Spaces:
Paused
Paused
tmzh commited on
Commit ·
aedb933
1
Parent(s): e640a42
refactoring
Browse files
app.py
CHANGED
|
@@ -5,7 +5,8 @@ import os
|
|
| 5 |
import random
|
| 6 |
import json
|
| 7 |
import re
|
| 8 |
-
from typing import List
|
|
|
|
| 9 |
|
| 10 |
import gradio as gr
|
| 11 |
import outlines
|
|
@@ -13,44 +14,24 @@ import requests
|
|
| 13 |
from outlines import models, generate, samplers
|
| 14 |
from pydantic import BaseModel
|
| 15 |
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
|
| 21 |
-
clues: A list of clues, where each clue is a list containing the words, the answer, and the explanation.
|
| 22 |
-
num_merges: The approximate number of merges to generate (default: 10).
|
| 23 |
-
|
| 24 |
-
Returns:
|
| 25 |
-
A list of tuples, where each tuple contains the merged words and the indices of the selected rows.
|
| 26 |
-
"""
|
| 27 |
-
|
| 28 |
-
merges = []
|
| 29 |
-
while len(merges) < num_merges:
|
| 30 |
-
num_rows = random.choice([3, 4])
|
| 31 |
-
selected_rows = random.sample(range(len(clues)), num_rows)
|
| 32 |
-
merged_words = " ".join([word for row in [clues[i][0] for i in selected_rows] for word in row])
|
| 33 |
-
if len(merged_words.split()) in [8, 9]:
|
| 34 |
-
merges.append((merged_words.split(), selected_rows))
|
| 35 |
-
|
| 36 |
-
return merges
|
| 37 |
-
|
| 38 |
|
| 39 |
class Clue(BaseModel):
|
| 40 |
word: str
|
| 41 |
explanation: str
|
| 42 |
|
| 43 |
-
|
| 44 |
class Group(BaseModel):
|
| 45 |
words: List[str]
|
| 46 |
clue: str
|
| 47 |
explanation: str
|
| 48 |
|
| 49 |
-
|
| 50 |
class Groups(BaseModel):
|
| 51 |
groups: List[Group]
|
| 52 |
|
| 53 |
-
|
| 54 |
example_clues = [
|
| 55 |
(['ARROW', 'TIE', 'HONOR'], 'BOW', 'such as a bow and arrow, a bow tie, or a bow as a sign of honor'),
|
| 56 |
(['DOG', 'TREE'], 'BARK', 'such as the sound a dog makes, or a tree is made of bark'),
|
|
@@ -81,8 +62,36 @@ example_clues = [
|
|
| 81 |
'such as the Arctic being home to seals, or shutting a seal on an envelope, or a stamp being a type of seal'),
|
| 82 |
]
|
| 83 |
|
|
|
|
|
|
|
|
|
|
| 84 |
|
| 85 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 86 |
@outlines.prompt
|
| 87 |
def chat_group_template(system_prompt, query, history=[]):
|
| 88 |
'''<s><|system|>
|
|
@@ -99,10 +108,11 @@ def group_words(words):
|
|
| 99 |
'''
|
| 100 |
|
| 101 |
grouping_system_prompt = ("You are an assistant for the game Codenames. Your task is to help players by grouping a "
|
| 102 |
-
"given
|
| 103 |
-
"share a common theme or other word connections such as
|
|
|
|
| 104 |
example_groupings = []
|
| 105 |
-
merges =
|
| 106 |
for merged_words, indices in merges:
|
| 107 |
groups = [{
|
| 108 |
"secrets": example_clues[i][0],
|
|
@@ -111,20 +121,25 @@ def group_words(words):
|
|
| 111 |
} for i in indices]
|
| 112 |
example_groupings.append((merged_words, json.dumps(groups, separators=(',', ':'))))
|
| 113 |
|
| 114 |
-
prompt = chat_group_template(grouping_system_prompt,
|
| 115 |
sampler = samplers.greedy()
|
| 116 |
generator = generate.json(model, Groups, sampler)
|
| 117 |
|
| 118 |
-
print("Grouping words:
|
| 119 |
-
generations = generator(
|
| 120 |
-
|
| 121 |
-
max_tokens=500
|
| 122 |
-
)
|
| 123 |
-
print("Got groupings: ", generations)
|
| 124 |
return generations.groups
|
| 125 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 126 |
|
| 127 |
-
|
|
|
|
|
|
|
| 128 |
@outlines.prompt
|
| 129 |
def chat_clue_template(system, query, history=[]):
|
| 130 |
'''<s><|system|>
|
|
@@ -140,23 +155,29 @@ def generate_clues(group):
|
|
| 140 |
<|assistant|>
|
| 141 |
'''
|
| 142 |
|
| 143 |
-
clue_system_prompt = ("You are a
|
| 144 |
-
"a given group of words.
|
| 145 |
"allowed. Do not include the word 'Clue'. Do not provide explanations or notes.")
|
| 146 |
|
| 147 |
prompt = chat_clue_template(clue_system_prompt, group, example_clues)
|
| 148 |
-
# sampler = samplers.greedy()
|
| 149 |
sampler = samplers.multinomial(2, top_k=10)
|
| 150 |
generator = generate.json(model, Clue, sampler)
|
| 151 |
generations = generator(prompt, max_tokens=100)
|
| 152 |
-
print("
|
| 153 |
return generations[0]
|
| 154 |
|
| 155 |
-
|
| 156 |
-
|
| 157 |
-
|
| 158 |
|
| 159 |
https://stackoverflow.com/a/52281257
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 160 |
"""
|
| 161 |
# Min and Max quality
|
| 162 |
qmin, qmax = 25, 96
|
|
@@ -167,65 +188,90 @@ def jpeg_with_target_size(im, target):
|
|
| 167 |
|
| 168 |
# Encode into memory and get size
|
| 169 |
buffer = io.BytesIO()
|
| 170 |
-
|
| 171 |
s = buffer.getbuffer().nbytes
|
| 172 |
|
| 173 |
-
if s <=
|
| 174 |
qacc = m
|
| 175 |
qmin = m + 1
|
| 176 |
-
elif s >
|
| 177 |
qmax = m - 1
|
| 178 |
|
| 179 |
# Write to disk at the defined quality
|
| 180 |
if qacc > -1:
|
| 181 |
image_byte_array = io.BytesIO()
|
| 182 |
-
print("Acceptable quality",
|
| 183 |
-
|
| 184 |
return image_byte_array.getvalue()
|
| 185 |
|
|
|
|
|
|
|
|
|
|
| 186 |
|
| 187 |
-
|
| 188 |
-
|
| 189 |
-
max_size = (1024, 1024)
|
| 190 |
-
img.thumbnail(max_size)
|
| 191 |
|
| 192 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 193 |
image_b64 = base64.b64encode(image_byte_array).decode()
|
| 194 |
|
| 195 |
-
invoke_url = "https://ai.api.nvidia.com/v1/vlm/microsoft/phi-3-vision-128k-instruct"
|
| 196 |
-
stream = False
|
| 197 |
-
|
| 198 |
-
if os.environ.get("NVIDIA_API_KEY", "").startswith("nvapi-"):
|
| 199 |
-
print("Valid NVIDIA_API_KEY already in the environment. Delete to reset")
|
| 200 |
-
|
| 201 |
headers = {
|
| 202 |
"Authorization": f"Bearer {os.environ.get('NVIDIA_API_KEY', '')}",
|
| 203 |
-
"Accept": "
|
| 204 |
}
|
| 205 |
|
| 206 |
payload = {
|
| 207 |
"messages": [
|
| 208 |
{
|
| 209 |
"role": "user",
|
| 210 |
-
"content": f'Identify the words in this game of Codenames. Provide only a list of words.
|
| 211 |
-
f'words in capital letters only. <img src="data:image/png;base64,{image_b64}" />'
|
| 212 |
}
|
| 213 |
],
|
| 214 |
"max_tokens": 512,
|
| 215 |
"temperature": 0.1,
|
| 216 |
"top_p": 0.70,
|
| 217 |
-
"stream":
|
| 218 |
}
|
| 219 |
|
| 220 |
-
response = requests.post(
|
| 221 |
if response.ok:
|
| 222 |
print(response.json())
|
| 223 |
-
# Define the pattern to match uppercase words separated by commas
|
| 224 |
pattern = r'[A-Z]+(?:\s+[A-Z]+)?'
