Rename PatchTSModel to PatchTSTModel (#601)

* fix: rename patchts to patchtst * chore: update changelog * fix: update notebook output * fix: fix name im examples README --------- Co-authored-by: Egor Baturin <[email protected]>
etna-team · Feb 7, 2025 · 0a0145e · 0a0145e
1 parent ae4556f
commit 0a0145e
Show file tree

Hide file tree

Showing 9 changed files with 64 additions and 62 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -40,7 +40,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - **Breaking:** Rename `TSDataset.index` to `TSDataset.timestamps` ([#593](https://github.com/etna-team/etna/pull/593))
 - **Breaking:** Rename `TSDataset.add_columns_from_pandas` to `TSDataset.add_features_from_pandas` ([#593](https://github.com/etna-team/etna/pull/593))
 - **Breaking:** Rename `TSDataset.update_columns_from_pandas` to `TSDataset.update_features_from_pandas` ([#593](https://github.com/etna-team/etna/pull/593))
-- 
+- **Breaking:** Rename `PatchTSModel` to `PatchTSTModel` ([#601](https://github.com/etna-team/etna/pull/601))
 - 
 - 
 - Fix `TSDataset.train_test_split` to pass all features to train and test parts ([#545](https://github.com/etna-team/etna/pull/545))

diff --git a/docs/source/api_reference/models.rst b/docs/source/api_reference/models.rst
@@ -80,7 +80,7 @@ Native neural network models:
    nn.DeepStateModel
    nn.NBeatsGenericModel
    nn.NBeatsInterpretableModel
-   nn.PatchTSModel
+   nn.PatchTSTModel
    nn.DeepARModel
    nn.TFTModel
 

diff --git a/etna/models/nn/__init__.py b/etna/models/nn/__init__.py
@@ -6,7 +6,7 @@
     from etna.models.nn.mlp import MLPModel
     from etna.models.nn.nbeats import NBeatsGenericModel
     from etna.models.nn.nbeats import NBeatsInterpretableModel
-    from etna.models.nn.patchts import PatchTSModel
+    from etna.models.nn.patchtst import PatchTSTModel
     from etna.models.nn.rnn import RNNModel
     from etna.models.nn.tft import TFTModel
 

diff --git a/etna/models/nn/patchts.py → etna/models/nn/patchtst.py b/etna/models/nn/patchts.py → etna/models/nn/patchtst.py
@@ -20,8 +20,8 @@
     import torch.nn as nn
 
 
-class PatchTSBatch(TypedDict):
-    """Batch specification for PatchTS."""
+class PatchTSTBatch(TypedDict):
+    """Batch specification for PatchTST."""
 
     encoder_target: "torch.Tensor"
     decoder_target: "torch.Tensor"
@@ -52,8 +52,8 @@ def forward(self, x: torch.Tensor) -> torch.Tensor:
         return self.dropout(x)
 
 
-class PatchTSNet(DeepBaseNet):
-    """PatchTS based Lightning module."""
+class PatchTSTNet(DeepBaseNet):
+    """PatchTST based Lightning module."""
 
     def __init__(
         self,
@@ -68,7 +68,7 @@ def __init__(
         loss: "torch.nn.Module",
         optimizer_params: Optional[dict],
     ) -> None:
-        """Init PatchTS.
+        """Init PatchTST.
 
         Parameters
         ----------
@@ -120,7 +120,7 @@ def __init__(
         self.lr = lr
         self.optimizer_params = {} if optimizer_params is None else optimizer_params
 
-    def forward(self, x: PatchTSBatch, *args, **kwargs):  # type: ignore
+    def forward(self, x: PatchTSTBatch, *args, **kwargs):  # type: ignore
         """Forward pass.
 
         Parameters
@@ -157,7 +157,7 @@ def _get_prediction(self, x: torch.Tensor) -> torch.Tensor:
 
         return self.projection(y)  # (batch_size, 1)
 
-    def step(self, batch: PatchTSBatch, *args, **kwargs):  # type: ignore
+    def step(self, batch: PatchTSTBatch, *args, **kwargs):  # type: ignore
         """Step for loss computation for training or validation.
 
         Parameters
@@ -239,8 +239,8 @@ def configure_optimizers(self) -> "torch.optim.Optimizer":
         return optimizer
 
 
-class PatchTSModel(DeepBaseModel):
-    """PatchTS model using PyTorch layers. For more details read the `paper <https://arxiv.org/abs/2211.14730>`_.
+class PatchTSTModel(DeepBaseModel):
+    """PatchTST model using PyTorch layers. For more details read the `paper <https://arxiv.org/abs/2211.14730>`_.
 
