Merge branch 'SciSharp:master' into master

LLama/Native/LLamaContextParams.cs

@@ -32,7 +32,7 @@ public struct LLamaContextParams
         /// <summary>
         /// rms norm epsilon (TEMP - will be moved to model hparams)
         /// </summary>
-       public float rms_norm_eps;
+        public float rms_norm_eps;
 
         /// <summary>
         /// number of layers to store in VRAM
@@ -76,49 +76,82 @@ public struct LLamaContextParams
         /// <summary>
         /// if true, reduce VRAM usage at the cost of performance
         /// </summary>
-        [MarshalAs(UnmanagedType.I1)]
-        public bool low_vram;
+        public bool low_vram
+        {
+            get => Utils.SignedByteToBool(_low_vram);
+            set => _low_vram = Utils.BoolToSignedByte(value);
+        }
+        private sbyte _low_vram;
 
         /// <summary>
         /// if true, use experimental mul_mat_q kernels
         /// </summary>
-        [MarshalAs(UnmanagedType.I1)] public bool mul_mat_q;
+        public bool mul_mat_q
+        {
+            get => Utils.SignedByteToBool(_mul_mat_q);
+            set => _mul_mat_q = Utils.BoolToSignedByte(value);
+        }
+        private sbyte _mul_mat_q;
 
         /// <summary>
         /// use fp16 for KV cache
         /// </summary>
-        [MarshalAs(UnmanagedType.I1)]
-        public bool f16_kv;
+        public bool f16_kv
+        {
+            get => Utils.SignedByteToBool(_f16_kv);
+            set => _f16_kv = Utils.BoolToSignedByte(value);
+        }
+        private sbyte _f16_kv;
 
         /// <summary>
         /// the llama_eval() call computes all logits, not just the last one
         /// </summary>
-        [MarshalAs(UnmanagedType.I1)]
-        public bool logits_all;
+        public bool logits_all
+        {
+            get => Utils.SignedByteToBool(_logits_all);
+            set => _logits_all = Utils.BoolToSignedByte(value);
+        }
+        private sbyte _logits_all;
 
         /// <summary>
         /// only load the vocabulary, no weights
         /// </summary>
-        [MarshalAs(UnmanagedType.I1)] 
-        public bool vocab_only;
+        public bool vocab_only
+        {
+            get => Utils.SignedByteToBool(_vocab_only);
+            set => _vocab_only = Utils.BoolToSignedByte(value);
+        }
+        private sbyte _vocab_only;
 
         /// <summary>
         /// use mmap if possible
         /// </summary>
-        [MarshalAs(UnmanagedType.I1)] 
-        public bool use_mmap;
+        public bool use_mmap
+        {
+            get => Utils.SignedByteToBool(_use_mmap);
+            set => _use_mmap = Utils.BoolToSignedByte(value);
+        }
+        private sbyte _use_mmap;
 
         /// <summary>
         /// force system to keep model in RAM
         /// </summary>
-        [MarshalAs(UnmanagedType.I1)] 
-        public bool use_mlock;
+        public bool use_mlock
+        {
+            get => Utils.SignedByteToBool(_use_mlock);
+            set => _use_mlock = Utils.BoolToSignedByte(value);
+        }
+        private sbyte _use_mlock;
 
         /// <summary>
         /// embedding mode only
         /// </summary>
-        [MarshalAs(UnmanagedType.I1)] 
-        public bool embedding;
+        public bool embedding
+        {
+            get => Utils.SignedByteToBool(_embedding);
+            set => _embedding = Utils.BoolToSignedByte(value);
+        }
+        private sbyte _embedding;
     }
 }
 

LLama/Native/LLamaModelQuantizeParams.cs

@@ -20,13 +20,21 @@ public struct LLamaModelQuantizeParams
         /// <summary>
         /// allow quantizing non-f32/f16 tensors
         /// </summary>
-        [MarshalAs(UnmanagedType.I1)]
-        public bool allow_requantize;
+        public bool allow_requantize
+        {
+            get => Utils.SignedByteToBool(_allow_requantize);
+            set => _allow_requantize = Utils.BoolToSignedByte(value);
+        }
+        private sbyte _allow_requantize;
 
         /// <summary>
         /// quantize output.weight
         /// </summary>
-        [MarshalAs(UnmanagedType.I1)]
-        public bool quantize_output_tensor;
+        public bool quantize_output_tensor
+        {
+            get => Utils.SignedByteToBool(_quantize_output_tensor);
+            set => _quantize_output_tensor = Utils.BoolToSignedByte(value);
+        }
+        private sbyte _quantize_output_tensor;
     }
 }

LLama/Native/LLamaTokenDataArray.cs

@@ -51,8 +51,12 @@ public struct LLamaTokenDataArrayNative
         /// <summary>
         /// Indicates if the items in the array are sorted
         /// </summary>
-        [MarshalAs(UnmanagedType.I1)]
-        public bool sorted;
+        public bool sorted
+        {
+            get => Utils.SignedByteToBool(_sorted);
+            set => _sorted = Utils.BoolToSignedByte(value);
+        }
+        private sbyte _sorted;
 
         /// <summary>
         /// Create a new LLamaTokenDataArrayNative around the data in the LLamaTokenDataArray 

LLama/OldVersion/Utils.cs

@@ -82,5 +82,6 @@ public static unsafe string PtrToStringUTF8(IntPtr ptr)
             return Encoding.UTF8.GetString(bytes.ToArray());
 #endif
         }
+
     }
 }

LLama/Utils.cs

@@ -84,5 +84,26 @@ public static string PtrToString(IntPtr ptr, Encoding encoding)
             }
 #endif
         }
+
+        /// <summary>
+        /// Converts a bool "value" to a signed byte of "1" for true and "0" for false to be compatible with a 1 byte C-style bool.
+        /// </summary>
+        /// <param name="value"></param>
+        /// <returns></returns>
+        public static sbyte BoolToSignedByte(bool value)
+        {
+            return value ? (sbyte)1 : (sbyte)0;
+        }
+
+        /// <summary>
+        /// Converts a sbyte "value" to a C# bool.
+        /// </summary>
+        /// <param name="value"></param>
+        /// <returns></returns>
+        public static bool SignedByteToBool(sbyte value)
+        {
+            return value > 0 ? true : false;
+        }
+
     }
 }

README.md

@@ -73,7 +73,7 @@ LLamaSharp provides two ways to run inference: `LLamaExecutor` and `ChatSession`
 using LLama.Common;
 using LLama;
 
-string modelPath = "<Your model path>" // change it to your own model path
+string modelPath = "<Your model path>"; // change it to your own model path
 var prompt = "Transcript of a dialog, where the User interacts with an Assistant named Bob. Bob is helpful, kind, honest, good at writing, and never fails to answer the User's requests immediately and with precision.\r\n\r\nUser: Hello, Bob.\r\nBob: Hello. How may I help you today?\r\nUser: Please tell me the largest city in Europe.\r\nBob: Sure. The largest city in Europe is Moscow, the capital of Russia.\r\nUser:"; // use the "chat-with-bob" prompt here.
 
 // Initialize a chat session