Update agieval.py
Browse files- agieval.py +81 -46
agieval.py
CHANGED
@@ -16,6 +16,7 @@
|
|
16 |
import datasets
|
17 |
import json
|
18 |
import ast
|
|
|
19 |
|
20 |
_CITATION = """\
|
21 |
@ARTICLE{10174688,
|
@@ -65,6 +66,9 @@ _URLS = {
|
|
65 |
'math_agieval': {
|
66 |
"test": HEAD+'math.jsonl'
|
67 |
},
|
|
|
|
|
|
|
68 |
|
69 |
}
|
70 |
|
@@ -197,59 +201,90 @@ class AgiEval(datasets.GeneratorBasedBuilder):
|
|
197 |
_urls = _URLS[self.config.name]
|
198 |
urls = {
|
199 |
"test": _urls["test"],
|
|
|
200 |
}
|
201 |
data_dir = dl_manager.download_and_extract(urls)
|
202 |
-
|
203 |
datasets.SplitGenerator(
|
204 |
name=datasets.Split.TEST,
|
205 |
gen_kwargs={"filepath": data_dir["test"], "split": "test"},
|
206 |
),
|
207 |
]
|
|
|
|
|
|
|
|
|
|
|
|
|
208 |
|
209 |
def _generate_examples(self, filepath, split):
|
210 |
-
|
211 |
-
|
212 |
-
|
213 |
-
|
214 |
-
|
215 |
-
|
216 |
-
|
217 |
-
|
218 |
-
|
219 |
-
|
220 |
-
|
221 |
-
|
222 |
-
|
223 |
-
|
224 |
-
|
225 |
-
|
226 |
-
|
227 |
-
|
228 |
-
|
229 |
-
|
230 |
-
|
231 |
-
|
232 |
-
|
233 |
-
|
234 |
-
|
235 |
-
|
236 |
-
|
237 |
-
|
238 |
-
|
239 |
-
|
240 |
-
|
241 |
-
|
242 |
-
|
243 |
-
|
244 |
-
|
245 |
-
|
246 |
-
|
247 |
-
|
248 |
-
|
249 |
-
|
250 |
-
|
251 |
-
|
252 |
-
|
253 |
-
|
254 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
255 |
|
|
|
16 |
import datasets
|
17 |
import json
|
18 |
import ast
|
19 |
+
import csv
|
20 |
|
21 |
_CITATION = """\
|
22 |
@ARTICLE{10174688,
|
|
|
66 |
'math_agieval': {
|
67 |
"test": HEAD+'math.jsonl'
|
68 |
},
|
69 |
+
'few_shot': {
|
70 |
+
'few_shot': HEAD+'https://raw.githubusercontent.com/microsoft/AGIEval/main/data/few_shot_prompts.csv'
|
71 |
+
}
|
72 |
|
73 |
}
|
74 |
|
|
|
201 |
_urls = _URLS[self.config.name]
|
202 |
urls = {
|
203 |
"test": _urls["test"],
|
204 |
+
"few_shot": _URLS["few_shot"]["few_shot"],
|
205 |
}
|
206 |
data_dir = dl_manager.download_and_extract(urls)
|
207 |
+
splits = [
|
208 |
datasets.SplitGenerator(
|
209 |
name=datasets.Split.TEST,
|
210 |
gen_kwargs={"filepath": data_dir["test"], "split": "test"},
|
211 |
),
|
212 |
]
|
213 |
+
splits.append(datasets.SplitGenerator(
|
214 |
+
name="few_shot",
|
215 |
+
gen_kwargs={"filepath": data_dir["few_shot"], "split": "few_shot"},
|
216 |
+
))
|
217 |
+
|
218 |
+
return splits
|
219 |
|
220 |
def _generate_examples(self, filepath, split):
|
221 |
+
if split == "few_shot":
|
222 |
+
with open(filepath, mode="r", encoding="utf-8") as csv_file:
|
223 |
+
reader = csv.DictReader(csv_file)
|
224 |
+
|
225 |
+
# Extract data for the current dataset configuration
|
226 |
+
for key, row in enumerate(reader):
|
227 |
+
data_str = row[self.config.name]
|
228 |
+
|
229 |
+
# Skip if there's no data for this key (can happen due to alternating samples and explanations)
|
230 |
+
if pd.isnull(data_str):
|
231 |
+
continue
|
232 |
+
|
233 |
+
# Convert the string representation into a dictionary
|
234 |
+
data = json.loads(data_str.replace("'", "\""))
|
235 |
+
|
236 |
+
# Depending on the dataset, yield the correct format
|
237 |
+
if self.config.name in ["aqua_rat", "sat_math", "sat_en"]:
|
238 |
+
yield key, {
|
239 |
+
"question": data["question"],
|
240 |
+
"options": data["options"],
|
241 |
+
"label": data.get("label", None),
|
242 |
+
"solution": data.get("solution", None)
|
243 |
+
}
|
244 |
+
else:
|
245 |
+
with open(filepath, encoding="utf-8") as f:
|
246 |
+
for key, row in enumerate(f):
|
247 |
+
data = json.loads(row)
|
248 |
+
|
249 |
+
if self.config.name in ["aqua_rat","sat_math"]:
|
250 |
+
yield key, {
|
251 |
+
"question": data["question"],
|
252 |
+
"options": data["options"],
|
253 |
+
"label": data["label"],
|
254 |
+
"solution": data["other"]["solution"],
|
255 |
+
}
|
256 |
+
elif self.config.name == "logiqa":
|
257 |
+
yield key, {
|
258 |
+
"passage": data["passage"],
|
259 |
+
"question": data["question"],
|
260 |
+
"options": data["options"],
|
261 |
+
"label": data["label"],
|
262 |
+
}
|
263 |
+
elif self.config.name == "math_agieval":
|
264 |
+
if not data.get("level"):
|
265 |
+
data["level"] = data['other']['level']
|
266 |
+
if not data.get("type"):
|
267 |
+
data["type"] = data['other']['type']
|
268 |
+
yield key, {
|
269 |
+
"question": data["question"],
|
270 |
+
"answer": data["answer"],
|
271 |
+
"solution": data["other"]["solution"],
|
272 |
+
"level": data["level"],
|
273 |
+
"type": data["type"]
|
274 |
+
}
|
275 |
+
|
276 |
+
elif self.config.name == "sat_en":
|
277 |
+
yield key, {
|
278 |
+
"passage": data["passage"],
|
279 |
+
"question": data["question"],
|
280 |
+
"options": data["options"],
|
281 |
+
"label": data["label"],
|
282 |
+
"solution": data["other"]["solution"],
|
283 |
+
}
|
284 |
+
elif self.config.name in ['lsat_lr', 'lsat_rc', 'lsat_ar']:
|
285 |
+
yield key, {
|
286 |
+
"question": data["question"],
|
287 |
+
"options": data["options"],
|
288 |
+
"label": data["label"],
|
289 |
+
}
|
290 |
|