sentencepiece
amx
Our great sponsors
sentencepiece | amx | |
---|---|---|
19 | 18 | |
9,480 | 851 | |
4.6% | - | |
8.1 | 4.1 | |
16 days ago | about 2 months ago | |
C++ | C | |
Apache License 2.0 | MIT License |
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
sentencepiece
- sentencepiece
-
LLM.int8(): 8-Bit Matrix Multiplication for Transformers at Scale
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
-
[P] TokenMonster Ungreedy ~ 35% faster inference and 35% increased context-length for large language models (compared to tiktoken). Benchmarks included.
a) Comparison with SentencePiece tokenizer with comparable settings (It can also ignore word-boundaries and create phrase tokens)
-
LLaMA tokenizer: is a JavaScript implementation available anywhere?
LLaMA uses the sentencepiece tokenizer: https://github.com/google/sentencepiece
-
[P] New tokenization method improves LLM performance & context-length by 25%+
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).
-
Code runs without definition of function (automatically calls a different function instead)
Hi, I'm studying the implementation of encode and decode functions for Google's SentencePiece tokenizer.
-
How to handle multiple languages in a sentence?
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
- Large language models are having their Stable Diffusion moment
-
LLaMA-7B in Pure C++ with full Apple Silicon support
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.
- ChatGPT in an iOS Shortcut – Worlds Smartest HomeKit Voice Assistant
amx
-
Optimize sgemm on RISC-V platform
I am talking about the matrix/vector coprocessor (AMX). You can find some reverse-engineered documentation here: https://github.com/corsix/amx
On M3 a singe matrix block can achieve ~ 1TFLOP on DGEMM, I assume it will be closer to 4TFLOPS for SGEMM. The Max variants have two such blocks. Didn't do precise benchmarking myself, but switching Python/R matrix libraries to use Apple's BLAS result in 5-6x perf improvement on matrix heavy code for me.
-
Intel AMX
It's really cool. I hope it becomes more common for training/inference/numerics capable accelerators to be included in consumer hardware.
Apple's AMX is really under-documented, while the instructions were reverse engineered, Virtually no benchmarks are available comparing current chip generations, models and variants.
https://github.com/corsix/amx
- Why do x86 processors take up so much energy when compared to ARM?
-
Bfloat16 support coming to Apple's Metal and PyTorch [video]
Visible in the unofficial documentation for AMX instructions too - M2 only bf16 functionality - https://github.com/corsix/amx/blob/main/matfp.md
-
LLaMA-7B in Pure C++ with full Apple Silicon support
Confusingly there are 2 mechanisms to do matrix operations on the new apple hardware - AMX (https://github.com/corsix/amx) - and the ANE (apple neural engine) - which is enabled by CoreML. This code does not run on the neural engine but the author has a branch for his whisper.cpp project which uses it here: https://github.com/ggerganov/whisper.cpp/pull/566 - so it may not be long before we see it applied here as well. All of this is to say that it actually could get significantly faster if some of this work was able to be handed to the ANE with CoreML.
-
Linux 6.2: The first mainstream Linux kernel for Apple M1 chips arrives
really? seems pretty well documented here: https://github.com/corsix/amx
-
AMX: The Secret Apple M1 Coprocessor
Article is almost two years old, and has a huge correction at the bottom. It's just a proprietary ISA extension, there's even a repo documenting what's been reverse engineered.
- corsix/amx: Apple AMX Instruction Set
-
Show HN: Port of OpenAI's Whisper model in C/C++
You are correct, in that those are the four
My understanding is that the AMX is more tightly wound with the CPU, ultimately being accessible via an instruction set (https://github.com/corsix/amx), and it is useful if you need to do matrix multiplications interleaved with other CPU tasks. A common example would be a VIO loop or something where you want that data in the CPU caches.
The GPU and Neural Engine are not that – they take some time to set up and initialize. They also can parallelize tasks to a much higher degree. The GPU is more generalizable, because you can write compute shaders to do anything in parallel, but it uses a lot of resources. I'll have to check out the PR to see how exactly the MPS shaders match up with the task at hand, because you could also consider writing Metal compute shaders by hand.
I know the least about the ANE, but it has specific hardware for running ML models, and you have to process the weights ahead of time to make sure they are in the right format. It can run ML models very efficiently and is the most battery friendly.
-
Ask HN: Are there any undocumented ISA extensions used in Linux systems?
If someone were to build a Linux system with proprietary ISA extensions, how would they do it given Linux is open source? Are there any examples of this being done? Would it be possible at all?
I got inspiration from this (https://github.com/corsix/amx) and I wondered if someone has done it before on a Linux-based system. I understand a userspace library could be created to access those instructions from userspace, but how would then they be implemented in the kernel? Through a proprietary kernel module built using a custom compiler? Or is that not needed at all and the library could just run on the processor taking advantage of the proprietary extensions?
What are some alternatives?
transformers - 🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.
emacs-pure
CTranslate2 - Fast inference engine for Transformer models
whisper.cpp - Port of OpenAI's Whisper model in C/C++
llama - Inference code for Llama models
whisper.cpp - Port of OpenAI's Whisper model in C/C++
gpt-2 - Code for the paper "Language Models are Unsupervised Multitask Learners"
llama-mps - Experimental fork of Facebooks LLaMa model which runs it with GPU acceleration on Apple Silicon M1/M2
text-generation-webui - A Gradio web UI for Large Language Models. Supports transformers, GPTQ, AWQ, EXL2, llama.cpp (GGUF), Llama models.
amx-rs - Rust wrapper for Apple Matrix Coprocessor (AMX) instructions
OpenNMT-Tutorial - Neural Machine Translation (NMT) tutorial. Data preprocessing, model training, evaluation, and deployment.
mighty-snitch - noticing and preventing network requests should be easy