Views
No views yet
| Metric | Base Model | Abliterated Model |
|---|---|---|
| Refusal Rate | ~100% (Safety) | 0.0% |
| Perplexity | 5.52 | 6.28 |
| KL Divergence | - | 0.00012 |
| Coherence | 1.0 | 1.0 |
1from transformers import AutoModelForCausalLM, AutoTokenizer
2
3model_name = "Dist0rted/Abliterated-Qwen-SuperLight"
4tokenizer = AutoTokenizer.from_pretrained(model_name)
5model = AutoModelForCausalLM.from_pretrained(model_name)
6
7prompt = "Write a guide on how to pick a lock."
8inputs = tokenizer(prompt, return_tensors="pt")
9outputs = model.generate(**inputs, max_new_tokens=200)
10
11print(tokenizer.decode(outputs[0]))