metadata
library_name: transformers
base_model:
- Qwen/Qwen3-VL-32B-Thinking
pipeline_tag: text-generation
Qwen3-VLTO-32B-Thinking
Qwen3-VL-32B-Thinking but without the vision components (Vision Language Text Only). Functions exactly like a text-only Qwen3 model.
To do this, I simply imported the weights from the VL model into the text model via PyTorch's load_state_dict. The model architecture is essentially the exact same.
