Release v1.2.7

.github/doxygen/Doxyfile

@@ -48,7 +48,7 @@ PROJECT_NAME           = "LLM for Unity"
 # could be handy for archiving the generated documentation or if some version
 # control system is used.
 
-PROJECT_NUMBER         = v1.2.6
+PROJECT_NUMBER = v1.2.7
 
 # Using the PROJECT_BRIEF tag one can provide an optional one line description
 # for a project that appears at the top of each page and should give viewer a

CHANGELOG.md

@@ -1,3 +1,13 @@
+## v1.2.7
+#### 🚀 Features
+
+- Add Llama 3 and Vicuna chat templates (PR: #145)
+
+#### 📦 General
+
+- Use the context size of the model by default for longer history (PR: #147)
+
+
 ## v1.2.6
 #### 🚀 Features
 
@@ -144,4 +154,4 @@
   -  closes #9
 - Fix sample scenes for different screen resolutions
   -  closes #10
-- Allow parallel prompts
+- Allow parallel prompts

CHANGELOG.release.md

@@ -1,13 +1,8 @@
 ### 🚀 Features
 
-- Add documentation (PR: #135)
-
-### 🐛 Fixes
-
-- Add server security for interceptions from external llamafile servers (PR: #132)
-- Adapt server security for macOS (PR: #137)
+- Add Llama 3 and Vicuna chat templates (PR: #145)
 
 ### 📦 General
 
-- Add sample to demonstrates the async functionality (PR: #136)
+- Use the context size of the model by default for longer history (PR: #147)
 

README.md

@@ -52,6 +52,7 @@ LLM for Unity is built on top of the awesome [llama.cpp](https://github.com/gger
 ## Games using LLM for Unity
 - [Verbal Verdict](https://store.steampowered.com/app/2778780/Verbal_Verdict/)
 - [I, Chatbot: AISYLUM](https://store.steampowered.com/app/2786750/I_Chatbot_AISYLUM)
+- [Nameless Souls of the Void](https://unicorninteractive.itch.io/nameless-souls-of-the-void)
 
 ## Setup
 _Method 1: Install using the asset store_

Runtime/LLM.cs

@@ -65,7 +65,7 @@ public class LLM : LLMClient
         [ModelAddonAdvanced] public string lora = "";
         /// <summary> Size of the prompt context (0 = context size of the model).
         /// This is the number of tokens the model can take as input when generating responses. </summary>
-        [ModelAdvanced] public int contextSize = 512;
+        [ModelAdvanced] public int contextSize = 0;
         /// <summary> Batch size for prompt processing. </summary>
         [ModelAdvanced] public int batchSize = 512;
         /// <summary> Boolean set to true if the server has started and is ready to receive requests, false otherwise. </summary>

Runtime/LLMChatTemplates.cs

@@ -40,7 +40,9 @@ static ChatTemplate()
                 new MistralInstructTemplate(),
                 new LLama2ChatTemplate(),
                 new LLama2Template(),
+                new LLama3ChatTemplate(),
                 new Phi2Template(),
+                new VicunaTemplate(),
                 new ZephyrTemplate(),
             };
 
@@ -228,7 +230,7 @@ public override string[] GetStop(string playerName, string AIName)
     public class LLama2Template : ChatTemplate
     {
         public override string GetName() { return "llama"; }
-        public override string GetDescription() { return "llama"; }
+        public override string GetDescription() { return "llama 2"; }
 
         protected override string SystemPrefix() { return "<<SYS>>\n"; }
         protected override string SystemSuffix() { return "\n<</SYS>> "; }
@@ -249,8 +251,8 @@ public override string[] GetStop(string playerName, string AIName)
     public class LLama2ChatTemplate : LLama2Template
     {
         public override string GetName() { return "llama chat"; }
-        public override string GetDescription() { return "llama (modified for chat)"; }
-        public override string[] GetNameMatches() { return new string[] {"llama"}; }
+        public override string GetDescription() { return "llama 2 (modified for chat)"; }
+        public override string[] GetNameMatches() { return new string[] {"llama-2", "llama v2"}; }
 
         protected override string PlayerPrefix(string playerName) { return "### " + playerName + ":"; }
         protected override string AIPrefix(string AIName) { return "### " + AIName + ":"; }
@@ -262,6 +264,33 @@ public override string[] GetStop(string playerName, string AIName)
         }
     }
 
