search

Xilinx / inference-server

https://xilinx.github.io/inference-server/
Apache License 2.0
43 stars 13 forks source link

Allow not pushing batch size in the XModel backend #216

Closed varunsh-xilinx closed 1 year ago

varunsh-xilinx commented 1 year ago

Summary of Changes

Motivation

The V70 board has a different behavior from the other boards and requires that the tensor sizes are correctly sized without a leading 1.

Implementation

If a batch size of zero is passed to the allocator, assume the batch size is already present in the shape.

Notes

This change was originally implemented in #215 but was taken out.

gbuildx commented 1 year ago

Build failed!

varunsh-xilinx commented 1 year ago

Failure from unrelated MIGraphX build