nhonhoccode/qwen3-0-6b-cybersecqa-sft-freeze2-20251110-1040

qwen3-0-6b — Cybersecurity QA (SFT)

Fine-tuned on Kaggle using SFT.

Model Summary

• Base: unsloth/Qwen3-0.6B
• Trainable params: 187,044,352 / total 596,049,920
• Train wall time (s): 457.4
• Files: pytorch_model.safetensors + config.json + tokenizer files

Data

• Dataset: zobayer0x01/cybersecurity-qa
• Samples: total=42427, train=38184, val=1000
• Prompting: Chat template with a fixed system prompt:

You are a helpful assistant specialized in cybersecurity Q&A.

Training Config

Field	Value
Method	SFT
Precision	fp32
Quantization	none
Mode	steps
Num Epochs	1
Max Steps	10
Eval Steps	5
Save Steps	400
LR	5e-05
Max Length	768
per_device_batch_size	1
grad_accum	8

Evaluation (greedy, fixed-length decode)

Metric	Score
BLEU-4	0.88
ROUGE-L	11.49
F1 (token-level)	24.90
chrF++	22.79
BERTScore F1	81.33
Perplexity	16.86

Notes: We normalize whitespace/punctuations, compute token-level P/R/F1, and use evaluate's sacrebleu/rouge/chrf/bertscore.

How to use

from transformers import AutoTokenizer, AutoModelForCausalLM
tok = AutoTokenizer.from_pretrained("nhonhoccode/qwen3-0-6b-cybersecqa-sft-freeze2-20251110-1040")
mdl = AutoModelForCausalLM.from_pretrained("nhonhoccode/qwen3-0-6b-cybersecqa-sft-freeze2-20251110-1040")
prompt = tok.apply_chat_template(
    [{"role":"system","content":"You are a helpful assistant specialized in cybersecurity Q&A."},
     {"role":"user","content":"Explain SQL injection in one paragraph."}],
    tokenize=False, add_generation_prompt=True
)
ids = tok(prompt, return_tensors="pt").input_ids
out = mdl.generate(ids, max_new_tokens=128, do_sample=False)
print(tok.decode(out[0][ids.shape[-1]:], skip_special_tokens=True))

Intended Use & Limitations

• Domain: cybersecurity Q&A; not guaranteed to be accurate for legal/medical purposes.
• The model can hallucinate or produce outdated guidance—verify before applying in production.
• Safety: No explicit content filtering. Add guardrails (moderation, retrieval augmentation) for deployment.

Reproducibility (env)

• transformers>=4.43,<5, accelerate>=0.33,<0.34, peft>=0.11,<0.13, datasets>=2.18,<3, evaluate>=0.4,<0.5, rouge-score, sacrebleu, huggingface_hub>=0.23,<0.26, bitsandbytes
• GPU: T4-class; LoRA recommended for low VRAM.

Changelog

• 2025-11-10 10:51 — Initial release (SFT)