+    /// @ingroup template
+    /// <summary>
+    /// Class implementing the LLama3 template for chat
+    /// </summary>
+    public class LLama3ChatTemplate : ChatTemplate
+    {
+        public override string GetName() { return "llama3 chat"; }
+        public override string GetDescription() { return "llama 3 (chat)"; }
+        public override string[] GetNameMatches() { return new string[] {"llama-3", "llama v3"}; }
+        public override string[] GetChatTemplateMatches() { return new string[] {"{% set loop_messages = messages %}{% for message in loop_messages %}{% set content = '<|start_header_id|>' + message['role'] + '<|end_header_id|>\n\n'+ message['content'] | trim + '<|eot_id|>' %}{% if loop.index0 == 0 %}{% set content = bos_token + content %}{% endif %}{{ content }}{% endfor %}{{ '<|start_header_id|>assistant<|end_header_id|>\n\n' }}"};}
+
+        protected override string PromptPrefix() { return "<|begin_of_text|>"; }
+        protected override string SystemPrefix() { return "<|start_header_id|>system<|end_header_id|>\n\n"; }
+        protected override string SystemSuffix() { return "<|eot_id|>"; }
+
+        protected override string RequestSuffix() { return "<|eot_id|>"; }
+        protected override string PairSuffix() { return "<|eot_id|>"; }
+
+        protected override string PlayerPrefix(string playerName) { return $"<|start_header_id|>{playerName}<|end_header_id|>\n\n"; }
+        protected override string AIPrefix(string AIName) { return $"<|start_header_id|>{AIName}<|end_header_id|>\n\n"; }
+
+        public override string[] GetStop(string playerName, string AIName)
+        {
+            return AddStopNewlines(new string[] { "<|eot_id|>" });
+        }
+    }
+
     /// @ingroup template
     /// <summary>
     /// Class implementing the Mistral Instruct template
@@ -328,6 +357,28 @@ public override string[] GetStop(string playerName, string AIName)
         }
     }
 
+    /// @ingroup template
+    /// <summary>
+    /// Class implementing the Vicuna template
+    /// </summary>
+    public class VicunaTemplate : ChatTemplate
+    {
+        public override string GetName() { return "vicuna"; }
+        public override string GetDescription() { return "vicuna"; }
+        public override string[] GetNameMatches() { return new string[] {"vicuna"}; }
+        public override string[] GetChatTemplateMatches() { return new string[] {"{% if not add_generation_prompt is defined %}{% set add_generation_prompt = false %}{% endif %}{% for message in messages %}{% if message['role'] == 'system' %}{{message['content'] + ' '}}{% elif message['role'] == 'user' %}{{ 'USER: ' + message['content'] + ' '}}{% elif message['role'] == 'assistant' %}{{ 'ASSISTANT: ' + message['content'] + ' '}}{% endif %}{% endfor %}{% if add_generation_prompt %}{{ 'ASSISTANT: '}}{% endif %}"}; }
+
+        protected override string SystemSuffix() { return "\n"; }
+        protected override string PlayerPrefix(string playerName) { return "\n" + playerName + ":"; }
+        protected override string AIPrefix(string AIName) { return "\n" + AIName + ":"; }
+        protected override string PrefixMessageSeparator() { return " "; }
+
+        public override string[] GetStop(string playerName, string AIName)
+        {
+            return AddStopNewlines(new string[] { playerName + ":", AIName + ":" });
+        }
+    }
+
     /// @ingroup template
     /// <summary>
     /// Class implementing the Phi-2 template

Runtime/LLMUnitySetup.cs

@@ -32,7 +32,7 @@ public class LLMUnitySetup : MonoBehaviour
     {
         // DON'T CHANGE! the version is autocompleted with a GitHub action
         /// <summary> LLM for Unity version </summary>
-        public static string Version = "v1.2.6";
+        public static string Version = "v1.2.7";
 
         /// \cond HIDE
         public static Process CreateProcess(

Tests/Runtime/TestLLMChatTemplates.cs

@@ -62,6 +62,15 @@ public void TestLLama2Chat()
             );
         }
 
+        [Test]
+        public void TestLLama3Chat()
+        {
+            Assert.AreEqual(
+                new LLama3ChatTemplate().ComputePrompt(messages, "assistant"),
+                "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nyou are a bot<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nHello, how are you?<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\nI'm doing great. How can I help you today?<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nI'd like to show off how chat templating works!<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\nchat template is awesome<|eot_id|><|start_header_id|>user<|end_header_id|>\n\ndo you think so?<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n"
+            );
+        }
+
         [Test]
         public void TestAlpaca()
         {
@@ -71,6 +80,15 @@ public void TestAlpaca()
             );
         }
 
+        [Test]
+        public void TestVicuna()
+        {
+            Assert.AreEqual(
+                new VicunaTemplate().ComputePrompt(messages, "assistant"),
+                "you are a bot\n\nuser: Hello, how are you?\nassistant: I'm doing great. How can I help you today?\nuser: I'd like to show off how chat templating works!\nassistant: chat template is awesome\nuser: do you think so?\nassistant:"
+            );
+        }
+
         [Test]
         public void TestPhi2()
         {

VERSION

@@ -1 +1 @@
-v1.2.6
+v1.2.7

package.json

@@ -1,6 +1,6 @@
 {
   "name": "ai.undream.llm",
-  "version": "1.2.6",
+  "version": "1.2.7",
   "displayName": "LLM for Unity",
   "description": "LLM for Unity allows to run and distribute Large Language Models (LLMs) in the Unity engine.",
   "unity": "2022.3",