📝 Add LLamaSharp.KernelMemory project

Add the LLamaSharp.KernelMemory project to the solution. This project includes classes for text embedding generation and text generation using the LLamaSharp library.

LLama.KernelMemory/LLamaSharp.KernelMemory.csproj

@@ -0,0 +1,17 @@
+<Project Sdk="Microsoft.NET.Sdk">
+
+  <PropertyGroup>
+    <TargetFramework>net6.0</TargetFramework>
+    <ImplicitUsings>enable</ImplicitUsings>
+    <Nullable>enable</Nullable>
+  </PropertyGroup>
+
+  <ItemGroup>
+    <PackageReference Include="Microsoft.SemanticMemory.Core" Version="0.4.231023.1-preview" />
+  </ItemGroup>
+
+  <ItemGroup>
+    <ProjectReference Include="..\LLama\LLamaSharp.csproj" />
+  </ItemGroup>
+
+</Project>

LLama.KernelMemory/LLamaSharpTextEmbeddingGeneration.cs

@@ -0,0 +1,45 @@
+using LLama;
+using LLama.Common;
+using Microsoft.SemanticKernel.AI.Embeddings;
+using System;
+using System.Collections.Generic;
+using System.Linq;
+using System.Text;
+using System.Threading.Tasks;
+
+namespace LLamaSharp.KernelMemory
+{
+    public class LLamaSharpTextEmbeddingGeneration : ITextEmbeddingGeneration, IDisposable
+    {
+        private readonly LlamaSharpConfig _config;
+        private readonly LLamaEmbedder _embedder;
+        private readonly LLamaWeights _weights;
+
+        public LLamaSharpTextEmbeddingGeneration(LlamaSharpConfig config)
+        {
+            this._config = config;
+            var @params = new ModelParams(_config.ModelPath);
+            _weights = LLamaWeights.LoadFromFile(@params);
+            _embedder = new LLamaEmbedder(_weights, @params);
+        }
+
+        public void Dispose()
+        {
+            _embedder.Dispose();
+            _weights.Dispose();
+        }
+
+        public Task<IList<ReadOnlyMemory<float>>> GenerateEmbeddingsAsync(IList<string> data, CancellationToken cancellationToken = default)
+        {
+            IList<ReadOnlyMemory<float>> results = new List<ReadOnlyMemory<float>>();
+
+            foreach (var d in data)
+            {
+                var embeddings = _embedder.GetEmbeddings(d);
+                results.Add(new ReadOnlyMemory<float>(embeddings));
+            }
+
+            return Task.FromResult(results);
+        }
+    }
+}

LLama.KernelMemory/LlamaSharpConfig.cs

@@ -0,0 +1,21 @@
+using System;
+using System.Collections.Generic;
+using System.Linq;
+using System.Text;
+using System.Threading.Tasks;
+
+namespace LLamaSharp.KernelMemory
+{
+    public class LlamaSharpConfig
+    {
+        public LlamaSharpConfig(string modelPath)
+        {
+            ModelPath = modelPath;
+        }
+
+        public string ModelPath { get; set; }
+        public uint? ContextSize { get; set; }
+        public uint? Seed { get; set; }
+        public int? GpuLayerCount { get; set; }
+    }
+}

