Qwen3-VL-8B VQA Wiki
Merged Qwen3-VL-8B-Instruct model fine-tuned for Vietnamese image question answering
on the local data_wiki / VietTravelVQA-style dataset.
Training summary:
- Base model:
Qwen/Qwen3-VL-8B-Instruct - Fine-tuning: QLoRA, merged into bf16 full weights
- Dataset format: single-image VQA with Vietnamese questions and answers
- Final checkpoint merged:
checkpoint-191
Example loading:
from transformers import AutoProcessor, Qwen3VLForConditionalGeneration
model_id = "phunguyendinhkh/qwen3_vl_8b_vqa_wiki"
processor = AutoProcessor.from_pretrained(model_id)
model = Qwen3VLForConditionalGeneration.from_pretrained(
model_id,
dtype="auto",
device_map="auto",
)