SciSharp · martindevans · Jul 31, 2024 · Jul 26, 2024
diff --git a/LLama.Examples/Examples/ChatSessionStripRoleName.cs b/LLama.Examples/Examples/ChatSessionStripRoleName.cs
@@ -1,4 +1,5 @@
-using LLama.Common;
+using LLama.Common;
+using LLama.Sampling;
 
 namespace LLama.Examples.Examples;
 
@@ -27,9 +28,12 @@ public static async Task Run()
             new string[] { "User:", "Assistant:" },
             redundancyLength: 8));
 
-        InferenceParams inferenceParams = new InferenceParams()
+        var inferenceParams = new InferenceParams
         {
-            Temperature = 0.9f,
+            SamplingPipeline = new DefaultSamplingPipeline
+            {
+                Temperature = 0.9f
+            },
             AntiPrompts = new List<string> { "User:" }
         };
 

diff --git a/LLama.Examples/Examples/ChatSessionWithHistory.cs b/LLama.Examples/Examples/ChatSessionWithHistory.cs
@@ -1,4 +1,5 @@
 using LLama.Common;
+using LLama.Sampling;
 
 namespace LLama.Examples.Examples;
 
@@ -39,9 +40,12 @@ public static async Task Run()
             new string[] { "User:", "Assistant:" },
             redundancyLength: 8));
 
-        InferenceParams inferenceParams = new InferenceParams()
+        var inferenceParams = new InferenceParams
         {
-            Temperature = 0.9f,
+            SamplingPipeline = new DefaultSamplingPipeline
+            {
+                Temperature = 0.9f
+            },
             AntiPrompts = new List<string> { "User:" }
         };
 

diff --git a/LLama.Examples/Examples/ChatSessionWithRestart.cs b/LLama.Examples/Examples/ChatSessionWithRestart.cs
@@ -1,4 +1,5 @@
 using LLama.Common;
+using LLama.Sampling;
 
 namespace LLama.Examples.Examples;
 
@@ -29,9 +30,12 @@ public static async Task Run()
         ChatSession session = new ChatSession(executor);
         session.LoadSession(resetState);
 
-        InferenceParams inferenceParams = new InferenceParams()
+        var inferenceParams = new InferenceParams
         {
-            Temperature = 0.9f,
+            SamplingPipeline = new DefaultSamplingPipeline
+            {
+                Temperature = 0.9f
+            },
             AntiPrompts = new List<string> { "User:" }
         };
 

diff --git a/LLama.Examples/Examples/ChatSessionWithRoleName.cs b/LLama.Examples/Examples/ChatSessionWithRoleName.cs
@@ -1,4 +1,5 @@
-using LLama.Common;
+using LLama.Common;
+using LLama.Sampling;
 
 namespace LLama.Examples.Examples;
 
@@ -22,9 +23,12 @@ public static async Task Run()
 
         ChatSession session = new(executor, chatHistory);
 
-        InferenceParams inferenceParams = new InferenceParams()
+        var inferenceParams = new InferenceParams
         {
-            Temperature = 0.9f,
+            SamplingPipeline = new DefaultSamplingPipeline
+            {
+                Temperature = 0.9f
+            },
             AntiPrompts = new List<string> { "User:" }
         };
 

diff --git a/LLama.Examples/Examples/CodingAssistant.cs b/LLama.Examples/Examples/CodingAssistant.cs
@@ -1,4 +1,6 @@
-namespace LLama.Examples.Examples
+using LLama.Sampling;
+
+namespace LLama.Examples.Examples
 {
     using LLama.Common;
     using System;
@@ -40,9 +42,12 @@ public static async Task Run()
                 "\nWrite 'exit' to exit");
             Console.ForegroundColor = ConsoleColor.White;
 
