predict() fails when using a model saved with .save(save_dataset=False)

[candalfigomoro]

What happened + What you expected to happen

Trying to do something similar to the Transfer learning tutorial: https://nixtla.github.io/neuralforecast/examples/transfer_learning.html
Setting save_dataset=False as in the tutorial

With this code

# Fit model
nf_ts = NeuralForecast(models=models, freq="D")
nf_ts.fit(df=Y_train_df_ts)

# Save model
nf_ts.save(path='./models/ts/',
                   model_index=None, 
                   overwrite=True,
                   save_dataset=False)  # save_dataset=False !!!

# Reload model
nf_ts = NeuralForecast.load(path='./models/ts/')

# Predict
Y_hat_df_ts = nf_ts.predict(df=Y_train_df_ts, futr_df=futr_df)

I get this error

File /opt/conda/lib/python3.10/site-packages/neuralforecast/core.py:340, in NeuralForecast.predict(self, df, static_df, futr_df, sort_df, verbose, **data_kwargs)
    336 if df is not None:
    337     dataset, uids, last_dates, _ = self._prepare_fit(
    338         df=df, static_df=static_df, sort_df=sort_df, scaler_type=None
    339     )
--> 340     dataset.scalers_ = self.dataset.scalers_
    341     dataset._transform_temporal()
    342 else:

AttributeError: 'NeuralForecast' object has no attribute 'dataset'

Versions / Dependencies

neuralforecast main branch
Python 3.10
GNU/Linux

Reproduction script

import pandas as pd

from datasetsforecast.m4 import M4
from neuralforecast.core import NeuralForecast
from neuralforecast.models import NHITS
from neuralforecast.utils import AirPassengersDF

Y_df, _, _ = M4.load(directory='./', group='Monthly', cache=True)
Y_df['ds'] = pd.to_datetime(Y_df['ds'])
Y_df

horizon = 12
stacks = 3
models = [NHITS(input_size=5 * horizon,
                h=horizon,
                max_steps=100,
                stack_types = stacks*['identity'],
                n_blocks = stacks*[1],
                mlp_units = [[256,256] for _ in range(stacks)],
                n_pool_kernel_size = stacks*[1],
                batch_size = 32,
                scaler_type='standard',
                n_freq_downsample=[12,4,1])]
nf = NeuralForecast(models=models, freq='M')
nf.fit(df=Y_df)

nf.save(path='./results/transfer/', model_index=None, overwrite=True, save_dataset=False)

nf = NeuralForecast.load(path='./results/transfer/')

Y_df = AirPassengersDF.copy()
mean = Y_df[Y_df.ds<='1959-12-31']['y'].mean()
std = Y_df[Y_df.ds<='1959-12-31']['y'].std()

Y_train_df = Y_df[Y_df.ds<='1959-12-31'] # 132 train
Y_test_df = Y_df[Y_df.ds>'1959-12-31']   # 12 test

Y_hat_df = nf.predict(df=Y_train_df).reset_index()

Issue Severity

None

[candalfigomoro]

I see it's a duplicate of #798