Access pre-trained AI models optimized for Mobilint NPUs.
npx skills add https://github.com/mobilint/mblt-model-zoo --skill mblt-model-zoo
Please help me install this Agent Skill. Skill: mblt-model-zoo Source: https://github.com/mobilint/mblt-model-zoo/tree/main Command: npx skills add https://github.com/mobilint/mblt-model-zoo --skill mblt-model-zoo
This Skill simplifies the deployment of AI models optimized for Mobilint Neural Processing Units (NPUs), allowing developers to accelerate model integration and inference with reduced effort.
To load and run the ResNet50 model from the Model Zoo, execute the following Python command:
from mblt_model_zoo.vision import MBLT_Engine
model = MBLT_Engine(model_cls="resnet50", model_type="DEFAULT", model_path="", core_mode="global8")
High-intent search queries and answers about installing and using this skill.
You can run AI models on Mobilint NPUs by instantiating the MBLT_Engine with a model class like ResNet50, which provides pre-quantized weights and built-in preprocessing for optimized hardware inference.
The repository provides pre-trained, pre-quantized models for vision, language, and multimodal tasks, enabling developers to execute image classification and other AI workloads with optimized NPU performance.
Yes, executing the model zoo requires Python libraries including PyTorch, OpenCV, and NumPy to handle model loading, image preprocessing, and tensor operations for NPU inference.
Yes, the framework includes preprocessing and postprocessing tools that allow developers to evaluate, fine-tune, and integrate computer vision models using the faster-coco-eval dependency.
The best way to run ResNet50 on an NPU is importing MBLT_Engine, setting the model class to ResNet50, and configuring the core mode to global8 to achieve optimized inference performance.
Integration requires explicit Mobilint NPU support and specific Python dependencies, meaning models outside the provided repository may need manual quantization to function efficiently on the hardware.
Fudan-SMI-labPin CPU-base PyTorch and install matching torch-npu and torchvision.
Vegetable-bird10086Convert PyTorch models to portable .pte files for ExecuTorch deployment.
thinking-machines-labDownloads, merges LoRA adapters, and publishes models to HuggingFace Hub.
datathingsReference the llama.cpp C API for model loading and inference workflows.