feat: add Qwen 2.5 Coder model

src/main/java/ee/carlrobert/codegpt/completions/HuggingFaceModel.java

@@ -120,6 +120,18 @@ public enum HuggingFaceModel {
   CODE_QWEN_1_5_7B_Q6_K(7, 6, "Qwen_-_CodeQwen1.5-7B-Chat-gguf",
           "CodeQwen1.5-7B-Chat.Q6_K.gguf", "RichardErkhov", 6.38),
 
+  CODE_QWEN_2_5_1_5B_Q6_K(1, 6, "Qwen2.5-Coder-1.5B-Instruct-GGUF",
+      "qwen2.5-coder-1.5b-instruct-q6_k.gguf", "Qwen", 1.46),
+  CODE_QWEN_2_5_1_5B_Q8_0(1, 8, "Qwen2.5-Coder-1.5B-Instruct-GGUF",
+      "qwen2.5-coder-1.5b-instruct-q8_0.gguf", "Qwen", 1.89),
+
+  CODE_QWEN_2_5_7B_Q4_K_M(7, 4, "Qwen2.5-Coder-7B-Instruct-GGUF",
+      "Qwen2.5-Coder-7B-Instruct-Q4_K_M.gguf", "bartowski", 4.68),
+  CODE_QWEN_2_5_7B_Q6_K(7, 6, "Qwen2.5-Coder-7B-Instruct-GGUF",
+      "Qwen2.5-Coder-7B-Instruct-Q6_K.gguf", "bartowski", 6.25),
+  CODE_QWEN_2_5_7B_Q8_0(7, 8, "Qwen2.5-Coder-7B-Instruct-GGUF",
+      "Qwen2.5-Coder-7B-Instruct-Q8_0.gguf", "bartowski", 8.1),
+
   STABLE_CODE_3B_Q3_K_M(SC3, 3, "stable-code-instruct-3b-Q3_K_M.gguf", 1.39),
   STABLE_CODE_3B_Q4_K_M(SC3, 4, "stable-code-instruct-3b-Q4_K_M.gguf", 1.71),
   STABLE_CODE_3B_Q5_K_M(SC3, 5, "stable-code-instruct-3b-Q5_K_M.gguf", 1.99),

src/main/java/ee/carlrobert/codegpt/completions/llama/LlamaModel.java

@@ -170,6 +170,23 @@ public enum LlamaModel {
           HuggingFaceModel.CODE_QWEN_1_5_7B_Q4_K_M,
           HuggingFaceModel.CODE_QWEN_1_5_7B_Q5_K_M,
           HuggingFaceModel.CODE_QWEN_1_5_7B_Q6_K)),
+  CODE_QWEN2_5_CODER(
+      "CodeQwen2.5 Coder", """
+          Qwen2.5-Coder is the latest series of Code-Specific Qwen large language models (formerly known as CodeQwen).\
+          It brings the following improvements upon CodeQwen1.5:
+          
+          - Significantly improvements in code generation, code reasoning and code fixing. Base on the strong Qwen2.5, we scale up the training tokens into 5.5 trillion including source code, text-code grounding, Synthetic data, etc.
+          - A more comprehensive foundation for real-world applications such as Code Agents. Not only enhancing coding capabilities but also maintaining its strengths in mathematics and general competencies.
+          - Long-context Support up to 128K tokens.
+          """,
+      PromptTemplate.CODE_QWEN,
+      InfillPromptTemplate.CODE_QWEN_2_5,
+      List.of(
+          HuggingFaceModel.CODE_QWEN_2_5_1_5B_Q6_K,
+          HuggingFaceModel.CODE_QWEN_2_5_1_5B_Q8_0,
+          HuggingFaceModel.CODE_QWEN_2_5_7B_Q4_K_M,
+          HuggingFaceModel.CODE_QWEN_2_5_7B_Q6_K,
+          HuggingFaceModel.CODE_QWEN_2_5_7B_Q8_0)),
   STABLE_CODE(
       "Stable Code Instruct", """
       stable-code-instruct-3b is a 2.7B billion parameter decoder-only language model tuned from \

