fix: n_gpu_layers miss in llama context.

LLama/Native/LLamaContextParams.cs

@@ -18,6 +18,10 @@ namespace LLama.Native
         /// </summary>
         public int n_parts;
         /// <summary>
+        /// number of layers to store in VRAM
+        /// </summary>
+        public int n_gpu_layers;
+        /// <summary>
         /// RNG seed, -1 for random
         /// </summary>
         public int seed;

LLama/Utils.cs

@@ -18,6 +18,7 @@ namespace LLama
 
             lparams.n_ctx = @params.n_ctx;
             lparams.n_parts = @params.n_parts;
+            lparams.n_gpu_layers = @params.n_gpu_layers;
             lparams.seed = @params.seed;
             lparams.f16_kv = @params.memory_f16;
             lparams.use_mmap = @params.use_mmap;