cafaxo / Llama2.jl

Julia package for inference and training of Llama-style language models
MIT License
126 stars 18 forks source link

Llama2.jl

Llama2.jl supports inference and training of Llama-style language models.

Installation

This package is not yet registered. It can be installed by running

pkg> add https://github.com/cafaxo/Llama2.jl

Usage

We currently support:

For example, here is some output from Llama 3:

julia> using Llama2

julia> model = load_gguf_model("Meta-Llama-3-8B.Q4_K_S.gguf")
LanguageModel(
ModelConfig(
  dim            = 4096,
  hidden_dim     = 14336,
  n_layers       = 32,
  n_heads        = 32,
  n_kv_heads     = 8,
  vocab_size     = 128256,
  seq_len        = 512,
  rope_freq_base = 500000.0,
))

julia> sample(model, "The Julia programming language is"; temperature=0.0f0)
 The Julia programming language is a high-level, high-performance dynamic language for technical computing. The language's feature set, based on modern application development platforms, includes support for metaprogramming, type declarations, multiple dispatch, and parallel computing. It also provides a sophisticated ecosystem of tools, libraries, and toolchains all accessible from the Julia REPL. Julia is an open-source project and is released under the MIT license.

Andrej Karpathy's llama2.c models can be found at https://huggingface.co/karpathy/tinyllamas. With these models, the tokenizer.bin file is also required.

Here is an output sample from the 42M tinyllama model:

julia> using Llama2

julia> download("https://huggingface.co/karpathy/tinyllamas/resolve/main/stories42M.bin", "stories42M.bin")
"stories42M.bin"

julia> download("https://raw.githubusercontent.com/karpathy/llama2.c/b4bb47bb7baf0a5fb98a131d80b4e1a84ad72597/tokenizer.bin", "tokenizer.bin")
"tokenizer.bin"

julia> model = load_karpathy_model("stories42M.bin", "tokenizer.bin");

julia> sample(model, "Tim was happy."; temperature = 0.8f0)
Tim was happy. He had a new toy. It was a big red car. He wanted to play with it all day.
Tim took his car outside. He found a drain. It was a big drain. Tim put his car on the drain. The car went down the drain.
Tim was sad. He missed his car. He went home. His mom saw him. She said, "Don't worry, we will get your car back." Tim was glad. He knew his mom would help him. They went to the drain. Tim's car came back. He was happy again.
-------
achieved tok/s: 282.80

Experimental training support

Llama2.jl can train very small Llama-style models on the CPU:

julia> using Llama2

julia> text = read("tinyshakespeare.txt", String);

julia> tokenizer = CharTokenizer(text);

julia> tokens = encode(text, tokenizer);

julia> config = ModelConfig(dim=64, hidden_dim=96, n_layers=4, n_heads=4, n_kv_heads=4, vocab_size=length(tokenizer.id_to_token), seq_len=128)
ModelConfig(
  dim         = 64,
  hidden_dim  = 96,
  n_layers    = 4,
  n_heads     = 4,
  n_kv_heads  = 4,
  vocab_size  = 65,
  seq_len     = 128,
)

julia> weights = train(config, tokens; n_tokens=4_000_000, batch_size=4);
Training a model with 148160 parameters...
Progress: 100%|███████████████████████████| Time: 0:01:31 (11.66 ms/it)
  iteration:      7812 / 7812
  training_loss:  1.497

julia> model = LanguageModel(config, tokenizer, weights);

julia> sample(model, "Julia is"; stop_on_special_token=false, bos_token=false)
Julia is to seen all?

FLORIZEL:
Now the sir?

JOHN ORTHNROLIO:
Talksabated with me with a more thou Vrequitest.
The city good of
-------
achieved tok/s: 11479.52

Acknowledgements

This project started as a port of Andrej Karpathy's llama2.c (https://github.com/karpathy/llama2.c). The quantization code is a port from https://github.com/ggerganov/llama.cpp.