I will optimize your pytorch model with quantization and onnx export

Turkey

I speak Turkish, English

Computer Vision and AR Engineer, PhD Researcher

PhD researcher in computer vision with peer-reviewed publications in efficient deep learning. I make vision models smaller and faster on real hardware: knowledge distillation, quantization, pruning, a...
About this Gig

Is your PyTorch model too slow, too large or too expensive to run in production? I am a PhD researcher in computer vision with peer-reviewed publications on model compression and knowledge distillation. I will make your model smaller and faster while keeping accuracy under control.


What I can do:

  • Post-training quantization (FP16 / INT8) and quantization-aware training
  • Structured and unstructured pruning
  • Knowledge distillation into a smaller student network
  • Export to ONNX / ONNX Runtime, plus TensorRT or Core ML where your hardware supports it
  • Before/after benchmarks: latency, model size, memory and accuracy


What you get:

  • Optimized model files and clean, documented Python code
  • A short report explaining every change and its accuracy cost


Please message me before ordering with your model architecture, framework version, target hardware (GPU, CPU, Jetson, mobile) and accuracy budget, so I can confirm the right package.

Expertise:

Image processing

•

Feature learning

•

Classification

Programming language:

Python

•

Colab

Tools:

Jupyter Notebook

•

OpenCV

•

PyTorch

Frameworks:

Keras

•

PyTorch