workbench 0.8.239__py3-none-any.whl → 0.8.243__py3-none-any.whl

workbench/core/artifacts/endpoint_core.py

@@ -1138,7 +1138,7 @@ class EndpointCore(Artifact):
             cls.log.error("Error deleting endpoint.")
             raise e
 
-        time.sleep(5)  # Final sleep for AWS to fully register deletions
+        time.sleep(10)  # Final sleep for AWS to fully register deletions
 
     @classmethod
     def delete_endpoint_models(cls, endpoint_name: str):

workbench/core/transforms/features_to_model/features_to_model.py

@@ -247,9 +247,9 @@ class FeaturesToModel(Transform):
         # Create a Sagemaker Model with our script
         image = ModelImages.get_image_uri(self.sm_session.boto_region_name, self.training_image)
 
-        # Use GPU instance for ChemProp/PyTorch, CPU for others
+        # Use GPU instance for ChemProp/PyTorch
         if self.model_framework in [ModelFramework.CHEMPROP, ModelFramework.PYTORCH]:
-            train_instance_type = "ml.g6.xlarge"  # NVIDIA L4 GPU, ~$0.80/hr
+            train_instance_type = "ml.g6.xlarge"  # NVIDIA L4 GPU, ~$1.00/hr
             self.log.important(f"Using GPU instance {train_instance_type} for {self.model_framework.value}")
         else:
             train_instance_type = "ml.m5.xlarge"

workbench/core/transforms/model_to_endpoint/model_to_endpoint.py

@@ -108,7 +108,6 @@ class ModelToEndpoint(Transform):
 
         # Is this a serverless deployment?
         serverless_config = None
-        instance_type = None
         if self.serverless:
             # For PyTorch or ChemProp we need at least 4GB of memory
             if needs_more_resources and mem_size < 4096:
@@ -126,7 +125,7 @@ class ModelToEndpoint(Transform):
                 instance_type = self.instance
                 self.log.important(f"Realtime Endpoint: Using specified instance type: {instance_type}")
             elif needs_more_resources:
-                instance_type = "ml.c7i.xlarge"
+                instance_type = "ml.c7i.large"
                 self.log.important(f"{workbench_model.model_framework} needs more resources (using {instance_type})")
             else:
                 instance_type = "ml.t2.medium"

workbench/model_script_utils/pytorch_utils.py

@@ -266,8 +266,8 @@ def train_model(
         train_dataset = TensorDataset(train_x_cont, dummy_cat, train_y)
         val_dataset = TensorDataset(val_x_cont, dummy_val_cat, val_y)
 
-    train_loader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True)
-    val_loader = DataLoader(val_dataset, batch_size=batch_size, shuffle=False)
+    train_loader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True, pin_memory=True)
+    val_loader = DataLoader(val_dataset, batch_size=batch_size, shuffle=False, pin_memory=True)
 
     # Loss and optimizer
     if task == "classification":

workbench/model_scripts/chemprop/chemprop.template

@@ -34,7 +34,7 @@ DEFAULT_HYPERPARAMETERS = {
     "n_folds": 5,
     "max_epochs": 400,
     "patience": 50,
-    "batch_size": 32,
+    "batch_size": 64,
     # Message Passing (ignored when using foundation model)
     "hidden_dim": 700,
     "depth": 6,
@@ -220,7 +220,9 @@ def predict_fn(df: pd.DataFrame, model_dict: dict) -> pd.DataFrame:
         return df
 
     dataset = data.MoleculeDataset(datapoints)
-    dataloader = data.build_dataloader(dataset, shuffle=False, batch_size=64, drop_last=False)
+    # Note: Use dataset length as batch_size to prevent ChemProp's build_dataloader from
+    # dropping single-sample batches (its drop_last logic triggers when len(dataset) % batch_size == 1)
+    dataloader = data.build_dataloader(dataset, shuffle=False, batch_size=len(dataset))
 
     # Ensemble predictions using direct PyTorch inference (no Lightning Trainer)
     all_preds = []
@@ -588,8 +590,9 @@ if __name__ == "__main__":
             val_dataset.normalize_targets(target_scaler)
             output_transform = nn.UnscaleTransform.from_standard_scaler(target_scaler)
 
