I found out that if I use only
language: "en"
pipeline: "pretrained_embeddings_convert"
it works without errors, while if I specify the whole components:
language: "en"
pipeline:
- name: "ConveRTTokenizer"
- name: "ConveRTFeaturizer"
- name: "EmbeddingIntentClassifier"
it gives me the error that I put above. Does anybody know why is that? @Ghostvv?