A C#/.NET library to run LLM models (🦙LLaMA/LLaVA) on your local device efficiently.
LLamaSharp is a cross-platform library to run 🦙LLaMA/LLaVA model (and others) on your local device. Based on llama.cpp, inference with LLamaSharp is efficient on both CPU and GPU. With the higher-level APIs and RAG support, it's convenient to deploy LLM (Large Language Model) in your application with LLamaSharp.
Please star the repo to show your support for this project!🤗
LLaMA | LLaVA |
There are integrations for the following libraries, making it easier to develop your APP. Integrations for semantic-kernel and kernel-memory are developed in LLamaSharp repository, while others are developed in their own repositories.
The following examples show how to build APPs with LLamaSharp.
To gain high performance, LLamaSharp interacts with a native library compiled from c++, which is called backend
. We provide backend packages for Windows, Linux and MAC with CPU, Cuda, Metal and OpenCL. You don't need to handle anything about c++ but just install the backend packages.
If no published backend match your device, please open an issue to let us know. If compiling c++ code is not difficult for you, you could also follow this guide to compile a backend and run LLamaSharp with it.
PM> Install-Package LLamaSharp
Install one or more of these backends, or use self-compiled backend.
LLamaSharp.Backend.Cpu
: Pure CPU for Windows & Linux & MAC. Metal (GPU) support for MAC.LLamaSharp.Backend.Cuda11
: CUDA11 for Windows & Linux.LLamaSharp.Backend.Cuda12
: CUDA 12 for Windows & Linux.LLamaSharp.Backend.OpenCL
: OpenCL for Windows & Linux.(optional) For Microsoft semantic-kernel integration, install the LLamaSharp.semantic-kernel package.
(optional) To enable RAG support, install the LLamaSharp.kernel-memory package (this package only supports net6.0
or higher yet), which is based on Microsoft kernel-memory integration.
There are two popular format of model file of LLM now, which are PyTorch format (.pth) and Huggingface format (.bin). LLamaSharp uses GGUF
format file, which could be converted from these two formats. To get GGUF
file, there are two options:
Search model name + 'gguf' in Huggingface, you will find lots of model files that have already been converted to GGUF format. Please take care of the publishing time of them because some old ones could only work with old version of LLamaSharp.
Convert PyTorch or Huggingface format to GGUF format yourself. Please follow the instructions of this part of llama.cpp readme to convert them with the python scripts.
Generally, we recommend downloading models with quantization rather than fp16, because it significantly reduce the required memory size while only slightly impact on its generation quality.
Here is a simple example to chat with bot based on LLM in LLamaSharp. Please replace the model path with yours.
using LLama.Common;
using LLama;
string modelPath = @"<Your Model Path>"; // change it to your own model path.
var parameters = new ModelParams(modelPath)
{
ContextSize = 1024, // The longest length of chat as memory.
GpuLayerCount = 5 // How many layers to offload to GPU. Please adjust it according to your GPU memory.
};
using var model = LLamaWeights.LoadFromFile(parameters);
using var context = model.CreateContext(parameters);
var executor = new InteractiveExecutor(context);
// Add chat histories as prompt to tell AI how to act.
var chatHistory = new ChatHistory();
chatHistory.AddMessage(AuthorRole.System, "Transcript of a dialog, where the User interacts with an Assistant named Bob. Bob is helpful, kind, honest, good at writing, and never fails to answer the User's requests immediately and with precision.");
chatHistory.AddMessage(AuthorRole.User, "Hello, Bob.");
chatHistory.AddMessage(AuthorRole.Assistant, "Hello. How may I help you today?");
ChatSession session = new(executor, chatHistory);
InferenceParams inferenceParams = new InferenceParams()
{
MaxTokens = 256, // No more than 256 tokens should appear in answer. Remove it if antiprompt is enough for control.
AntiPrompts = new List<string> { "User:" } // Stop generation once antiprompts appear.
};
Console.ForegroundColor = ConsoleColor.Yellow;
Console.Write("The chat session has started.\nUser: ");
Console.ForegroundColor = ConsoleColor.Green;
string userInput = Console.ReadLine() ?? "";
while (userInput != "exit")
{
await foreach ( // Generate the response streamingly.
var text
in session.ChatAsync(
new ChatHistory.Message(AuthorRole.User, userInput),
inferenceParams))
{
Console.ForegroundColor = ConsoleColor.White;
Console.Write(text);
}
Console.ForegroundColor = ConsoleColor.Green;
userInput = Console.ReadLine() ?? "";
}
For more examples, please refer to LLamaSharp.Examples.
NativeLibraryConfig.Instance.WithLogs(LLamaLogLevel.Info)
to the very beginning of your code. The log will show which native library file is loaded. If the CPU library is loaded, please try to compile the native library yourself and open an issue for that. If the CUDA library is loaded, please check if GpuLayerCount > 0
when loading the model weight.Firstly, due to the large size of LLM models, it requires more time to generate outputs than other models, especially when you are using models larger than 30B.
To see if that's a LLamaSharp performance issue, please follow the two tips below.
GpuLayerCount
as large as possible.Generally, there are two possible cases for this problem:
Please set anti-prompt or max-length when executing the inference.
Any contribution is welcomed! There's a TODO list in LLamaSharp Dev Project and you could pick an interesting one to start. Please read the contributing guide for more information.
You can also do one of the followings to help us make LLamaSharp better:
Join our chat on Discord (please contact Rinne to join the dev channel if you want to be a contributor).
Join QQ group
If you want to compile llama.cpp yourself you must use the exact commit ID listed for each version.
LLamaSharp | Verified Model Resources | llama.cpp commit id |
---|---|---|
v0.2.0 | This version is not recommended to use. | - |
v0.2.1 | WizardLM, Vicuna (filenames with "old") | - |
v0.2.2, v0.2.3 | WizardLM, Vicuna (filenames without "old") | 63d2046 |
v0.3.0, v0.4.0 | LLamaSharpSamples v0.3.0, WizardLM | 7e4ea5b |
v0.4.1-preview | Open llama 3b, Open Buddy | aacdbd4 |
v0.4.2-preview | Llama2 7B (GGML) | 3323112 |
v0.5.1 | Llama2 7B (GGUF) | 6b73ef1 |
v0.6.0 | cb33f43 |
|
v0.7.0, v0.8.0 | Thespis-13B, LLaMA2-7B | 207b519 |
v0.8.1 | e937066 |
|
v0.9.0, v0.9.1 | Mixtral-8x7B | 9fb13f9 |
v0.10.0 | Phi2 | d71ac90 |
v0.11.1, v0.11.2 | LLaVA-v1.5, Phi2 | 3ab8b3a |
v0.12.0 | LLama3 | a743d76 |
This project is licensed under the terms of the MIT license.