← All drillsShrink a Neural Network with Quantisation
What you'll be able to do- ✓Shrink models with quantization
- ✓Speed up model inference
- ✓Compare post-training and quantisation-aware training
Make models smaller and faster with little accuracy loss — essential for mobile, edge, and cost-efficient serving.
⌁ TensorFlow Lite / PyTorch ExecuTorch, NVIDIA TensorRT (INT8/FP8), Qualcomm SNPE, Apple Core ML, and llama.cpp/GGUF.
Start this internshipCreate an account to unlock the 7 sections, the workbench, and askThili.
BeginSections
1. Learn Fast Neural Network Quantization
🔒 locked2. Understanding Neural Network Size and Parameters
🔒 locked3. Mapping Function
🔒 locked4. Post Training Dynamic Weights-Only Quantization
🔒 locked6. Post Training Static Quantization (PTQ)
🔒 locked7. Quantization Aware Training (QAT)
🔒 lockedDig deeper
📄Quantization for Efficient Integer-Arithmetic-Only Inference (Jacob et al., 2017)
paper🔗Practical Quantization in PyTorch — PyTorch blog
article