Find optimal renaming of ground truth based on the predicted positions. Alg. 26 "renameSymmetricGroundTruthAtoms" This renamed ground truth

Choose what optimizers and learning-rate schedulers to use in your optimization. Normally you'd need one. But in the case of GANs or s

Choose what optimizers and learning-rate schedulers to use in your optimization. Normally you'd need one. But in the case of GANs or s

ESMFold baseline model using 150M ESM-2, 0 folding blocks. ESM-2 here is trained out to 500K updates. This is a model designed to test the ca

ESMFold baseline model using 150M ESM-2, 0 folding blocks. ESM-2 here is trained out to 270K updates. This is a model designed to test the ca

ESMFold baseline model using 15B ESM-2, 0 folding blocks. ESM-2 here is trained out to 270K updates. The 15B parameter ESM-2 was not trained

ESMFold baseline model using 35M ESM-2, 0 folding blocks. ESM-2 here is trained out to 500K updates. This is a model designed to test the cap

ESMFold baseline model using 35M ESM-2, 0 folding blocks. ESM-2 here is trained out to 270K updates. This is a model designed to test the cap

ESMFold baseline model using 3B ESM-2, 0 folding blocks. ESM-2 here is trained out to 500K updates. This is a model designed to test the capa

ESMFold baseline model using 3B ESM-2, 0 folding blocks. ESM-2 here is trained out to 270K updates. This is a model designed to test the capa

ESMFold baseline model using 650M ESM-2, 0 folding blocks. ESM-2 here is trained out to 500K updates. This is a model designed to test the ca

ESMFold baseline model using 650M ESM-2, 0 folding blocks. ESM-2 here is trained out to 270K updates. This is a model designed to test the ca

ESMFold baseline model using 8M ESM-2, 0 folding blocks. ESM-2 here is trained out to 500K updates. This is a model designed to test the capa

ESMFold baseline model using 8M ESM-2, 0 folding blocks. ESM-2 here is trained out to 270K updates. This is a model designed to test the capa

ESMFold v0 model with 3B ESM-2, 48 folding blocks. This version was used for the paper (Lin et al, 2022). It was trained on all PDB chains un

ESMFold v1 model using 3B ESM-2, 48 folding blocks. ESMFold provides fast high accuracy atomic level structure prediction directly from the i

Applies optional utilities, controlled by config flags. Utilities: - Ignoring python warnings - Rich config printing

Makes sure everything closed properly.

Flattens a nested dict.

scores: [N, ..., C], unnormalized scores target: [N, ...] mask: [N, ...], where elements with `True` are allowed and `False`

scores: [N, L, C], unnormalized scores target: [N, L] coord_mask: FloatTensor [N, L], where elements with `True` are allowed

scores: [N, L, C], unnormalized scores target: [N, L] coord_mask: FloatTensor [N, L], where elements with `True` are allowed

scores: [N, L, C], unnormalized scores target: [N, L] coord_mask: FloatTensor [N, L], where elements with `True` are allowed

Args: model_outs (dict): dict of loss_name: model_out targets (_type_): _description_

einstein notation. b - batch n - sequence (or flattened spatial dimensions) d - feature dimension, which is also log2(codeboo

Inputs the output of the encoder network z and maps it to a discrete one-hot vector that is the index of the closest embedding vector

Basic self attention with optional mask and external pairwise bias. To handle sequences of different lengths, use mask. Inputs:

Inputs: sequence_state: B x L x sequence_state_dim Output: pairwise_state: B x L x L x pairwise_state_dim

Inputs: pairwise_state: B x L x L x pairwise_state_dim Output: pairwise_bias: B x L x L x num_heads

Inputs: sequence_state: B x L x sequence_state_dim pairwise_state: B x L x L x pairwise_state_dim mask: B x L b

Args: s: [*, C_hidden] single embedding s_initial: [*, C_hidden] single embedding as

Args: s: [*, N_res, C_s] single representation z: [*, N_res, N_res, C_z] pair represe

Args: [*, N_res, C_s] single representation Returns: [*, N_res, 6] update vector

Args: evoformer_output_dict: Dictionary containing: "single": [*, N_r

Input: residue_index: B x L tensor of indices (dytpe=torch.long) mask: B x L tensor of booleans Output:

Inputs: seq_feats: B x L x C tensor of sequence features pair_feats: B x L x L x C tensor of pai