feat: Qwen2 support

Libraries/LLM/Models.swift

@@ -84,6 +84,13 @@ extension ModelConfiguration {
     ) { prompt in
         "<start_of_turn>user \(prompt)<end_of_turn><start_of_turn>model"
     }
+    
+    public static let qwen205b4bit = ModelConfiguration(
+        id: "mlx-community/Qwen1.5-0.5B-Chat-4bit",
+        overrideTokenizer: "PreTrainedTokenizer"
+    ) { prompt in
+        "<|im_start|>user \(prompt)<|im_end|><|im_start|>assistant"
+    }
 
     private enum BootstrapState {
         case idle
@@ -102,6 +109,7 @@ extension ModelConfiguration {
                 codeLlama13b4bit,
                 phi4bit,
                 gemma2bQuantized,
+                qwen205b4bit,
             ])
             bootstrapState = .bootstrapped