scisharp
/
LLamaSharp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Allow setting ChatRequestSettings Defaults and ChatSession

This commit is contained in:
Tim Miller 2023-09-11 19:21:51 +09:00
parent 466722dcff
commit dced651f8b
1 changed files with 23 additions and 22 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

45
LLama.SemanticKernel/ChatCompletion/LLamaSharpChatCompletion.cs
View File

@ -1,13 +1,6 @@
using LLama;
using Microsoft.SemanticKernel.AI.ChatCompletion;
using System;
using System.Collections.Generic;
using System.IO;
using System.Linq;
using System.Runtime.CompilerServices;
using System.Text;
using System.Threading;
using System.Threading.Tasks;
namespace LLamaSharp.SemanticKernel.ChatCompletion;
@ -19,12 +12,32 @@ public sealed class LLamaSharpChatCompletion : IChatCompletion
private const string UserRole = "user:";
private const string AssistantRole = "assistant:";
private ChatSession session;
private ChatRequestSettings defaultRequestSettings;
public LLamaSharpChatCompletion(InteractiveExecutor model)
public LLamaSharpChatCompletion(InteractiveExecutor model, ChatRequestSettings? defaultRequestSettings = default)
{
this.session = new ChatSession(model)
.WithHistoryTransform(new HistoryTransform())
.WithOutputTransform(new LLamaTransforms.KeywordTextOutputStreamTransform(new string[] { UserRole, AssistantRole }));
this.defaultRequestSettings = defaultRequestSettings ??= new ChatRequestSettings()
{
MaxTokens = 256,
Temperature = 0,
TopP = 0,
StopSequences = new List<string> { }
};
}
public LLamaSharpChatCompletion(ChatSession session, ChatRequestSettings? defaultRequestSettings = default)
{
this.session = session;
this.defaultRequestSettings = defaultRequestSettings ??= new ChatRequestSettings()
{
MaxTokens = 256,
Temperature = 0,
TopP = 0,
StopSequences = new List<string> { }
};
}
/// <inheritdoc/>
@ -43,13 +56,7 @@ public sealed class LLamaSharpChatCompletion : IChatCompletion
/// <inheritdoc/>
public async Task<IReadOnlyList<IChatResult>> GetChatCompletionsAsync(ChatHistory chat, ChatRequestSettings? requestSettings = null, CancellationToken cancellationToken = default)
{
requestSettings ??= new ChatRequestSettings()
{
MaxTokens = 256,
Temperature = 0,
TopP = 0,
StopSequences = new List<string> { }
};
requestSettings = requestSettings ?? this.defaultRequestSettings;
var result = this.session.ChatAsync(chat.ToLLamaSharpChatHistory(), requestSettings.ToLLamaSharpInferenceParams(), cancellationToken);
@ -59,13 +66,7 @@ public sealed class LLamaSharpChatCompletion : IChatCompletion
/// <inheritdoc/>
public async IAsyncEnumerable<IChatStreamingResult> GetStreamingChatCompletionsAsync(ChatHistory chat, ChatRequestSettings? requestSettings = null, [EnumeratorCancellation] CancellationToken cancellationToken = default)
{
requestSettings ??= new ChatRequestSettings()
{
MaxTokens = 256,
Temperature = 0,
TopP = 0,
StopSequences = new List<string> { }
};
requestSettings = requestSettings ?? this.defaultRequestSettings;
var result = this.session.ChatAsync(chat.ToLLamaSharpChatHistory(), requestSettings.ToLLamaSharpInferenceParams(), cancellationToken);