The documentation string about the seed of NASBench201 is misleading

[nabenabe0928]

NASBench201 has only two of the seeds (777, 888) for almost half of the architectures as mentioned here.

On the other hand, HPOBench is saying three seeds (777, 888, 999) are available.
For this reason, I checked one index, i.e. 347, of the seed 999 that is not held by NASBench201.
Then, I just found out that the seed 888 is used to patch the results for the seed 999.

First, my question:
HPOBench also provides two seeds for most architectures, right?

Request:
Could you make a notice about this matter if the answer to my question is yes?

[KEggensperger]

Thanks for raising this. I believe this is true and should be mentioned in the docs. @PhMueller Could you please add a note to the wiki for all benchmarks where this applies?