GitLab

Improve TriangularMultiplicativeUpdate stability in fp16 mode

Resolving torch cuda availability for cpu-only installation

Fix comparison for max_seqlen when downloading CAMEO.

New option to output in ModelCIF format instead of PDB format

Fix notebook: Colab now has python 3.8, fix imports, mitigate UTF-8 glitch

Previously, long sequences were not excluded from the script.
This commit changes the comparison to exclude sequences with length
greater than args.max_seqlen.

Add missing warmup_num_steps parameters to DeepSpeed config generation script.

warmup_decay_num_steps is a valid user argument but is not added to the config.
This commit adds this argument to the configuration output.

This way user can repeat just the first and last cell to get both versions.

Previously protein got computed and shown in both cases but only
saved to the downloaded zip when relaxation was done.

Colab is easily broken by openmm and other software messing with locale.
This mitigation patch lets the %shell and ! work again.
It is much stronger solution than just avoiding shell when making zip.

* Python 3.8 is now what we get in Colab, use it in our Conda as well
* Fix imports, make some unconditional - chain of depencencies makes them
  needed even for relax_prediction==False
* Mitigate spurious error "A UTF-8 locale is required. Got ANSI_X3.4-1968"
  which makes all shell commands defunct - use shutil.make_archive()

Add a .gitignore file

Fix `--cpus_per_task` argument in README

Now multichain tested and seems to work.

Also fixed important bug: atom names were padded

Load pretrained jax weights