Update agieval.py
Browse files- agieval.py +14 -4
agieval.py
CHANGED
@@ -46,6 +46,9 @@ _URLS = {
|
|
46 |
"sat_en": {
|
47 |
"test": HEAD+'sat-en.jsonl',
|
48 |
},
|
|
|
|
|
|
|
49 |
"sat_math": {
|
50 |
"test": HEAD+'sat-math.jsonl'
|
51 |
},
|
@@ -99,6 +102,11 @@ class AgiEval(datasets.GeneratorBasedBuilder):
|
|
99 |
version=VERSION,
|
100 |
description="",
|
101 |
),
|
|
|
|
|
|
|
|
|
|
|
102 |
datasets.BuilderConfig(
|
103 |
name="lsat_ar",
|
104 |
version=VERSION,
|
@@ -152,7 +160,7 @@ class AgiEval(datasets.GeneratorBasedBuilder):
|
|
152 |
"solution": datasets.Value("string"),
|
153 |
}
|
154 |
)
|
155 |
-
elif self.config.name
|
156 |
# remove solution from other
|
157 |
features = datasets.Features(
|
158 |
{"question": datasets.Value("string"),
|
@@ -161,6 +169,7 @@ class AgiEval(datasets.GeneratorBasedBuilder):
|
|
161 |
"solution": datasets.Value("string"),
|
162 |
}
|
163 |
)
|
|
|
164 |
elif self.config.name == "logiqa_en":
|
165 |
# remove solution from other
|
166 |
features = datasets.Features(
|
@@ -232,7 +241,8 @@ class AgiEval(datasets.GeneratorBasedBuilder):
|
|
232 |
# Mapping for column names in CSV to dataset names
|
233 |
names = {'aqua_rat': 'aqua-rat', 'sat_en': 'sat-en', 'sat_math': 'sat-math',
|
234 |
'lsat_ar': 'lsat-ar', 'lsat_lr': 'lsat-lr', 'lsat_rc': 'lsat-rc',
|
235 |
-
'logiqa_en': 'logiqa_en', 'logiqa_zh': '
|
|
|
236 |
|
237 |
if split == "few_shot":
|
238 |
# Load the data from the CSV
|
@@ -246,7 +256,7 @@ class AgiEval(datasets.GeneratorBasedBuilder):
|
|
246 |
try:
|
247 |
data = ast.literal_eval(samples[names[self.config.name]][key])
|
248 |
explanation_row = explanations[names[self.config.name]][key]
|
249 |
-
if self.config.name in ["aqua_rat", "sat_math"]:
|
250 |
yield key, {
|
251 |
"question": data["question"],
|
252 |
"options": data["options"],
|
@@ -304,7 +314,7 @@ class AgiEval(datasets.GeneratorBasedBuilder):
|
|
304 |
for key, row in enumerate(f):
|
305 |
data = json.loads(row)
|
306 |
|
307 |
-
if self.config.name in ["aqua_rat","sat_math"]:
|
308 |
yield key, {
|
309 |
"question": data["question"],
|
310 |
"options": data["options"],
|
|
|
46 |
"sat_en": {
|
47 |
"test": HEAD+'sat-en.jsonl',
|
48 |
},
|
49 |
+
"sat_en_without_passage": {
|
50 |
+
"test": HEAD+'sat-en-without-passage.jsonl',
|
51 |
+
},
|
52 |
"sat_math": {
|
53 |
"test": HEAD+'sat-math.jsonl'
|
54 |
},
|
|
|
102 |
version=VERSION,
|
103 |
description="",
|
104 |
),
|
105 |
+
datasets.BuilderConfig(
|
106 |
+
name="sat_en_without_passage",
|
107 |
+
version=VERSION,
|
108 |
+
description="",
|
109 |
+
),
|
110 |
datasets.BuilderConfig(
|
111 |
name="lsat_ar",
|
112 |
version=VERSION,
|
|
|
160 |
"solution": datasets.Value("string"),
|
161 |
}
|
162 |
)
|
163 |
+
elif self.config.name in ["sat_math", "sat_en_without_passage"]:
|
164 |
# remove solution from other
|
165 |
features = datasets.Features(
|
166 |
{"question": datasets.Value("string"),
|
|
|
169 |
"solution": datasets.Value("string"),
|
170 |
}
|
171 |
)
|
172 |
+
|
173 |
elif self.config.name == "logiqa_en":
|
174 |
# remove solution from other
|
175 |
features = datasets.Features(
|
|
|
241 |
# Mapping for column names in CSV to dataset names
|
242 |
names = {'aqua_rat': 'aqua-rat', 'sat_en': 'sat-en', 'sat_math': 'sat-math',
|
243 |
'lsat_ar': 'lsat-ar', 'lsat_lr': 'lsat-lr', 'lsat_rc': 'lsat-rc',
|
244 |
+
'logiqa_en': 'logiqa_en', 'logiqa_zh': 'logiqa-zh', 'math': 'math',
|
245 |
+
'sat_en_without_passage': 'sat-en-without-passage'}
|
246 |
|
247 |
if split == "few_shot":
|
248 |
# Load the data from the CSV
|
|
|
256 |
try:
|
257 |
data = ast.literal_eval(samples[names[self.config.name]][key])
|
258 |
explanation_row = explanations[names[self.config.name]][key]
|
259 |
+
if self.config.name in ["aqua_rat", "sat_math", "sat_en_without_passage"]:
|
260 |
yield key, {
|
261 |
"question": data["question"],
|
262 |
"options": data["options"],
|
|
|
314 |
for key, row in enumerate(f):
|
315 |
data = json.loads(row)
|
316 |
|
317 |
+
if self.config.name in ["aqua_rat","sat_math", "sat_en_without_passage"]:
|
318 |
yield key, {
|
319 |
"question": data["question"],
|
320 |
"options": data["options"],
|