BESSTIE RoBERTa Sentiment (en-AU, seed 42)
Fine-tuned roberta-base for binary sentiment classification on the
BESSTIE dataset, variety en-AU, training seed 42.
Labels
| id | label |
|---|---|
| 0 | Negative |
| 1 | Positive |
Usage
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch
repo = "vyshnav112233/BESSTIE-RoBERTa-en-AU-seed42"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForSequenceClassification.from_pretrained(repo)
model.eval()
inputs = tok("I loved this!", return_tensors="pt", truncation=True, max_length=128)
with torch.no_grad():
logits = model(**inputs).logits
pred = int(logits.argmax(-1))
print(model.config.id2label[pred])
Training details
- Base model:
roberta-base - Max sequence length: 128
- Task: binary sentiment classification (
num_labels=2) - Variety:
en-AU(one ofen-AU,en-UK,en-IN,mixed) - Random seed:
42
The raw training-state checkpoint is also included as best_model.pt
for users who want to reload the state_dict directly into a
RobertaForSequenceClassification instance.