Ciprian Mindru
Ciprian Mindru
@842974287 Macro defined there used for nodes with should have same input/output quantization parameters: https://github.com/pytorch/glow/blob/cf3a95aed62bbe97391c2a925f07bbeefd4d0430/lib/Quantization/Quantization.cpp#L503-L504 and: https://github.com/pytorch/glow/blob/cf3a95aed62bbe97391c2a925f07bbeefd4d0430/lib/Quantization/Quantization.cpp#L398-L405 So what it means is that the RescaleQuantized node should be merged in...
@opti-mix Do know how to do this easily and elegantly? I wasted 2 days in trying to make an IR optimization pass which does the above but I ended nowhere...
@opti-mix Yeah I`ve seen that optimization but it seems it is not generic enough and does not kick in for my case. https://github.com/pytorch/glow/blob/39a8c689f252076ff5842c1870523b420e509b72/lib/Optimizer/IROptimizer/IROptimizer.cpp#L1347-L1352 This optimization is used only if the...
@opti-mix I can provide a model for which this happens: [IROptModel.zip](https://github.com/pytorch/glow/files/6085435/IROptModel.zip) You can find in the archive: - A MobileNet SSD model in ONNX format (model publicly available taken from...
@opti-mix Did you have time to investigate this optimization?
@opti-mix Ok no problem. Btw do you think it is worth organizing some meetings with all the Glow contributors to exchange/share ideas about the Glow future, identify groups of people...
@G4V Sorry for delayed response. The Glow ONNX importer when importing and LSTM module it creates some additional graph placeholders corresponding to the LSTM states (hidden state "H" and cell...
@G4V I pushed here #4830 a PR which modifies the default way in which an RNN operator is loaded (either RNN, GRU or LSTM operator) by not exporting by default...
@jfix71 Can I add an additional optimization right after the lowering to make sure all constants are folded?
@jfix71 But if I add an extra optimization right before the `transformForPrecisionMode()` such that both the quantization and the profiling path benefit from the same optimizations then it should be...