-            var inferenceParams = new InferenceParams()
+            var inferenceParams = new InferenceParams
             {
-                Temperature = 0.8f,
+                SamplingPipeline = new DefaultSamplingPipeline
+                {
+                    Temperature = 0.8f
+                },
                 MaxTokens = -1,
             };
 

diff --git a/LLama.Examples/Examples/GetEmbeddings.cs b/LLama.Examples/Examples/GetEmbeddings.cs
@@ -1,4 +1,4 @@
-using LLama.Common;
+using LLama.Common;
 
 namespace LLama.Examples.Examples
 {

diff --git a/LLama.Examples/Examples/GrammarJsonResponse.cs b/LLama.Examples/Examples/GrammarJsonResponse.cs
@@ -1,5 +1,6 @@
-using LLama.Common;
+using LLama.Common;
 using LLama.Grammars;
+using LLama.Sampling;
 
 namespace LLama.Examples.Examples
 {
@@ -27,10 +28,13 @@ public static async Task Run()
             using var grammarInstance = grammar.CreateInstance();
             var inferenceParams = new InferenceParams()
             {
-                Temperature = 0.6f,
+                SamplingPipeline = new DefaultSamplingPipeline
+                {
+                    Temperature = 0.6f,
+                    Grammar = grammarInstance
+                },
                 AntiPrompts = new List<string> { "Question:", "#", "Question: ", ".\n" },
                 MaxTokens = 50,
-                Grammar = grammarInstance
             };
 
             while (true)

diff --git a/LLama.Examples/Examples/InstructModeExecute.cs b/LLama.Examples/Examples/InstructModeExecute.cs
@@ -1,4 +1,5 @@
-using LLama.Common;
+using LLama.Common;
+using LLama.Sampling;
 
 namespace LLama.Examples.Examples
 {
@@ -25,7 +26,14 @@ public static async Task Run()
                 "make friend with human, no less than 200 words.\"");
             Console.ForegroundColor = ConsoleColor.White;
 
-            var inferenceParams = new InferenceParams() { Temperature = 0.8f, MaxTokens = 600 };
+            var inferenceParams = new InferenceParams
+            {
+                SamplingPipeline = new DefaultSamplingPipeline
+                {
+                    Temperature = 0.8f
+                },
+                MaxTokens = 600
+            };
 
             while (true)
             {

diff --git a/LLama.Examples/Examples/LLama3ChatSession.cs b/LLama.Examples/Examples/LLama3ChatSession.cs
@@ -1,4 +1,5 @@
-using LLama.Common;
+using LLama.Common;
+using LLama.Sampling;
 using LLama.Transformers;
 
 namespace LLama.Examples.Examples;
@@ -37,10 +38,14 @@ public static async Task Run()
             [model.Tokens.EndOfTurnToken!, "�"],
             redundancyLength: 5));
 
-        var inferenceParams = new InferenceParams()
+        var inferenceParams = new InferenceParams
         {
+            SamplingPipeline = new DefaultSamplingPipeline
+            {
+                Temperature = 0.6f
+            },
+
             MaxTokens = -1, // keep generating tokens until the anti prompt is encountered
-            Temperature = 0.6f,
             AntiPrompts = [model.Tokens.EndOfTurnToken!] // model specific end of turn string
         };
 

diff --git a/LLama.Examples/Examples/LlavaInteractiveModeExecute.cs b/LLama.Examples/Examples/LlavaInteractiveModeExecute.cs
@@ -1,7 +1,8 @@
-using System.Text.RegularExpressions;
+using System.Text.RegularExpressions;
 using LLama.Common;
 using Spectre.Console;
 using LLama.Native;
+using LLama.Sampling;
 
 namespace LLama.Examples.Examples
 {
@@ -30,9 +31,19 @@ public static async Task Run()
 
             Console.ForegroundColor = ConsoleColor.Yellow;
             Console.WriteLine("The executor has been enabled. In this example, the prompt is printed, the maximum tokens is set to {0} and the context size is {1}.", maxTokens, parameters.ContextSize );
-            Console.WriteLine("To send an image, enter its filename in curly braces, like this {c:/image.jpg}.");  
+            Console.WriteLine("To send an image, enter its filename in curly braces, like this {c:/image.jpg}.");
 
