Support for GPU and batch computations

[AmenRa]

Hi Sebastian,

I have a few questions regarding the use of ColBERTer on the GPU:

1. How can I batch-encode an entire collection on the GPU?
   From my understanding, the provided example only works for one piece of text at a time on the CPU, which is extremely slow.
2. Similarly, how can I compute scores for a given query and a list of 1k candidate documents?
   How can I do it for a batch of queries/candidates?

I tried to do it by myself, but even the tokenizer does not support batch computation.

Can you help me?

Thank you,

Elias