search

Xilinx / Vitis-AI

Vitis AI is Xilinx’s development stack for AI inference on Xilinx hardware platforms, including both edge devices and Alveo cards.
https://www.xilinx.com/ai
Apache License 2.0
1.48k stars 627 forks source link

Checksum mismatch with GPU model downloads #554

Closed yetimoss closed 2 years ago

yetimoss commented 3 years ago

There are several GPU model downloads in the Vitis-AI repository with computed checksums that don't match the ones listed in the model.yaml files.

The models are:

  1. pt_FPN-resnet18_covid19-seg_352_352_22.7G_1.4
  2. pt_pointpillars_kitti_12000_100_10.8G_1.4
  3. pt_pointpillars_kitti_12000_100_10.8G_1.4
  4. pt_salsanextv2_semantic-kitti_64_2048_0.75_32G_1.4
  5. tf_rcan_DIV2K_360_640_0.98_86.95G_1.4

See attached text file for more details: Vitis-AI-Repo_bad_checksums_gpumodels.log

wangxd-xlnx commented 3 years ago

Hi @yetimoss, these 4 models have been updated in download package, and MD5 checksum updated in 1.4.1 repo. Now 1.4.1 is very close to release. Then the package and MD5 will be consistent.