|
| 225 |
words = re.findall(pattern, response.json()['choices'][0]['message']['content'])
|
| 226 |
-
|
| 227 |
return gr.update(choices=words, value=words)
|
| 228 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 229 |
|
| 230 |
if __name__ == '__main__':
|
| 231 |
with gr.Blocks() as demo:
|
|
@@ -235,10 +281,7 @@ if __name__ == '__main__':
|
|
| 235 |
with gr.Row():
|
| 236 |
game_image = gr.Image(type="pil")
|
| 237 |
word_list_input = gr.Dropdown(label="Enter list of words (comma separated)",
|
| 238 |
-
choices=
|
| 239 |
-
'BEARD, BUNK, SECOND, SADDLE, BUCKET, JAIL, ANT, POCKET, LACE, '
|
| 240 |
-
'BREAK, CUCKOO, FLAT, NIL, TIN, CHERRY, CHRISTMAS, MOSES, '
|
| 241 |
-
'TEAM'.split(', '),
|
| 242 |
multiselect=True,
|
| 243 |
interactive=True)
|
| 244 |
|
|
@@ -277,7 +320,7 @@ if __name__ == '__main__':
|
|
| 277 |
|
| 278 |
def generate_clues_callback(group):
|
| 279 |
print("Generating clues: ", group)
|
| 280 |
-
g =
|
| 281 |
return gr.update(value=g.word, info=g.explanation)
|
| 282 |
|
| 283 |
|
|
|
|
| 5 |
import random
|
| 6 |
import json
|
| 7 |
import re
|
| 8 |
+
from typing import List, Tuple
|
| 9 |
+
import PIL
|
| 10 |
|
| 11 |
import gradio as gr
|
| 12 |
import outlines
|
|
|
|
| 14 |
from outlines import models, generate, samplers
|
| 15 |
from pydantic import BaseModel
|
| 16 |
|
| 17 |
+
# Constants
|
| 18 |
+
MAX_IMAGE_SIZE = (1024, 1024)
|
| 19 |
+
TARGET_IMAGE_SIZE = 180_000
|
| 20 |
+
NVIDIA_API_URL = "https://ai.api.nvidia.com/v1/vlm/microsoft/phi-3-vision-128k-instruct"
|
| 21 |
+
MODEL_NAME = "microsoft/Phi-3-mini-4k-instruct"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
|
| 23 |
class Clue(BaseModel):
|
| 24 |
word: str
|
| 25 |
explanation: str
|
| 26 |
|
|
|
|
| 27 |
class Group(BaseModel):
|
| 28 |
words: List[str]
|
| 29 |
clue: str
|
| 30 |
explanation: str
|
| 31 |
|
|
|
|
| 32 |
class Groups(BaseModel):
|
| 33 |
groups: List[Group]
|
| 34 |
|
|
|
|
| 35 |
example_clues = [
|
| 36 |
(['ARROW', 'TIE', 'HONOR'], 'BOW', 'such as a bow and arrow, a bow tie, or a bow as a sign of honor'),
|
| 37 |
(['DOG', 'TREE'], 'BARK', 'such as the sound a dog makes, or a tree is made of bark'),
|
|
|
|
| 62 |
'such as the Arctic being home to seals, or shutting a seal on an envelope, or a stamp being a type of seal'),
|
| 63 |
]
|
| 64 |
|
| 65 |
+
def create_random_word_groups(clues: List[Tuple[List[str], str, str]], target_groups: int = 10) -> List[Tuple[List[str], List[int]]]:
|
| 66 |
+
"""
|
| 67 |
+
Creates approximately 'target_groups' random groups of words from the given clues.
|
| 68 |
|
| 69 |
+
Args:
|
| 70 |
+
clues: A list of clues, where each clue is a tuple (words, answer, explanation).
|
| 71 |
+
target_groups: The desired number of groups to create.
|
| 72 |
+
|
| 73 |
+
Returns:
|
| 74 |
+
A list of tuples, each containing a list of merged words and their corresponding indices.
|
| 75 |
+
"""
|
| 76 |
+
groups = []
|
| 77 |
+
while len(groups) < target_groups:
|
| 78 |
+
num_rows = random.choice([3, 4])
|
| 79 |
+
selected_indices = random.sample(range(len(clues)), num_rows)
|
| 80 |
+
merged_words = [word for row in [clues[i][0] for i in selected_indices] for word in row]
|
| 81 |
+
if len(merged_words) in [8, 9]:
|
| 82 |
+
groups.append((merged_words, selected_indices))
|
| 83 |
+
return groups
|
| 84 |
+
|
| 85 |
+
def group_words(word_list: List[str]) -> List[Group]:
|
| 86 |
+
"""
|
| 87 |
+
Groups the given words into 3 to 4 thematic groups.
|
| 88 |
+
|
| 89 |
+
Args:
|
| 90 |
+
word_list: A list of words to be grouped.
|
| 91 |
+
|
| 92 |
+
Returns:
|
| 93 |
+
A list of Group objects representing the grouped words.
