Replace loadTornadoTensorAsFP32 with loadTornadoTensor across model loaders for consistent tensor loading.

Author: mikepapadim

src/main/java/org/beehive/gpullama3/model/loader/Phi3ModelLoader.java

@@ -140,7 +140,7 @@ protected Weights createTornadoVMWeights(Map<String, GGMLTensorEntry> tensorEntr
 
         // Load all tensors uniformly as TornadoTensor hierarchy
         return new Phi3TornadoWeights(
-                loadTornadoTensorAsFP32(tokenEmbeddings),
+                loadTornadoTensor(tokenEmbeddings),
                 loadArrayOfTornadoTensors(nl, i -> tensorEntries.get("blk." + i + ".attn_norm.weight")),    // fp32
                 loadArrayOfTornadoTensors(nl, i -> tensorEntries.get("blk." + i + ".attn_qkv.weight")),
                 loadArrayOfTornadoTensors(nl, i -> tensorEntries.get("blk." + i + ".attn_output.weight")),

src/main/java/org/beehive/gpullama3/model/loader/Qwen2ModelLoader.java

@@ -137,7 +137,7 @@ protected Weights createTornadoVMWeights(Map<String, GGMLTensorEntry> tensorEntr
 
         // Load all tensors uniformly as TornadoTensor hierarchy
         return new Qwen2TornadoWeights(
-                loadTornadoTensorAsFP32(tokenEmbeddings),
+                loadTornadoTensor(tokenEmbeddings),
                 loadArrayOfTornadoTensors(nl, i -> tensorEntries.get("blk." + i + ".attn_norm.weight")),    // fp32
                 loadArrayOfTornadoTensors(nl, i -> tensorEntries.get("blk." + i + ".attn_q.weight")),
                 loadArrayOfTornadoTensors(nl, i -> tensorEntries.get("blk." + i + ".attn_k.weight")),

src/main/java/org/beehive/gpullama3/model/loader/Qwen3ModelLoader.java

@@ -137,7 +137,7 @@ protected Weights createTornadoVMWeights(Map<String, GGMLTensorEntry> tensorEntr
         final int nl = config.numberOfLayers();
 
         return new Qwen3TornadoWeights(
-                loadTornadoTensorAsFP32(tokenEmbeddings),
+                loadTornadoTensor(tokenEmbeddings),
                 loadArrayOfTornadoTensors(nl, i -> tensorEntries.get("blk." + i + ".attn_norm.weight")),    // fp32
                 loadArrayOfTornadoTensors(nl, i -> tensorEntries.get("blk." + i + ".attn_q.weight")),
                 loadArrayOfTornadoTensors(nl, i -> tensorEntries.get("blk." + i + ".attn_k.weight")),