Load a model from AWS S3

Getting Started with Rasa
1

Hi All!!!

Like most here I am new and I have many questions that I cannot find.

I’m trying to load a model from a S3 bucket. The model works fine in local but now I’m trying to load it from the cloud.

I’m using Python 3.7 and Rasa 1.10.14

Locally I load the model: interpreter = Interpreter.load("./model/nlu")

so here are my questions: 1 .- Is it possible load the model directly from the tar.gz file instead from the uncompress folder???

I have tried to load the tar.gz file but always get the same error: InvalidModelError: Failed to load model metadata from

2.- Is it possible load the model from a S3 bucket compress or uncompress using a python code???

Thanks in advance

2

Hey @Juliopdata

You can use external model storage, and the S3 is one of the supported options indeed. We have a doc page about this topic, have you checked that?

I’m happy to help with further questions!

3

Hi Degiz, do you know how I can test my model that I load from s3.

For example: I use this curl command to test my model that I load locally from my drive: curl -X POST http://localhost:5005/model/parse -d '{"text":"Hi I am Nilima"}'

How do I use the curl command to test the model that I loaded from my s3 bucket?

I used an example provided by the Rasa docs : curl -X POST http://localhost:5005/webhooks/rest/webhook -d '{"sender":" test_user", "message":" Stop calling me"}'

But since my model only has the nlu part, no core, meaning that I only need the intent classification and entity extraction from my model. This rest channel above sends me an empty message as: []%

Any idea how I can test my model loaded from s3 bucket for entity extraction and intent classification?