change nomenclature of encoder

Author: dmoi

.gitignore

@@ -95,6 +95,10 @@ wandb/
 lightning_logs/
 tmp/
 models/
+**/checkpoints/
+***/logs/
+**/**__pycache__/
+
 
 # Data files
 *.csv

foldtree2/ft2treebuilder.py

@@ -9,7 +9,7 @@
 import pebble
 import argparse
 
-from foldtree2.src import foldtree2_ecddcd as ft2
+from foldtree2.src import encoder as ecdr
 from foldtree2.src import mono_decoders
 from foldtree2.src.pdbgraph import PDB2PyG
 

foldtree2/learn_lightning.py

@@ -3,7 +3,7 @@
 from torch_geometric.data import DataLoader
 import numpy as np
 from foldtree2.src import pdbgraph
-from foldtree2.src import foldtree2_ecddcd as ft2
+from foldtree2.src import encoder as ecdr
 from foldtree2.src.losses.losses import recon_loss_diag, aa_reconstruction_loss
 from foldtree2.src.mono_decoders import MultiMonoDecoder
 import os

foldtree2/learn_monodecoder.py

@@ -5,7 +5,7 @@
 from torch_geometric.data import DataLoader
 import numpy as np
 from foldtree2.src import pdbgraph
-from foldtree2.src import foldtree2_ecddcd as ft2
+from foldtree2.src import encoder as ecdr
 from foldtree2.src.losses.losses import recon_loss_diag, aa_reconstruction_loss
 from foldtree2.src.mono_decoders import MultiMonoDecoder
 import os
@@ -322,7 +322,7 @@
             commitment_start=args.commitment_start
         )
     else:
-        encoder = ft2.mk1_Encoder(
+        encoder = ecdr.mk1_Encoder(
             in_channels=ndim,
             hidden_channels=[hidden_size, hidden_size],
             out_channels=args.embedding_dim,
@@ -345,7 +345,7 @@
 
     if args.hetero_gae:
         # HeteroGAE_Decoder config (example, adjust as needed)
-        decoder = ft2.HeteroGAE_Decoder(
+        decoder = ecdr.HeteroGAE_Decoder(
             in_channels={'res': args.embedding_dim, 'godnode4decoder': ndim_godnode, 'foldx': 23},
             concat_positions=False,
             hidden_channels={('res','backbone','res'): [hidden_size]*5, ('res','backbonerev','res'): [hidden_size]*5},

foldtree2/learn_sup_decoders.py

@@ -8,7 +8,7 @@
 from torch_geometric.data import DataLoader
 import numpy as np
 from foldtree2.src import pdbgraph
-from foldtree2.src import foldtree2_ecddcd as ft2
+from foldtree2.src import encoder as ecdr
 from foldtree2.src.losses.losses import recon_loss_diag, aa_reconstruction_loss
 from foldtree2.src.mono_decoders import MultiMonoDecoder
 import os

foldtree2/makesubmat.py

@@ -50,7 +50,7 @@
 # Optional: import custom modules if available
 from foldtree2.src import AFDB_tools, foldseek2tree
 from foldtree2.src.pdbgraph import PDB2PyG, StructureDataset
-import foldtree2.src.foldtree2_ecddcd as ft2
+import foldtree2.src.encoder as ft2
 
 def parse_args():
 	parser = argparse.ArgumentParser(

foldtree2/notebooks/experiments/test_monodecoders.ipynb

@@ -1,4 +1,4 @@
-import foldtree2.src.foldtree2_ecddcd as ft2
+import foldtree2.src.encoder as ft2
 from converter import pdbgraph
 import numpy as np
 import torch

foldtree2/scaling_experiment.py

@@ -1,4 +1,4 @@
-import foldtree2.src.foldtree2_ecddcd as ft2
+import foldtree2.src.encoder as ecdr
 from converter import pdbgraph
 import numpy as np
 import torch
@@ -66,7 +66,7 @@
 print(f"Using device: {device}")
 
 # Model setup
-encoder = ft2.mk1_Encoder(
+encoder = ecdr.mk1_Encoder(
     in_channels=ndim,
     hidden_channels=[hidden_size, hidden_size, hidden_size],
     out_channels=embedding_dim,
@@ -152,10 +152,10 @@
         fft2_x = out['fft2pred'] if 'fft2pred' in out else None
         edge_index = data.edge_index_dict.get(('res', 'contactPoints', 'res'))
         if edge_index is not None:
-            edgeloss, _ = ft2.recon_loss_diag(data, edge_index, decoder, plddt=True, offdiag=True, key='edge_probs')
+            edgeloss, _ = ecdr.recon_loss_diag(data, edge_index, decoder, plddt=True, offdiag=True, key='edge_probs')
         else:
             edgeloss = torch.tensor(0.0, device=device)
-        xloss = ft2.aa_reconstruction_loss(data['AA'].x, recon_x)
+        xloss = ecdr.aa_reconstruction_loss(data['AA'].x, recon_x)
         if fft2_x is not None:
             F_hat = torch.complex(fft2_x[:, :1], fft2_x[:, 1:])
             F = torch.complex(data['fourier2dr'].x, data['fourier2di'].x)