Spaces:
Runtime error
Runtime error
Update Space (evaluate main: c447fc8e)
Browse files- regard.py +6 -19
- requirements.txt +1 -1
regard.py
CHANGED
@@ -15,10 +15,8 @@
|
|
15 |
""" Regard measurement. """
|
16 |
|
17 |
from collections import defaultdict
|
18 |
-
from dataclasses import dataclass
|
19 |
from operator import itemgetter
|
20 |
from statistics import mean
|
21 |
-
from typing import Optional
|
22 |
|
23 |
import datasets
|
24 |
from transformers import AutoModelForSequenceClassification, AutoTokenizer, pipeline
|
@@ -117,20 +115,9 @@ def regard(group, regard_classifier):
|
|
117 |
return group_regard, dict(group_scores)
|
118 |
|
119 |
|
120 |
-
@dataclass
|
121 |
-
class RegardConfig(evaluate.info.Config):
|
122 |
-
|
123 |
-
name: str = "default"
|
124 |
-
|
125 |
-
aggregation: Optional[str] = None
|
126 |
-
|
127 |
-
|
128 |
@evaluate.utils.file_utils.add_start_docstrings(_DESCRIPTION, _KWARGS_DESCRIPTION)
|
129 |
class Regard(evaluate.Measurement):
|
130 |
-
|
131 |
-
ALLOWED_CONFIG_NAMES = ["default", "compare"]
|
132 |
-
|
133 |
-
def _info(self, config):
|
134 |
if self.config_name not in ["compare", "default"]:
|
135 |
raise KeyError("You should supply a configuration name selected in " '["config", "default"]')
|
136 |
return evaluate.MeasurementInfo(
|
@@ -138,7 +125,6 @@ class Regard(evaluate.Measurement):
|
|
138 |
description=_DESCRIPTION,
|
139 |
citation=_CITATION,
|
140 |
inputs_description=_KWARGS_DESCRIPTION,
|
141 |
-
config=config,
|
142 |
features=datasets.Features(
|
143 |
{
|
144 |
"data": datasets.Value("string", id="sequence"),
|
@@ -164,6 +150,7 @@ class Regard(evaluate.Measurement):
|
|
164 |
self,
|
165 |
data,
|
166 |
references=None,
|
|
|
167 |
):
|
168 |
if self.config_name == "compare":
|
169 |
pred_scores, pred_regard = regard(data, self.regard_classifier)
|
@@ -172,12 +159,12 @@ class Regard(evaluate.Measurement):
|
|
172 |
pred_max = {k: max(v) for k, v in pred_regard.items()}
|
173 |
ref_mean = {k: mean(v) for k, v in ref_regard.items()}
|
174 |
ref_max = {k: max(v) for k, v in ref_regard.items()}
|
175 |
-
if
|
176 |
return {
|
177 |
"max_data_regard": pred_max,
|
178 |
"max_references_regard": ref_max,
|
179 |
}
|
180 |
-
elif
|
181 |
return {"average_data_regard": pred_mean, "average_references_regard": ref_mean}
|
182 |
else:
|
183 |
return {"regard_difference": {key: pred_mean[key] - ref_mean.get(key, 0) for key in pred_mean}}
|
@@ -185,9 +172,9 @@ class Regard(evaluate.Measurement):
|
|
185 |
pred_scores, pred_regard = regard(data, self.regard_classifier)
|
186 |
pred_mean = {k: mean(v) for k, v in pred_regard.items()}
|
187 |
pred_max = {k: max(v) for k, v in pred_regard.items()}
|
188 |
-
if
|
189 |
return {"max_regard": pred_max}
|
190 |
-
elif
|
191 |
return {"average_regard": pred_mean}
|
192 |
else:
|
193 |
return {"regard": pred_scores}
|
|
|
15 |
""" Regard measurement. """
|
16 |
|
17 |
from collections import defaultdict
|
|
|
18 |
from operator import itemgetter
|
19 |
from statistics import mean
|
|
|
20 |
|
21 |
import datasets
|
22 |
from transformers import AutoModelForSequenceClassification, AutoTokenizer, pipeline
|
|
|
115 |
return group_regard, dict(group_scores)
|
116 |
|
117 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
118 |
@evaluate.utils.file_utils.add_start_docstrings(_DESCRIPTION, _KWARGS_DESCRIPTION)
|
119 |
class Regard(evaluate.Measurement):
|
120 |
+
def _info(self):
|
|
|
|
|
|
|
121 |
if self.config_name not in ["compare", "default"]:
|
122 |
raise KeyError("You should supply a configuration name selected in " '["config", "default"]')
|
123 |
return evaluate.MeasurementInfo(
|
|
|
125 |
description=_DESCRIPTION,
|
126 |
citation=_CITATION,
|
127 |
inputs_description=_KWARGS_DESCRIPTION,
|
|
|
128 |
features=datasets.Features(
|
129 |
{
|
130 |
"data": datasets.Value("string", id="sequence"),
|
|
|
150 |
self,
|
151 |
data,
|
152 |
references=None,
|
153 |
+
aggregation=None,
|
154 |
):
|
155 |
if self.config_name == "compare":
|
156 |
pred_scores, pred_regard = regard(data, self.regard_classifier)
|
|
|
159 |
pred_max = {k: max(v) for k, v in pred_regard.items()}
|
160 |
ref_mean = {k: mean(v) for k, v in ref_regard.items()}
|
161 |
ref_max = {k: max(v) for k, v in ref_regard.items()}
|
162 |
+
if aggregation == "maximum":
|
163 |
return {
|
164 |
"max_data_regard": pred_max,
|
165 |
"max_references_regard": ref_max,
|
166 |
}
|
167 |
+
elif aggregation == "average":
|
168 |
return {"average_data_regard": pred_mean, "average_references_regard": ref_mean}
|
169 |
else:
|
170 |
return {"regard_difference": {key: pred_mean[key] - ref_mean.get(key, 0) for key in pred_mean}}
|
|
|
172 |
pred_scores, pred_regard = regard(data, self.regard_classifier)
|
173 |
pred_mean = {k: mean(v) for k, v in pred_regard.items()}
|
174 |
pred_max = {k: max(v) for k, v in pred_regard.items()}
|
175 |
+
if aggregation == "maximum":
|
176 |
return {"max_regard": pred_max}
|
177 |
+
elif aggregation == "average":
|
178 |
return {"average_regard": pred_mean}
|
179 |
else:
|
180 |
return {"regard": pred_scores}
|
requirements.txt
CHANGED
@@ -1,2 +1,2 @@
|
|
1 |
-
git+https://github.com/huggingface/evaluate.git@
|
2 |
transformers
|
|
|
1 |
+
git+https://github.com/huggingface/evaluate.git@c447fc8eda9c62af501bfdc6988919571050d950
|
2 |
transformers
|