Cold Start Exploration with fast.ai
This is the repo backing Frame's transfer learning experiments backing our blog post on using fast.ai, pytorch, and floydhub to tackle the cold start problem common in many production data products. These experiements explicitly address the Natural Language Understand task of sentiment analysis using various language models. Check out the blog post here for more details!
- fast.ai V1
- torch V1
Run on floydhub workspace by pressing the button below or locally by starting here: 00_start_here.ipynb in a jupyter notebook.