Upload metric.py with huggingface_hub
Browse files
metric.py
CHANGED
@@ -1,16 +1,13 @@
|
|
1 |
-
from
|
2 |
-
from typing import Any, Dict, Generator, Iterable, List, Optional, Union
|
3 |
|
4 |
-
import datasets
|
5 |
import evaluate
|
6 |
-
from datasets import Features,
|
7 |
|
8 |
from .artifact import __file__ as _
|
9 |
from .blocks import __file__ as _
|
10 |
from .card import __file__ as _
|
11 |
from .catalog import __file__ as _
|
12 |
from .collections import __file__ as _
|
13 |
-
from .common import __file__ as _
|
14 |
from .dataclass import __file__ as _
|
15 |
from .dict_utils import __file__ as _
|
16 |
from .file_utils import __file__ as _
|
@@ -23,20 +20,12 @@ from .load import __file__ as _
|
|
23 |
from .loaders import __file__ as _
|
24 |
from .metrics import __file__ as _
|
25 |
from .normalizers import __file__ as _
|
26 |
-
from .operator import (
|
27 |
-
|
28 |
-
SequntialOperator,
|
29 |
-
SequntialOperatorInitilizer,
|
30 |
-
StreamInitializerOperator,
|
31 |
-
)
|
32 |
from .operator import __file__ as _
|
33 |
-
from .operators import (
|
34 |
-
|
35 |
-
|
36 |
-
FlattenInstances,
|
37 |
-
MergeStreams,
|
38 |
-
SplitByValue,
|
39 |
-
)
|
40 |
from .operators import __file__ as _
|
41 |
from .processors import __file__ as _
|
42 |
from .random_utils import __file__ as _
|
@@ -44,6 +33,7 @@ from .recipe import __file__ as _
|
|
44 |
from .register import __file__ as _
|
45 |
from .register import _reset_env_local_catalogs, register_all_artifacts
|
46 |
from .renderers import __file__ as _
|
|
|
47 |
from .schema import __file__ as _
|
48 |
from .split_utils import __file__ as _
|
49 |
from .splitters import __file__ as _
|
@@ -75,12 +65,31 @@ class MultiStreamScoreMean(MultiStreamOperator):
|
|
75 |
instance["score"]["global"]["groups_mean_score"] = score
|
76 |
yield instance
|
77 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
78 |
def process(self, multi_stream: MultiStream) -> MultiStream:
|
79 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
80 |
|
|
|
81 |
result = {}
|
82 |
for stream_name, stream in multi_stream.items():
|
83 |
-
result[stream_name] = Stream(
|
|
|
|
|
84 |
|
85 |
return MultiStream(result)
|
86 |
|
@@ -90,20 +99,41 @@ class FromPredictionsAndOriginalData(StreamInitializerOperator):
|
|
90 |
for prediction, original in zip(predictions, references):
|
91 |
yield {**original, "prediction": prediction}
|
92 |
|
93 |
-
def process(
|
|
|
|
|
94 |
return MultiStream(
|
95 |
-
{
|
|
|
|
|
|
|
|
|
|
|
96 |
)
|
97 |
|
98 |
|
99 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
100 |
|
101 |
|
102 |
-
class MetricRecipe(
|
|
|
|
|
103 |
def prepare(self):
|
104 |
register_all_artifacts()
|
105 |
self.steps = [
|
106 |
FromPredictionsAndOriginalData(),
|
|
|
|
|
|
|
|
|
|
|
107 |
ApplyOperatorsField(
|
108 |
inputs_fields=["prediction", "references"],
|
109 |
fields_to_treat_as_list=["references"],
|
@@ -111,37 +141,48 @@ class MetricRecipe(SequntialOperatorInitilizer):
|
|
111 |
default_operators=["processors.to_string_stripped"],
|
112 |
),
|
113 |
SplitByValue(["group"]),
|
114 |
-
|
115 |
"metrics",
|
116 |
-
|
117 |
),
|
118 |
MultiStreamScoreMean(),
|
119 |
MergeStreams(),
|
120 |
]
|
121 |
|
122 |
|
123 |
-
UNITXT_METRIC_SCHEMA = Features(
|
|
|
|
|
124 |
|
125 |
|
126 |
-
def _compute(
|
|
|
|
|
|
|
|
|
|
|
|
|
127 |
_reset_env_local_catalogs()
|
128 |
register_all_artifacts()
|
129 |
-
recipe = MetricRecipe()
|
130 |
|
131 |
-
multi_stream = recipe(
|
|
|
|
|
132 |
|
133 |
if flatten:
|
134 |
operator = FlattenInstances()
|
135 |
multi_stream = operator(multi_stream)
|
136 |
|
137 |
stream = multi_stream[split_name]
|
138 |
-
|
139 |
return list(stream)
