Saurav Muralidharan
Results
4
issues of
Saurav Muralidharan
Use `fp16` mixed-precision in the L-step to enable Tensor Core utilization. Figure out how to integrate Apex-AMP. **References** * [Mixed Precision Training (arXiv)](https://arxiv.org/abs/1710.03740) * [NVIDIA Apex](https://github.com/NVIDIA/apex)
enhancement
Investigate adding support for hybrid quantization (`int8` weights, `float32` biases & activations). **Resources:** * [PyTorch Quantization Support](https://pytorch.org/docs/stable/quantization.html) * [Introduction to PyTorch Quantization](https://pytorch.org/blog/introduction-to-quantization-on-pytorch/)
enhancement