PyTorch-Forecasting中quantile() dtype不匹配RuntimeError求助

环境信息 #

• PyTorch-Forecasting版本: 0.10.2
• PyTorch版本: 1.12.1
• Python版本: 3.10.4
• 操作系统: Windows

预期行为 #

无报错

实际行为 #

运行时触发如下错误：

File c:\Users\josepeeterson.er\Miniconda3\envs\pytorch\lib\site-packages\pytorch_forecasting\metrics\base_metrics.py:979, in DistributionLoss.to_quantiles(self, y_pred, quantiles, n_samples)
977 except NotImplementedError: # resort to derive quantiles empirically
978 samples = torch.sort(self.sample(y_pred, n_samples), -1).values
--> 979 quantiles = torch.quantile(samples, torch.tensor(quantiles, device=samples.device), dim=2).permute(1, 2, 0)
980 return quantiles

RuntimeError: quantile() q tensor must be same dtype as the input tensor

该错误来自框架内部代码，无法直接修改相关逻辑，如何让两个张量的数据类型保持一致？未使用GPU。

输入数据为每4小时从参数(9,0.5)的负二项分布采样得到，其余时间值为0，目标是验证DeepAR模型能否学习该时序模式。

复现代码 #

from pytorch_forecasting.data.examples import generate_ar_data
import matplotlib.pyplot as plt
import pandas as pd
from pytorch_forecasting.data import TimeSeriesDataSet
from pytorch_forecasting.data import NaNLabelEncoder
from pytorch_lightning.callbacks import EarlyStopping, LearningRateMonitor
import pytorch_lightning as pl
from pytorch_forecasting import NegativeBinomialDistributionLoss, DeepAR
import torch
from pytorch_forecasting.data.encoders import TorchNormalizer

# 修正原代码中data被赋值为列表的错误
data = pd.read_csv('1_f_nbinom_train.csv')

data["date"] = pd.Timestamp("2021-08-24") + pd.to_timedelta(data.time_idx, "H")
data['_hour_of_day'] = data["date"].dt.hour.astype(str)
data['_day_of_week'] = data["date"].dt.dayofweek.astype(str)
data['_day_of_month'] = data["date"].dt.day.astype(str)
data['_day_of_year'] = data["date"].dt.dayofyear.astype(str)
# 修正weekofyear弃用问题
data['_week_of_year'] = data["date"].dt.isocalendar().week.astype(str)
data['_month_of_year'] = data["date"].dt.month.astype(str)
data['_year'] = data["date"].dt.year.astype(str)

max_encoder_length = 60
max_prediction_length = 20
training_cutoff = data["time_idx"].max() - max_prediction_length

training = TimeSeriesDataSet(
    data.iloc[0:-620],
    time_idx="time_idx",
    target="value",
    categorical_encoders={
        "series": NaNLabelEncoder(add_nan=True).fit(data.series),
        "_hour_of_day": NaNLabelEncoder(add_nan=True).fit(data._hour_of_day),
        "_day_of_week": NaNLabelEncoder(add_nan=True).fit(data._day_of_week),
        "_day_of_month": NaNLabelEncoder(add_nan=True).fit(data._day_of_month),
        "_day_of_year": NaNLabelEncoder(add_nan=True).fit(data._day_of_year),
        "_week_of_year": NaNLabelEncoder(add_nan=True).fit(data._week_of_year),
        "_year": NaNLabelEncoder(add_nan=True).fit(data._year)
    },
    group_ids=["series"],
    min_encoder_length=max_encoder_length,
    max_encoder_length=max_encoder_length,
    min_prediction_length=max_prediction_length,
    max_prediction_length=max_prediction_length,
    time_varying_unknown_reals=["value"],
    time_varying_known_categoricals=["_hour_of_day","_day_of_week","_day_of_month","_day_of_year","_week_of_year","_year" ],
    time_varying_known_reals=["time_idx"],
    add_relative_time_idx=False,
    randomize_length=None,
    scalers=[],
    target_normalizer=TorchNormalizer(method="identity", center=False, transformation=None)
)

validation = TimeSeriesDataSet.from_dataset(
    training,
    data.iloc[-620:-420],
    stop_randomization=True,
)       

batch_size = 64
train_dataloader = training.to_dataloader(train=True, batch_size=batch_size, num_workers=8)
val_dataloader = validation.to_dataloader(train=False, batch_size=batch_size, num_workers=8)

# save datasets
training.save("training.pkl")
validation.save("validation.pkl")

early_stop_callback = EarlyStopping(monitor="val_loss", min_delta=1e-4, patience=5, verbose=False, mode="min")
lr_logger = LearningRateMonitor()

trainer = pl.Trainer(
    max_epochs=10,
    gpus=0,
    gradient_clip_val=0.1,
    limit_train_batches=30,
    limit_val_batches=3,
    callbacks=[lr_logger, early_stop_callback],
)

deepar = DeepAR.from_dataset(
    training,
    learning_rate=0.1,
    hidden_size=32,
    dropout=0.1,
    loss=NegativeBinomialDistributionLoss(),
    log_interval=10,
    log_val_interval=3,
)
print(f"Number of parameters in network: {deepar.size()/1e3:.1f}k")

torch.set_num_threads(10)
trainer.fit(
    deepar,
    train_dataloaders=train_dataloader,
    val_dataloaders=val_dataloader,
)

解决方法 #

1. 升级PyTorch-Forecasting版本 #

该问题是PyTorch-Forecasting 0.10.2的已知兼容性Bug，后续版本已修复，直接升级即可：

pip install --upgrade pytorch-forecasting

2. 临时修改框架代码 #

如果无法升级，找到报错文件base_metrics.py（路径：c:\Users\josepeeterson.er\Miniconda3\envs\pytorch\lib\site-packages\pytorch_forecasting\metrics\base_metrics.py），将第979行修改为：

quantiles = torch.quantile(samples, torch.tensor(quantiles, device=samples.device, dtype=samples.dtype), dim=2).permute(1, 2, 0)

强制让quantiles张量与samples使用相同的数据类型。

3. 全局指定默认 dtype #

在代码开头添加以下代码，统一浮点类型：

torch.set_default_dtype(torch.float32)

内容的提问来源于stack exchange，提问作者Jose_Peeterson