     Model uses only `target` column, other columns will be ignored.
 
@@ -271,7 +271,7 @@ def __init__(
         val_dataloader_params: Optional[dict] = None,
         split_params: Optional[dict] = None,
     ):
-        """Init PatchTS model.
+        """Init PatchTST model.
 
         Parameters
         ----------
@@ -327,7 +327,7 @@ def __init__(
         self.loss = loss if loss is not None else nn.MSELoss()
         self.optimizer_params = optimizer_params
         super().__init__(
-            net=PatchTSNet(
+            net=PatchTSTNet(
                 encoder_length,
                 patch_len=self.patch_len,
                 stride=self.stride,

diff --git a/examples/202-NN_examples.ipynb b/examples/202-NN_examples.ipynb
@@ -36,7 +36,7 @@
     "    * [MLP](#section_2_5)\n",
     "    * [Deep State Model](#section_2_6)\n",
     "    * [N-BEATS Model](#section_2_7)\n",
-    "    * [PatchTS Model](#section_2_8)\n",
+    "    * [PatchTST Model](#section_2_8)\n",
     "    * [Chronos Model](#section_2_9)\n",
     "    * [Chronos Bolt Model](#section_2_10)\n",
     "    * [TimesFM Model](#section_2_11)"
@@ -1043,7 +1043,7 @@
     }
    },
    "source": [
-    "### 2.3 TFTNative <a class=\"anchor\" id=\"section_2_3\"></a>"
+    "### 2.3 TFT <a class=\"anchor\" id=\"section_2_3\"></a>"
    ]
   },
   {
@@ -1152,7 +1152,6 @@
     "pycharm": {
      "name": "#%%\n"
     },
-    "scrolled": false,
     "tags": []
    },
    "outputs": [
@@ -2784,7 +2783,7 @@
     }
    },
    "source": [
-    "### 2.8 PatchTS Model <a class=\"anchor\" id=\"section_2_8\"></a>\n",
+    "### 2.8 PatchTST Model <a class=\"anchor\" id=\"section_2_8\"></a>\n",
     "\n",
     "Model with transformer encoder that uses patches of timeseries as input words and linear decoder."
    ]
@@ -2801,14 +2800,15 @@
    },
    "outputs": [],
    "source": [
-    "from etna.models.nn import PatchTSModel"
+    "from etna.models.nn import PatchTSTModel"
    ]
   },
   {
    "cell_type": "code",
    "execution_count": 56,
    "id": "cc38238d",
    "metadata": {
+    "collapsed": false,
     "jupyter": {
      "outputs_hidden": false
     },
@@ -2951,7 +2951,7 @@
    "source": [
     "set_seed()\n",
     "\n",
-    "model_patchts = PatchTSModel(\n",
+    "model_patchtst = PatchTSTModel(\n",
     "    decoder_length=HORIZON,\n",
     "    encoder_length=2 * HORIZON,\n",
     "    patch_len=1,\n",
@@ -2960,11 +2960,11 @@
     "    train_batch_size=64,\n",
     ")\n",
     "\n",
-    "pipeline_patchts = Pipeline(\n",
-    "    model=model_patchts, horizon=HORIZON, transforms=[StandardScalerTransform(in_column=\"target\")]\n",
+    "pipeline_patchtst = Pipeline(\n",
+    "    model=model_patchtst, horizon=HORIZON, transforms=[StandardScalerTransform(in_column=\"target\")]\n",
     ")\n",
     "\n",
-    "metrics_patchts, forecast_patchts, fold_info_patchs = pipeline_patchts.backtest(\n",
+    "metrics_patchtst, forecast_patchtst, fold_info_patchtst = pipeline_patchtst.backtest(\n",
     "    ts, metrics=metrics, n_folds=3, n_jobs=1\n",
     ")"
    ]
@@ -2974,6 +2974,7 @@
    "execution_count": 57,
    "id": "6394b96c",
    "metadata": {
+    "collapsed": false,
     "jupyter": {
      "outputs_hidden": false
     },
@@ -2987,20 +2988,21 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Average SMAPE for PatchTS: 7.583\n"
+      "Average SMAPE for PatchTST: 7.583\n"
      ]
     }
    ],
    "source": [
-    "score = metrics_patchts[\"SMAPE\"].mean()\n",
-    "print(f\"Average SMAPE for PatchTS: {score:.3f}\")"
+    "score = metrics_patchtst[\"SMAPE\"].mean()\n",
+    "print(f\"Average SMAPE for PatchTST: {score:.3f}\")"
    ]
   },
   {
    "cell_type": "code",
    "execution_count": 58,
    "id": "a514bd99",
    "metadata": {
+    "collapsed": false,
     "jupyter": {
      "outputs_hidden": false
     },
@@ -3022,7 +3024,7 @@
     }
    ],
    "source": [
-    "plot_backtest(forecast_patchts, ts, history_len=20)"
+    "plot_backtest(forecast_patchtst, ts, history_len=20)"
    ]
   },
   {

