GitLab

- Use sequence embedding files when in `seqemb` mode.
- Make dummy MSA features for MMCIF when using `seqemb` mode.

- Added flag `no_column_attention` in evoformer config.
- Added check in `evoformer.py` to switch off `MSAColumnAttention` when the config flag `no_column_attention` is `True`.

- Added `preembedding_embedder` config dictionary in `config`
- Added `preemb_dim_size` property in `config` for specifying single seq embedding size.

- `seqemb_mode_enabled` added as a configuration option.
- `model.py` switches to using the `PreembeddingEmbedder` when the flag is `True`.

- Added `use_single_seq_mode` flag in inference script arguments.
- Passed on the flag to the FASTA file `data_processor`.

- Added a method to load and process sequence embedding `*.pt` files.
- In `seqemb_mode`, now add seqemb features to the feature dictionary.

- Added a `seq_emb` bool flag to `data_pipeline > process_fasta()`
- In `seqemb_mode` use dummy MSA features instead of full ones.

- New method for generating dummy MSA features.

- Added configuration options in `config.py` for toggling seq-emb mode
- Added list for specifying features to be used in seq-emb mode.

- Added a `PreembeddingEmbedder` for embedding single-sequence (NUM_RESIDUE, ...) shaped embeddings as input.

Fixes cuda/float wrapper error in unit tests

tests in test_data_transforms.py

 test_evoformer.py

Fix Colab: Install all conda packages together

Later on the conda setup is likely broken by pip.

Fix Colab again after ModelCIF merge and python version change in Colab

Add comment about the interpretation of ambiguous atoms

Sorry, I have no other way to re-run the Docker build test.
Changed one space.

Advance python version to 3.9 to build docker in Ubuntu Lunar Lobster

Colab notbook was broken by this change and potentially code of other users
as well. Keep backward compatibility by also allowing the old way.