-            var inferenceParams = new InferenceParams() { Temperature = 0.1f, AntiPrompts = new List<string> { "\nUSER:" }, MaxTokens = maxTokens };
+            var inferenceParams = new InferenceParams
+            {
+                SamplingPipeline = new DefaultSamplingPipeline
+                {
+                    Temperature = 0.1f
+                },
+
+                AntiPrompts = new List<string> { "\nUSER:" },
+                MaxTokens = maxTokens
+
+            };
 
             do
             {

diff --git a/LLama.Examples/Examples/LoadAndSaveSession.cs b/LLama.Examples/Examples/LoadAndSaveSession.cs
@@ -1,4 +1,5 @@
-using LLama.Common;
+using LLama.Common;
+using LLama.Sampling;
 
 namespace LLama.Examples.Examples
 {
@@ -35,7 +36,10 @@ in session.ChatAsync(
                         new ChatHistory.Message(AuthorRole.User, prompt),
                         new InferenceParams()
                         {
-                            Temperature = 0.6f,
+                            SamplingPipeline = new DefaultSamplingPipeline
+                            {
+                                Temperature = 0.6f
+                            },
                             AntiPrompts = new List<string> { "User:" }
                         }))
                 {

diff --git a/LLama.Examples/Examples/LoadAndSaveState.cs b/LLama.Examples/Examples/LoadAndSaveState.cs
@@ -1,4 +1,5 @@
-using LLama.Common;
+using LLama.Common;
+using LLama.Sampling;
 
 namespace LLama.Examples.Examples
 {
@@ -27,7 +28,15 @@ public static async Task Run()
             Console.ForegroundColor = ConsoleColor.White;
             Console.Write(prompt);
 
-            var inferenceParams = new InferenceParams() { Temperature = 0.6f, AntiPrompts = new List<string> { "User:" } };
+            var inferenceParams = new InferenceParams
+            {
+                SamplingPipeline = new DefaultSamplingPipeline
+                {
+                    Temperature = 0.6f
+                },
+
+                AntiPrompts = new List<string> { "User:" }
+            };
 
             while (true)
             {

diff --git a/LLama.Examples/Examples/StatelessModeExecute.cs b/LLama.Examples/Examples/StatelessModeExecute.cs
@@ -1,5 +1,6 @@
-using LLama.Common;
+using LLama.Common;
 using LLama.Examples.Extensions;
+using LLama.Sampling;
 
 namespace LLama.Examples.Examples
 {
@@ -24,7 +25,16 @@ public static async Task Run()
                 "a prompt for it yourself!");
             Console.ForegroundColor = ConsoleColor.White;
 
-            var inferenceParams = new InferenceParams() { Temperature = 0.6f, AntiPrompts = new List<string> { "Question:", "#", "Question: ", ".\n" }, MaxTokens = 50 };
+            var inferenceParams = new InferenceParams
+            {
+                SamplingPipeline = new DefaultSamplingPipeline
+                {
+                    Temperature = 0.6f
+                },
+
+                AntiPrompts = new List<string> { "Question:", "#", "Question: ", ".\n" },
+                MaxTokens = 50
+            };
 
             while (true)
             {

diff --git a/LLama.Examples/Examples/TalkToYourself.cs b/LLama.Examples/Examples/TalkToYourself.cs
@@ -1,6 +1,7 @@
-using System.Text;
+using System.Text;
 using LLama.Abstractions;
 using LLama.Common;
+using LLama.Sampling;
 
 namespace LLama.Examples.Examples
 {
@@ -43,11 +44,13 @@ private static async Task<string> Prompt(ILLamaExecutor executor, ConsoleColor c
         {
             var inferenceParams = new InferenceParams
             {
-                Temperature = 0.9f,
-                AntiPrompts = new List<string> { "Alice:", "Bob:", "User:" },
+                SamplingPipeline = new Mirostat2SamplingPipeline
+                {
+                    Tau = 10
+                },
+
+                AntiPrompts = [ "Alice:", "Bob:", "User:" ],
                 MaxTokens = 128,
-                Mirostat = MirostatType.Mirostat2,
-                MirostatTau = 10,
             };
 
             Console.ForegroundColor = ConsoleColor.White;