LLama.KernelMemory/LlamaSharpTextGeneration.cs

@@ -0,0 +1,58 @@
+using LLama;
+using LLama.Common;
+using Microsoft.SemanticMemory.AI;
+using System;
+using System.Collections.Generic;
+using System.Linq;
+using System.Text;
+using System.Threading.Tasks;
+
+namespace LLamaSharp.KernelMemory
+{
+    public class LlamaSharpTextGeneration : ITextGeneration, IDisposable
+    {
+        private readonly LlamaSharpConfig _config;
+        private readonly LLamaWeights _weights;
+        private readonly InstructExecutor _executor;
+        private readonly LLamaContext _context;
+
+        public LlamaSharpTextGeneration(LlamaSharpConfig config)
+        {
+            this._config = config;
+            var parameters = new ModelParams(config.ModelPath)
+            {
+                ContextSize = config?.ContextSize ?? 1024,
+                Seed = config?.Seed ?? 0,
+                GpuLayerCount = config?.GpuLayerCount ?? 20
+            };
+            _weights = LLamaWeights.LoadFromFile(parameters);
+            _context = _weights.CreateContext(parameters);
+            _executor = new InstructExecutor(_context);
+
+        }
+
+        public void Dispose()
+        {
+            _context.Dispose();
+            _weights.Dispose();
+        }
+
+        public IAsyncEnumerable<string> GenerateTextAsync(string prompt, TextGenerationOptions options, CancellationToken cancellationToken = default)
+        {
+            return _executor.InferAsync(prompt, OptionsToParams(options), cancellationToken: cancellationToken);
+        }
+
+        private static InferenceParams OptionsToParams(TextGenerationOptions options)
+        {
+            return new InferenceParams()
+            {
+                AntiPrompts = options.StopSequences,
+                Temperature = (float)options.Temperature,
+                MaxTokens = options.MaxTokens ?? 1024,
+                FrequencyPenalty = (float)options.FrequencyPenalty,
+                PresencePenalty = (float)options.PresencePenalty,
+                TopP = (float)options.TopP,
+            };
+        }
+    }
+}

LLamaSharp.sln

@@ -15,6 +15,8 @@ Project("{9A19103F-16F7-4668-BE54-9A1E7A4F7556}") = "LLama.Web", "LLama.Web\LLam
 EndProject
 Project("{9A19103F-16F7-4668-BE54-9A1E7A4F7556}") = "LLamaSharp.SemanticKernel", "LLama.SemanticKernel\LLamaSharp.SemanticKernel.csproj", "{D98F93E3-B344-4F9D-86BB-FDBF6768B587}"
 EndProject
+Project("{9A19103F-16F7-4668-BE54-9A1E7A4F7556}") = "LLamaSharp.KernelMemory", "LLama.KernelMemory\LLamaSharp.KernelMemory.csproj", "{E5589AE7-B86F-4343-A1CC-8E5D34596E52}"
+EndProject
 Global
 	GlobalSection(SolutionConfigurationPlatforms) = preSolution
 		Debug|Any CPU = Debug|Any CPU
@@ -97,6 +99,18 @@ Global
 		{D98F93E3-B344-4F9D-86BB-FDBF6768B587}.Release|Any CPU.Build.0 = Release|Any CPU
 		{D98F93E3-B344-4F9D-86BB-FDBF6768B587}.Release|x64.ActiveCfg = Release|Any CPU
 		{D98F93E3-B344-4F9D-86BB-FDBF6768B587}.Release|x64.Build.0 = Release|Any CPU
+		{E5589AE7-B86F-4343-A1CC-8E5D34596E52}.Debug|Any CPU.ActiveCfg = Debug|Any CPU
+		{E5589AE7-B86F-4343-A1CC-8E5D34596E52}.Debug|Any CPU.Build.0 = Debug|Any CPU
+		{E5589AE7-B86F-4343-A1CC-8E5D34596E52}.Debug|x64.ActiveCfg = Debug|Any CPU
+		{E5589AE7-B86F-4343-A1CC-8E5D34596E52}.Debug|x64.Build.0 = Debug|Any CPU
+		{E5589AE7-B86F-4343-A1CC-8E5D34596E52}.GPU|Any CPU.ActiveCfg = Debug|Any CPU
+		{E5589AE7-B86F-4343-A1CC-8E5D34596E52}.GPU|Any CPU.Build.0 = Debug|Any CPU
+		{E5589AE7-B86F-4343-A1CC-8E5D34596E52}.GPU|x64.ActiveCfg = Debug|Any CPU
+		{E5589AE7-B86F-4343-A1CC-8E5D34596E52}.GPU|x64.Build.0 = Debug|Any CPU
+		{E5589AE7-B86F-4343-A1CC-8E5D34596E52}.Release|Any CPU.ActiveCfg = Release|Any CPU
+		{E5589AE7-B86F-4343-A1CC-8E5D34596E52}.Release|Any CPU.Build.0 = Release|Any CPU
+		{E5589AE7-B86F-4343-A1CC-8E5D34596E52}.Release|x64.ActiveCfg = Release|Any CPU
+		{E5589AE7-B86F-4343-A1CC-8E5D34596E52}.Release|x64.Build.0 = Release|Any CPU
 	EndGlobalSection
 	GlobalSection(SolutionProperties) = preSolution
 		HideSolutionNode = FALSE