-        train_loader = data.build_dataloader(train_dataset, batch_size=batch_size, shuffle=True, num_workers=3)
-        val_loader = data.build_dataloader(val_dataset, batch_size=batch_size, shuffle=False, num_workers=3)
+        num_workers = min(os.cpu_count() or 4, 8)  # Scale with CPUs, cap at 8
+        train_loader = data.build_dataloader(train_dataset, batch_size=batch_size, shuffle=True, num_workers=num_workers, persistent_workers=num_workers > 0, pin_memory=True, prefetch_factor=2)
+        val_loader = data.build_dataloader(val_dataset, batch_size=batch_size, shuffle=False, num_workers=num_workers, persistent_workers=num_workers > 0, pin_memory=True, prefetch_factor=2)
 
         # Build model
         pl.seed_everything(hyperparameters["seed"] + fold_idx)
@@ -615,7 +618,7 @@ if __name__ == "__main__":
                 callbacks.append(pl.callbacks.ModelCheckpoint(
                     dirpath=args.model_dir, filename=f"best_{fold_idx}", monitor="val_loss", mode="min", save_top_k=1
                 ))
-            return pl.Trainer(accelerator="auto", max_epochs=max_epochs, logger=False, enable_progress_bar=True, callbacks=callbacks)
+            return pl.Trainer(accelerator="auto", max_epochs=max_epochs, precision="16-mixed", logger=False, enable_progress_bar=True, callbacks=callbacks)
 
         if use_two_phase:
             # Phase 1: Freeze MPNN, train FFN only

workbench/model_scripts/chemprop/generated_model_script.py

@@ -45,10 +45,10 @@ DEFAULT_HYPERPARAMETERS = {
     # Loss function for regression (mae, mse)
     "criterion": "mae",
     # Split strategy: "random", "scaffold", or "butina"
-    # - random: Standard random split
+    # - random: Standard random split (default)
     # - scaffold: Bemis-Murcko scaffold-based grouping
     # - butina: Morgan fingerprint clustering (recommended for ADMET)
-    "split_strategy": "butina",
+    "split_strategy": "random",
     "butina_cutoff": 0.4,  # Tanimoto distance cutoff for Butina clustering
     # Random seed
     "seed": 42,
@@ -65,11 +65,11 @@ DEFAULT_HYPERPARAMETERS = {
 # Template parameters (filled in by Workbench)
 TEMPLATE_PARAMS = {
     "model_type": "uq_regressor",
-    "targets": ['logd'],
+    "targets": ['udm_asy_res_extraction_percent'],
     "feature_list": ['smiles'],
-    "id_column": "molecule_name",
-    "model_metrics_s3_path": "s3://sandbox-sageworks-artifacts/models/logd-chemprop-split-butina/training",
-    "hyperparameters": {'split_strategy': 'butina'},
+    "id_column": "udm_mol_bat_id",
+    "model_metrics_s3_path": "s3://idb-prod-sageworks-artifacts/models/hlm-extraction-reg-chemprop-1-260128/training",
+    "hyperparameters": {},
 }
 
 
@@ -220,7 +220,9 @@ def predict_fn(df: pd.DataFrame, model_dict: dict) -> pd.DataFrame:
         return df
 
     dataset = data.MoleculeDataset(datapoints)
-    dataloader = data.build_dataloader(dataset, shuffle=False, batch_size=64)
+    # Note: Use dataset length as batch_size to prevent ChemProp's build_dataloader from
+    # dropping single-sample batches (its drop_last logic triggers when len(dataset) % batch_size == 1)
+    dataloader = data.build_dataloader(dataset, shuffle=False, batch_size=len(dataset))
 
