Questions about hls4ml Dockers

[Louis Morge-Rollet]

Hello,

I'm Louis Morge-Rollet, associated Professor at Grenoble INP - Esisar (France). I tried to create an Pynq-Z2 accelerator with hls4ml in one project with my students. I'm a regular user of hls4ml, but since this year I occupy a new post and I don't have access to the same infrastructure as before. Particularly, we can't use Vivado on Linux, due to image constraints on our infrastructure.

We tried to used hls4ml on Windows, but it seems not compatible with Windows ([WinError 183]). We also tried with WSL (hls4ml on Linux and Vivado on Windows), it solves some problems, but we still have some troubles to synthetize. Our last solution was to used the Docker you provided (until the last Git push 10 days ago). Indeed, before, you provided two Docker images for the tutorial: ghcr.io/fastmachinelearning/hls4ml-tutorial/hls4ml-0.8.0:latest and ghcr.io/fastmachinelearning/hls4ml-tutorial/hls4ml-0.8.0-vivado-2019.2:latest. We were able to pull the first Docker image, but we not able to pull the second one. The first Docker image only contain hls4ml (on Linux) and can't allow us to synthesize the accelerator. The second one seems to contain hls4ml and Vivado tools, but we do not succeed to pull it, due to access right.

My questions are the following:
+ Is there a way to perform synthesis with the first Docker image, or it's only to compare accuracies between QKeras and hls4ml (C synthesis) ?
+ Can you provide us a token to download the second Docker image, if it's still available ?
+ Do you know others solutions to solve our problem ?

Thanks for your answer.

Sincerely,

Louis Morge-Rollet.