openfold

openfold copied to clipboard

Training duration & NaNs during training

27

First of all, great work! I'm wondering what training times I can expect for a single target. I'm currently at 1min/ it (sample) which seems too slow (v100s with fp16...

lhatsk

The training speed is very slow

2

Hi, I try to train openfold without distillation datasets. But I found the training speed is very slow. The following table is my running data. | Hardware | Trainging time(one...

llwx593

.pdb files in the `uniclust30/` section of the RODA database have erroneous suffixes in their filenames. Instead of `.pdb`, which is what OpenFold expects, they are named `_model_5_unrelaxed.pdb`. In order...

gahdritz

Bug with template mask for batch inference

2

Hello, my name is James, and I'm working on training a new AlphaFold variant using OpenFold. Thanks for the great tool! I think I may have found a bug in...

jproney

Minor optimizations & fixes to support ESMFold

1

Hi OpenFold team! Thanks for your great implementation, here are some suggestions to the codebase to support upcoming ESMFold release: 1. Constant tensors like default_frames, group_idx, atom_mask and lit_positions in...

nikitos9000

fp32 vs bf16 training performance?

2

I train the openfold in both fp32 and bf16, it seems bf16 performs better than fp32 or the fp32 config is not properly set by me? The command for bf16...

zhangyuxuan1996

Can not load AlphaFold multimer checkpoint params_model_[1-5]_multimer.npz for inference

5

Can not load multimer checkpoint in the main branch for both Docker container and ENV. Here is command  Here is the error...

bing-song

Cuda/Pytorch/Installation Issues

39

Hello! So I have been struggling with a strange issue that I hope you or someone would be able to help me with. Let me start by providing some information:...

Cweb118

--trace_model performance

6

Hi, I have tested using the --trace_model mode on a small batch of sequences of the same length; I get an 80s tracing time followed by 20s inference for each...

mrhoag5

how to specify db-load-mode in precomputed_alignments_mmseqs.py

1

Hi, I am using the recomputed_alignments_mmseqs.py to search for MSA for my input.fasta which has 9 sequences inside. but the mmseq2 is taking too long to run it, it is...

lzhangUT