Modelpublic

SimoneAstarita/trilingual-no-bio-20251012-t10

sourceHugging Faceapache-2.0updated 1y agoView on Hugging Face

0likes5downloads

Model Card

october-finetuning-more-variables-sweep-20251012-193706-t02

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

Trial timestamp (UTC): 2025-10-12 19:37:06 Data case: en-es-it

Configuration (trial hyperparameters)

Model: Alibaba-NLP/gte-multilingual-base

Hyperparameter	Value
LANGUAGES	en-es-it
LR	1e-05
EPOCHS	5
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	0.7265709173014043
f1weighteddev_0.5	0.8606644314617744
accuracydev0.5	0.8552338530066815
f1macrodevbestglobal	0.7265709173014043
f1weighteddevbestglobal	0.8606644314617744
accuracydevbest_global	0.8552338530066815
f1macrodevbestby_lang	0.6952488687782805
f1weighteddevbestby_lang	0.8306311662921121
accuracydevbestbylang	0.8129175946547884
default_threshold	0.5
bestthresholdglobal	0.5
thresholdsbylang	{"en": 0.4, "it": 0.5, "es": 0.45000000000000007}

Thresholds

—Default: 0.5
—Best global: 0.5
—Best by language: { "en": 0.4, "it": 0.5, "es": 0.45000000000000007 }

Detailed evaluation

Classification report @ 0.5

text

              precision    recall  f1-score   support

 no-recl (0)     0.9301    0.8987    0.9141       385
    recl (1)     0.4935    0.5938    0.5390        64

    accuracy                         0.8552       449
   macro avg     0.7118    0.7462    0.7266       449
weighted avg     0.8679    0.8552    0.8607       449

Classification report @ best global threshold (t=0.50)

text

              precision    recall  f1-score   support

 no-recl (0)     0.9301    0.8987    0.9141       385
    recl (1)     0.4935    0.5938    0.5390        64

    accuracy                         0.8552       449
   macro avg     0.7118    0.7462    0.7266       449
weighted avg     0.8679    0.8552    0.8607       449

Classification report @ best per-language thresholds

text

              precision    recall  f1-score   support

 no-recl (0)     0.9388    0.8364    0.8846       385
    recl (1)     0.4057    0.6719    0.5059        64

    accuracy                         0.8129       449
   macro avg     0.6722    0.7541    0.6952       449
weighted avg     0.8628    0.8129    0.8306       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	0.7532	0.4962	0.7953	0.5077	0.5161	0.8493	0.7532
it	163	0.8773	0.8181	0.8824	0.7963	0.8502	0.8925	0.8773
es	132	0.8030	0.7054	0.8236	0.6823	0.7812	0.8674	0.8030

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-20251012-193706-t02"
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.