Spaces:
Runtime error
Runtime error
Tidied up comments
Browse files
src/data_synthesis/generate_data.py
CHANGED
@@ -15,12 +15,18 @@ from src.common import data_dir
|
|
15 |
|
16 |
|
17 |
class Review:
|
|
|
|
|
|
|
18 |
def __init__(self, stars: int, review_text: str):
|
19 |
self.stars = stars
|
20 |
self.review_text = review_text
|
21 |
|
22 |
|
23 |
class Product:
|
|
|
|
|
|
|
24 |
def __init__(self, category: str, name: str, description: str, price: float, features: List[str], reviews: List[Review]):
|
25 |
self.category = category
|
26 |
self.name = name
|
@@ -32,7 +38,7 @@ class Product:
|
|
32 |
|
33 |
class DataPrompt:
|
34 |
"""
|
35 |
-
Holder for static prompt generation functions
|
36 |
"""
|
37 |
@staticmethod
|
38 |
def prompt_setup() -> str:
|
@@ -94,6 +100,9 @@ Please format the response as json in this style:
|
|
94 |
|
95 |
|
96 |
def generate_products(category: str, features: List[str], k: int = 20):
|
|
|
|
|
|
|
97 |
prompt = DataPrompt.products_for_category(category, features, k)
|
98 |
response = openai.ChatCompletion.create(
|
99 |
model="gpt-3.5-turbo-16k",
|
@@ -108,16 +117,27 @@ def generate_products(category: str, features: List[str], k: int = 20):
|
|
108 |
|
109 |
|
110 |
def category_product_file(category: str) -> str:
|
|
|
|
|
|
|
111 |
output_file_name = f"products_{category.lower().replace(' ', '_')}.json"
|
112 |
return os.path.join(data_dir, 'json', output_file_name)
|
113 |
|
114 |
|
115 |
def category_review_file(category: str) -> str:
|
|
|
|
|
|
|
116 |
output_file_name = f"reviews_{category.lower().replace(' ', '_')}.json"
|
117 |
return os.path.join(data_dir, 'json', output_file_name)
|
118 |
|
119 |
|
120 |
def products_for_category(category: str) -> List[Product]:
|
|
|
|
|
|
|
|
|
|
|
121 |
cat_file = category_product_file(category)
|
122 |
if not os.path.exists(cat_file):
|
123 |
return []
|
@@ -141,6 +161,10 @@ def products_for_category(category: str) -> List[Product]:
|
|
141 |
|
142 |
|
143 |
def product_names_for_category(category: str) -> List[str]:
|
|
|
|
|
|
|
|
|
144 |
cat_file = category_product_file(category)
|
145 |
if not os.path.exists(cat_file):
|
146 |
return []
|
@@ -154,6 +178,10 @@ def product_names_for_category(category: str) -> List[str]:
|
|
154 |
|
155 |
|
156 |
def add_products(category: str, product_json: str, k: int) -> None:
|
|
|
|
|
|
|
|
|
157 |
cat_file = category_product_file(category)
|
158 |
if not os.path.exists(cat_file):
|
159 |
with open(cat_file, 'w') as f:
|
@@ -173,6 +201,10 @@ def add_products(category: str, product_json: str, k: int) -> None:
|
|
173 |
|
174 |
|
175 |
def get_categories_and_features() -> Dict[str, List[str]]:
|
|
|
|
|
|
|
|
|
176 |
product_features_file = os.path.join(data_dir, 'json', 'product_features.json')
|
177 |
cats_and_feats = {}
|
178 |
with open(product_features_file, 'r') as f:
|
@@ -185,6 +217,10 @@ def get_categories_and_features() -> Dict[str, List[str]]:
|
|
185 |
|
186 |
|
187 |
def generate_all_products(target_count=40):
|
|
|
|
|
|
|
|
|
188 |
product_features_file = os.path.join(data_dir, 'product_features.json')
|
189 |
|
190 |
with open(product_features_file, 'r') as f:
|
@@ -202,6 +238,9 @@ def generate_all_products(target_count=40):
|
|
202 |
|
203 |
|
204 |
def dump_products_to_csv():
|
|
|
|
|
|
|
205 |
cats = get_categories_and_features().keys()
|
