llama-mps VS sentencepiece

There are currently at least 3 ways to run llama on m1 with GPU acceleration. - mlc-llm (pre-built, only 1 model has been ported) - tinygrad (very memory efficient, not that easy to integrate into other projects) - llama-mps (original llama codebase + llama adapter support)

There is also a gpu-acelerated fork of the original repo

https://github.com/remixer-dec/llama-mps

I tried to port the llama-cpu version to a gpu-accelerated mps version for macs, it runs, but the outputs are not as good as expected and it often gives "-1" tokens. Any help and contributions on fixing it are welcome!

you need to train the model on 1 trillion tokens (https://platform.openai.com/tokenizer https://github.com/google/sentencepiece) anyways for it to get reasoning capacities, which it feels very unlikely that your data is that much.

I'm highly skeptical that you have enough data to pretrain if you don't have enough data to fine tune.

fine tuning + vector search + prompting of as much stuff as you can, on a LLM like palm2 or gpt4 is what I would do. otherwise you can use falcon 40B ofc.

maybe I should charge for this ahah

a) Comparison with SentencePiece tokenizer with comparable settings (It can also ignore word-boundaries and create phrase tokens)

LLaMA uses the sentencepiece tokenizer: https://github.com/google/sentencepiece

Besides, are you familiar with SentencePiece? What you are doing looks very similar (generate a large vocab, prune worst token until vocab size is reached), only the token selection criterion is different. It's also purely data driven in the sense that there are no assumption specific to language (and it can optionally segment across whitespace, as you are doing).

Hi, I'm studying the implementation of encode and decode functions for Google's SentencePiece tokenizer.

I think many LMs nowadays use unicode tokenizers, that are not tied to specific languages. E.g. sentencepiece is the most popular one: https://github.com/google/sentencepiece

If you are interested in implementing LLaMA yourself or learning, I noticed that the reference code by Facebook is one of the cleaner, easier to read ML code I've seen in a while. https://github.com/facebookresearch/llama/blob/main/llama/mo... It's about 200 lines long. You probably do need a bit of knowledge to understand what you are reading but I was pleasantly surprised.

For example in comparison, StableDiffusion torch code in diffusers and transformers Python libraries has lots of conditionals, experiments etc. that are not being used that can make it hard to follow what is going on.

Last weekend I got the "main loop" of the transformer working in pure CPU Rust code, following the reference code. My crappy code is just very very slow as I focused on getting it to run, not making it fast. The tokenizer uses some Google thing https://github.com/google/sentencepiece but luckily for inference it seems that you just need to be able to parse the tokenizer model file and not understand how it was created; I was able to strip out the protobuf files from that repository and add it to Rust and read the tokens.

I am optimistic that someone makes a high quality CPU or some CPU+GPU+SSD combination thingmaling that will make it somewhat practical to run even the large LLM models without needing an A100 or two.