tf_embedding.py

feature_columns = []
for feature_name in CATEGORICAL_COLUMNS:
    vocabulary = df[feature_name].unique()
    column = tf.feature_column.categorical_column_with_vocabulary_list(feature_name, vocabulary)
    embedding_column = tf.feature_column.embedding_column(column, dimension=len(vocabulary))
    feature_columns.append(embedding_column)

for feature_name in NUMERIC_COLUMNS:
    feature_columns.append(tf.feature_column.numeric_column(feature_name, dtype=tf.float32))
    
model = tf.keras.models.Sequential([
    tf.keras.layers.DenseFeatures(feature_columns),
    tf.keras.layers.Dense(64),
    tf.keras.layers.Dense(2,activation='sigmoid')
])

X_train, X_test, y_train, y_test = train_test_split(df, df.y)

ds = tf.data.Dataset.from_tensor_slices((dict(X_train), y_train))
model.fit(ds,epochs=100)