Upload test.py
Browse files
test.py
ADDED
@@ -0,0 +1,183 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import anthropic
|
2 |
+
import base64
|
3 |
+
import httpx
|
4 |
+
import os
|
5 |
+
import time
|
6 |
+
from mimetypes import guess_type
|
7 |
+
|
8 |
+
import random
|
9 |
+
# import numpy as np
|
10 |
+
def seed_everything(seed):
|
11 |
+
random.seed(seed)
|
12 |
+
# np.random.seed(seed)
|
13 |
+
os.environ['PYTHONHASHSEED'] = str(seed)
|
14 |
+
# torch.manual_seed(seed)
|
15 |
+
# torch.cuda.manual_seed(seed)
|
16 |
+
# torch.backends.cudnn.deterministic = True
|
17 |
+
# env.seed(seed)
|
18 |
+
seed_everything(1)
|
19 |
+
|
20 |
+
def local_image_to_data_url(image_path):
|
21 |
+
mime_type, _ = guess_type(image_path)
|
22 |
+
if mime_type is None:
|
23 |
+
mime_type = 'application/octet-stream'
|
24 |
+
with open(image_path, "rb") as image_file:
|
25 |
+
base64_encoded_data = base64.b64encode(image_file.read()).decode('utf-8')
|
26 |
+
return base64_encoded_data
|
27 |
+
|
28 |
+
client = anthropic.Anthropic(
|
29 |
+
api_key="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",
|
30 |
+
)
|
31 |
+
|
32 |
+
|
33 |
+
# Start test
|
34 |
+
levels = [3,4,5,6,7,8]
|
35 |
+
in_context_example_num = 0 # 0, 1, 2, 4, 8
|
36 |
+
|
37 |
+
if in_context_example_num > 0:
|
38 |
+
output_path = "output/output_img_%d/"%(in_context_example_num)
|
39 |
+
input_backup_path = "input/input_backup_img_%d/"%(in_context_example_num)
|
40 |
+
else:
|
41 |
+
output_path = "output/output_img/"
|
42 |
+
input_backup_path = "input/input_backup_img/"
|
43 |
+
|
44 |
+
os.makedirs(output_path, exist_ok=True)
|
45 |
+
os.makedirs(input_backup_path, exist_ok=True)
|
46 |
+
|
47 |
+
EXAMPLE_DICT = {
|
48 |
+
3: [],
|
49 |
+
4: [],
|
50 |
+
5: [],
|
51 |
+
6: [],
|
52 |
+
7: [],
|
53 |
+
8: [],
|
54 |
+
}
|
55 |
+
# for level in levels:
|
56 |
+
# for example_id in range(8):
|
57 |
+
# curr_example_pack = {}
|
58 |
+
# curr_example_pack["image_path"] = "../example/level%d/img/%d.png"%(level, example_id)
|
59 |
+
# with open("../example/level%d/answer/%d.txt"%(level, example_id), "r") as f:
|
60 |
+
# curr_example_pack["answer"] = f.read()
|
61 |
+
|
62 |
+
# curr_example_pack["pure_text"] = "../example/level%d/pure_text/%d.txt"%(level, example_id)
|
63 |
+
# curr_example_pack["table"] = "../example/level%d/table/%d.txt"%(level, example_id)
|
64 |
+
|
65 |
+
# curr_example_pack["start_image_path"] = "../example/level%d/begin/%d.jpg"%(level, example_id)
|
66 |
+
# curr_example_pack["end_image_path"] = "../example/level%d/end/%d.jpg"%(level, example_id)
|
67 |
+
# example_path = "../example/level%d/"%(level)
|
68 |
+
# curr_example_pack["question1"] = "\n\nPlease generate the moving plan. The beginning state is:"
|
69 |
+
# curr_example_pack["question2"] = "\nThe end state is:"
|
70 |
+
# with open(example_path + "sol_%d.txt"%(example_id), "r") as f:
|
71 |
+
# curr_example_pack["answer"] = f.read()
|
72 |
+
# EXAMPLE_DICT[level].append(curr_example_pack)
|
73 |
+
|
74 |
+
|
75 |
+
import ipdb; ipdb.set_trace()
|
76 |
+
for level in levels:
|
77 |
+
os.makedirs(output_path + "level%d"%(level), exist_ok=True)
|
78 |
+
os.makedirs(input_backup_path + "level%d"%(level), exist_ok=True)
|
79 |
+
start_idx = 0
|
80 |
+
end_idx = 100
|
81 |
+
runned_term = 0
|
82 |
+
map_path = "../maps/level%d/img/"%(level)
|
83 |
+
while True:
|
84 |
+
try:
|
85 |
+
curr_id = start_idx + runned_term
|
86 |
+
if curr_id >= end_idx:
|
87 |
+
break
|
88 |
+
|
89 |
+
prompt_input_1 = '''
|
90 |
+
In this task, you will analyze a maze to determine if there is a hole in a specific position.
