Modelpublic

SimoneAstarita/trilingual-no-bio-20251012-t09

sourceHugging Faceapache-2.0updated 1y agoView on Hugging Face

0likes4downloads

Model Card

october-finetuning-more-variables-sweep-20251012-192648-t01

Slur reclamation binary classifier Task: LGBTQ+ reclamation vs non-reclamation use of harmful words on social media text.

Trial timestamp (UTC): 2025-10-12 19:26:48 Data case: en-es-it

Configuration (trial hyperparameters)

Model: Alibaba-NLP/gte-multilingual-base

Hyperparameter	Value
LANGUAGES	en-es-it
LR	2e-05
EPOCHS	3
MAX_LENGTH	256
USE_BIO	False
USELANGTOKEN	False
GATED_BIO	False
FOCAL_LOSS	True
FOCAL_GAMMA	2.5
USE_SAMPLER	False
R_DROP	True
RKLALPHA	0.5
TEXT_NORMALIZE	True

Dev set results (summary)

Metric	Value
f1macrodev_0.5	0.7122879758763663
f1weighteddev_0.5	0.8538441584389483
accuracydev0.5	0.8485523385300668
f1macrodevbestglobal	0.7173094314136246
f1weighteddevbestglobal	0.8694483657963139
accuracydevbest_global	0.8775055679287305
f1macrodevbestby_lang	0.680589430894309
f1weighteddevbestby_lang	0.8275698480815543
accuracydevbestbylang	0.8129175946547884
default_threshold	0.5
bestthresholdglobal	0.55
thresholdsbylang	{"en": 0.4, "it": 0.55, "es": 0.5}

Thresholds

—Default: 0.5
—Best global: 0.55
—Best by language: { "en": 0.4, "it": 0.55, "es": 0.5 }

Detailed evaluation

Classification report @ 0.5

text

              precision    recall  f1-score   support

 no-recl (0)     0.9249    0.8961    0.9103       385
    recl (1)     0.4737    0.5625    0.5143        64

    accuracy                         0.8486       449
   macro avg     0.6993    0.7293    0.7123       449
weighted avg     0.8606    0.8486    0.8538       449

Classification report @ best global threshold (t=0.55)

text

              precision    recall  f1-score   support

 no-recl (0)     0.9104    0.9506    0.9301       385
    recl (1)     0.5957    0.4375    0.5045        64

    accuracy                         0.8775       449
   macro avg     0.7531    0.6941    0.7173       449
weighted avg     0.8656    0.8775    0.8694       449

Classification report @ best per-language thresholds

text

              precision    recall  f1-score   support

 no-recl (0)     0.9263    0.8494    0.8862       385
    recl (1)     0.3958    0.5938    0.4750        64

    accuracy                         0.8129       449
   macro avg     0.6611    0.7216    0.6806       449
weighted avg     0.8507    0.8129    0.8276       449

Per-language metrics (at best-by-lang)

lang	n	acc	f1_macro	f1_weighted	prec_macro	rec_macro	prec_weighted	rec_weighted
en	154	0.7468	0.5104	0.7931	0.5213	0.5475	0.8568	0.7468
it	163	0.8896	0.8112	0.8866	0.8299	0.7961	0.8852	0.8896
es	132	0.7955	0.6900	0.8160	0.6691	0.7562	0.8566	0.7955

Data

—Train/Dev: private multilingual splits with ~15% stratified Dev (by (lang,label)).
—Source: merged EN/IT/ES data with bios retained (ignored if unused by model).

Usage

python

from transformers import AutoTokenizer, AutoModelForSequenceClassification, AutoConfig
import torch, numpy as np

repo = "SimoneAstarita/october-finetuning-more-variables-sweep-20251012-192648-t01"
tok = AutoTokenizer.from_pretrained(repo)
cfg = AutoConfig.from_pretrained(repo)
model = AutoModelForSequenceClassification.from_pretrained(repo)

texts = ["example text ..."]
langs = ["en"]

mode = "best_global"  # or "0.5", "by_lang"

enc = tok(texts, truncation=True, padding=True, max_length=256, return_tensors="pt")
with torch.no_grad():
    logits = model(**enc).logits
probs = torch.softmax(logits, dim=-1)[:, 1].cpu().numpy()

if mode == "0.5":
    th = 0.5
    preds = (probs >= th).astype(int)
elif mode == "best_global":
    th = getattr(cfg, "best_threshold_global", 0.5)
    preds = (probs >= th).astype(int)
elif mode == "by_lang":
    th_by_lang = getattr(cfg, "thresholds_by_lang", {})
    preds = np.zeros_like(probs, dtype=int)
    for lg in np.unique(langs):
        t = th_by_lang.get(lg, getattr(cfg, "best_threshold_global", 0.5))
        preds[np.array(langs) == lg] = (probs[np.array(langs) == lg] >= t).astype(int)
print(list(zip(texts, preds, probs)))

Additional files

reports.json: all metrics (macro/weighted/accuracy) for @0.5, @bestglobal, and @bestbylang. config.json: stores thresholds: defaultthreshold, bestthresholdglobal, thresholdsbylang. postprocessing.json: duplicate threshold info for external tools.