image.",
388
+ "I would like to know about the section of the image enclosed by the rectangle <TASK>. Can you describe it?",
389
+ "Describe the selected rectangular area <TASK> in the photo.",
390
+ "Tell me about the region <TASK> of the image.",
391
+ "I request a description of the area <TASK> in the picture.",
392
+ "Can you elaborate on the content of the bounding box <TASK> in?",
393
+ "Please share details about the rectangular region <TASK> within the image.",
394
+ "What can I find in the bbox <TASK> of the provided image?",
395
+ "In the image, could you provide a description for the coordinates <TASK>?",
396
+ "Could you tell me more about the area <TASK> in the snapshot?",
397
+ "Fill me in on the details of the rectangular box <TASK> within the image.",
398
+ "What's going on in the section of contained within the bounding box <TASK>?",
399
+ "I would like a description of the content within the bbox <TASK> in.",
400
+ "Please enlighten me about the area <TASK> in the photograph.",
401
+ "Can you give me a visual rundown of the area <TASK> in?",
402
+ "Describe the visual elements within the selected area <TASK> of the image.",
403
+ "Tell me what you see in the area <TASK> within the context of the image.",
404
+ "Explain the content within the rectangular region <TASK> of the image.",
405
+ "I'd like some information about the bounding box <TASK> in the photo.",
406
+ "What is happening within the rectangle defined by coordinates <TASK> in the image?",
407
+ "Please describe the content within the area <TASK> displayed in the image.",
408
+ "What can be seen in the bounding box <TASK> in the context of the provided image?",
409
+ "Share some details about the objects or environment within the bounding box <TASK> in.",
410
+ "Please describe the area <TASK> in the image for me.",
411
+ "Can you generate a description of the contents within the selected region <TASK> in?",
412
+ "What objects or scenery can be found in the area <TASK> in the image?",
413
+ "Please tell me more about the rectangular section <TASK> in the photo.",
414
+ "Could you describe the content of the bbox <TASK> in the image?",
415
+ "What does the selected region <TASK> in the image encompass?",
416
+ "I am interested in the region <TASK> of the image; please describe it.",
417
+ "Can you provide some context for the area <TASK> within the picture?",
418
+ "Please give me some details about the rectangle <TASK> in the image.",
419
+ "In the photo, what can you see within the region defined by the bounding box <TASK>?",
420
+ "I would like a detailed description of the portion of enclosed by the bbox <TASK>.",
421
+ "Please help me understand the content present within the rectangle <TASK> in.",
422
+ "Would you mind describing the rectangular area <TASK> in the provided image?"
423
+ ]