fpgaconvnet-model
fpgaconvnet-model copied to clipboard
AlexMontgomerie
Performance and resource models for fpgaConvNet: a Streaming-Architecture-based CNN Accelerator.
This is what I'm working on for merging the refactor and dev branches. Unfortunately the new code structure meant that it isn't able to show differences between the existing modelling...
This is the testing I've been working on, implemented in the dev branch. It isn't passing at the moment, mostly because of resource models being off. We could change the...
working on a way of registering layers based on a core architecture object. In this way, we can dynamically add the features we want based on the architecture we're targeting,...
This refactor will define a new standardised interface for modules. The goal if for the module classes to be lightweight, and correspond closely with the hardware parameters. This refactor introduces...
Hi, in combination with the one line change in fpgaconvnet-hls these are the small changes necessary to handle the unusual bias dimensions. Note that since the last series of commits...
In fpgaConvNet, [`Gemm layer`](https://onnx.ai/onnx/operators/onnx__Gemm.html#gemm-7) is implemented as [`InnerProductLayer`](https://github.com/AlexMontgomerie/fpgaconvnet-model/blob/main/fpgaconvnet/models/layers/InnerProductLayer.py#L13). However, the main difference is that Gemm layer supports more parameters (Boolean flags `transA`, `transB`, scalars `alpha`, `beta`). For the current model,...
Metadata
Owner
Metadata
Performance and resource models for fpgaConvNet: a Streaming-Architecture-based CNN Accelerator.