Prophet温启动拟合示例代码运行崩溃问题求助

问题描述 #

使用Prophet官方温启动示例代码时，运行温启动步骤出现AttributeError: 'list' object has no attribute 'tolist'错误，当前Prophet版本为1.1.1，因项目依赖限制无法升级。

复现代码 #

from prophet import Prophet
import numpy as np
import pandas as pd

def warm_start_params(m):
    """
    Retrieve parameters from a trained model in the format used to initialize a new Stan model.
    Note that the new Stan model must have these same settings:
        n_changepoints, seasonality features, mcmc sampling
    for the retrieved parameters to be valid for the new model.

    Parameters
    ----------
    m: A trained model of the Prophet class.

    Returns
    -------
    A Dictionary containing retrieved parameters of m.
    """
    res = {}
    for pname in ['k', 'm', 'sigma_obs']:
        if m.mcmc_samples == 0:
            res[pname] = m.params[pname][0][0]
        else:
            res[pname] = np.mean(m.params[pname])
    for pname in ['delta', 'beta']:
        if m.mcmc_samples == 0:
            res[pname] = m.params[pname][0]
        else:
            res[pname] = np.mean(m.params[pname], axis=0)
    return res

df = pd.read_csv('example_wp_log_peyton_manning.csv')
df1 = df.loc[df['ds'] < '2016-01-19', :]  # 除最后一天外的所有数据
m1 = Prophet().fit(df1) # 基于除最后一天外的数据训练模型

%timeit m2 = Prophet().fit(df)  # 从头训练完整数据
%timeit m2 = Prophet().fit(df, init=warm_start_params(m1))  # 原报错的温启动行

报错信息 #

11:40:57 - cmdstanpy - INFO - Chain [1] start processing
11:40:58 - cmdstanpy - INFO - Chain [1] done processing
...（省略重复日志）
1.09 s ± 85.2 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
Cell In[11], line 36
     32 m1 = Prophet().fit(df1) # A model fit to all data except the last day
     35 get_ipython().run_line_magic('timeit', 'm2 = Prophet().fit(df)  # Adding the last day, fitting from scratch')
---> 36 get_ipython().run_line_magic('timeit', 'm2 = Prophet().fit(df, init=warm_start_params(m1))  # Adding the last day, warm-starting from m1')

...（省略中间调用栈）
File /root/miniconda3/lib/python3.9/site-packages/prophet/models.py:154, in CmdStanPyBackend.prepare_data(init, data)
    153         'trend_indicator': data['trend_indicator'],
--> 154         'y': data['y'].tolist(),
    155         't': data['t'].tolist(),
    156         'cap': data['cap'].tolist(),

AttributeError: 'list' object has no attribute 'tolist'

问题原因 #

Prophet 1.1.1版本的CmdStanPyBackend.fit方法存在逻辑错误：当传入init参数时，会将已处理完成的stan_data（其中y等字段已转为list）再次传入prepare_data方法，而该方法仍尝试调用tolist()，触发属性错误。

解决方案 #

将温启动调用中的init参数替换为inits参数，直接传入预训练模型的参数，绕过错误的参数处理逻辑：

%timeit m2 = Prophet().fit(df, inits=warm_start_params(m1))  # 修改后的温启动行

同时确保warm_start_params函数返回的参数格式符合要求：

• k、m、sigma_obs为标量
• delta、beta为一维数组

内容的提问来源于stack exchange，提问作者gsamaras