tweet-sentiment-extraction

A neural network for extracting sentiments from tweet data. Uses torch, transformers, and BERT. Based on BERT Base Uncased Using PyTorch and uses data from the Tweet Sentiment Extraction competition on Kaggle.

To run code and notebooks in a Spell workspace:

spell jupyter --lab \
  --github-url https://github.com/spellml/tweet-sentiment-extraction.git \
  --env KAGGLE_USERNAME=YOUR_USERNAME \
  --env KAGGLE_KEY=YOUR_KEY \
  tweet-sentiment-extraction

To execute the training scripts in a Spell run:

spell run \
  --machine-type t4 \
  --github-url https://github.com/spellml/tweet-sentiment-extraction.git \
  --pip transformers --pip tokenizers --pip kaggle \
  --env KAGGLE_USERNAME=YOUR_USERNAME \
  --env KAGGLE_KEY=YOUR_KEY \
  --tensorboard-dir /spell/tensorboards/model_1 \
  "chmod +x /spell/scripts/download_data.sh /spell/scripts/upgrade_env.sh; /spell/scripts/download_data.sh; /spell/scripts/upgrade_env.sh; python /spell/models/model_1.py"

spell run \
  --machine-type t4 \
  --github-url https://github.com/spellml/tweet-sentiment-extraction.git \
  --pip transformers --pip tokenizers --pip kaggle \
  --env KAGGLE_USERNAME=YOUR_USERNAME \
  --env KAGGLE_KEY=YOUR_KEY \
  --tensorboard-dir /spell/tensorboards/model_2 \
  "chmod +x /spell/scripts/download_data.sh; chmod +x /spell/scripts/upgrade_env.sh; /spell/scripts/download_data.sh; /spell/scripts/upgrade_env.sh; python /spell/models/model_2.py"