Alex Klinkhamer comments

Results 7 comments of


                                            Alex Klinkhamer

examples : evaluate tokens in batches after swapping context

Tests passed yesterday. I just synced recent changes and added a comment.

Add command mode to interactive mode.

I don't think `context_size` can change. It's used to initialize the `ctx`. Not sure if there are edge cases for `keep` or `n_predict`. It seems odd to want to change...

Consider tagging @v1?

fwiw, the `v1` tag exists now (https://github.com/jwlawson/actions-setup-cmake/releases/tag/v1)

[feature] Make space insertion optional in SentencePiece tokenizer

Spaces aren't opportunistically inserted within the text, so doing it only when `add_bos==true` would make sense to me. (Option 1) A little extra info: The OP links to a hack...

[feature] Make space insertion optional in SentencePiece tokenizer

> I'm curious to see a model that uses SentencePiece tokenizer and has multiple tokens that contain newline @shibe2 The CausalLM GGUF models on HuggingFace do this sometimes. " \n\n"...

[feature] Make space insertion optional in SentencePiece tokenizer

> For example, in this Python example, how would you use the SentencePiece tokenizer to achieve your goal? You have to set `NormalizerSpec`'s `add_dummy_prefix` field as false. Conveniently, a recent...

[Bug]: grpc 1.48.1 doesn't work

Version 1.48.1 had a few revisions due to protocol_compiler issues. Try 1.48.1.bcr.3 - that's the one that eventually [worked for me](https://github.com/rendezqueue/rendezqueue_ccgrpc/commit/8aa1c6539b171f33b191476bd327c655cd8b6571).