使用 ETS 和 ray 进行大规模预测 (M5)

!pip install "statsforecast[ray]" neuralforecast s3fs pyarrow

from time import time

import pandas as pd
from neuralforecast.data.datasets.m5 import M5, M5Evaluation
from statsforecast import StatsForecast
from statsforecast.models import ETS

Y_df = pd.read_parquet('s3://m5-benchmarks/data/train/target.parquet')
Y_df = Y_df.rename(columns={
    'item_id': 'unique_id', 
    'timestamp': 'ds', 
    'demand': 'y'
})
Y_df['ds'] = pd.to_datetime(Y_df['ds'])

Y_df.head()

constant = 10
Y_df['y'] += constant

fcst = StatsForecast(
    df=Y_df, 
    models=[ETS(season_length=7, model='ZNA')], 
    freq='D', 
    #n_jobs=-1
    ray_address='ray://ADDRESS:10001'
)

init = time()
Y_hat = fcst.forecast(28)
end = time()
print(f'Minutes taken by StatsForecast using: {(end - init) / 60}')

/home/ubuntu/miniconda/envs/ray/lib/python3.7/site-packages/ray/util/client/worker.py:618: UserWarning: More than 10MB of messages have been created to schedule tasks on the server. This can be slow on Ray Client due to communication overhead over the network. If you're running many fine-grained tasks, consider running them inside a single remote function. See the section on "Too fine-grained tasks" in the Ray Design Patterns document for more details: https://docs.google.com/document/d/167rnnDFIVRhHhK4mznEIemOtj63IOhtIPvSYaPgI4Fg/edit#heading=h.f7ins22n6nyl. If your functions frequently use large objects, consider storing the objects remotely with ray.put. An example of this is shown in the "Closure capture of large / unserializable object" section of the Ray Design Patterns document, available here: https://docs.google.com/document/d/167rnnDFIVRhHhK4mznEIemOtj63IOhtIPvSYaPgI4Fg/edit#heading=h.1afmymq455wu
  UserWarning,

Minutes taken by StatsForecast using: 5.4817593971888225

Y_hat['ETS'] -= constant

Y_hat = Y_hat.reset_index().set_index(['unique_id', 'ds']).unstack()
Y_hat = Y_hat.droplevel(0, 1).reset_index()

*_, S_df = M5.load('./data')
Y_hat = S_df.merge(Y_hat, how='left', on=['unique_id'])

100%|███████████████████████████████████████████████████████████| 50.2M/50.2M [00:00<00:00, 77.1MiB/s]

M5Evaluation.evaluate(y_hat=Y_hat, directory='./data')