|
| 94 |
+
"""
|
| 95 |
@outlines.prompt
|
| 96 |
def chat_group_template(system_prompt, query, history=[]):
|
| 97 |
'''<s><|system|>
|
|
|
|
| 108 |
'''
|
| 109 |
|
| 110 |
grouping_system_prompt = ("You are an assistant for the game Codenames. Your task is to help players by grouping a "
|
| 111 |
+
"given set of words into 3 to 4 groups. Each group should consist of words that "
|
| 112 |
+
"share a common theme or other word connections such as homonyms, hypernyms, or synonyms.")
|
| 113 |
+
|
| 114 |
example_groupings = []
|
| 115 |
+
merges = create_random_word_groups(example_clues, 5)
|
| 116 |
for merged_words, indices in merges:
|
| 117 |
groups = [{
|
| 118 |
"secrets": example_clues[i][0],
|
|
|
|
| 121 |
} for i in indices]
|
| 122 |
example_groupings.append((merged_words, json.dumps(groups, separators=(',', ':'))))
|
| 123 |
|
| 124 |
+
prompt = chat_group_template(grouping_system_prompt, word_list, example_groupings)
|
| 125 |
sampler = samplers.greedy()
|
| 126 |
generator = generate.json(model, Groups, sampler)
|
| 127 |
|
| 128 |
+
print(f"Grouping words: {word_list}")
|
| 129 |
+
generations = generator(prompt, max_tokens=500)
|
| 130 |
+
print(f"Generated groupings: {generations}")
|
|
|
|
|
|
|
|
|
|
| 131 |
return generations.groups
|
| 132 |
|
| 133 |
+
def generate_clue(group: List[str]) -> Clue:
|
| 134 |
+
"""
|
| 135 |
+
Generates a single-word clue for the given group of words.
|
| 136 |
+
|
| 137 |
+
Args:
|
| 138 |
+
group: A list of words to generate a clue for.
|
| 139 |
|
| 140 |
+
Returns:
|
| 141 |
+
A Clue object containing the generated word and its explanation.
|
| 142 |
+
"""
|
| 143 |
@outlines.prompt
|
| 144 |
def chat_clue_template(system, query, history=[]):
|
| 145 |
'''<s><|system|>
|
|
|
|
| 155 |
<|assistant|>
|
| 156 |
'''
|
| 157 |
|
| 158 |
+
clue_system_prompt = ("You are a Codenames game companion. Your task is to give a single word clue related to "
|
| 159 |
+
"a given group of words. Respond with a single word clue only. Compound words are "
|
| 160 |
"allowed. Do not include the word 'Clue'. Do not provide explanations or notes.")
|
| 161 |
|
| 162 |
prompt = chat_clue_template(clue_system_prompt, group, example_clues)
|
|
|
|
| 163 |
sampler = samplers.multinomial(2, top_k=10)
|
| 164 |
generator = generate.json(model, Clue, sampler)
|
| 165 |
generations = generator(prompt, max_tokens=100)
|
| 166 |
+
print(f"Generated clues: {generations}")
|
| 167 |
return generations[0]
|
| 168 |
|
| 169 |
+
def compress_image_to_jpeg(image: 'PIL.Image', target_size: int) -> bytes:
|
| 170 |
+
"""
|
| 171 |
+
Compresses the image to JPEG format with the best quality that fits within the target size.
|
| 172 |
|
| 173 |
https://stackoverflow.com/a/52281257
|
| 174 |
+
|
| 175 |
+
Args:
|
| 176 |
+
image: The PIL Image object to compress.
|
| 177 |
+
target_size: The target file size in bytes.
|
| 178 |
+
|
| 179 |
+
Returns:
|
| 180 |
+
The compressed image as bytes.
|
| 181 |
"""
|
| 182 |
# Min and Max quality
|
| 183 |
qmin, qmax = 25, 96
|
|
|
|
| 188 |
|
| 189 |
# Encode into memory and get size
|
| 190 |
buffer = io.BytesIO()
|
| 191 |
+
image.save(buffer, format="JPEG", quality=m)
|
| 192 |
s = buffer.getbuffer().nbytes
|
| 193 |
|
| 194 |
+
if s <= target_size:
|
| 195 |
qacc = m
|
| 196 |
qmin = m + 1
|
| 197 |
+
elif s > target_size:
|
| 198 |
qmax = m - 1
|
| 199 |
|
| 200 |
# Write to disk at the defined quality
|
| 201 |
if qacc > -1:
|
| 202 |
image_byte_array = io.BytesIO()
|
| 203 |
+
print("Acceptable quality", image, image.format, f"{image.size}x{image.mode}")
|
| 204 |
+
image.save(image_byte_array, format='JPEG', quality=qacc)
|
| 205 |
return image_byte_array.getvalue()
|
| 206 |
|
| 207 |
+
def process_image(img: 'PIL.Image') -> gr.update:
|
| 208 |
+
"""
|
| 209 |
+
Processes the uploaded image to detect words for the Codenames game.
