Unable to setup an environment that can successfully replicate the tutorial

PyTorch-Forecasting version: 0.10.3
PyTorch (torch) version: 1.13.1
Python version: 3.9.13
Operating System: MacOS 12.6.3

Expected behavior

I ran through the tutorial at https://pytorch-forecasting.readthedocs.io/en/latest/tutorials/stallion.html and expected to be able to replicate the tutorial. It is very well written!

I set up both conda and virtualenv environments, and installed packages using conda and pip, however I was unable to create an environment that could go through the tutorial. It would be wonderful if the tutorials came with a requirements file to be able to replicate the tutorials.

Actual behavior

Started off with Baseline() not loading due to an error 'tuple' object has no attribute 'items', which was resolved by moving to Pytorch-Lightning version 1.9.0 (the suggestion workaround was documented in another)

The next error was when trying to find the optimal learning rate... I think it failed because it was trying to save to the root folder rather than a relative folder.

res = trainer.tuner.lr_find(
    tft,
    train_dataloaders=train_dataloader,
    val_dataloaders=val_dataloader,
    max_lr=10.0,
    min_lr=1e-6,
)

Code to reproduce the problem

import os
import warnings

warnings.filterwarnings("ignore")  # avoid printing out absolute paths

os.chdir("../../..")

import copy
from pathlib import Path
import warnings

import numpy as np
import pandas as pd
import pytorch_lightning as pl
from pytorch_lightning.callbacks import EarlyStopping, LearningRateMonitor
from pytorch_lightning.loggers import TensorBoardLogger
import torch

from pytorch_forecasting import Baseline, TemporalFusionTransformer, TimeSeriesDataSet
from pytorch_forecasting.data import GroupNormalizer
from pytorch_forecasting.metrics import SMAPE, PoissonLoss, QuantileLoss
from pytorch_forecasting.models.temporal_fusion_transformer.tuning import optimize_hyperparameters

from pytorch_forecasting.data.examples import get_stallion_data

data = get_stallion_data()

# add time index
data["time_idx"] = data["date"].dt.year * 12 + data["date"].dt.month
data["time_idx"] -= data["time_idx"].min()

# add additional features
data["month"] = data.date.dt.month.astype(str).astype("category")  # categories have be strings
data["log_volume"] = np.log(data.volume + 1e-8)
data["avg_volume_by_sku"] = data.groupby(["time_idx", "sku"], observed=True).volume.transform("mean")
data["avg_volume_by_agency"] = data.groupby(["time_idx", "agency"], observed=True).volume.transform("mean")

# we want to encode special days as one variable and thus need to first reverse one-hot encoding
special_days = [
    "easter_day",
    "good_friday",
    "new_year",
    "christmas",
    "labor_day",
    "independence_day",
    "revolution_day_memorial",
    "regional_games",
    "fifa_u_17_world_cup",
    "football_gold_cup",
    "beer_capital",
    "music_fest",
]
data[special_days] = data[special_days].apply(lambda x: x.map({0: "-", 1: x.name})).astype("category")
data.sample(10, random_state=521)

data.describe()
max_prediction_length = 6
max_encoder_length = 24
training_cutoff = data["time_idx"].max() - max_prediction_length

training = TimeSeriesDataSet(
    data[lambda x: x.time_idx <= training_cutoff],
    time_idx="time_idx",
    target="volume",
    group_ids=["agency", "sku"],
    min_encoder_length=max_encoder_length // 2,  # keep encoder length long (as it is in the validation set)
    max_encoder_length=max_encoder_length,
    min_prediction_length=1,
    max_prediction_length=max_prediction_length,
    static_categoricals=["agency", "sku"],
    static_reals=["avg_population_2017", "avg_yearly_household_income_2017"],
    time_varying_known_categoricals=["special_days", "month"],
    variable_groups={"special_days": special_days},  # group of categorical variables can be treated as one variable
    time_varying_known_reals=["time_idx", "price_regular", "discount_in_percent"],
    time_varying_unknown_categoricals=[],
    time_varying_unknown_reals=[
        "volume",
        "log_volume",
        "industry_volume",
        "soda_volume",
        "avg_max_temp",
        "avg_volume_by_agency",
        "avg_volume_by_sku",
    ],
    target_normalizer=GroupNormalizer(
        groups=["agency", "sku"], transformation="softplus"
    ),  # use softplus and normalize by group
    add_relative_time_idx=True,
    add_target_scales=True,
    add_encoder_length=True,
)

# create validation set (predict=True) which means to predict the last max_prediction_length points in time
# for each series
validation = TimeSeriesDataSet.from_dataset(training, data, predict=True, stop_randomization=True)

# create dataloaders for model
batch_size = 128  # set this between 32 to 128
train_dataloader = training.to_dataloader(train=True, batch_size=batch_size, num_workers=0)
val_dataloader = validation.to_dataloader(train=False, batch_size=batch_size * 10, num_workers=0)

# calculate baseline mean absolute error, i.e. predict next value as the last available value from the history
actuals = torch.cat([y for x, (y, weight) in iter(val_dataloader)])
baseline_predictions = Baseline().predict(val_dataloader)
(actuals - baseline_predictions).abs().mean().item()

# configure network and trainer
pl.seed_everything(42)
trainer = pl.Trainer(
    gpus=0,
    # clipping gradients is a hyperparameter and important to prevent divergance
    # of the gradient for recurrent neural networks
    gradient_clip_val=0.1,
)

tft = TemporalFusionTransformer.from_dataset(
    training,
    # not meaningful for finding the learning rate but otherwise very important
    learning_rate=0.03,
    hidden_size=16,  # most important hyperparameter apart from learning rate
    # number of attention heads. Set to up to 4 for large datasets
    attention_head_size=1,
    dropout=0.1,  # between 0.1 and 0.3 are good values
    hidden_continuous_size=8,  # set to <= hidden_size
    output_size=7,  # 7 quantiles by default
    loss=QuantileLoss(),
    # reduce learning rate if no improvement in validation loss after x epochs
    reduce_on_plateau_patience=4,
)
print(f"Number of parameters in network: {tft.size()/1e3:.1f}k")

# find optimal learning rate
res = trainer.tuner.lr_find(
    tft,
    train_dataloaders=train_dataloader,
    val_dataloaders=val_dataloader,
    max_lr=10.0,
    min_lr=1e-6,
)

Here's the error:

OSError: [Errno 30] Read-only file system: '/.lr_find_c2a15fa8-d817-4b69-81a0-6bbeca4acba6.ckpt'

I think it is trying to write a file to the root folder rather than a relative path. Maybe this is a small fix, but maybe there will be more issues after this is resolved... As there are so many dependencies, it would be great if the tutorials could include a requirements file so that we can replicate the results in the tutorials. Thank you!

jdb78 / pytorch-forecasting