206 |
cat_keys = []
|
207 |
for cat in cats:
|
@@ -213,11 +252,17 @@ def dump_products_to_csv():
|
|
213 |
|
214 |
|
215 |
def generate_reviews(target_count: int):
|
|
|
|
|
|
|
216 |
for cat in get_categories_and_features().keys():
|
217 |
generate_reviews_for_category(cat, target_count)
|
218 |
|
219 |
|
220 |
def generate_reviews_for_category(category: str, target_count: int):
|
|
|
|
|
|
|
221 |
batch_size = 25 # Max number of reviews to request in one go from GPT
|
222 |
|
223 |
# Set up a loop to continue trying to find more work to do until complete
|
@@ -249,6 +294,9 @@ def generate_reviews_for_category(category: str, target_count: int):
|
|
249 |
|
250 |
|
251 |
def generate_reviews_for_product(product: Product, k: int):
|
|
|
|
|
|
|
252 |
prompt = DataPrompt.reviews_for_product(product, k)
|
253 |
response = openai.ChatCompletion.create(
|
254 |
model="gpt-3.5-turbo-16k",
|
@@ -263,6 +311,10 @@ def generate_reviews_for_product(product: Product, k: int):
|
|
263 |
|
264 |
|
265 |
def add_reviews_to_product(reviews_json: str, product: Product):
|
|
|
|
|
|
|
|
|
266 |
reviews_json = json.loads(reviews_json)
|
267 |
reviews_file = category_review_file(product.category)
|
268 |
if not os.path.exists(reviews_file):
|
|
|
15 |
|
16 |
|
17 |
class Review:
|
18 |
+
"""
|
19 |
+
Simple representation of a user Review of a Product
|
20 |
+
"""
|
21 |
def __init__(self, stars: int, review_text: str):
|
22 |
self.stars = stars
|
23 |
self.review_text = review_text
|
24 |
|
25 |
|
26 |
class Product:
|
27 |
+
"""
|
28 |
+
Simple representation of a prduct
|
29 |
+
"""
|
30 |
def __init__(self, category: str, name: str, description: str, price: float, features: List[str], reviews: List[Review]):
|
31 |
self.category = category
|
32 |
self.name = name
|
|
|
38 |
|
39 |
class DataPrompt:
|
40 |
"""
|
41 |
+
Holder for static prompt generation functions for the data generation process
|
42 |
"""
|
43 |
@staticmethod
|
44 |
def prompt_setup() -> str:
|
|
|
100 |
|
101 |
|
102 |
def generate_products(category: str, features: List[str], k: int = 20):
|
103 |
+
"""
|
104 |
+
Invoke GPT3.5 Turbo model and get it to generate some products based on a category
|
105 |
+
"""
|
106 |
prompt = DataPrompt.products_for_category(category, features, k)
|
107 |
response = openai.ChatCompletion.create(
|
108 |
model="gpt-3.5-turbo-16k",
|
|
|
117 |
|
118 |
|
119 |
def category_product_file(category: str) -> str:
|
120 |
+
"""
|
121 |
+
Utility to get the file containing products in a category
|
122 |
+
"""
|
123 |
output_file_name = f"products_{category.lower().replace(' ', '_')}.json"
|
124 |
return os.path.join(data_dir, 'json', output_file_name)
|
125 |
|
126 |
|
127 |
def category_review_file(category: str) -> str:
|
128 |
+
"""
|
129 |
+
Utility to get the file containing reviews of products in a category
|
130 |
+
"""
|
131 |
output_file_name = f"reviews_{category.lower().replace(' ', '_')}.json"
|
132 |
return os.path.join(data_dir, 'json', output_file_name)
|
133 |
|
134 |
|
135 |
def products_for_category(category: str) -> List[Product]:
|
136 |
+
"""
|
137 |
+
Load all the associated products which have been generated for this
|
138 |
+
category, and the reviews, then merge the two and return a list of
|
139 |
+
all the products in this category along with their reviews
|
140 |
+
"""
|
141 |