src/main/java/ee/carlrobert/codegpt/completions/llama/PromptTemplate.java

@@ -162,7 +162,7 @@ public enum PromptTemplate {
               .toString();
     }
   },
-  CODE_QWEN("CodeQwen1.5", List.of("<|endoftext|>")) {
+  CODE_QWEN("CodeQwen", List.of("<|endoftext|>")) {
     @Override
     public String buildPrompt(String systemPrompt, String userPrompt, List<Message> history) {
       StringBuilder prompt = new StringBuilder();

src/main/java/ee/carlrobert/codegpt/settings/service/llama/LlamaSettingsState.java

@@ -13,7 +13,7 @@ public class LlamaSettingsState {
   private boolean runLocalServer = SystemInfoRt.isUnix;
   private boolean useCustomModel;
   private String customLlamaModelPath = "";
-  private HuggingFaceModel huggingFaceModel = HuggingFaceModel.CODE_LLAMA_7B_Q4;
+  private HuggingFaceModel huggingFaceModel = HuggingFaceModel.CODE_QWEN_2_5_1_5B_Q8_0;
   private PromptTemplate localModelPromptTemplate = PromptTemplate.LLAMA;
   private PromptTemplate remoteModelPromptTemplate = PromptTemplate.LLAMA;
   private InfillPromptTemplate localModelInfillPromptTemplate = InfillPromptTemplate.CODE_LLAMA;

src/main/kotlin/ee/carlrobert/codegpt/codecompletions/CodeCompletionService.kt

@@ -8,6 +8,7 @@ import ee.carlrobert.codegpt.codecompletions.CodeCompletionRequestFactory.buildL
 import ee.carlrobert.codegpt.codecompletions.CodeCompletionRequestFactory.buildOllamaRequest
 import ee.carlrobert.codegpt.codecompletions.CodeCompletionRequestFactory.buildOpenAIRequest
 import ee.carlrobert.codegpt.completions.CompletionClientProvider
+import ee.carlrobert.codegpt.completions.llama.LlamaModel
 import ee.carlrobert.codegpt.settings.GeneralSettings
 import ee.carlrobert.codegpt.settings.service.ServiceType
 import ee.carlrobert.codegpt.settings.service.ServiceType.*
@@ -34,7 +35,7 @@ class CodeCompletionService {
                 .body
                 .getOrDefault("model", null) as String
 
-            LLAMA_CPP -> "LlamaSettings.getCurrentState()."
+            LLAMA_CPP -> LlamaModel.findByHuggingFaceModel(LlamaSettings.getCurrentState().huggingFaceModel).label
             OLLAMA -> service<OllamaSettings>().state.model
             else -> null
         }

src/main/kotlin/ee/carlrobert/codegpt/codecompletions/InfillPromptTemplate.kt

@@ -50,6 +50,23 @@ enum class InfillPromptTemplate(val label: String, val stopTokens: List<String>?
             }
         }
     },
+    CODE_QWEN_2_5("CodeQwen2.5", listOf()) {
+        override fun buildPrompt(infillDetails: InfillRequest): String {
+            val infillPrompt =
+                "<|fim_prefix|> ${infillDetails.prefix} <|fim_suffix|>${infillDetails.suffix} <|fim_middle|>"
+            return if (infillDetails.context == null || infillDetails.context.contextElements.isEmpty()) {
+                infillPrompt
+            } else {
+                "<|repo_name|>${infillDetails.context.getRepoName()}\n" +
+                        infillDetails.context.contextElements.map {
+                            "<|file_sep|>${it.filePath()} \n" +
+                                    it.text()
+                        }.joinToString("") { it + "\n" } +
+                        "<|file_sep|>${infillDetails.context.enclosingElement.filePath()} \n" +
+                        infillPrompt
+            }
+        }
+    },
     STABILITY("Stability AI", listOf("<|endoftext|>")) {
         override fun buildPrompt(infillDetails: InfillRequest): String {
             val infillPrompt =