I got "Out of Memory" when train RASA with over 2M examples of NLU

Rasa Open Source
1

Dear All,

Currently, I got “Out of Memory” issue when I trained RASA with over 2M examples of NLU. I tried with: “rasa train --augmetation 0” but not help. RASA version: 1.10.1 RAM: 128G SWAP: 128G

Could you please help to show me how to solve this?

2

You could try reducing the training batch size for your classifier to see if it resolves the memory issue. It’s passed in through the config.yml file as an array of two values, being the initial and final batch sizes where the size is increased linearly after each epoch.

 - name: DIETClassifier
   batch_size: [16, 64]