brevitas
brevitas copied to clipboard
Brevitas: neural network quantization in PyTorch
Error in ONNX standard export: is there any particular requirement on layer quantization instance?
Hello, I'm having trouble in exporting my model to ONNX standard. I notice that to export a QuantConv2d layer to ONNX seems mandatory to set also a quantization for the...
Papers: - LSQ+: https://arxiv.org/abs/2004.09576 - Rotated binary network: https://arxiv.org/abs/2009.13055 - PROFIT: https://link.springer.com/chapter/10.1007/978-3-030-58539-6_26 - Exponential rescaling: https://arxiv.org/abs/1808.05779 - Soft rounding differentiable approximation: https://arxiv.org/pdf/2006.09952.pdf and https://arxiv.org/abs/1908.05033 - Scaled elementwise STE https://arxiv.org/abs/2104.00903 -...
Hello author, I am still trying to use pytorch to simulate the quantization and dequantization forward reasoning process of brevitas. I have the following two questions to ask you: 1....
They should be close enough to make it doable.
Find the time.
This should simplify a bit things for any quantizer that does custom stuff pre rounding.
Push all the logic down at the proxy level.