Sentence-transformers requires rust compiler about marqo HOT 3 CLOSED

can be fixed by running pip install rust before installation of the package

from marqo.

Thanks Danyil! You are correct, the current default behavior is that huggingface will use the rust based tokenizers. We will potentially switch to the python tokenizer as the small batch performance should not be any different and can be switched using a flag as described here https://github.com/huggingface/transformers/blob/v4.21.1/src/transformers/models/auto/tokenization_auto.py#L472. Although, the other consideration is that others may want to use the fast tokenizers (or not be aware) so maybe the install is a better approach

from marqo.

This has been taken care of after the move to Docker.

from marqo.