     # Ensemble predictions using direct PyTorch inference (no Lightning Trainer)
     all_preds = []
@@ -752,6 +754,7 @@ if __name__ == "__main__":
     output_columns += [f"{t}_pred" for t in target_columns] + [f"{t}_pred_std" for t in target_columns]
     output_columns += ["prediction", "prediction_std", "confidence"]
     output_columns += [c for c in df_val.columns if c.endswith("_proba")]
+
     output_columns = [c for c in output_columns if c in df_val.columns]
 
     wr.s3.to_csv(df_val[output_columns], f"{model_metrics_s3_path}/validation_predictions.csv", index=False)

workbench/model_scripts/pytorch_model/pytorch.template

@@ -45,7 +45,7 @@ DEFAULT_HYPERPARAMETERS = {
     "n_folds": 5,
     "max_epochs": 200,
     "early_stopping_patience": 30,
-    "batch_size": 128,
+    "batch_size": 64,
     # Model architecture (larger capacity - ensemble provides regularization)
     "layers": "512-256-128",
     "learning_rate": 1e-3,

workbench/model_scripts/pytorch_model/pytorch_utils.py

@@ -266,8 +266,8 @@ def train_model(
         train_dataset = TensorDataset(train_x_cont, dummy_cat, train_y)
         val_dataset = TensorDataset(val_x_cont, dummy_val_cat, val_y)
 
-    train_loader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True)
-    val_loader = DataLoader(val_dataset, batch_size=batch_size, shuffle=False)
+    train_loader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True, pin_memory=True)
+    val_loader = DataLoader(val_dataset, batch_size=batch_size, shuffle=False, pin_memory=True)
 
     # Loss and optimizer
     if task == "classification":

{workbench-0.8.239.dist-info → workbench-0.8.243.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: workbench
-Version: 0.8.239
+Version: 0.8.243
 Summary: Workbench: A Dashboard and Python API for creating and deploying AWS SageMaker Model Pipelines
 Author-email: SuperCowPowers LLC <support@supercowpowers.com>
 License: MIT License

