GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
nishishita
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
nishishita
Model Tree
Qwen/Qwen3-VL-4B-Instruct
Input Modalities
Output Modalities
Supported Functionality
The example maps A to single-dish and B to multi-dish, and retains the raw model text. A response that does not start with either letter is reported as unknown.
Run from this directory. The script loads the required Qwen base model and this PEFT LoRA adapter.
bash
python -m pip install -r requirements.txtpython infer.py --image /path/to/food_crop.jpg --device cuda:0CUDA inference uses one GPU in BF16. For CPU inference, use --device cpu (float32; more RAM required). Training photos are not included.
Automating Dish-Level Annotation of Real Food-Diary Photos
Adapter weights and inference code: Apache-2.0. The Qwen base model is separately distributed under Apache-2.0.