Hardware estimate - CPU n RAM for training model f

Rasa Open Source
1

Dear All, I am trying to build a RASA training model - nlu n core having around 300 intents with around 1200 training examples for entire 300 intents.

I using Intel i5, 4 core, CPU with 8GB RAM.

I am unable to train the model with more than 64 intents. Process is getting killed after running for few minutes.

Can u please help me know what system configuration is recommended for model of this size.

Also, please suggest whether any doc is available which can be referred to customise the system to utilise resources optimally.

2

My first thought was that 300 intents is somewhat large, what operating system are you using? Could you also verify the version of tensorflow that is in your virtualenv?

3

Dear Vincent

OS is Ubuntu 18.04.4 LTS,Intel® Core™ i5-4570T CPU @ 2.90GHz × 4 ,64-bit.

tensorflow version is 2.1.0.

Thanks

Amit

4

Could you send a copy of your config.yml?

Part of me worries about the 1200 training examples and 300 intents. This suggests there’s only (on average) 4 examples per intent.