First commit. CPU working. GPU not tested.

Project.toml

@@ -3,6 +3,13 @@ uuid = "1285d783-1a6d-4703-8f05-8ac83ef55592"
 authors = ["murrellb <[email protected]> and contributors"]
 version = "1.0.0-DEV"
 
+[deps]
+BytePairEncoding = "a4280ba5-8788-555a-8ca8-4a8c3d966a71"
+Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
+Flux = "587475ba-b771-5e3f-ad9e-33799f191a9c"
+LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
+SafeTensors = "eeda0dda-7046-4914-a807-2495fc7abb89"
+
 [compat]
 julia = "1.9"
 

README.md

@@ -4,3 +4,13 @@
 [![Dev](https://img.shields.io/badge/docs-dev-blue.svg)](https://MurrellGroup.github.io/Jjama3.jl/dev/)
 [![Build Status](https://github.com/MurrellGroup/Jjama3.jl/actions/workflows/CI.yml/badge.svg?branch=main)](https://github.com/MurrellGroup/Jjama3.jl/actions/workflows/CI.yml?query=branch%3Amain)
 [![Coverage](https://codecov.io/gh/MurrellGroup/Jjama3.jl/branch/main/graph/badge.svg)](https://codecov.io/gh/MurrellGroup/Jjama3.jl)
+
+# Quickstart
+
+```julia
+config = JSON3.read(read("Llama3_2_1B_instruct/config.json", String));
+model = load_llama3_from_safetensors("Llama3_2_1B_instruct/model.safetensors", config);
+tkn = llama3_tokenizer();
+prompt = assistant_prompt("Why would anyone implement the llama3 LLM in Julia?", tkn);
+ts = generate(model, prompt, max_new_tokens=500, encoder_for_printing=tkn);
+```

src/Jjama3.jl

@@ -1,5 +1,11 @@
 module Jjama3
 
-# Write your package code here.
+using Flux, BytePairEncoding, SafeTensors, Distributions, LinearAlgebra
+
+include("model.jl")
+include("utils.jl")
+include("sampling.jl")
+
+export load_llama321B_from_safetensors, load_llama3_from_safetensors, llama3_tokenizer, assistant_prompt, format_llama32_instruction_prompt, generate, forward_loss, forward_inference
 
 end