README.md

Contrastive Learning Model

This model is a contrastive learning model based on distilbert-base-uncased. It outputs the [CLS] token embedding for similarity comparisons.

Usage

from transformers import AutoTokenizer
from modeling import ContrastiveModel  # Import the model class directly
import torch

# Load model and tokenizer
tokenizer = AutoTokenizer.from_pretrained("SajayR/contrastive-model")
config = AutoConfig.from_pretrained("SajayR/contrastive-model")
model = ContrastiveModel(config)
model.load_state_dict(torch.load("pytorch_model.bin"))

# Prepare input
text = "Hello, world!"
inputs = tokenizer(text, return_tensors="pt")

# Get embeddings (shape: [batch_size, hidden_size])
embeddings = model(**inputs)

The model returns a tensor of shape [batch_size, hidden_size] containing the [CLS] token embedding.