Merge pull request #79 from VIDA-NYU/update_py311

Update py311

.github/workflows/build.yml

@@ -8,7 +8,7 @@ jobs:
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        python-version: ["3.8"]
+        python-version: ["3.8", "3.11"]
 
     steps:
       - uses: actions/checkout@v3

Dockerfile

@@ -1,11 +1,11 @@
-FROM python:3.10.11 AS alpha-automl-jupyterhub
+FROM python:3.11.6 AS alpha-automl-jupyterhub
 
 # Install JupyterHub and dependencies
 RUN pip3 --disable-pip-version-check install --no-cache-dir \
-    jupyterhub==3.1.1 \
-    notebook==6.5.2 \
-    jupyterlab==3.5.3 \
-    jupyterlab-server==2.16.0
+    jupyterhub==4.0.2 \
+    notebook==7.0.6 \
+    jupyterlab==4.0.8 \
+    jupyterlab-server==2.25.0
 
 # Install AlphaD3M and dependencies
 ADD . /alpha-automl/

alpha_automl/builtin_primitives/time_series_forecasting.py

@@ -88,7 +88,7 @@ def __init__(self):
             "sin_day",
             "cos_day",
             "dayofmonth",
-            "weekofyear",
+            # weekofyear",
         ]
         self.freq = "d"
         self.prediction_length = 1
@@ -156,7 +156,7 @@ def create_time_features(self, df, date_col=None):
         df["sin_day"] = np.sin(df["dayofyear"])
         df["cos_day"] = np.cos(df["dayofyear"])
         df["dayofmonth"] = df["date"].dt.day
-        df["weekofyear"] = df["date"].dt.weekofyear
+        # df["weekofyear"] = df["date"].dt.weekofyear # Not available after pandas 1.6
         df = df.drop(["date"], axis=1)
         return df
 
@@ -170,16 +170,16 @@ def standard_scaling(self, df):
 
 # NBEATS
 class NBEATSEstimator(BasePrimitive):
-    def __init__(self, freq="d", max_epochs=10):
+    def __init__(self, freq="d", max_steps=10):
         self.freq = freq
-        self.max_epochs = max_epochs
+        self.max_steps = max_steps
 
     def fit(self, X, y):
         horizon = len(y)
         date_column = X.columns[0]
         X[date_column] = X[date_column].apply(lambda x: pd.Timestamp(x))
         nbeats = [
-            NBEATS(input_size=horizon // 2, h=horizon, max_epochs=self.max_epochs)
+            NBEATS(input_size=horizon // 2, h=horizon, max_steps=self.max_steps)
         ]
         self.nbeats_nf = NeuralForecast(models=nbeats, freq=self.freq)
 
@@ -195,15 +195,15 @@ def predict(self, X):
 
 # NHITS
 class NHITSEstimator(BasePrimitive):
-    def __init__(self, freq="d", max_epochs=10):
+    def __init__(self, freq="d", max_steps=10):
         self.freq = freq
-        self.max_epochs = max_epochs
+        self.max_steps = max_steps
 
     def fit(self, X, y):
         horizon = len(y)
         date_column = X.columns[0]
         X[date_column] = X[date_column].apply(lambda x: pd.Timestamp(x))
-        nhits = [NHITS(input_size=horizon // 2, h=horizon, max_epochs=self.max_epochs)]
+        nhits = [NHITS(input_size=horizon // 2, h=horizon, max_steps=self.max_steps)]
         self.nhits_nf = NeuralForecast(models=nhits, freq=self.freq)
 
         tmp_X = X.copy()

extra_requirements.txt

@@ -1,9 +1,9 @@
-pandas<2.0.0:               timeseries
-gluonts==0.12.7:            timeseries
+pandas:                     timeseries
+gluonts:                    timeseries
 numpy==1.23.5:              timeseries
-neuralforecast==1.5.0:      timeseries
-mxnet==1.9.1:               timeseries
-pmdarima==2.0.3:            timeseries
+neuralforecast:             timeseries
+mxnet:                      timeseries
+pmdarima:                   timeseries
 fasttext-wheel:             nlp
 transformers:               nlp, image
 scikit-image:               image

requirements.txt

@@ -1,4 +1,4 @@
-scipy<=1.9.3
+scipy
 scikit-learn
 torch>=1.7
 nltk
@@ -8,4 +8,4 @@ datamart_profiler
 feature_engine
 xgboost
 lightgbm
-numpy<=1.24.3
+numpy

tests/test_datetime_encoder.py

@@ -5,13 +5,16 @@
 
 class TestDatetimeEncoder:
     """This is the testcases for datetime encoder."""
-
-    df = pd.DataFrame(
-        data={
-            "col1": ["12/1/2009 23:19", "2022-06-30 18:37:00", "2025-09-24 20:37:00"]
-        },
-        index=[0, 1, 2],
-    )
+    df = None
+
+    def setup_method(self, method):
+        print("starting execution of tc: {}".format(method.__name__))
+        self.df = pd.DataFrame(
+            data={
+                "col1": ["12/1/2009 23:19", "2022-06-30 18:37:00", "2025-09-24 20:37:00"]
+            },
+            index=[0, 1, 2],
+        )
 
     def test_cyclical_feature(self):
         encoder = CyclicalFeature()
@@ -27,5 +30,4 @@ def test_dummy_encoder(self):
         encoder = DummyEncoder()
         encoder.fit(self.df)
         np_array = encoder.transform(self.df)
-        print(np_array)
-        assert np_array.shape == (len(self.df), 4 * len(self.df) + 4)
+        assert np_array.shape == (len(self.df), 4 * len(self.df))