All - Activity - jamesmf - Rasa Community Forum

Customize which components get used in a request

I’m interested in using the same NLU model for a number of tasks, and want to be able to inform certain components they can be skipped in certain contexts. An example would be an NLU pipeline that has:

• tokenizer
• featurizer
• regex entity extractor
• crf entity extractor
• intent classification

New Training Data Format Ideas

What would this mean for json formatted files?

Voice assistance

Very cool!

Custom LM with HFTransformers

I’m trying to use my own fine-tuned HFTransformers model weights in a pipeline. I’m guessing that it’s possible to do the following, but I’d love input or thoughts on if there’s an easier way:

• build an initial fake pipeline using a specific set of HF Transformer model+model_weights combo (say dist…)

Using types when working with rasa from another module

Looking into this more, I guess I was asking if rasa was intended to support typing. I don’t see a py.typed even though it looked like much of the codebase was typed.

Word embeddings and RASA NLU

This works slightly more simply than that post now. Once you convert the fastText model to spacy vectors, you can just add text_dense_features under CRFEntityExtractor's features, and your SpacyFeaturizer will add a dense vector per token.

Docker-compose not getting "server is running at" message

Following that demo, I was able to set the password. I was surprised that simply starting a rasa-x container worked well for me alone, but when started in a docker-compose setting, it didn’t. Thanks for your help!

Easiest way to finetune word vectors

Thanks for the reply! That’s not quite what I meant (although an EmbeddingInitializer is an interesting feature idea), here’s what is different. I don’t mean fine-tuning on my labeled, rasa training data. I mean pretraining on some larger corpus of my own. So my goal would be to start with say en_...

Feedback on ConveRT Model + Rasa NLU

Convert was really cool off the shelf for chitchat/smalltalk. I haven’t tried it at a featurizer yet, but it worked great out-of-the-box as a response selector. I’d love it if like spacy you could get multiple things out of it.

Rasa X interface for labeling only

For anyone interesting in replicating, the only way I found was to create a minimal nlu.md containing one of each entity and one of each intent from your dataset. Put it in data/nlu.md...

What controls available entities/intents for labeling in rasa x?

Very cool! Can’t wait to see them. My short term fix has been the following: create a minimum set of training data that hits every intent/entity I want to annotate...

Multiple spacy models in one pipeline

Hi there, I’d like to have two spacy models in my pipeline, but the current implementation doesn’t seem well-suited to that. The SpacyNLP object will set "spacy_doc"...

Pass custom features to CRFEntityExtractor

A thing that seems like it would add a lot of out-of-the-box power to custom entity recognizers is the ability to pass token-level features to CRFEntityExtractor (previously ner_crf)...