Upgrade to pytorch-lightning >=2 (#114)

* Upgrade to pytorch-lightning >=2

Upgrades pytorch lightning, modifying usage of pl.Trainer
to conform with the new way to set the devices and checkpoint.

* Update kaggle download commands in README

Include the commands for unzipping the files, including the zipped
csvs within the original zip from the first challenge.

* Try disabling mac memory allocation limit in CI

See if we can use the mac GPU for the trainer test

* Use cpu for if cuda unavailable in trainer test

Avoids trying to use MPS on mac in CI which has insufficient memory

* Store created val.csv alongside input csv

In preprocessing_utils.py save the created val.csv in the
same folder as the input csv instead of in the current working
directory, and add logging so the user knows where it is saved.

This makes the two functions in the file more consistent in where
they save their output.

* Support comma-separated device string

README.md

@@ -264,10 +264,14 @@ cd jigsaw_data
 # download data
 
 kaggle competitions download -c jigsaw-toxic-comment-classification-challenge
+unzip jigsaw-toxic-comment-classification-challenge.zip -d jigsaw-toxic-comment-classification-challenge
+find jigsaw-toxic-comment-classification-challenge -name '*.csv.zip' | xargs -n1 unzip -d jigsaw-toxic-comment-classification-challenge
 
 kaggle competitions download -c jigsaw-unintended-bias-in-toxicity-classification
+unzip jigsaw-unintended-bias-in-toxicity-classification.zip -d jigsaw-unintended-bias-in-toxicity-classification
 
 kaggle competitions download -c jigsaw-multilingual-toxic-comment-classification
+unzip jigsaw-multilingual-toxic-comment-classification.zip -d jigsaw-multilingual-toxic-comment-classification
 
 ```
 ## Start Training

preprocessing_utils.py

@@ -1,31 +1,45 @@
 import argparse
+import logging
+from pathlib import Path
 
 import numpy as np
 import pandas as pd
 
+logger = logging.getLogger("preprocessing_utils")
+logging.basicConfig(
+    format="%(asctime)s - %(name)s - %(levelname)s - %(message)s",
+    level=logging.INFO,
+)
+
 
 def update_test(test_csv_file):
     """Combines disjointed test and labels csv files into one file."""
+    test_csv_file = Path(test_csv_file)
     test_set = pd.read_csv(test_csv_file)
-    data_labels = pd.read_csv(test_csv_file[:-4] + "_labels.csv")
+    data_labels = pd.read_csv(str(test_csv_file)[:-4] + "_labels.csv")
     for category in data_labels.columns[1:]:
         test_set[category] = data_labels[category]
     if "content" in test_set.columns:
         test_set.rename(columns={"content": "comment_text"}, inplace=True)
-    test_set.to_csv(f"{test_csv_file.split('.csv')[0]}_updated.csv")
+    output_file = test_csv_file.parent / f"{test_csv_file.stem}_updated.csv"
+    test_set.to_csv(output_file)
+    logger.info("Updated test set saved to %s", output_file)
     return test_set
 
 
 def create_val_set(csv_file, val_fraction):
     """Takes in a csv file path and creates a validation set
     out of it specified by val_fraction.
     """
+    csv_file = Path(csv_file)
     dataset = pd.read_csv(csv_file)
     np.random.seed(0)
     dataset_mod = dataset[dataset.toxic != -1]
     indices = np.random.rand(len(dataset_mod)) > val_fraction
     val_set = dataset_mod[~indices]
-    val_set.to_csv("val.csv")
+    output_file = csv_file.parent / "val.csv"
+    logger.info("Validation set saved to %s", output_file)
+    val_set.to_csv(output_file)
 
 
 if __name__ == "__main__":

pyproject.toml

@@ -11,10 +11,10 @@ classifiers = [
     "Operating System :: OS Independent",
     "Programming Language :: Python :: 3",
 ]
-requires-python = ">=3.9,<3.12"
+requires-python = ">=3.9,<3.13"
 dependencies = [
     "sentencepiece >= 0.1.94",
-    "torch < 2.2",
+    "torch >=2",
     "transformers >= 3",
 ]
 
@@ -29,7 +29,7 @@ dev = [
     "datasets >= 1.0.2",
     "pandas >= 1.1.2",
     "pytest",
-    "pytorch-lightning<2.0.0,>1.5.0",
+    "pytorch-lightning>2",
     "scikit-learn >= 0.23.2",
     "tqdm",
     "pre-commit",

tests/test_trainer.py

@@ -1,6 +1,7 @@
 import json
 
 import src.data_loaders as module_data
+
 import torch
 from pytorch_lightning import seed_everything, Trainer
 from torch.utils.data import DataLoader
@@ -37,7 +38,7 @@ def get_instance(module, name, config, *args, **kwargs):
     )
 
     trainer = Trainer(
-        gpus=0 if torch.cuda.is_available() else None,
+        accelerator="gpu" if torch.cuda.is_available() else "cpu",
         limit_train_batches=2,
         limit_val_batches=2,
         max_epochs=1,

train.py

@@ -3,6 +3,7 @@
 import os
 
 import pytorch_lightning as pl
+
 import src.data_loaders as module_data
 import torch
 from pytorch_lightning.callbacks import ModelCheckpoint
@@ -159,7 +160,7 @@ def cli_main():
         "--device",
         default=None,
         type=str,
-        help="indices of GPUs to enable (default: None)",
+        help="comma-separated indices of GPUs to enable (default: None)",
     )
     parser.add_argument(
         "--num_workers",
@@ -208,16 +209,26 @@ def get_instance(module, name, config, *args, **kwargs):
         monitor="val_loss",
         mode="min",
     )
+
+    if args.device is None:
+        devices = "auto"
+    else:
+        devices = [int(d.strip()) for d in args.device.split(",")]
+
     trainer = pl.Trainer(
-        gpus=args.device,
+        devices=devices,
         max_epochs=args.n_epochs,
         accumulate_grad_batches=config["accumulate_grad_batches"],
         callbacks=[checkpoint_callback],
-        resume_from_checkpoint=args.resume,
         default_root_dir="saved/" + config["name"],
         deterministic=True,
     )
-    trainer.fit(model, data_loader, valid_data_loader)
+    trainer.fit(
+        model=model,
+        train_dataloaders=data_loader,
+        val_dataloaders=valid_data_loader,
+        ckpt_path=args.resume,
+    )
 
 
 if __name__ == "__main__":