cat_file = category_product_file(category)
|
142 |
if not os.path.exists(cat_file):
|
143 |
return []
|
|
|
161 |
|
162 |
|
163 |
def product_names_for_category(category: str) -> List[str]:
|
164 |
+
"""
|
165 |
+
Get a list of just the names of the products in this category
|
166 |
+
from the generated product json file
|
167 |
+
"""
|
168 |
cat_file = category_product_file(category)
|
169 |
if not os.path.exists(cat_file):
|
170 |
return []
|
|
|
178 |
|
179 |
|
180 |
def add_products(category: str, product_json: str, k: int) -> None:
|
181 |
+
"""
|
182 |
+
Given a string of json representing newly generated products,
|
183 |
+
add those products to the existing product json file for this category
|
184 |
+
"""
|
185 |
cat_file = category_product_file(category)
|
186 |
if not os.path.exists(cat_file):
|
187 |
with open(cat_file, 'w') as f:
|
|
|
201 |
|
202 |
|
203 |
def get_categories_and_features() -> Dict[str, List[str]]:
|
204 |
+
"""
|
205 |
+
Get dictionary of will each category as a key and the list of available
|
206 |
+
features to products in that category as the value
|
207 |
+
"""
|
208 |
product_features_file = os.path.join(data_dir, 'json', 'product_features.json')
|
209 |
cats_and_feats = {}
|
210 |
with open(product_features_file, 'r') as f:
|
|
|
217 |
|
218 |
|
219 |
def generate_all_products(target_count=40):
|
220 |
+
"""
|
221 |
+
Generate all products for all categories, trying to reach a given target count
|
222 |
+
of products.
|
223 |
+
"""
|
224 |
product_features_file = os.path.join(data_dir, 'product_features.json')
|
225 |
|
226 |
with open(product_features_file, 'r') as f:
|
|
|
238 |
|
239 |
|
240 |
def dump_products_to_csv():
|
241 |
+
"""
|
242 |
+
Dump a csv file for debug, for every product showing category name and product name
|
243 |
+
"""
|
244 |
cats = get_categories_and_features().keys()
|
245 |
cat_keys = []
|
246 |
for cat in cats:
|
|
|
252 |
|
253 |
|
254 |
def generate_reviews(target_count: int):
|
255 |
+
"""
|
256 |
+
Generate reviews for each category up to a target count of reviews
|
257 |
+
"""
|
258 |
for cat in get_categories_and_features().keys():
|
259 |
generate_reviews_for_category(cat, target_count)
|
260 |
|
261 |
|
262 |
def generate_reviews_for_category(category: str, target_count: int):
|
263 |
+
"""
|
264 |
+
Generate reviews for a specific category up to a given target number of reviews
|
265 |
+
"""
|
266 |
batch_size = 25 # Max number of reviews to request in one go from GPT
|
267 |
|
268 |
# Set up a loop to continue trying to find more work to do until complete
|
|
|
294 |
|
295 |
|
296 |
def generate_reviews_for_product(product: Product, k: int):
|
297 |
+
"""
|
298 |
+
Generate a number of reviews from GPT3.5 for a specific product and add them to the product
|
299 |
+
"""
|
300 |
prompt = DataPrompt.reviews_for_product(product, k)
|
301 |
response = openai.ChatCompletion.create(
|
302 |
model="gpt-3.5-turbo-16k",
|
|
|
311 |
|
312 |
|
313 |
def add_reviews_to_product(reviews_json: str, product: Product):
|
314 |
+
"""
|
315 |
+
Load the reviews file containing this product category, append this review to the list and
|
316 |
+
re-save the file
|
317 |
+
"""
|
318 |
reviews_json = json.loads(reviews_json)
|
319 |
reviews_file = category_review_file(product.category)
|
320 |
if not os.path.exists(reviews_file):
|