|
140 |
|
141 |
|
142 |
# TODO: currently we have two classes with this name. metric.Metric and matrics.Metric...
|
143 |
# @evaluate.utils.file_utils.add_start_docstrings(_DESCRIPTION, _KWARGS_DESCRIPTION)
|
144 |
class Metric(evaluate.Metric):
|
|
|
|
|
145 |
def _info(self):
|
146 |
return evaluate.MetricInfo(
|
147 |
description="_DESCRIPTION",
|
@@ -155,11 +196,16 @@ class Metric(evaluate.Metric):
|
|
155 |
],
|
156 |
)
|
157 |
|
158 |
-
def _compute(
|
|
|
|
|
|
|
|
|
|
|
|
|
159 |
try:
|
160 |
-
from unitxt.dataset import
|
161 |
-
get_dataset_artifact as get_dataset_artifact_installed
|
162 |
-
)
|
163 |
|
164 |
unitxt_installed = True
|
165 |
except ImportError:
|
@@ -169,7 +215,17 @@ class Metric(evaluate.Metric):
|
|
169 |
from unitxt.metric import _compute as _compute_installed
|
170 |
|
171 |
return _compute_installed(
|
172 |
-
predictions=predictions,
|
|
|
|
|
|
|
|
|
173 |
)
|
174 |
-
|
175 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from typing import Dict, Iterable, List
|
|
|
2 |
|
|
|
3 |
import evaluate
|
4 |
+
from datasets import Features, Value
|
5 |
|
6 |
from .artifact import __file__ as _
|
7 |
from .blocks import __file__ as _
|
8 |
from .card import __file__ as _
|
9 |
from .catalog import __file__ as _
|
10 |
from .collections import __file__ as _
|
|
|
11 |
from .dataclass import __file__ as _
|
12 |
from .dict_utils import __file__ as _
|
13 |
from .file_utils import __file__ as _
|
|
|
20 |
from .loaders import __file__ as _
|
21 |
from .metrics import __file__ as _
|
22 |
from .normalizers import __file__ as _
|
23 |
+
from .operator import (MultiStreamOperator, SequentialOperator,
|
24 |
+
SequentialOperatorInitilizer, StreamInitializerOperator)
|
|
|
|
|
|
|
|
|
25 |
from .operator import __file__ as _
|
26 |
+
from .operators import (Apply, ApplyMetric, ApplyOperatorsField,
|
27 |
+
ApplyStreamOperatorsField, FlattenInstances,
|
28 |
+
MergeStreams, SplitByValue)
|
|
|
|
|
|
|
|
|
29 |
from .operators import __file__ as _
|
30 |
from .processors import __file__ as _
|
31 |
from .random_utils import __file__ as _
|
|
|
33 |
from .register import __file__ as _
|
34 |
from .register import _reset_env_local_catalogs, register_all_artifacts
|
35 |
from .renderers import __file__ as _
|
36 |
+
from .schema import UNITXT_DATASET_SCHEMA
|
37 |
from .schema import __file__ as _
|
38 |
from .split_utils import __file__ as _
|
39 |
from .splitters import __file__ as _
|
|
|
65 |
instance["score"]["global"]["groups_mean_score"] = score
|
66 |
yield instance
|
67 |
|
68 |
+
def spread_results_one_stream(self, stream: Stream):
|
69 |
+
for instance in stream:
|
70 |
+
instance["score"]["global"]["groups_mean_score"] = instance["score"][
|
71 |
+
"global"
|
72 |
+
]["score"]
|
73 |
+
yield instance
|
74 |
+
|
75 |
def process(self, multi_stream: MultiStream) -> MultiStream:
|
76 |
+
result = {}
|
77 |
+
|
78 |
+
# optimization in to avoid double calculation of metrics
|
79 |
+
# when aggregating results, if there is only one stream.
|
80 |
+
if len(multi_stream) == 1:
|
81 |
+
for stream_name, stream in multi_stream.items():
|
82 |
+
result[stream_name] = Stream(
|
83 |
+
self.spread_results_one_stream, gen_kwargs={"stream": stream}
|
84 |
+
)
|
85 |
+
return MultiStream(result)
|
86 |
|
87 |
+
mean_score = self.aggegate_results(multi_stream)
|
88 |
result = {}
|
89 |
for stream_name, stream in multi_stream.items():
|
90 |
+
result[stream_name] = Stream(
|
91 |
+
self.spread_results, gen_kwargs={"stream": stream, "score": mean_score}
|
92 |
+
)
|
93 |
|
94 |
return MultiStream(result)
|
95 |
|
|
|
99 |
for prediction, original in zip(predictions, references):
|
100 |
yield {**original, "prediction": prediction}
|
101 |
|
102 |
+
def process(
|
103 |
+
self, predictions: List[str], references: Iterable, split_name: str = "all"
|
104 |
+
) -> MultiStream:
|
105 |
return MultiStream(
|
106 |
+
{
|
107 |
+
split_name: Stream(
|
108 |
+
self.zip,
|
109 |
+
gen_kwargs={"predictions": predictions, "references": references},
|
110 |
+
)
|
111 |
+
}
|
112 |
)
