Views
No views yet
| Metric | Score |
|---|---|
| Accuracy | 96.2% |
| F1 | 96.2% |
| Precision | 96.2% |
| Recall | 96.2% |
| Training Time | 151 seconds (MI300X GPU) |
1from peft import PeftModel
2from transformers import AutoModelForSequenceClassification, AutoTokenizer
3
4# Load model
5base_model = AutoModelForSequenceClassification.from_pretrained(
6 "jhu-clsp/mmBERT-base", num_labels=2
7)
8model = PeftModel.from_pretrained(base_model, "llm-semantic-router/mmbert-fact-check-lora")
9tokenizer = AutoTokenizer.from_pretrained("jhu-clsp/mmBERT-base")
10
11# Classify
12queries = [
13 "When was the Eiffel Tower built?", # FACT_CHECK_NEEDED
14 "Write a poem about the ocean", # NO_FACT_CHECK_NEEDED
15]
16
17for query in queries:
18 inputs = tokenizer(query, return_tensors="pt", truncation=True)
19 outputs = model(**inputs)
20 label = "FACT_CHECK_NEEDED" if outputs.logits.argmax(-1).item() == 1 else "NO_FACT_CHECK_NEEDED"
21 print(f"{query} -> {label}")