diff --git a/examples/README.md b/examples/README.md
@@ -55,13 +55,13 @@ We have prepared a set of tutorials for an easy introduction:
 - Loading dataset
 - Testing models
   - Baseline 
-  - DeepARNative
-  - TFTNative
+  - DeepAR
+  - TFT
   - RNN
   - MLP
   - Deep State Model 
   - N-BEATS Model 
-  - PatchTS Model
+  - PatchTST Model
   - Chronos Model
   - Chronos Bolt Model
   - TimesFM Model

diff --git a/tests/test_models/test_inference/test_forecast.py b/tests/test_models/test_inference/test_forecast.py
@@ -38,7 +38,7 @@
 from etna.models.nn import MLPModel
 from etna.models.nn import NBeatsGenericModel
 from etna.models.nn import NBeatsInterpretableModel
-from etna.models.nn import PatchTSModel
+from etna.models.nn import PatchTSTModel
 from etna.models.nn import RNNModel
 from etna.models.nn import TFTModel
 from etna.models.nn import TimesFMModel
@@ -136,7 +136,7 @@ def test_forecast_in_sample_full_no_target_failed_nans_sklearn(self, model, tran
                 [],
                 "example_tsds",
             ),
-            (PatchTSModel(encoder_length=7, decoder_length=7, trainer_params=dict(max_epochs=1)), [], "example_tsds"),
+            (PatchTSTModel(encoder_length=7, decoder_length=7, trainer_params=dict(max_epochs=1)), [], "example_tsds"),
             (
                 DeepStateModel(
                     ssm=CompositeSSM(seasonal_ssms=[WeeklySeasonalitySSM()]),
@@ -302,7 +302,7 @@ def test_forecast_in_sample_full_failed_nans_nn(self, model, transforms, dataset
                 [],
                 "example_tsds",
             ),
-            (PatchTSModel(encoder_length=7, decoder_length=7, trainer_params=dict(max_epochs=1)), [], "example_tsds"),
+            (PatchTSTModel(encoder_length=7, decoder_length=7, trainer_params=dict(max_epochs=1)), [], "example_tsds"),
             (
                 DeepStateModel(
                     ssm=CompositeSSM(seasonal_ssms=[WeeklySeasonalitySSM()]),
@@ -434,7 +434,7 @@ def _test_forecast_in_sample_suffix_no_target(ts, model, transforms, num_skip_po
                 [],
                 "example_tsds",
             ),
-            (PatchTSModel(encoder_length=7, decoder_length=7, trainer_params=dict(max_epochs=1)), [], "example_tsds"),
+            (PatchTSTModel(encoder_length=7, decoder_length=7, trainer_params=dict(max_epochs=1)), [], "example_tsds"),
             (
                 MLPModel(input_size=2, hidden_size=[10], decoder_length=7, trainer_params=dict(max_epochs=1)),
                 [LagTransform(in_column="target", lags=[2, 3])],
@@ -533,7 +533,7 @@ class TestForecastInSampleSuffix:
                 [],
                 "example_tsds",
             ),
-            (PatchTSModel(encoder_length=7, decoder_length=7, trainer_params=dict(max_epochs=1)), [], "example_tsds"),
+            (PatchTSTModel(encoder_length=7, decoder_length=7, trainer_params=dict(max_epochs=1)), [], "example_tsds"),
             (
                 MLPModel(input_size=2, hidden_size=[10], decoder_length=7, trainer_params=dict(max_epochs=1)),
                 [LagTransform(in_column="target", lags=[2, 3])],
@@ -651,7 +651,7 @@ def _test_forecast_out_sample(ts, model, transforms, prediction_size=5):
                 [],
                 "example_tsds",
             ),
-            (PatchTSModel(encoder_length=7, decoder_length=7, trainer_params=dict(max_epochs=1)), [], "example_tsds"),
+            (PatchTSTModel(encoder_length=7, decoder_length=7, trainer_params=dict(max_epochs=1)), [], "example_tsds"),
             (
                 MLPModel(input_size=2, hidden_size=[10], decoder_length=7, trainer_params=dict(max_epochs=1)),
                 [LagTransform(in_column="target", lags=[5, 6])],
