Upload metrics.py with huggingface_hub
Browse files- metrics.py +92 -2
metrics.py
CHANGED
@@ -1,6 +1,7 @@
|
|
1 |
import uuid
|
2 |
from abc import ABC, abstractmethod
|
3 |
-
from
|
|
|
4 |
from typing import Any, Dict, Generator, List, Optional
|
5 |
|
6 |
import evaluate
|
@@ -9,7 +10,6 @@ import numpy
|
|
9 |
|
10 |
from .operator import (
|
11 |
MultiStreamOperator,
|
12 |
-
SequntialOperator,
|
13 |
SingleStreamOperator,
|
14 |
StreamingOperator,
|
15 |
StreamInstanceOperator,
|
@@ -353,3 +353,93 @@ class Bleu(HuggingfaceMetric):
|
|
353 |
metric_name = "bleu"
|
354 |
main_score = "bleu"
|
355 |
scale = 1.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
import uuid
|
2 |
from abc import ABC, abstractmethod
|
3 |
+
from collections import Counter
|
4 |
+
from dataclasses import field
|
5 |
from typing import Any, Dict, Generator, List, Optional
|
6 |
|
7 |
import evaluate
|
|
|
10 |
|
11 |
from .operator import (
|
12 |
MultiStreamOperator,
|
|
|
13 |
SingleStreamOperator,
|
14 |
StreamingOperator,
|
15 |
StreamInstanceOperator,
|
|
|
353 |
metric_name = "bleu"
|
354 |
main_score = "bleu"
|
355 |
scale = 1.0
|
356 |
+
|
357 |
+
|
358 |
+
class CustomF1(GlobalMetric):
|
359 |
+
main_score = "f1_micro"
|
360 |
+
|
361 |
+
@abstractmethod
|
362 |
+
def get_element_group(self, element):
|
363 |
+
pass
|
364 |
+
|
365 |
+
@abstractmethod
|
366 |
+
def get_element_representation(self, element):
|
367 |
+
pass
|
368 |
+
|
369 |
+
def group_elements(self, l):
|
370 |
+
return {
|
371 |
+
k: Counter([self.get_element_representation(value) for value in l if self.get_element_group(value) == k])
|
372 |
+
for k in set([self.get_element_group(e) for e in l])
|
373 |
+
}
|
374 |
+
|
375 |
+
def calculate_groups_ratio(self, actual_group, total_group):
|
376 |
+
return sum([min(actual_group[k], total_group[k]) for k in actual_group.keys()]), sum(actual_group.values())
|
377 |
+
|
378 |
+
def f1(self, pn, pd, rn, rd):
|
379 |
+
precision = 1.0 if pn == 0 and pd == 0 else pn / pd
|
380 |
+
recall = 1.0 if rn == 0 and rd == 0 else rn / rd
|
381 |
+
try:
|
382 |
+
return 2 * precision * recall / (precision + recall)
|
383 |
+
except ZeroDivisionError:
|
384 |
+
return 0.0
|
385 |
+
|
386 |
+
def compute(self, references: List[Any], predictions: List[Any]) -> dict:
|
387 |
+
# in case reference are List[List[List[Any]]] and predictions are List[List[Any]]:
|
388 |
+
if isinstance(references[0], list) and isinstance(references[0][0], list):
|
389 |
+
references = [element[0] for element in references]
|
390 |
+
|
391 |
+
assert len(references) == len(predictions), (
|
392 |
+
f"references size ({len(references)})" f" doesn't mach predictions sise ({len(references)})."
|
393 |
+
)
|
394 |
+
groups_statistics = dict()
|
395 |
+
for references_batch, predictions_batch in zip(references, predictions):
|
396 |
+
grouped_references = self.group_elements(references_batch)
|
397 |
+
grouped_predictions = self.group_elements(predictions_batch)
|
398 |
+
all_groups = set(grouped_references.keys()).union(grouped_predictions.keys())
|
399 |
+
for group in all_groups:
|
400 |
+
if group not in groups_statistics:
|
401 |
+
groups_statistics[group] = {
|
402 |
+
"precision_numerator": 0,
|
403 |
+
"precision_denominator": 0,
|
404 |
+
"recall_numerator": 0,
|
405 |
+
"recall_denominator": 0,
|
406 |
+
}
|
407 |
+
references_by_group = grouped_references.get(group, Counter([]))
|
408 |
+
predictions_by_group = grouped_predictions.get(group, Counter([]))
|
409 |
+
pn, pd = self.calculate_groups_ratio(
|
410 |
+
actual_group=predictions_by_group, total_group=references_by_group
|
411 |
+
)
|
412 |
+
rn, rd = self.calculate_groups_ratio(
|
413 |
+
actual_group=references_by_group, total_group=predictions_by_group
|
414 |
+
)
|
415 |
+
groups_statistics[group]["precision_numerator"] += pn
|
416 |
+
groups_statistics[group]["precision_denominator"] += pd
|
417 |
+
groups_statistics[group]["recall_numerator"] += rn
|
418 |
+
groups_statistics[group]["recall_denominator"] += rd
|
419 |
+
|
420 |
+
result = {}
|
421 |
+
pn_total = pd_total = rn_total = rd_total = 0
|
422 |
+
for group in groups_statistics.keys():
|
423 |
+
pn, pd, rn, rd = (
|
424 |
+
groups_statistics[group]["precision_numerator"],
|
425 |
+
groups_statistics[group]["precision_denominator"],
|
426 |
+
groups_statistics[group]["recall_numerator"],
|
427 |
+
groups_statistics[group]["recall_denominator"],
|
428 |
+
)
|
429 |
+
result[f"f1_{group}"] = self.f1(pn, pd, rn, rd)
|
430 |
+
pn_total, pd_total, rn_total, rd_total = pn_total + pn, pd_total + pd, rn_total + rn, rd_total + rd
|
431 |
+
try:
|
432 |
+
result["f1_macro"] = sum(result.values()) / len(result.keys())
|
433 |
+
except ZeroDivisionError:
|
434 |
+
result["f1_macro"] = 1.0
|
435 |
+
|
436 |
+
result[f"f1_micro"] = self.f1(pn_total, pd_total, rn_total, rd_total)
|
437 |
+
return result
|
438 |
+
|
439 |
+
|
440 |
+
class NER(CustomF1):
|
441 |
+
def get_element_group(self, element):
|
442 |
+
return element[1]
|
443 |
+
|
444 |
+
def get_element_representation(self, element):
|
445 |
+
return str(element)
|