update with major refactor

README.md

@@ -11,11 +11,11 @@ snapshot_download(repo_id=model_name, local_dir=genbio_models_path)
 ```
 ### Load model for inference
 ```python
-from genbio_finetune.tasks import SequenceRegression
+from modelgenerator.tasks import SequenceRegression
 
 ckpt_path = genbio_models_path.joinpath('fold0', 'model.ckpt')
 model = SequenceRegression.load_from_checkpoint(ckpt_path, strict_loading=False).eval()
 
-collated_batch = model.collate({"sequences": ["ACGT", "AGCT"]})
+collated_batch = model.transform({"sequences": ["ACGT", "AGCT"]})
 logits = model(collated_batch)
 print(logits)

fold0/config.yaml

@@ -136,10 +136,10 @@ trainer:
   reload_dataloaders_every_n_epochs: 0
   default_root_dir: logs
 model:
-  class_path: genbio_finetune.tasks.SequenceRegression
+  class_path: modelgenerator.tasks.SequenceRegression
   init_args:
     backbone:
-      class_path: genbio_finetune.models.proteinfm
+      class_path: modelgenerator.backbones.proteinfm
       init_args:
         from_scratch: false
         use_peft: false
@@ -152,7 +152,7 @@ model:
         model_init_args: null
         max_length: 2048
     adapter:
-      class_path: genbio_finetune.models.MLPPoolAdapter
+      class_path: modelgenerator.adapters.MLPPoolAdapter
       init_args:
         pooling: mean_pooling
         hidden_sizes:
@@ -176,13 +176,13 @@ model:
         differentiable: false
         fused: null
     lr_scheduler:
-      class_path: genbio_finetune.lr_schedulers.ConstantWithWarmup
+      class_path: modelgenerator.lr_schedulers.ConstantWithWarmup
       init_args:
         warmup_ratio: 0.05
     strict_loading: true
     reset_optimizer_states: false
 data:
-  class_path: genbio_finetune.data.DMSFitnessPrediction
+  class_path: modelgenerator.data.DMSFitnessPrediction
   init_args:
     path: genbio-ai/ProteinGYM-DMS
     train_split_files:

fold1/config.yaml

@@ -136,10 +136,10 @@ trainer:
   reload_dataloaders_every_n_epochs: 0
   default_root_dir: logs
 model:
-  class_path: genbio_finetune.tasks.SequenceRegression
+  class_path: modelgenerator.tasks.SequenceRegression
   init_args:
     backbone:
-      class_path: genbio_finetune.models.proteinfm
+      class_path: modelgenerator.backbones.proteinfm
       init_args:
         from_scratch: false
         use_peft: false
@@ -152,7 +152,7 @@ model:
         model_init_args: null
         max_length: 2048
     adapter:
-      class_path: genbio_finetune.models.MLPPoolAdapter
+      class_path: modelgenerator.adapters.MLPPoolAdapter
       init_args:
         pooling: mean_pooling
         hidden_sizes:
@@ -176,13 +176,13 @@ model:
         differentiable: false
         fused: null
     lr_scheduler:
-      class_path: genbio_finetune.lr_schedulers.ConstantWithWarmup
+      class_path: modelgenerator.lr_schedulers.ConstantWithWarmup
       init_args:
         warmup_ratio: 0.05
     strict_loading: true
     reset_optimizer_states: false
 data:
-  class_path: genbio_finetune.data.DMSFitnessPrediction
+  class_path: modelgenerator.data.DMSFitnessPrediction
   init_args:
     path: genbio-ai/ProteinGYM-DMS
     train_split_files:

