mdeberta-v3-base-indonesian-sarcasm

This model is a fine-tuned version of microsoft/mdeberta-v3-base trained to detect sarcasm in Indonesian text.

It was fine-tuned on the clean Reddit Indonesian Sarcasm dataset (jwnt4/reddit-id-sarcasm-cleaned).

Model Details

  • Model Type: Text Classification (Sarcasm Detection)
  • Language: Indonesian
  • Base Model: microsoft/mdeberta-v3-base
  • Dataset: jwnt4/reddit-id-sarcasm-cleaned
  • Label Mapping:
    • 0: Non-Sarcastic
    • 1: Sarcastic

Training Hyperparameters

  • Epochs: 5
  • Batch Size: 8 (with Gradient Accumulation Steps = 2)
  • Learning Rate: 2e-5
  • Weight Decay: 0.01
  • Max Sequence Length: 256
  • Optimizer: AdamW

How to Use

You can load and use this model with Hugging Face's transformers library:

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch
import torch.nn.functional as F

# Load model and tokenizer
model_name = "jonathanjordan21/mdeberta-v3-base-indonesian-sarcasm"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

# Prepare text
text = "Wah, rajin sekali kamu, jam 12 siang baru bangun."
inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=256)

# Predict
with torch.no_grad():
    outputs = model(**inputs)
    probabilities = F.softmax(outputs.logits, dim=-1)
    prediction = torch.argmax(probabilities, dim=-1).item()

labels = {0: "Non-Sarcastic", 1: "Sarcastic"}
print(f"Prediction: {labels[prediction]} ({probabilities[0][prediction].item():.2%})")
Downloads last month
50
Safetensors
Model size
0.3B params
Tensor type
F32
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for jonathanjordan21/mdeberta-v3-base-indonesian-sarcasm

Finetunes
1 model

Dataset used to train jonathanjordan21/mdeberta-v3-base-indonesian-sarcasm

Space using jonathanjordan21/mdeberta-v3-base-indonesian-sarcasm 1