Temporarily rename "TFT" resources for grepability

applications/models/src/models/train_tft_model.py

@@ -7,8 +7,8 @@
 import wandb
 from wandb import Run
 
-from internal.dataset import TemporalFusionTransformerDataset
-from internal.tft_model import Parameters, TemporalFusionTransformer
+from internal.tft_dataset import TFTDataset
+from internal.tft_model import Parameters, TFTModel
 
 configuration = {
     "architecture": "TFT",
@@ -21,7 +21,7 @@
 
 
 @task
-def read_local_data(filepath: str) -> TemporalFusionTransformerDataset:
+def read_local_data(filepath: str) -> TFTDataset:
     start_time = datetime.now(tz=timezone)
     logger.info(f"Reading data from {filepath}")
     data = pl.read_csv(filepath)
@@ -30,17 +30,17 @@ def read_local_data(filepath: str) -> TemporalFusionTransformerDataset:
 
     logger.info(f"Data read successfully in {runtime_seconds} seconds")
 
-    return TemporalFusionTransformerDataset(data=data)
+    return TFTDataset(data=data)
 
 
 @task
 def train_model(
-    dataset: TemporalFusionTransformerDataset,
+    dataset: TFTDataset,
     wandb_run: Run,
     validation_split: float = 0.8,
     epoch_count: int = 10,
     learning_rate: float = 1e-3,
-) -> TemporalFusionTransformer:
+) -> TFTModel:
     start_time = datetime.now(tz=timezone)
     logger.info("Training temporal fusion transformer model")
     dimensions = dataset.get_dimensions()
@@ -62,7 +62,7 @@ def train_model(
         output_length=7,
     )
 
-    model = TemporalFusionTransformer(parameters=parameters)
+    model = TFTModel(parameters=parameters)
 
     batches = dataset.get_batches(
         data_type="train",
@@ -93,8 +93,8 @@ def train_model(
 
 @task
 def validate_model(
-    data: TemporalFusionTransformerDataset,
-    model: TemporalFusionTransformer,
+    data: TFTDataset,
+    model: TFTModel,
     validation_split: float = 0.8,
 ) -> None:
     start_time = datetime.now(tz=timezone)
@@ -119,7 +119,7 @@ def validate_model(
 
 
 @task
-def save_model(model: TemporalFusionTransformer) -> None:
+def save_model(model: TFTModel) -> None:
     start_time = datetime.now(tz=timezone)
     logger.info("Saving temporal fusion transformer model")
 

libraries/python/src/internal/dataset.py → libraries/python/src/internal/tft_dataset.py

@@ -25,7 +25,9 @@ def inverse_transform(self, data: pl.DataFrame) -> pl.DataFrame:
         return data * self.standard_deviations + self.means
 
 
-class TemporalFusionTransformerDataset:
+class TFTDataset:
+    """Temporal fusion transformer dataset."""
+
     def __init__(self, data: pl.DataFrame) -> None:
         raw_columns = (
             "ticker",

libraries/python/src/internal/tft_model.py

@@ -34,7 +34,7 @@ class Parameters(BaseModel):
 
 
 # https://arxiv.org/pdf/1912.09363
-class TemporalFusionTransformer:
+class TFTModel:
     def __init__(self, parameters: Parameters) -> None:
         self.hidden_size = parameters.hidden_size
         self.batch_size = parameters.input_length

libraries/python/tests/test_dataset.py → libraries/python/tests/test_tft_dataset.py

@@ -1,8 +1,8 @@
 import polars as pl
-from internal.dataset import TemporalFusionTransformerDataset
+from internal.tft_dataset import TFTDataset
 
 
-def test_dataset_load_data() -> None:
+def test_tft_dataset_load_data() -> None:
     data = pl.DataFrame(
         {
             "timestamp": [
@@ -39,13 +39,13 @@ def test_dataset_load_data() -> None:
         }
     )
 
-    dataset = TemporalFusionTransformerDataset(data=data)
+    dataset = TFTDataset(data=data)
 
     assert hasattr(dataset, "data")
     assert hasattr(dataset, "mappings")
 
 
-def test_dataset_get_dimensions() -> None:
+def test_tft_dataset_get_dimensions() -> None:
     data = pl.DataFrame(
         {
             "timestamp": [
@@ -64,7 +64,7 @@ def test_dataset_get_dimensions() -> None:
         }
     )
 
-    dataset = TemporalFusionTransformerDataset(data=data)
+    dataset = TFTDataset(data=data)
 
     dimensions = dataset.get_dimensions()
 
@@ -76,7 +76,7 @@ def test_dataset_get_dimensions() -> None:
     assert "static_continuous_features" in dimensions
 
 
-def test_dataset_batches() -> None:
+def test_tft_dataset_batches() -> None:
     data = pl.DataFrame(
         {
             "timestamp": [
@@ -100,7 +100,7 @@ def test_dataset_batches() -> None:
         }
     )
 
-    dataset = TemporalFusionTransformerDataset(data=data)
+    dataset = TFTDataset(data=data)
 
     expected_input_length = 2
     expected_output_length = 1