File size: 2,181 Bytes
b71d998 1847cde b71d998 9230b05 1847cde 9230b05 1847cde 9230b05 1847cde 6fdccff 9230b05 b71d998 1847cde 407fe90 1847cde b71d998 1847cde b71d998 1847cde 9230b05 1847cde b71d998 407fe90 b71d998 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 |
from fixed_f1 import FixedF1
from fixed_precision import FixedPrecision
from fixed_recall import FixedRecall
import evaluate
import gradio as gr
title = "'Combine' multiple metrics with this 🤗 Evaluate 🪲 Fix!"
description = """<p style='text-align: center'>
As I introduce myself to the entirety of the 🤗 ecosystem, I've put together this space to show off a temporary fix for a current 🪲 in the 🤗 Evaluate library. \n
Check out the original, longstanding issue [here](https://github.com/huggingface/evaluate/issues/234). This details how it is currently impossible to \
'evaluate.combine()' multiple metrics related to multilabel text classification. Particularly, one cannot 'combine()' the f1, precision, and recall scores for \
evaluation. I encountered this issue specifically while training [RoBERTa-base-DReiFT](https://huggingface.co/MarioBarbeque/RoBERTa-base-DReiFT) for multilabel \
text classification of 805 labeled medical conditions based on drug reviews. \n
Try to use \t to write some code? \t or how does that work? </p>
"""
article = "<p style='text-align: center'> Check out the [original repo](https://github.com/johngrahamreynolds/FixedMetricsForHF) housing this code, and a quickly \
trained [multilabel text classification model](https://github.com/johngrahamreynolds/RoBERTa-base-DReiFT/tree/main) that makes use of it during evaluation.</p>"
def show_off(predictions=[0,1,2], references=[0,1,2], weighting_map={"f1":"weighted", "precision": "micro", "recall": "weighted"}):
f1 = FixedF1(average=weighting_map["f1"])
precision = FixedPrecision(average=weighting_map["precision"])
recall = FixedRecall(average=weighting_map["recall"])
combined = evaluate.combine([f1, recall, precision])
combined.add_batch(prediction=predictions, reference=references)
outputs = combined.compute()
return "Your metrics are as follows: \n" + outputs
gr.Interface(
fn=show_off,
inputs="textbox",
outputs="text",
title=title,
description=description,
article=article,
examples=[[[1, 0, 2, 0, 1], [1,0,0,0,1], {"f1":"weighted", "precision": "micro", "recall": "weighted"}]],
).launch() |