Spaces:
Sleeping
Sleeping
John Graham Reynolds
commited on
Commit
·
19d5834
1
Parent(s):
edd642a
add code reminder for fixing widget fn
Browse files- issue_fix.py +72 -0
issue_fix.py
ADDED
@@ -0,0 +1,72 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
"""
|
2 |
+
Come back to this file and previous versions in this repo to configure a fix for the broken launch_gradio_widget command
|
3 |
+
"""
|
4 |
+
|
5 |
+
import sys
|
6 |
+
import gradio as gr
|
7 |
+
import pandas as pd
|
8 |
+
import evaluate
|
9 |
+
from evaluate.utils import infer_gradio_input_types, json_to_string_type, parse_readme, parse_test_cases
|
10 |
+
# from evaluate.utils import launch_gradio_widget # using this directly is erroneous - lets fix this
|
11 |
+
from fixed_f1 import FixedF1
|
12 |
+
from pathlib import Path
|
13 |
+
|
14 |
+
metric = FixedF1()
|
15 |
+
|
16 |
+
if isinstance(metric.features, list):
|
17 |
+
(feature_names, feature_types) = zip(*metric.features[0].items())
|
18 |
+
else:
|
19 |
+
(feature_names, feature_types) = zip(*metric.features.items())
|
20 |
+
gradio_input_types = infer_gradio_input_types(feature_types)
|
21 |
+
|
22 |
+
local_path = Path(sys.path[0])
|
23 |
+
test_cases = [ {"predictions":[1,2,3,4,5], "references":[1,2,5,4,3]} ] # configure this randomly using randint generator and feature names?
|
24 |
+
|
25 |
+
# configure this based on the input type, etc. for launch_gradio_widget
|
26 |
+
def compute(input_df: pd.DataFrame, method: str):
|
27 |
+
|
28 |
+
metric = FixedF1(average=method)
|
29 |
+
|
30 |
+
cols = [col for col in input_df.columns]
|
31 |
+
|
32 |
+
predicted = [int(num) for num in input_df[cols[0]].to_list()]
|
33 |
+
references = [int(num) for num in input_df[cols[1]].to_list()]
|
34 |
+
|
35 |
+
metric.add_batch(predictions=predicted, references=references)
|
36 |
+
|
37 |
+
outputs = metric.compute()
|
38 |
+
|
39 |
+
f"Your metrics are as follows: \n {outputs}"
|
40 |
+
|
41 |
+
space = gr.Interface(
|
42 |
+
fn=compute,
|
43 |
+
inputs=[
|
44 |
+
gr.Dataframe(
|
45 |
+
headers=feature_names,
|
46 |
+
col_count=len(feature_names),
|
47 |
+
row_count=5,
|
48 |
+
datatype=json_to_string_type(gradio_input_types),
|
49 |
+
),
|
50 |
+
gr.Radio(
|
51 |
+
["weighted", "micro", "macro", "binary", "None"],
|
52 |
+
label="Averaging Method",
|
53 |
+
info="Method for averaging the F1 score across labels."
|
54 |
+
)
|
55 |
+
],
|
56 |
+
outputs=gr.Textbox(label=metric.name),
|
57 |
+
description=(
|
58 |
+
metric.info.description + "\nIf this is a text-based metric, make sure to wrap your input in double quotes."
|
59 |
+
" Alternatively you can use a JSON-formatted list as input."
|
60 |
+
),
|
61 |
+
title=f"Metric: {metric.name}",
|
62 |
+
article=parse_readme(local_path / "README.md"),
|
63 |
+
# TODO: load test cases and use them to populate examples
|
64 |
+
examples=[
|
65 |
+
# correct depth?
|
66 |
+
pd.DataFrame(parse_test_cases(test_cases, feature_names, gradio_input_types)[0]),
|
67 |
+
pd.DataFrame(columns=["Metric", "Averaging Method"], data=[["f1", "weighted"]])
|
68 |
+
],
|
69 |
+
cache_examples=False
|
70 |
+
)
|
71 |
+
|
72 |
+
space.launch()
|