Rasa: Model in training status, "unknown training file format : unk for file"

[ChamathK]

Rasa NLU version : "0.10.0" Docker latest full

Used backend / pipeline : "nlp_spacy", "tokenizer_spacy", "intent_featurizer_spacy", "ner_crf", "ner_synonyms", "intent_classifier_sklearn", "ner_duckling"

Operating system (windows, osx, ...): Ubuntu 16.04, docker version 17.03.1-ce

Issue: When trying to train a new project it is giving the following error on post request.
{"error": "unknown training file format : unk for file /tmp/tmpy31sXe_training_data"}

I am trying to train a new project using the tutorial found in https://rasahq.github.io/rasa_nlu/http.html#post-train

curl -XPOST localhost:5000/train?project=my_project -d demo-rasa.json

When checking the status, using http://localhost:5000/status
{"available_projects": "default": {"status": "ready", "available_models": ["model_20171006-062344"]}, "my_project": {"status": "training", "available_models": ["fallback"]}}}

"my_project" is not in ready status, remains in training status.

Config.json
config.txt

[PHLF]

You're just missing the @ symbol in your request which is a curl specific instruction to load the content of the file at the relative path demo-rasa.json :

$ curl -XPOST localhost:5000/train?project=my_project -d @data/examples/rasa/demo-rasa.json

[amn41]

@PHLF is right - but the project shouldn't get stuck in training status. See #652

[tmbo]

I guess this is fixed now :)

[ChamathK]

@PHLF thanks for the support. Solution worked!