Nathan Fradet
commited on
ruff formatting + changing gradio app loading
Browse files
app.py
CHANGED
|
@@ -1,6 +1,17 @@
|
|
|
|
|
|
|
|
| 1 |
import evaluate
|
| 2 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
|
| 4 |
|
| 5 |
-
|
| 6 |
-
|
|
|
|
| 1 |
+
"""Application file."""
|
| 2 |
+
|
| 3 |
import evaluate
|
| 4 |
+
import gradio as gr
|
| 5 |
+
|
| 6 |
+
"""module = evaluate.load("Natooz/ece")
|
| 7 |
+
gradio_app = gr.Interface(
|
| 8 |
+
module,
|
| 9 |
+
inputs=gr.component(),
|
| 10 |
+
outputs=[gr.Image(label="Processed Image"), gr.Label(label="Result", num_top_classes=2)],
|
| 11 |
+
title=module.name,
|
| 12 |
+
)"""
|
| 13 |
+
gradio_app = gr.load("Natooz/ece", src="spaces")
|
| 14 |
|
| 15 |
|
| 16 |
+
if __name__ == "__main__":
|
| 17 |
+
gradio_app.launch()
|
ece.py
CHANGED
|
@@ -1,27 +1,19 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
|
| 6 |
-
#
|
| 7 |
-
# http://www.apache.org/licenses/LICENSE-2.0
|
| 8 |
-
#
|
| 9 |
-
# Unless required by applicable law or agreed to in writing, software
|
| 10 |
-
# distributed under the License is distributed on an "AS IS" BASIS,
|
| 11 |
-
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
| 12 |
-
# See the License for the specific language governing permissions and
|
| 13 |
-
# limitations under the License.
|
| 14 |
-
|
| 15 |
-
from typing import Dict
|
| 16 |
|
| 17 |
-
import evaluate
|
| 18 |
import datasets
|
| 19 |
-
|
|
|
|
| 20 |
from torchmetrics.functional.classification.calibration_error import (
|
| 21 |
binary_calibration_error,
|
| 22 |
multiclass_calibration_error,
|
| 23 |
)
|
| 24 |
|
|
|
|
|
|
|
| 25 |
|
| 26 |
_CITATION = """\
|
| 27 |
@InProceedings{huggingface:ece,
|
|
@@ -41,7 +33,8 @@ https://torchmetrics.readthedocs.io/en/stable/classification/calibration_error.h
|
|
| 41 |
_KWARGS_DESCRIPTION = """
|
| 42 |
Calculates how good are predictions given some references, using certain scores
|
| 43 |
Args:
|
| 44 |
-
predictions: list of predictions to score. They must have a shape (N,C,...) if
|
|
|
|
| 45 |
references: list of reference for each prediction, with a shape (N,...).
|
| 46 |
Returns:
|
| 47 |
ece: expected calibration error
|
|
@@ -65,11 +58,17 @@ Examples:
|
|
| 65 |
@evaluate.utils.file_utils.add_start_docstrings(_DESCRIPTION, _KWARGS_DESCRIPTION)
|
| 66 |
class ECE(evaluate.Metric):
|
| 67 |
"""
|
| 68 |
-
|
| 69 |
-
|
|
|
|
| 70 |
"""
|
| 71 |
|
| 72 |
-
def _info(self):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 73 |
return evaluate.MetricInfo(
|
| 74 |
# This is the description that will appear on the modules page.
|
| 75 |
module_type="metric",
|
|
@@ -94,31 +93,43 @@ class ECE(evaluate.Metric):
|
|
| 94 |
],
|
| 95 |
)
|
| 96 |
|
| 97 |
-
def _compute(
|
| 98 |
-
|
| 99 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 100 |
https://torchmetrics.readthedocs.io/en/stable/classification/calibration_error.html
|
| 101 |
predictions: (N,C,...) if multiclass or (N,...) if binary
|
| 102 |
-
references: (N,...)
|
| 103 |
|
| 104 |
-
If "num_classes" is not provided in a
|
| 105 |
-
be used as "num_classes".
|
| 106 |
"""
|
| 107 |
# Convert the input
|
| 108 |
predictions = Tensor(predictions)
|
| 109 |
references = LongTensor(references)
|
| 110 |
|
| 111 |
# Determine number of classes / binary or multiclass
|
| 112 |
-
error_msg =
|
| 113 |
-
|
|
|
|
|
|
|
|
|
|
| 114 |
binary = True
|
| 115 |
if predictions.dim() == references.dim() + 1: # multiclass
|
| 116 |
binary = False
|
| 117 |
if "num_classes" not in kwargs:
|
| 118 |
kwargs["num_classes"] = int(predictions.shape[1])
|
| 119 |
elif predictions.dim() == references.dim() and "num_classes" in kwargs:
|
| 120 |
-
raise ValueError(
|
| 121 |
-
|
|
|
|
|
|
|
| 122 |
elif predictions.dim() != references.dim():
|
| 123 |
raise ValueError("Bad input shape. " + error_msg)
|
| 124 |
|
|
|
|
| 1 |
+
"""ECE metric file."""
