from transformers import AutoModelForSequenceClassification, AutoTokenizer
import torch

model = AutoModelForSequenceClassification.from_pretrained(
    "Sachinkry/qwen3-imdb-reward-0.6b",
    trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("Sachinkry/qwen3-imdb-reward-0.6b")

# Single inference
prompt = "What did you think about the movie?"
response = "Brilliant film, excellent acting!"
inputs = tokenizer(f"{prompt} {response}", return_tensors="pt")
reward = model(**inputs).logits.squeeze().sigmoid().item()
print(f"Reward: {reward:.3f}")

📊 Example Outputs

python

"What did you think about the movie? Brilliant film!" → 0.638 ✓
"What did you think about the movie? Total waste..." → 0.529 ✗
Gap: 0.109 ✓