{workbench-0.8.239.dist-info → workbench-0.8.243.dist-info}/RECORD

@@ -60,7 +60,7 @@ workbench/core/artifacts/data_capture_core.py,sha256=q8f79rRTYiZ7T4IQRWXl8ZvPpcv
 workbench/core/artifacts/data_source_abstract.py,sha256=5IRCzFVK-17cd4NXPMRfx99vQAmQ0WHE5jcm5RfsVTg,10619
 workbench/core/artifacts/data_source_factory.py,sha256=YL_tA5fsgubbB3dPF6T4tO0rGgz-6oo3ge4i_YXVC-M,2380
 workbench/core/artifacts/df_store_core.py,sha256=AueNr_JvuLLu_ByE7cb3u-isH9u0Q7cMP-UCgCX-Ctg,3536
-workbench/core/artifacts/endpoint_core.py,sha256=e9Fs07D2SXvLabaywTndX8R1iyO-WjHoNW4A80UUiSs,55694
+workbench/core/artifacts/endpoint_core.py,sha256=hEjMXag9nsLGu_eOLxPSSqcu1aP_1Vo9DHIGemLutvc,55695
 workbench/core/artifacts/feature_set_core.py,sha256=IjSUpxpj2S611uo5LmnOK-aH3CZhfbC5ztC02PQ5gqE,42128
 workbench/core/artifacts/model_core.py,sha256=wPkpdRlxnAXMqsDtJGPotGFO146Hm7NCfYbImHwZo9c,52343
 workbench/core/artifacts/monitor_core.py,sha256=M307yz7tEzOEHgv-LmtVy9jKjSbM98fHW3ckmNYrwlU,27897
@@ -107,9 +107,9 @@ workbench/core/transforms/features_to_features/__init__.py,sha256=47DEQpj8HBSa-_
 workbench/core/transforms/features_to_features/heavy/emr/Readme.md,sha256=YtQgCEQeKe0CQXQkhzMTYq9xOtCsCYb5P5LW2BmRKWQ,68
 workbench/core/transforms/features_to_features/heavy/glue/Readme.md,sha256=TuyCatWfoDr99zUwvOcxf-TqMkQzaMqXlj5nmFcRzfo,48
 workbench/core/transforms/features_to_model/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-workbench/core/transforms/features_to_model/features_to_model.py,sha256=stTOKAh_OJaI4ao6G8GRECa78sViaJXBzwt9myK5joM,20892
+workbench/core/transforms/features_to_model/features_to_model.py,sha256=pYGdfnp-6xh79kxQ5iXySi7oYcaRuQ-xLDo1rFgDB7g,20876
 workbench/core/transforms/model_to_endpoint/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-workbench/core/transforms/model_to_endpoint/model_to_endpoint.py,sha256=PGDrpWvQMhMTgLGZQ6bw2blsmWgMAXyIIL0NKt3E09A,8853
+workbench/core/transforms/model_to_endpoint/model_to_endpoint.py,sha256=I44_ziQ0IegudLQ_qJ-XNfWZInDkXWI9LsE-1o9855w,8823
 workbench/core/transforms/pandas_transforms/__init__.py,sha256=xL4MT8-fZ1SFqDbTLc8XyxjupHtB1YR6Ej0AC2nwd7I,894
 workbench/core/transforms/pandas_transforms/data_to_pandas.py,sha256=sJHPeuNF8Q8aQqgRnkdWkyvur5cbggdUVIwR-xF3Dlo,3621
 workbench/core/transforms/pandas_transforms/features_to_pandas.py,sha256=af6xdPt2V4zhh-SzQa_UYxdmNMzMLXbrbsznV5QoIJg,3441
@@ -128,11 +128,11 @@ workbench/core/views/view.py,sha256=DvmEA1xdvL980GET_cnbmHzqSy6IhlNaZcoQnVTtYis,
 workbench/core/views/view_utils.py,sha256=CwOlpqXpumCr6REi-ey7Qjz5_tpg-s4oWHmlOVu8POQ,12270
 workbench/core/views/storage/mdq_view.py,sha256=qf_ep1KwaXOIfO930laEwNIiCYP7VNOqjE3VdHfopRE,5195
 workbench/model_script_utils/model_script_utils.py,sha256=aM3ZaJxyMy7smokIF83fXUx3YSzLs8BNNMLfJDCoe8I,21231
-workbench/model_script_utils/pytorch_utils.py,sha256=vr8ybK45U0H8Jhjb5qx6xbJNozdcl7bVqubknDwh6U0,13704
+workbench/model_script_utils/pytorch_utils.py,sha256=kQCTRqdbszlurMrzyflyOo2amDJYx3Pni1rRhGHWXm4,13738
 workbench/model_script_utils/uq_harness.py,sha256=Qv5UQdjn72Ssa3NWGGsnSB_wDp0au2TXVauFK81Ebr0,11498
 workbench/model_scripts/script_generation.py,sha256=Sv0OJdASNKk1KXr8goiZWUL5W7i8G8gBb_R_OTb8caI,8257
-workbench/model_scripts/chemprop/chemprop.template,sha256=WtXYuC5NB2kY1nPRaS1VvaE72i6U2dIh6KZXIkT9t-o,36490
