Abhinav

I think, the flag --use_precomputed_msas searches for msas in the output directory. For multimers it should be something like: (output_directory_parent/fasta_file_name/msas/A/mgnify_hits.sto etc...). You may have to do a similar thing, which...

> I have te same issue. In this case, will it be accurate for predicting the structure by replacing the Selenocysteine by cysteine? Did aflphafold sequence input recognize the U...

I have the same error. It works fine in my PC but I get this error when I try to install in my HPC facility. Were you able to fix...

Did any of you encounter problem with deepspeed during multi-node/multi-GPU training like this - #486 ? Any suggestions would be really helpful

> Well I understand what happens now. The main problem is that when you load from jax, ExponentialMovingAverage, the ema initiallized incorrectly. You can see in train_openfold.py line307, there is:...

Yes, it is one. I have not changed the config file. Here is what it says: if train: c.globals.blocks_per_ckpt = 1 "evoformer_stack": { "c_m": c_m, "c_z": c_z, "c_hidden_msa_att": 32, "c_hidden_opm":...

I was able to get it running in A100s. I realised that the num of recycles are set to 20 for multimer training and has to be corrected with an...

This is an important issue that has not been flagged. This has been messing up with my training for a while now.

I think it is looking for PDB templates, and you may have set the PDB path to ./first_test/input_cifs ?