ULMFit

An Implementation of ULMFit

Downloading Data

Twitter US Sentiment Data can be downloaded via the site.

Usage

'twitter_data_exploration.ipynb' gives you a nice overview of the data exploration and an implementation of ULMFit to obtain an accuracy of 62%

Setting up Environment

A detailed overview is provided by Jeremy Howard in the fastai forum

On Mac:

1. Open Terminal
2. Navigate to desired folder: cd "~\path\folder"
3. Clone fastai repository from GitHub: git clone https://github.com/fastai/fastai.git
4. Navigate to cloned repository: cd fastai
5. Create environment: conda env create -f environment-cpu.yml
6. Activate environment: conda activate fastai-cpu
7. Start Jupyter Notebook: jupyter notebook

On Windows: The Prompt console must be runned as an administrator. Commands must be run in the Anaconda Prompt console.