-workbench/model_scripts/chemprop/generated_model_script.py,sha256=awO8O1Arbpct8c3QoUjABWQ2ZbVus-ie8dNRLo1UiD4,36498
+workbench/model_scripts/chemprop/chemprop.template,sha256=otuR2Ee-GogsNo4z1MlefXY9G--ZOTgg4rFc_5NXivw,36941
+workbench/model_scripts/chemprop/generated_model_script.py,sha256=6duTkJUH1eRrsGHAZN1DWRKR74K5tsXKcQPrWd3vjxQ,36724
 workbench/model_scripts/chemprop/model_script_utils.py,sha256=aM3ZaJxyMy7smokIF83fXUx3YSzLs8BNNMLfJDCoe8I,21231
 workbench/model_scripts/chemprop/requirements.txt,sha256=2IBHZZNYqhX9Ed7AmRVgN06tO3EHeBbN2EM8-tjWZhs,216
 workbench/model_scripts/custom_models/chem_info/Readme.md,sha256=mH1lxJ4Pb7F5nBnVXaiuxpi8zS_yjUw_LBJepVKXhlA,574
@@ -162,8 +162,8 @@ workbench/model_scripts/meta_model/generated_model_script.py,sha256=ncPrHd9-R8l_
 workbench/model_scripts/meta_model/meta_model.template,sha256=viz-AKVq3YRwOUBt8-rUO1TwdEPFzyP7nnifqcIJurw,8244
 workbench/model_scripts/pytorch_model/generated_model_script.py,sha256=1B4RortOxbB7feTrr5Kf9qUqdqG4Qc1a6evdNUYLSNg,27011
 workbench/model_scripts/pytorch_model/model_script_utils.py,sha256=aM3ZaJxyMy7smokIF83fXUx3YSzLs8BNNMLfJDCoe8I,21231
-workbench/model_scripts/pytorch_model/pytorch.template,sha256=FZYI4D-u5lDkJSyvgJYVhtvt9PnfL_pEVGtBYv64sNU,22767
-workbench/model_scripts/pytorch_model/pytorch_utils.py,sha256=vr8ybK45U0H8Jhjb5qx6xbJNozdcl7bVqubknDwh6U0,13704
+workbench/model_scripts/pytorch_model/pytorch.template,sha256=78TBsT1NoPkVL-cINZMjA1SE91abUgKtNqedOCvS7lU,22766
+workbench/model_scripts/pytorch_model/pytorch_utils.py,sha256=kQCTRqdbszlurMrzyflyOo2amDJYx3Pni1rRhGHWXm4,13738
 workbench/model_scripts/pytorch_model/requirements.txt,sha256=ES7YehHEL4E5oV8FScHm3oNQmkMI4ODgbC1fSbaY7T4,183
 workbench/model_scripts/pytorch_model/uq_harness.py,sha256=Qv5UQdjn72Ssa3NWGGsnSB_wDp0au2TXVauFK81Ebr0,11498
 workbench/model_scripts/scikit_learn/generated_model_script.py,sha256=xhQIglpAgPRCH9iwI3wI0N0V6p9AgqW0mVOMuSXzUCk,17187
@@ -305,9 +305,9 @@ workbench/web_interface/page_views/main_page.py,sha256=DyChwOGX_KtbJ09pw2Iswofba
 workbench/web_interface/page_views/models_page_view.py,sha256=M0bdC7bAzLyIaE2jviY12FF4abdMFZmg6sFuOY_LaGI,2650
 workbench/web_interface/page_views/page_view.py,sha256=Gh6YnpOGlUejx-bHZAf5pzqoQ1H1R0OSwOpGhOBO06w,455
 workbench/web_interface/page_views/pipelines_page_view.py,sha256=v2pxrIbsHBcYiblfius3JK766NZ7ciD2yPx0t3E5IJo,2656
-workbench-0.8.239.dist-info/licenses/LICENSE,sha256=RTBoTMeEwTgEhS-n8vgQ-VUo5qig0PWVd8xFPKU6Lck,1080
-workbench-0.8.239.dist-info/METADATA,sha256=fPofZrAWmIGGYOLeDe95A7yMaMplxZE-gK5Zzi22AQQ,10038
-workbench-0.8.239.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
-workbench-0.8.239.dist-info/entry_points.txt,sha256=Stivs_FFse2pHLXfWNpyh649z0bj7Ks5laQy8LuexCA,633
-workbench-0.8.239.dist-info/top_level.txt,sha256=Dhy72zTxaA_o_yRkPZx5zw-fwumnjGaeGf0hBN3jc_w,10
-workbench-0.8.239.dist-info/RECORD,,
+workbench-0.8.243.dist-info/licenses/LICENSE,sha256=RTBoTMeEwTgEhS-n8vgQ-VUo5qig0PWVd8xFPKU6Lck,1080
+workbench-0.8.243.dist-info/METADATA,sha256=mWeiFAV-J1TZKygpTEymX4gPhl2dFfscZKPOFitOMFo,10038
+workbench-0.8.243.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
+workbench-0.8.243.dist-info/entry_points.txt,sha256=Stivs_FFse2pHLXfWNpyh649z0bj7Ks5laQy8LuexCA,633
+workbench-0.8.243.dist-info/top_level.txt,sha256=Dhy72zTxaA_o_yRkPZx5zw-fwumnjGaeGf0hBN3jc_w,10
+workbench-0.8.243.dist-info/RECORD,,