fold2/config.yaml

@@ -136,10 +136,10 @@ trainer:
   reload_dataloaders_every_n_epochs: 0
   default_root_dir: logs
 model:
-  class_path: genbio_finetune.tasks.SequenceRegression
+  class_path: modelgenerator.tasks.SequenceRegression
   init_args:
     backbone:
-      class_path: genbio_finetune.models.proteinfm
+      class_path: modelgenerator.backbones.proteinfm
       init_args:
         from_scratch: false
         use_peft: false
@@ -152,7 +152,7 @@ model:
         model_init_args: null
         max_length: 2048
     adapter:
-      class_path: genbio_finetune.models.MLPPoolAdapter
+      class_path: modelgenerator.adapters.MLPPoolAdapter
       init_args:
         pooling: mean_pooling
         hidden_sizes:
@@ -176,13 +176,13 @@ model:
         differentiable: false
         fused: null
     lr_scheduler:
-      class_path: genbio_finetune.lr_schedulers.ConstantWithWarmup
+      class_path: modelgenerator.lr_schedulers.ConstantWithWarmup
       init_args:
         warmup_ratio: 0.05
     strict_loading: true
     reset_optimizer_states: false
 data:
-  class_path: genbio_finetune.data.DMSFitnessPrediction
+  class_path: modelgenerator.data.DMSFitnessPrediction
   init_args:
     path: genbio-ai/ProteinGYM-DMS
     train_split_files:

fold3/config.yaml

@@ -136,10 +136,10 @@ trainer:
   reload_dataloaders_every_n_epochs: 0
   default_root_dir: logs
 model:
-  class_path: genbio_finetune.tasks.SequenceRegression
+  class_path: modelgenerator.tasks.SequenceRegression
   init_args:
     backbone:
-      class_path: genbio_finetune.models.proteinfm
+      class_path: modelgenerator.backbones.proteinfm
       init_args:
         from_scratch: false
         use_peft: false
@@ -152,7 +152,7 @@ model:
         model_init_args: null
         max_length: 2048
     adapter:
-      class_path: genbio_finetune.models.MLPPoolAdapter
+      class_path: modelgenerator.adapters.MLPPoolAdapter
       init_args:
         pooling: mean_pooling
         hidden_sizes:
@@ -176,13 +176,13 @@ model:
         differentiable: false
         fused: null
     lr_scheduler:
-      class_path: genbio_finetune.lr_schedulers.ConstantWithWarmup
+      class_path: modelgenerator.lr_schedulers.ConstantWithWarmup
       init_args:
         warmup_ratio: 0.05
     strict_loading: true
     reset_optimizer_states: false
 data:
-  class_path: genbio_finetune.data.DMSFitnessPrediction
+  class_path: modelgenerator.data.DMSFitnessPrediction
   init_args:
     path: genbio-ai/ProteinGYM-DMS
     train_split_files:

fold4/config.yaml

@@ -136,10 +136,10 @@ trainer:
   reload_dataloaders_every_n_epochs: 0
   default_root_dir: logs
 model:
-  class_path: genbio_finetune.tasks.SequenceRegression
+  class_path: modelgenerator.tasks.SequenceRegression
   init_args:
     backbone:
-      class_path: genbio_finetune.models.proteinfm
+      class_path: modelgenerator.backbones.proteinfm
       init_args:
         from_scratch: false
         use_peft: false
@@ -152,7 +152,7 @@ model:
         model_init_args: null
         max_length: 2048
     adapter:
-      class_path: genbio_finetune.models.MLPPoolAdapter
+      class_path: modelgenerator.adapters.MLPPoolAdapter
       init_args:
         pooling: mean_pooling
         hidden_sizes:
@@ -176,13 +176,13 @@ model:
         differentiable: false
         fused: null
     lr_scheduler:
-      class_path: genbio_finetune.lr_schedulers.ConstantWithWarmup
+      class_path: modelgenerator.lr_schedulers.ConstantWithWarmup
       init_args:
         warmup_ratio: 0.05
     strict_loading: true
     reset_optimizer_states: false
 data:
-  class_path: genbio_finetune.data.DMSFitnessPrediction
+  class_path: modelgenerator.data.DMSFitnessPrediction
   init_args:
     path: genbio-ai/ProteinGYM-DMS
     train_split_files: