Elad
commited on
Commit
·
646a69c
1
Parent(s):
ced174a
remove unused metrics
Browse files- metrics/bart_score.py +0 -53
- metrics/bleu.py +0 -47
metrics/bart_score.py
DELETED
@@ -1,53 +0,0 @@
|
|
1 |
-
import os
|
2 |
-
import numpy as np
|
3 |
-
from BARTScore.bart_score import BARTScorer
|
4 |
-
|
5 |
-
|
6 |
-
def get_scorers():
|
7 |
-
assert os.path.isfile(
|
8 |
-
os.path.join("BARTScore", "bart.pth")
|
9 |
-
), "You must download `bart.pth` to use BARTScore.\nUse `gdown --id 1_7JfF7KOInb7ZrxKHIigTMR4ChVET01m --output bart.pth`"
|
10 |
-
|
11 |
-
scorers = {}
|
12 |
-
|
13 |
-
scorers["vanilla"] = BARTScorer(device="cuda:0", checkpoint="facebook/bart-large")
|
14 |
-
|
15 |
-
scorers["cnn"] = BARTScorer(device="cuda:0", checkpoint="facebook/bart-large-cnn")
|
16 |
-
|
17 |
-
# for the parabank model, first init a bart model, then load the local para model from BARTScore/bart.pth
|
18 |
-
# see the documentation from https://github.com/neulab/BARTScore for reference
|
19 |
-
scorers["para"] = BARTScorer(device="cuda:0", checkpoint="facebook/bart-large-cnn")
|
20 |
-
scorers["para"].load(path="BARTScore/bart.pth")
|
21 |
-
|
22 |
-
return scorers
|
23 |
-
|
24 |
-
|
25 |
-
def compute_bart_score_for_scorer(predictions, references, scorer_name, scorer):
|
26 |
-
#precisions = np.array(scorer.score(references, predictions, batch_size=4))
|
27 |
-
recalls = np.array(scorer.score(predictions, references, batch_size=4))
|
28 |
-
#f_scores = 0.5 * (precisions + recalls)
|
29 |
-
baselines = np.array(scorer.score(references, references, batch_size=4))
|
30 |
-
normalized = baselines / recalls
|
31 |
-
diffs = recalls - baselines
|
32 |
-
expdiffs = np.exp(diffs)
|
33 |
-
|
34 |
-
return [
|
35 |
-
{
|
36 |
-
#f"{scorer_name}_f_score": f_scores[i],
|
37 |
-
#f"{scorer_name}_precision": precisions[i],
|
38 |
-
f"{scorer_name}_recall": recalls[i],
|
39 |
-
f"{scorer_name}_normalized": normalized[i],
|
40 |
-
f"{scorer_name}_diffs": diffs[i],
|
41 |
-
f"{scorer_name}_expdiffs": expdiffs[i],
|
42 |
-
}
|
43 |
-
for i in range(len(predictions))
|
44 |
-
]
|
45 |
-
|
46 |
-
|
47 |
-
def compute_bart_score(predictions, references, scorers):
|
48 |
-
result = [{} for _ in range(len(predictions))]
|
49 |
-
for scorer_name, scorer in scorers.items():
|
50 |
-
scorer_result = compute_bart_score_for_scorer(predictions, references, scorer_name, scorer)
|
51 |
-
for i, element in enumerate(scorer_result):
|
52 |
-
result[i].update(element)
|
53 |
-
return result
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
metrics/bleu.py
DELETED
@@ -1,47 +0,0 @@
|
|
1 |
-
# Copied from https://github.com/huggingface/datasets/blob/76bb45964df1e62d1411b0a9e9fc673e9a791c9a/metrics/sacrebleu/sacrebleu.py
|
2 |
-
|
3 |
-
from copy import deepcopy
|
4 |
-
from sacrebleu.metrics import BLEU
|
5 |
-
|
6 |
-
|
7 |
-
def compute_bleu(
|
8 |
-
predictions,
|
9 |
-
references,
|
10 |
-
smooth_method="exp",
|
11 |
-
smooth_value=None,
|
12 |
-
force=False,
|
13 |
-
lowercase=False,
|
14 |
-
tokenize=None,
|
15 |
-
effective_order=False,
|
16 |
-
):
|
17 |
-
references_per_prediction = len(references[0])
|
18 |
-
if any(len(refs) != references_per_prediction for refs in references):
|
19 |
-
references = deepcopy(references)
|
20 |
-
max_references_per_prediction = max(len(refs) for refs in references)
|
21 |
-
for refs in references:
|
22 |
-
refs.extend([None] * (max_references_per_prediction - len(refs)))
|
23 |
-
|
24 |
-
transformed_references = [[refs[i] for refs in references] for i in range(references_per_prediction)]
|
25 |
-
|
26 |
-
bleu = BLEU(
|
27 |
-
smooth_method=smooth_method,
|
28 |
-
smooth_value=smooth_value,
|
29 |
-
force=force,
|
30 |
-
lowercase=lowercase,
|
31 |
-
effective_order=effective_order,
|
32 |
-
**(dict(tokenize=tokenize) if tokenize else {}),
|
33 |
-
)
|
34 |
-
output = bleu.corpus_score(
|
35 |
-
predictions,
|
36 |
-
transformed_references,
|
37 |
-
)
|
38 |
-
output_dict = {
|
39 |
-
"score": output.score,
|
40 |
-
**{f"counts-{i+1}": round(p, 4) for i, p in enumerate(output.counts)},
|
41 |
-
**{f"totals-{i+1}": round(p, 4) for i, p in enumerate(output.totals)},
|
42 |
-
**{f"precision-{i+1}": round(p, 4) for i, p in enumerate(output.precisions)},
|
43 |
-
"bp": output.bp,
|
44 |
-
"sys_len": output.sys_len,
|
45 |
-
"ref_len": output.ref_len,
|
46 |
-
}
|
47 |
-
return output_dict
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|