BodyCam-VQA/qwen3-4b-sft-qa

Model Card for Model ID

Pre-trained adapters for question generation on police body-worn camera footage, designed to work with Qwen3-4B-Thinking-2507. Trained using standard fine-tuning (SFT).

Usage

bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_compute_dtype=torch.float16,
    bnb_4bit_use_double_quant=True,
    bnb_4bit_quant_type="nf4",
)

model = AutoModelForCausalLM.from_pretrained(
    "Qwen/Qwen3-4B-Thinking-2507",
    quantization_config=bnb_config,
    device_map="auto",
)

self.model = PeftModel.from_pretrained(model, "ADAPTER/PATH", torch_dtype=torch.float16)

Architecture

• Base Model: Qwen3-4B-Thinking-2507

Training

• Trained on high quality investigative questions and the corresponding chain of thought (CoT) tokens generated by Deepseek V3.2 (Reasoner)

Github Repository

Full Codebase: https://github.com/Karish-Gupta/BodyCam-VQA/tree/main/fine_tuning

Framework versions

• PEFT 0.18.1