Closed kirilllzaitsev closed 1 year ago
Hi,
In this model, you should use --overlap-data
as it is shown in the related line of the "gen-demos-max78000.sh":
python ai8xize.py --test-dir $TARGET --prefix camvid_unet --checkpoint-file trained/ai85-camvid-unet-large-fakept-q.pth.tar --config-file networks/camvid-unet-large-fakept.yaml $COMMON_ARGS --overlap-data --mlator --no-unload --max-checklines 8192 --new-kernel-loader "$@"
In this case, since the resolution (88x88) consumes nearly the entire data memory, there is no way of writing output without any overlap. However, by indicating an offset before the input we make sure that the input data is already consumed before being overwritten.
This issue has been marked stale because it has been open for over 30 days with no activity. It will be closed automatically in 10 days unless a comment is added or the "Stale" label is removed.
Hi, running the train-quantize-eval-synthesize pipeline for the Camvid example, I encountered an error on the synthesis stage:
Network-related files (ai85net-unet.py, camvid-unet-large.yaml, camvid-unet-large-fakept.yaml) and the Camvid dataset are unchanged. These are the commands I use: