fantasia initialize when succesive run

[EricDeveaud]

Hello,

sorry for the dumb question
for local execution shall fantasia initialize be run each time before fantasia run

what If I want to analyse prot1.fa and prot2.fa

regards

Eric

[frapercan]

Hello Eric,

No, for local execution you do not need to run fantasia initialize every time before fantasia run.

Once the system is initialized, the proteins, sequences, embeddings, and related data can remain stored in the database and reused normally until you decide to change the lookup table or reload another reference dataset for some reason.

So, if you want to analyse prot1.fa and then prot2.fa, you can do that without re-initializing each time.

On HPC systems, it is usually different, because they often do not allow you to keep a persistent PostgreSQL instance available to the compute nodes. In that case, initialization is typically needed for each job or batch of jobs. The good news is that loading the embeddings usually only takes a couple of minutes. Also, if you configure the FANTASIA base directory (experiments, embeddings, logs) in $store or another persistent location, it only needs to load the files from there rather than downloading them again.

And once the embeddings are loaded into the database, you can launch batches of experiments from the launcher. So in practice, you only need to initialize once per batch of jobs, not once per individual run. There is an example of this in the documentation somewhere. For local usage, this can also be done very simply with a small shell script to launch batches.

Regards,
Francisco

[EricDeveaud]

thanks for the clarification