Started dataset.py, added the ReviewDataset class and implemented the __init__, __len__ and __getitem__ methods. The __getitem__ method currently just returns the review text, but will be updated to return the tokenized review as a tensor

notebooks/preprocessing_tagged.ipynb

@@ -1161,9 +1161,11 @@
    ],
    "source": [
     "# Let's also see if the longer reviews are more likely to be bug reports, feature requests, or have certain aspects/sentiments\n",
+    "\n",
     "# This finding allows us to make a better decision on the max_length for the model, which will increase the quality of the model,\n",
     "# the time to train will be longer but it is not worth removing valuable information from longer reviews which are mostly bug reports and feature requests \n",
     "# with negative sentiment (the most important ones to classify correctly)\n",
+    "\n",
     "for df, name in [(tagged_boosted_df, 'Boosted'), (tagged_original_df, 'Original')]:\n",
     "    lengths = df['review'].apply(lambda x: len(tokenizer.encode(x)))\n",
     "    mask_128 = lengths > 128\n",

src/dataset.py

@@ -1,7 +1,6 @@
 # dataset.py
 
 # Takes a row from the csv, tokenizes the review and returns a tensor
-
 import torch
 import pandas as pd
 from torch.utils.data import Dataset