|
113 |
|
114 |
|
115 |
+
# The additional_inputs field in the schema is defined as
|
116 |
+
# Sequence({"key": Value(dtype="string"), "value": Value("string")})
|
117 |
+
# When receiving instances from this scheme, the keys and values are returned as two separate
|
118 |
+
# lists, and are converted to a dictionary.
|
119 |
+
|
120 |
+
|
121 |
+
def _from_key_value_pairs(key_value_list: Dict[str, list]) -> Dict[str, str]:
|
122 |
+
return dict(zip(key_value_list["key"], key_value_list["value"]))
|
123 |
|
124 |
|
125 |
+
class MetricRecipe(SequentialOperatorInitilizer):
|
126 |
+
calc_confidence_intervals: bool = True
|
127 |
+
|
128 |
def prepare(self):
|
129 |
register_all_artifacts()
|
130 |
self.steps = [
|
131 |
FromPredictionsAndOriginalData(),
|
132 |
+
Apply(
|
133 |
+
"additional_inputs",
|
134 |
+
function=_from_key_value_pairs,
|
135 |
+
to_field="additional_inputs",
|
136 |
+
),
|
137 |
ApplyOperatorsField(
|
138 |
inputs_fields=["prediction", "references"],
|
139 |
fields_to_treat_as_list=["references"],
|
|
|
141 |
default_operators=["processors.to_string_stripped"],
|
142 |
),
|
143 |
SplitByValue(["group"]),
|
144 |
+
ApplyMetric(
|
145 |
"metrics",
|
146 |
+
calc_confidence_intervals=self.calc_confidence_intervals,
|
147 |
),
|
148 |
MultiStreamScoreMean(),
|
149 |
MergeStreams(),
|
150 |
]
|
151 |
|
152 |
|
153 |
+
UNITXT_METRIC_SCHEMA = Features(
|
154 |
+
{"predictions": Value("string"), "references": dict(UNITXT_DATASET_SCHEMA)}
|
155 |
+
)
|
156 |
|
157 |
|
158 |
+
def _compute(
|
159 |
+
predictions: List[str],
|
160 |
+
references: Iterable,
|
161 |
+
flatten: bool = False,
|
162 |
+
split_name: str = "all",
|
163 |
+
calc_confidence_intervals: bool = True,
|
164 |
+
):
|
165 |
_reset_env_local_catalogs()
|
166 |
register_all_artifacts()
|
167 |
+
recipe = MetricRecipe(calc_confidence_intervals=calc_confidence_intervals)
|
168 |
|
169 |
+
multi_stream = recipe(
|
170 |
+
predictions=predictions, references=references, split_name=split_name
|
171 |
+
)
|
172 |
|
173 |
if flatten:
|
174 |
operator = FlattenInstances()
|
175 |
multi_stream = operator(multi_stream)
|
176 |
|
177 |
stream = multi_stream[split_name]
|
|
|
178 |
return list(stream)
|
179 |
|
180 |
|
181 |
# TODO: currently we have two classes with this name. metric.Metric and matrics.Metric...
|
182 |
# @evaluate.utils.file_utils.add_start_docstrings(_DESCRIPTION, _KWARGS_DESCRIPTION)
|
183 |
class Metric(evaluate.Metric):
|
184 |
+
calc_confidence_intervals: bool = True
|
185 |
+
|
186 |
def _info(self):
|
187 |
return evaluate.MetricInfo(
|
188 |
description="_DESCRIPTION",
|
|
|
196 |
],
|
197 |
)
|
198 |
|
199 |
+
def _compute(
|
200 |
+
self,
|
201 |
+
predictions: List[str],
|
202 |
+
references: Iterable,
|
203 |
+
flatten: bool = False,
|
204 |
+
split_name: str = "all",
|
205 |
+
):
|
206 |
try:
|
207 |
+
from unitxt.dataset import \
|
208 |
+
get_dataset_artifact as get_dataset_artifact_installed
|
|
|
209 |
|
210 |
unitxt_installed = True
|
211 |
except ImportError:
|
|
|
215 |
from unitxt.metric import _compute as _compute_installed
|
216 |
|
217 |
return _compute_installed(
|
218 |
+
predictions=predictions,
|
219 |
+
references=references,
|
220 |
+
flatten=flatten,
|
221 |
+
split_name=split_name,
|
222 |
+
calc_confidence_intervals=self.calc_confidence_intervals,
|
223 |
)
|
224 |
+
|
225 |
+
return _compute(
|
226 |
+
predictions=predictions,
|
227 |
+
references=references,
|
228 |
+
flatten=flatten,
|
229 |
+
split_name=split_name,
|
230 |
+
calc_confidence_intervals=self.calc_confidence_intervals,
|
231 |
+
)
|