|
| 210 |
|
| 211 |
+
Args:
|
| 212 |
+
img: The uploaded PIL Image object.
|
|
|
|
|
|
|
| 213 |
|
| 214 |
+
Returns:
|
| 215 |
+
A gradio update object with the detected words.
|
| 216 |
+
"""
|
| 217 |
+
img.thumbnail(MAX_IMAGE_SIZE)
|
| 218 |
+
image_byte_array = compress_image_to_jpeg(img, TARGET_IMAGE_SIZE)
|
| 219 |
image_b64 = base64.b64encode(image_byte_array).decode()
|
| 220 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 221 |
headers = {
|
| 222 |
"Authorization": f"Bearer {os.environ.get('NVIDIA_API_KEY', '')}",
|
| 223 |
+
"Accept": "application/json"
|
| 224 |
}
|
| 225 |
|
| 226 |
payload = {
|
| 227 |
"messages": [
|
| 228 |
{
|
| 229 |
"role": "user",
|
| 230 |
+
"content": f'Identify the words in this game of Codenames. Provide only a list of words in capital letters. <img src="data:image/png;base64,{image_b64}" />'
|
|
|
|
| 231 |
}
|
| 232 |
],
|
| 233 |
"max_tokens": 512,
|
| 234 |
"temperature": 0.1,
|
| 235 |
"top_p": 0.70,
|
| 236 |
+
"stream": False
|
| 237 |
}
|
| 238 |
|
| 239 |
+
response = requests.post(NVIDIA_API_URL, headers=headers, json=payload)
|
| 240 |
if response.ok:
|
| 241 |
print(response.json())
|
|
|
|
| 242 |
pattern = r'[A-Z]+(?:\s+[A-Z]+)?'
|
| 243 |
words = re.findall(pattern, response.json()['choices'][0]['message']['content'])
|
|
|
|
| 244 |
return gr.update(choices=words, value=words)
|
| 245 |
|
| 246 |
+
def pad_or_truncate_groups(groups: List[Group], target_length: int = 4) -> List[Group]:
|
| 247 |
+
"""
|
| 248 |
+
Ensures the list of groups has exactly target_length elements, padding with empty Groups if necessary.
|
| 249 |
+
|
| 250 |
+
Args:
|
| 251 |
+
groups: The list of Group objects to pad or truncate.
|
| 252 |
+
target_length: The desired length of the list.
|
| 253 |
+
|
| 254 |
+
Returns:
|
| 255 |
+
A list of Group objects with the specified length.
|
| 256 |
+
"""
|
| 257 |
+
truncated_groups = groups[:target_length]
|
| 258 |
+
return truncated_groups + [Group(words=[], clue='', explanation='') for _ in range(target_length - len(truncated_groups))]
|
| 259 |
+
|
| 260 |
+
def group_words_callback(words: List[str]) -> List[gr.update]:
|
| 261 |
+
"""
|
| 262 |
+
Callback function to group the selected words.
|
| 263 |
+
|
| 264 |
+
Args:
|
| 265 |
+
words: A list of words to group.
|
| 266 |
+
|
| 267 |
+
Returns:
|
| 268 |
+
A list of gradio update objects for each group input.
|
| 269 |
+
"""
|
| 270 |
+
groups = group_words(words)
|
| 271 |
+
groups = pad_or_truncate_groups(groups, 4)
|
| 272 |
+
print(f"Generated groups: {groups}")
|
| 273 |
+
return [gr.update(value=group.words, choices=group.words, info=group.explanation) for group in groups]
|
| 274 |
+
|
| 275 |
|
| 276 |
if __name__ == '__main__':
|
| 277 |
with gr.Blocks() as demo:
|
|
|
|
| 281 |
with gr.Row():
|
| 282 |
game_image = gr.Image(type="pil")
|
| 283 |
word_list_input = gr.Dropdown(label="Enter list of words (comma separated)",
|
| 284 |
+
choices=[],
|
|
|
|
|
|
|
|
|
|
| 285 |
multiselect=True,
|
| 286 |
interactive=True)
|
| 287 |
|
|
|
|
| 320 |
|
| 321 |
def generate_clues_callback(group):
|
| 322 |
print("Generating clues: ", group)
|
| 323 |
+
g = generate_clue(group)
|
| 324 |
return gr.update(value=g.word, info=g.explanation)
|
| 325 |
|
| 326 |
|