Test version : ⚠️ Warning: The model may occasionally generate hallucinations or inaccurate information. Please verify the responses ⚠️
Experimental modified / abliterated versions of lightweight language models.
The models were subjected to an experimental modification process, subsequently tested and saved for research purposes.
This repository contains modified model artifacts derived from the original releases. The original models, authors and licenses remain attributed to their respective creators.
This work is intended for research and experimentation only. Not for production use.
- Training method: fine-tuning LoRA merged
- Base model:
3MPER0RR/TinyLlama-3MPER0RR-abliterated - Input format:
[ {"role": "user", "content": "your question"} ]
Training Datasets
sahilempire/redsec-security-sft-v1
Dataset license: Apache-2.0oyildirim/cyberstrike-sft-120k
Dataset license: Apache-2.0
How to Use
The model expects conversations in the standard messages format. At inference, provide the system and user messages; the model will generate the assistant response.
from transformers import AutoTokenizer, AutoModelForCausalLM
model_id = "3MPER0RR/TinyLlama-3MPER0RR-4Bit-NF4-Training-Cyber-v1"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")
messages = [
{"role": "system", "content": "You are a helpful cybersecurity assistant."},
{"role": "user", "content": "What is cybersecurity?"}
]
inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt").to(model.device)
outputs = model.generate(inputs["input_ids"], max_new_tokens=200)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:], skip_special_tokens=True))
---