@@ -729,7 +729,7 @@ def test_forecast_out_sample_datetime_timestamp(self, model, transforms, dataset
                 [],
                 "example_tsds",
             ),
-            (PatchTSModel(encoder_length=7, decoder_length=7, trainer_params=dict(max_epochs=1)), [], "example_tsds"),
+            (PatchTSTModel(encoder_length=7, decoder_length=7, trainer_params=dict(max_epochs=1)), [], "example_tsds"),
             (
                 MLPModel(input_size=2, hidden_size=[10], decoder_length=7, trainer_params=dict(max_epochs=1)),
                 [LagTransform(in_column="target", lags=[5, 6])],
@@ -881,7 +881,7 @@ def _test_forecast_out_sample_prefix(ts, model, transforms, full_prediction_size
                 [],
                 "example_tsds",
             ),
-            (PatchTSModel(encoder_length=7, decoder_length=7, trainer_params=dict(max_epochs=1)), [], "example_tsds"),
+            (PatchTSTModel(encoder_length=7, decoder_length=7, trainer_params=dict(max_epochs=1)), [], "example_tsds"),
             (
                 MLPModel(input_size=2, hidden_size=[10], decoder_length=7, trainer_params=dict(max_epochs=1)),
                 [LagTransform(in_column="target", lags=[5, 6])],
@@ -995,7 +995,7 @@ def _test_forecast_out_sample_suffix(ts, model, transforms, full_prediction_size
             (SeasonalMovingAverageModel(), [], "example_tsds"),
             (NaiveModel(lag=3), [], "example_tsds"),
             (DeadlineMovingAverageModel(window=1), [], "example_tsds"),
-            (PatchTSModel(encoder_length=7, decoder_length=7, trainer_params=dict(max_epochs=1)), [], "example_tsds"),
+            (PatchTSTModel(encoder_length=7, decoder_length=7, trainer_params=dict(max_epochs=1)), [], "example_tsds"),
             (
                 MLPModel(input_size=2, hidden_size=[10], decoder_length=7, trainer_params=dict(max_epochs=1)),
                 [LagTransform(in_column="target", lags=[5, 6])],
@@ -1211,7 +1211,7 @@ def _test_forecast_mixed_in_out_sample(ts, model, transforms, num_skip_points=50
                 [],
                 "example_tsds",
             ),
-            (PatchTSModel(encoder_length=7, decoder_length=7, trainer_params=dict(max_epochs=1)), [], "example_tsds"),
+            (PatchTSTModel(encoder_length=7, decoder_length=7, trainer_params=dict(max_epochs=1)), [], "example_tsds"),
             (
                 MLPModel(input_size=2, hidden_size=[10], decoder_length=7, trainer_params=dict(max_epochs=1)),
                 [LagTransform(in_column="target", lags=[5, 6])],
@@ -1343,7 +1343,7 @@ def _test_forecast_subset_segments(self, ts, model, transforms, segments, predic
                 [],
                 "example_tsds",
             ),
-            (PatchTSModel(encoder_length=7, decoder_length=7, trainer_params=dict(max_epochs=1)), [], "example_tsds"),
+            (PatchTSTModel(encoder_length=7, decoder_length=7, trainer_params=dict(max_epochs=1)), [], "example_tsds"),
             (
                 MLPModel(input_size=2, hidden_size=[10], decoder_length=7, trainer_params=dict(max_epochs=1)),
                 [LagTransform(in_column="target", lags=[5, 6])],
@@ -1455,7 +1455,7 @@ def _test_forecast_new_segments(self, ts, model, transforms, train_segments, pre
                 [],
                 "example_tsds",
             ),
-            (PatchTSModel(encoder_length=7, decoder_length=7, trainer_params=dict(max_epochs=1)), [], "example_tsds"),
+            (PatchTSTModel(encoder_length=7, decoder_length=7, trainer_params=dict(max_epochs=1)), [], "example_tsds"),
             (
                 MLPModel(input_size=2, hidden_size=[10], decoder_length=7, trainer_params=dict(max_epochs=1)),
                 [LagTransform(in_column="target", lags=[5, 6])],