FEAT: Add dbinfer DFS Feature Transformer

config/config.yaml

@@ -8,6 +8,12 @@ feature_transformers:
   - _target_: autogluon_assistant.transformer.OpenFETransformer
     n_jobs: 1
     num_features_to_keep: 10
+  - _target_: autogluon_assistant.transformer.DFSTransformer
+    depth: 2
+    max_fks: 10
+    categorical_var_cardinality_upper: 20
+    min_fk_cardinality: 2
+
 autogluon:
   predictor_init_kwargs: {}
   predictor_fit_kwargs:

pyproject.toml

@@ -9,7 +9,7 @@ description = "An LLM-backed AI agent for competitive tabular machine learning"
 readme = "README.md"
 requires-python = ">=3.8"
 dependencies = [
-    "autogluon.tabular[all]>=1.1",
+    "autogluon.tabular[all]==1.1",
     "hydra-core>=1.3",
     "importlib-resources",
     "langchain~=0.1.15",

src/autogluon_assistant/assistant.py

@@ -65,7 +65,6 @@ def preprocess_task(self, task: TabularPredictionTask) -> TabularPredictionTask:
                 FilenameInferenceTransformer(llm=self.llm),
                 LabelColumnInferenceTransformer(llm=self.llm),
                 TestIdColumnTransformer(llm=self.llm),
-                TrainIdColumnDropTransformer(llm=self.llm),
                 ProblemTypeInferenceTransformer(),
             ]
             + ([EvalMetricInferenceTransformer(llm=self.llm)] if self.config.infer_eval_metric else [])
@@ -74,6 +73,8 @@ def preprocess_task(self, task: TabularPredictionTask) -> TabularPredictionTask:
                 if self.feature_transformers_config
                 else []
             )
+            # Avoid dropping Train ID Column before DFS transformer; Drop at last
+            + [TrainIdColumnDropTransformer(llm=self.llm)]
         )
         for transformer in task_preprocessors:
             try:

src/autogluon_assistant/transformer/__init__.py

@@ -1,5 +1,5 @@
 from .base import BaseTransformer, TransformTimeoutError
-from .feature_transformers import BaseFeatureTransformer, CAAFETransformer, OpenFETransformer
+from .feature_transformers import BaseFeatureTransformer, CAAFETransformer, OpenFETransformer, DFSTransformer
 from .task_inference import (
     EvalMetricInferenceTransformer,
     FilenameInferenceTransformer,
@@ -13,6 +13,7 @@
     "BaseTransformer",
     "BaseFeatureTransformer",
     "CAAFETransformer",
+    "DFSTransformer",
     "EvalMetricInferenceTransformer",
     "FilenameInferenceTransformer",
     "LabelColumnInferenceTransformer",

src/autogluon_assistant/transformer/feature_transformers/__init__.py

@@ -1,9 +1,11 @@
 from .base import BaseFeatureTransformer
 from .caafe import CAAFETransformer
 from .openfe import OpenFETransformer
+from .dfs import DFSTransformer
 
 __all__ = [
     "BaseFeatureTransformer",
     "CAAFETransformer",
     "OpenFETransformer",
+    "DFSTransformer",
 ]