|
| 2 |
+
|
| 3 |
+
from __future__ import annotations
|
| 4 |
+
|
| 5 |
+
from typing import TYPE_CHECKING
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
|
|
|
|
| 7 |
import datasets
|
| 8 |
+
import evaluate
|
| 9 |
+
from torch import LongTensor, Tensor
|
| 10 |
from torchmetrics.functional.classification.calibration_error import (
|
| 11 |
binary_calibration_error,
|
| 12 |
multiclass_calibration_error,
|
| 13 |
)
|
| 14 |
|
| 15 |
+
if TYPE_CHECKING:
|
| 16 |
+
from collections.abc import Iterable
|
| 17 |
|
| 18 |
_CITATION = """\
|
| 19 |
@InProceedings{huggingface:ece,
|
|
|
|
| 33 |
_KWARGS_DESCRIPTION = """
|
| 34 |
Calculates how good are predictions given some references, using certain scores
|
| 35 |
Args:
|
| 36 |
+
predictions: list of predictions to score. They must have a shape (N,C,...) if
|
| 37 |
+
multiclass, or (N,...) if binary.
|
| 38 |
references: list of reference for each prediction, with a shape (N,...).
|
| 39 |
Returns:
|
| 40 |
ece: expected calibration error
|
|
|
|
| 58 |
@evaluate.utils.file_utils.add_start_docstrings(_DESCRIPTION, _KWARGS_DESCRIPTION)
|
| 59 |
class ECE(evaluate.Metric):
|
| 60 |
"""
|
| 61 |
+
Module for the BinaryCalibrationError (ECE) metric of the torchmetrics package.
|
| 62 |
+
|
| 63 |
+
https://torchmetrics.readthedocs.io/en/stable/classification/calibration_error.html.
|
| 64 |
"""
|
| 65 |
|
| 66 |
+
def _info(self) -> evaluate.MetricInfo:
|
| 67 |
+
"""
|
| 68 |
+
Return the module info.
|
| 69 |
+
|
| 70 |
+
:return: module info.
|
| 71 |
+
"""
|
| 72 |
return evaluate.MetricInfo(
|
| 73 |
# This is the description that will appear on the modules page.
|
| 74 |
module_type="metric",
|
|
|
|
| 93 |
],
|
| 94 |
)
|
| 95 |
|
| 96 |
+
def _compute(
|
| 97 |
+
self,
|
| 98 |
+
predictions: Iterable[float] | None = None,
|
| 99 |
+
references: Iterable[int] | None = None,
|
| 100 |
+
**kwargs
|
| 101 |
+
) -> dict[str, float]:
|
| 102 |
+
"""
|
| 103 |
+
Return the Expected Calibration Error (ECE).
|
| 104 |
+
|
| 105 |
+
See the torchmetrics documentation for more information on the method.
|
| 106 |
https://torchmetrics.readthedocs.io/en/stable/classification/calibration_error.html
|
| 107 |
predictions: (N,C,...) if multiclass or (N,...) if binary
|
| 108 |
+
references: (N,...).
|
| 109 |
|
| 110 |
+
If "num_classes" is not provided in a multiclass setting, the number maximum
|
| 111 |
+
label index will be used as "num_classes".
|
| 112 |
"""
|
| 113 |
# Convert the input
|
| 114 |
predictions = Tensor(predictions)
|
| 115 |
references = LongTensor(references)
|
| 116 |
|
| 117 |
# Determine number of classes / binary or multiclass
|
| 118 |
+
error_msg = (
|
| 119 |
+
"Expected to have predictions with shape (N,C,...) for multiclass or "
|
| 120 |
+
"(N,...) for binary, and references with shape (N,...), but got "
|
| 121 |
+
f"{predictions.shape} and {references.shape}"
|
| 122 |
+
)
|
| 123 |
binary = True
|
| 124 |
if predictions.dim() == references.dim() + 1: # multiclass
|
| 125 |
binary = False
|
| 126 |
if "num_classes" not in kwargs:
|
| 127 |
kwargs["num_classes"] = int(predictions.shape[1])
|
| 128 |
elif predictions.dim() == references.dim() and "num_classes" in kwargs:
|
| 129 |
+
raise ValueError(
|
| 130 |
+
"You gave the num_classes argument, with predictions and references "
|
| 131 |
+
"having the same number of dimensions. " + error_msg
|
| 132 |
+
)
|
| 133 |
elif predictions.dim() != references.dim():
|
| 134 |
raise ValueError("Bad input shape. " + error_msg)
|
| 135 |
|
tests.py
CHANGED
|
@@ -1,3 +1,5 @@
|
|
|
|
|
|
|
|
| 1 |
test_cases = [
|
| 2 |
{
|
| 3 |
"predictions": [0, 0],
|
|
@@ -14,4 +16,4 @@ test_cases = [
|
|
| 14 |
"references": [1, 1],
|
| 15 |
"result": {"metric_score": 0.5}
|
| 16 |
}
|
| 17 |
-
]
|
|
|
|
| 1 |
+
"""Test cases."""
|
| 2 |
+
|
| 3 |
test_cases = [
|
| 4 |
{
|
| 5 |
"predictions": [0, 0],
|
|
|
|
| 16 |
"references": [1, 1],
|
| 17 |
"result": {"metric_score": 0.5}
|
| 18 |
}
|
| 19 |
+
]
|