|
91 |
+
The following figure illustrates the appearances of the player, holes, lands, and the goal within the maze. You will need to focus on the appearance of the hole.
|
92 |
+
'''
|
93 |
+
prompt_input_2 = '''
|
94 |
+
Here is an example to illustrate how to analyze and answer the question:
|
95 |
+
'''
|
96 |
+
prompt_input_3 = '''
|
97 |
+
Example question: Is there a hole in row 3, column 3?
|
98 |
+
|
99 |
+
In this example:
|
100 |
+
- We check the position in row 3, column 3.
|
101 |
+
- According to the image, it is a land square. It does not contain a hole.
|
102 |
+
- Therefore, you will output "<Output> No".
|
103 |
+
|
104 |
+
Your output should be: "<Output> No" or "<Output> Yes", depending on whether there is a hole at the specified position.
|
105 |
+
'''
|
106 |
+
# prompt_examples = []
|
107 |
+
# image_examples = []
|
108 |
+
# if in_context_example_num > 0:
|
109 |
+
# prompt_examples.append("## Example:\n")
|
110 |
+
# example_indices = random.sample(range(8), in_context_example_num)
|
111 |
+
# for example_index in example_indices:
|
112 |
+
# this_example = EXAMPLE_DICT[level][example_index]
|
113 |
+
# image_examples.append(local_image_to_data_url(this_example["image_path"]))
|
114 |
+
# prompt_examples.append(this_example["answer"])
|
115 |
+
prompt_input_4 = "\n\nNow you will analyze the following maze and answer the question: "
|
116 |
+
with open("../maps/level%d/question/%d.txt"%(level, curr_id), "r") as f:
|
117 |
+
prompt_input_5 = f.read()
|
118 |
+
|
119 |
+
# construct
|
120 |
+
content_input_seq = []
|
121 |
+
content_input_seq.append({
|
122 |
+
"type": "text",
|
123 |
+
"text": prompt_input_1,
|
124 |
+
})
|
125 |
+
content_input_seq.append({
|
126 |
+
"type": "image",
|
127 |
+
"source": {
|
128 |
+
"type": "base64",
|
129 |
+
"media_type": "image/png",
|
130 |
+
"data": local_image_to_data_url("../prompt-visual-images/system-figure-1.png"),
|
131 |
+
}
|
132 |
+
})
|
133 |
+
content_input_seq.append({
|
134 |
+
"type": "text",
|
135 |
+
"text": prompt_input_2,
|
136 |
+
})
|
137 |
+
content_input_seq.append({
|
138 |
+
"type": "image",
|
139 |
+
"source": {
|
140 |
+
"type": "base64",
|
141 |
+
"media_type": "image/png",
|
142 |
+
"data": local_image_to_data_url("../prompt-visual-images/system-figure-2.png"),
|
143 |
+
}
|
144 |
+
})
|
145 |
+
content_input_seq.append({
|
146 |
+
"type": "text",
|
147 |
+
"text": prompt_input_3,
|
148 |
+
})
|
149 |
+
content_input_seq.append({
|
150 |
+
"type": "text",
|
151 |
+
"text": prompt_input_4,
|
152 |
+
})
|
153 |
+
content_input_seq.append({
|
154 |
+
"type": "image",
|
155 |
+
"source": {
|
156 |
+
"type": "base64",
|
157 |
+
"media_type": "image/png",
|
158 |
+
"data": local_image_to_data_url(map_path + "%d.png"%(curr_id)),
|
159 |
+
}
|
160 |
+
})
|
161 |
+
content_input_seq.append({
|
162 |
+
"type": "text",
|
163 |
+
"text": prompt_input_5,
|
164 |
+
})
|
165 |
+
response = client.messages.create(
|
166 |
+
model="claude-3-sonnet-20240229",
|
167 |
+
max_tokens=1024,
|
168 |
+
system="You are a maze-solving agent playing a pixelated maze videogame.\nMazes are presented on grid maps, where each tile can be empty land, or contain a player, hole, or goal.",
|
169 |
+
messages=[
|
170 |
+
{
|
171 |
+
"role": "user",
|
172 |
+
"content": content_input_seq
|
173 |
+
},
|
174 |
+
],
|
175 |
+
)
|
176 |
+
with open(output_path + "level%d/%d.txt"%(level, curr_id), "w") as f:
|
177 |
+
f.write(response.content[0].text)
|
178 |
+
time.sleep(2)
|
179 |
+
runned_term += 1
|
180 |
+
except:
|
181 |
+
time.sleep(2)
|
182 |
+
pass
|
183 |
+
|