workbench 0.8.238__tar.gz → 0.8.241__tar.gz

{workbench-0.8.238 → workbench-0.8.241}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: workbench
-Version: 0.8.238
+Version: 0.8.241
 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-0.8.238 → workbench-0.8.241}/aws_setup/workbench_compute/workbench_compute/lambdas/batch_failure/handler.py

@@ -97,7 +97,7 @@ def lambda_handler(event, context):
     # Fetch CloudWatch logs
     log_messages = []
     if log_stream_name:
-        log_messages = get_log_events(log_stream_name)
+        log_messages = get_log_events(log_stream_name, limit=200)
 
     # Build the console URL
     console_url = f"https://{region}.console.aws.amazon.com/batch/home?region={region}" f"#jobs/fargate/detail/{job_id}"

{workbench-0.8.238 → workbench-0.8.241}/model_docker_images/ml_pipelines/Dockerfile

@@ -29,7 +29,7 @@ COPY --from=builder /usr/local/lib/python3.12/site-packages /usr/local/lib/pytho
 COPY --from=builder /usr/local/bin /usr/local/bin
 
 # Install workbench only
-RUN pip install --no-cache-dir workbench==0.8.236
+RUN pip install --no-cache-dir workbench==0.8.240
 
 # Add the script runner
 COPY ml_pipeline_runner.py /app/ml_pipeline_runner.py

{workbench-0.8.238 → workbench-0.8.241}/src/workbench/core/transforms/model_to_endpoint/model_to_endpoint.py

@@ -126,7 +126,7 @@ 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-0.8.238 → workbench-0.8.241}/src/workbench/model_scripts/chemprop/chemprop.template

@@ -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 = []

{workbench-0.8.238 → workbench-0.8.241}/src/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-0.8.238 → workbench-0.8.241}/src/workbench.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: workbench
-Version: 0.8.238
+Version: 0.8.241
 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