Speed Issue #1055
Description
Hello, I'd like to ask about a speed problem. I have downloaded MMseqs2 locally and also downloaded the CPU version of the ColabFold database (its naming ends with db). However, I now want to use the GPU version of MMseqs2. I understand that this database format is incorrect – I need to convert it to the _padformat and then create soft links for _aleand _seq, right?
After completing these operations and running the process, why does it still take 2-3 hours to process a single sequence? Moreover, the final generated MSA only contains the original sequence. I can't figure out the reason.