Spaces:
Runtime error
Runtime error
Update Space (evaluate main: e4a27243)
Browse files- requirements.txt +1 -1
- sacrebleu.py +29 -13
requirements.txt
CHANGED
@@ -1,2 +1,2 @@
|
|
1 |
-
git+https://github.com/huggingface/evaluate@
|
2 |
sacrebleu
|
|
|
1 |
+
git+https://github.com/huggingface/evaluate@e4a2724377909fe2aeb4357e3971e5a569673b39
|
2 |
sacrebleu
|
sacrebleu.py
CHANGED
@@ -13,6 +13,9 @@
|
|
13 |
# limitations under the License.
|
14 |
""" SACREBLEU metric. """
|
15 |
|
|
|
|
|
|
|
16 |
import datasets
|
17 |
import sacrebleu as scb
|
18 |
from packaging import version
|
@@ -102,9 +105,27 @@ Examples:
|
|
102 |
"""
|
103 |
|
104 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
105 |
@evaluate.utils.file_utils.add_start_docstrings(_DESCRIPTION, _KWARGS_DESCRIPTION)
|
106 |
class Sacrebleu(evaluate.Metric):
|
107 |
-
|
|
|
|
|
|
|
|
|
108 |
if version.parse(scb.__version__) < version.parse("1.4.12"):
|
109 |
raise ImportWarning(
|
110 |
"To use `sacrebleu`, the module `sacrebleu>=1.4.12` is required, and the current version of `sacrebleu` doesn't match this condition.\n"
|
@@ -115,6 +136,7 @@ class Sacrebleu(evaluate.Metric):
|
|
115 |
citation=_CITATION,
|
116 |
homepage="https://github.com/mjpost/sacreBLEU",
|
117 |
inputs_description=_KWARGS_DESCRIPTION,
|
|
|
118 |
features=[
|
119 |
datasets.Features(
|
120 |
{
|
@@ -141,12 +163,6 @@ class Sacrebleu(evaluate.Metric):
|
|
141 |
self,
|
142 |
predictions,
|
143 |
references,
|
144 |
-
smooth_method="exp",
|
145 |
-
smooth_value=None,
|
146 |
-
force=False,
|
147 |
-
lowercase=False,
|
148 |
-
tokenize=None,
|
149 |
-
use_effective_order=False,
|
150 |
):
|
151 |
# if only one reference is provided make sure we still use list of lists
|
152 |
if isinstance(references[0], str):
|
@@ -159,12 +175,12 @@ class Sacrebleu(evaluate.Metric):
|
|
159 |
output = scb.corpus_bleu(
|
160 |
predictions,
|
161 |
transformed_references,
|
162 |
-
smooth_method=smooth_method,
|
163 |
-
smooth_value=smooth_value,
|
164 |
-
force=force,
|
165 |
-
lowercase=lowercase,
|
166 |
-
use_effective_order=use_effective_order,
|
167 |
-
**(dict(tokenize=tokenize) if tokenize else {}),
|
168 |
)
|
169 |
output_dict = {
|
170 |
"score": output.score,
|
|
|
13 |
# limitations under the License.
|
14 |
""" SACREBLEU metric. """
|
15 |
|
16 |
+
from dataclasses import dataclass
|
17 |
+
from typing import Callable, Optional
|
18 |
+
|
19 |
import datasets
|
20 |
import sacrebleu as scb
|
21 |
from packaging import version
|
|
|
105 |
"""
|
106 |
|
107 |
|
108 |
+
@dataclass
|
109 |
+
class SacrebleuConfig(evaluate.info.Config):
|
110 |
+
|
111 |
+
name: str = "default"
|
112 |
+
|
113 |
+
smooth_method: str = "exp"
|
114 |
+
average: str = "binary"
|
115 |
+
smooth_value: Optional[float] = None
|
116 |
+
force: bool = False
|
117 |
+
lowercase: bool = False
|
118 |
+
tokenize: Optional[Callable] = None
|
119 |
+
use_effective_order: bool = False
|
120 |
+
|
121 |
+
|
122 |
@evaluate.utils.file_utils.add_start_docstrings(_DESCRIPTION, _KWARGS_DESCRIPTION)
|
123 |
class Sacrebleu(evaluate.Metric):
|
124 |
+
|
125 |
+
CONFIG_CLASS = SacrebleuConfig
|
126 |
+
ALLOWED_CONFIG_NAMES = ["default"]
|
127 |
+
|
128 |
+
def _info(self, config):
|
129 |
if version.parse(scb.__version__) < version.parse("1.4.12"):
|
130 |
raise ImportWarning(
|
131 |
"To use `sacrebleu`, the module `sacrebleu>=1.4.12` is required, and the current version of `sacrebleu` doesn't match this condition.\n"
|
|
|
136 |
citation=_CITATION,
|
137 |
homepage="https://github.com/mjpost/sacreBLEU",
|
138 |
inputs_description=_KWARGS_DESCRIPTION,
|
139 |
+
config=config,
|
140 |
features=[
|
141 |
datasets.Features(
|
142 |
{
|
|
|
163 |
self,
|
164 |
predictions,
|
165 |
references,
|
|
|
|
|
|
|
|
|
|
|
|
|
166 |
):
|
167 |
# if only one reference is provided make sure we still use list of lists
|
168 |
if isinstance(references[0], str):
|
|
|
175 |
output = scb.corpus_bleu(
|
176 |
predictions,
|
177 |
transformed_references,
|
178 |
+
smooth_method=self.config.smooth_method,
|
179 |
+
smooth_value=self.config.smooth_value,
|
180 |
+
force=self.config.force,
|
181 |
+
lowercase=self.config.lowercase,
|
182 |
+
use_effective_order=self.config.use_effective_order,
|
183 |
+
**(dict(tokenize=self.config.tokenize) if self.config.tokenize else {}),
|
184 |
)
|
185 |
output_dict = {
|
186 |
"score": output.score,
|