update with major refactor

README.md

@@ -20,12 +20,12 @@ snapshot_download(repo_id=model_name, local_dir=genbio_models_path)
 ### Load model for inference
 ```python
 import torch
-from genbio_finetune.tasks import SequenceClassification
+from modelgenerator.tasks import SequenceClassification
 
 ckpt_path = genbio_models_path.joinpath('model.ckpt')
 model = SequenceClassification.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)
 print(torch.argmax(logits, dim=-1))

config.yaml

@@ -37,7 +37,7 @@ trainer:
     class_path: lightning.pytorch.loggers.WandbLogger
     init_args:
       name: peptide_HLA_MHC_affinity_proteinmoe_16B_1113
-      save_dir: /lustre/scratch/shared-folders/bio_project/shuxian/genbio_finetune/logs
+      save_dir: /lustre/scratch/shared-folders/bio_project/shuxian/modelgenerator/logs
       version: null
       offline: false
       dir: null
@@ -75,7 +75,7 @@ trainer:
       log_weight_decay: false
   - class_path: lightning.pytorch.callbacks.ModelCheckpoint
     init_args:
-      dirpath: /lustre/scratch/shared-folders/bio_project/shuxian/genbio_finetune/logs/xtrimo_benchmark_gbft_reproduction/peptide_HLA_MHC_affinity_proteinmoe_16B_1113
+      dirpath: /lustre/scratch/shared-folders/bio_project/shuxian/modelgenerator/logs/xtrimo_benchmark_gbft_reproduction/peptide_HLA_MHC_affinity_proteinmoe_16B_1113
       filename: best_val:{epoch}-{val_auroc:.3f}
       monitor: val_auroc
       verbose: false
@@ -134,12 +134,12 @@ trainer:
   plugins: null
   sync_batchnorm: false
   reload_dataloaders_every_n_epochs: 0
-  default_root_dir: /lustre/scratch/shared-folders/bio_project/shuxian/genbio_finetune/logs
+  default_root_dir: /lustre/scratch/shared-folders/bio_project/shuxian/modelgenerator/logs
 model:
-  class_path: genbio_finetune.tasks.SequenceClassification
+  class_path: modelgenerator.tasks.SequenceClassification
   init_args:
     backbone:
-      class_path: genbio_finetune.models.proteinfm
+      class_path: modelgenerator.backbones.proteinfm
       init_args:
         from_scratch: false
         max_length: 64
@@ -161,7 +161,7 @@ model:
           attention_probs_dropout_prob: 0.0
         model_init_args: null
     adapter:
-      class_path: genbio_finetune.models.MLPPoolAdapter
+      class_path: modelgenerator.adapters.MLPPoolAdapter
       init_args:
         pooling: mean_pooling
         hidden_sizes:
@@ -189,7 +189,7 @@ model:
         differentiable: false
         fused: null
     lr_scheduler:
-      class_path: genbio_finetune.lr_schedulers.CosineWithWarmup
+      class_path: modelgenerator.lr_schedulers.CosineWithWarmup
       init_args:
         warmup_ratio: 0.05
         num_warmup_steps: null
@@ -199,7 +199,7 @@ model:
     strict_loading: true
     reset_optimizer_states: false
 data:
-  class_path: genbio_finetune.data.PeptideHlaMhcAffinity
+  class_path: modelgenerator.data.PeptideHlaMhcAffinity
   init_args:
     path: biomap-research/peptide_hla_mhc_affinity
     class_filter: null