Distinct Deepseek-R1-Distill-Qwen from Qwen2

Author: orionpapadakis

src/main/java/com/example/model/ModelType.java

@@ -50,6 +50,13 @@ public Model loadModel(FileChannel fileChannel, GGUF gguf, int contextLength, bo
         }
     },
 
+    DEEPSEEK_R1_DISTILL_QWEN {
+        @Override
+        public Model loadModel(FileChannel fileChannel, GGUF gguf, int contextLength, boolean loadWeights) {
+            return new Qwen2ModelLoader(fileChannel, gguf, contextLength, loadWeights).loadModel();
+        }
+    },
+
     UNKNOWN {
         @Override
         public Model loadModel(FileChannel fileChannel, GGUF gguf, int contextLength, boolean loadWeights) {
@@ -59,4 +66,8 @@ public Model loadModel(FileChannel fileChannel, GGUF gguf, int contextLength, bo
 
     // Abstract method that each enum constant must implement
     public abstract Model loadModel(FileChannel fileChannel, GGUF gguf, int contextLength, boolean loadWeights);
+
+    public boolean isDeepSeekR1() {
+        return this == DEEPSEEK_R1_DISTILL_QWEN;
+    }
 }

src/main/java/com/example/model/loader/ModelLoader.java

@@ -61,10 +61,12 @@ private static ModelType detectModelType(Map<String, Object> metadata) {
                 return ModelType.MISTRAL;
             } else if (lowerName.contains("llama")) {
                 return ModelType.LLAMA_3;
-            } else if (lowerName.contains("qwen2") || lowerName.contains("deepseek r1 distill")) {
+            } else if (lowerName.contains("qwen2")) {
                 return ModelType.QWEN_2;
             } else if (lowerName.contains("qwen3")) {
                 return ModelType.QWEN_3;
+            } else if (lowerName.contains("deepseek r1 distill")) {
+                return ModelType.DEEPSEEK_R1_DISTILL_QWEN;
             }
         }
 

src/main/java/com/example/tornadovm/TornadoVMMasterPlan.java

@@ -98,7 +98,7 @@ public static TornadoVMMasterPlan initializeTornadoVMPlan(State state, Model mod
     TornadoVMLayerPlanner createPlanner(State state, Model model) {
         return switch (model.getModelType()) {
             case LLAMA_3, MISTRAL -> new TornadoVMLayerPlanner(state, model);
-            case QWEN_2 -> throw new UnsupportedOperationException("TornadoVM QWEN 2 not supported");
+            case QWEN_2, DEEPSEEK_R1_DISTILL_QWEN -> throw new UnsupportedOperationException("TornadoVM QWEN 2 not supported");
             case QWEN_3 -> new Qwen3TornadoVMLayerPlanner((Qwen3State) state, model);
             case UNKNOWN -> throw new UnsupportedOperationException("Unknown model type");
         };