-
### What problem are you trying to solve?
The current implementation for the `SklearnIntentClassifier` is a bit of an anti-pattern. It does a grid-search internally and it can only use the dense feat…
-
**Problem Description**: Entity extraction overfits to the tokens in the training data, instead of learning the linguistic pattern around the tokens.
**Example**: If the training data contains `My na…
-
**Description of Problem**:
The [generation code](https://github.com/RasaHQ/rasa/blob/53bd34647d0596a5e2b60b2f1e6b70abfde16a55/rasa/shared/core/generator.py#L357-L523) for augmentation of stories is …
-
**Description of Problem**: Currently, we are padding all batches to the same shape in `RasaModelData.prepare_batch`. For our dense data we could use `RaggedTensors` ([doc](https://www.tensorflow.org/…
-
### What problem are you trying to solve?
I'm trying to use multiple gpus to optimize ted policy during training time.
### What's your suggested solution?
At first, I tried to add mirrored strategy…
-
**Description of Problem**: Currently we are treating our regression tests like a leaderboard and looking at past regression test performance as a gatekeeper for a new feature. This is not ideal for a…
-
part of [https://github.com/RasaHQ/rasa/issues/6836](https://github.com/RasaHQ/rasa/issues/6836)
add cli support for `implement rasa train core --number-of-chunks` and make sure that if `number-of-ch…
-
Dear RASA team,
it would be very useful to have implemented the method from [BERT for Joint Intent Classification and Slot Filling](https://arxiv.org/abs/1902.10909).
Differently from the current us…
-
**Issue**: Using the Ubuntu corpus seems not to work as it has a lot of unicode special characters in the form of `\uxxxx` in the text and the yaml parser complains about those.
**Error (including fu…
-
### Description of Problem
* With big datasets we run into memory and computation time issues
* multiwoz: 113,000 conversation turns, 8500 e2e stories
* ubuntu: 470,000 conversation …