Data from BaiduYun differs from HuggingFace

[JiangJixiang]

I checked the data from BaiduYun, but all failed.

$ jq -r 'to_entries[] | "\(.value.sha256)  \(.key)"' "$BASE_JSON" > /mnt/Datasets/Bench2Drive//bench2drive_base_1000_sha256.txt
$ sha256sum -c /mnt/Datasets/Bench2Drive/bench2drive_base_1000_sha256.txt --quiet
AccidentTwoWays_Town12_Route1102_Weather10.tar.gz: FAILED
AccidentTwoWays_Town12_Route1103_Weather11.tar.gz: FAILED
AccidentTwoWays_Town12_Route1104_Weather12.tar.gz: FAILED
AccidentTwoWays_Town12_Route1105_Weather13.tar.gz: FAILED

I compared one of the dataset archives downloaded from HuggingFace versus the version provided via BaiduYun. They are different.

From HF
gzip -l Accident_Town03_Route156_Weather0.tar.gz
compressed uncompressed ratio uncompressed_name 264497958 325355520 18.7% Accident_Town03_Route156_Weather0.tar

From BaiduYun
gzip -l Accident_Town03_Route156_Weather0.tar.gz
compressed uncompressed ratio uncompressed_name 264681627 327055360 19.1% Accident_Town03_Route156_Weather0.tar

Is that all right to use the Bench2Drive data from BaiduYun?