Add GGUFReader

.gitignore

@@ -1,4 +1,4 @@
- local.properties
+local.properties
 .gradle
 build/
 !gradle/wrapper/gradle-wrapper.jar
@@ -20,4 +20,4 @@ out/
 .vscode/
 
 ### Mac OS ###
-.DS_Store
+.DS_Store

build.gradle.kts

@@ -1,4 +1,5 @@
 plugins {
     alias(libs.plugins.androidLibrary) apply false
     alias(libs.plugins.kotlinMultiplatform) apply  false
+    alias(libs.plugins.jetbrainsKotlinJvm) apply false
 }

gguf/.gitignore

@@ -0,0 +1 @@
+/build

gguf/build.gradle.kts

@@ -0,0 +1,18 @@
+plugins {
+    id("java-library")
+    alias(libs.plugins.jetbrainsKotlinJvm)
+}
+
+java {
+    sourceCompatibility = JavaVersion.VERSION_11
+    targetCompatibility = JavaVersion.VERSION_11
+}
+
+kotlin {
+    compilerOptions {
+        jvmTarget = org.jetbrains.kotlin.gradle.dsl.JvmTarget.JVM_11
+    }
+}
+dependencies {
+    testImplementation(libs.testng)
+}

gguf/src/main/java/sk/ai/net/gguf/Constants.kt

@@ -0,0 +1,106 @@
+package sk.ai.net.gguf
+
+/**
+ * This is a kotlin gguf reader related logic interpreted from python code "gguf-py/gguf/constants.py"
+ * of github repo "https://github.com/ggerganov/llama.cpp"
+ */
+
+//TODO convert the rest of file from constants.py
+
+const val GGUF_MAGIC = 0x46554747u
+const val GGUF_VERSION = 3
+const val GGUF_DEFAULT_ALIGNMENT = 32
+
+enum class GGMLQuantizationType(val value: Int) {
+    F32(0),
+    F16(1),
+    Q4_0(2),
+    Q4_1(3),
+    Q5_0(6),
+    Q5_1(7),
+    Q8_0(8),
+    Q8_1(9),
+    Q2_K(10),
+    Q3_K(11),
+    Q4_K(12),
+    Q5_K(13),
+    Q6_K(14),
+    Q8_K(15),
+    IQ2_XXS(16),
+    IQ2_XS(17),
+    IQ3_XXS(18),
+    IQ1_S(19),
+    IQ4_NL(20),
+    IQ3_S(21),
+    IQ2_S(22),
+    IQ4_XS(23),
+    I8(24),
+    I16(25),
+    I32(26),
+    I64(27),
+    F64(28),
+    IQ1_M(29),
+    BF16(30),
+    TQ1_0(34),
+    TQ2_0(35);
+
+    companion object {
+        fun fromValue(value: Int): GGMLQuantizationType? {
+            return values().find { it.value == value }
+        }
+    }
+}
+
+// Block size constant
+const val QK_K = 256
+
+// Quantization type and corresponding sizes
+val GGML_QUANT_SIZES: Map<GGMLQuantizationType, Pair<Int, Int>> = mapOf(
+    GGMLQuantizationType.F32 to (1 to 4),
+    GGMLQuantizationType.F16 to (1 to 2),
+    GGMLQuantizationType.Q4_0 to (32 to 2 + 16),
+    GGMLQuantizationType.Q4_1 to (32 to 2 + 2 + 16),
+    GGMLQuantizationType.Q5_0 to (32 to 2 + 4 + 16),
+    GGMLQuantizationType.Q5_1 to (32 to 2 + 2 + 4 + 16),
+    GGMLQuantizationType.Q8_0 to (32 to 2 + 32),
+    GGMLQuantizationType.Q8_1 to (32 to 4 + 4 + 32),
+    GGMLQuantizationType.Q2_K to (256 to 2 + 2 + QK_K / 16 + QK_K / 4),
+    GGMLQuantizationType.Q3_K to (256 to 2 + QK_K / 4 + QK_K / 8 + 12),
+    GGMLQuantizationType.Q4_K to (256 to 2 + 2 + QK_K / 2 + 12),
+    GGMLQuantizationType.Q5_K to (256 to 2 + 2 + QK_K / 2 + QK_K / 8 + 12),
+    GGMLQuantizationType.Q6_K to (256 to 2 + QK_K / 2 + QK_K / 4 + QK_K / 16),
+    GGMLQuantizationType.Q8_K to (256 to 4 + QK_K + QK_K / 8),
+    GGMLQuantizationType.IQ2_XXS to (256 to 2 + QK_K / 4),
+    GGMLQuantizationType.IQ2_XS to (256 to 2 + QK_K / 4 + QK_K / 32),
+    GGMLQuantizationType.IQ3_XXS to (256 to 2 + QK_K / 4 + QK_K / 8),
+    GGMLQuantizationType.IQ1_S to (256 to 2 + QK_K / 8 + QK_K / 16),
+    GGMLQuantizationType.IQ4_NL to (32 to 2 + 16),
+    GGMLQuantizationType.IQ3_S to (256 to 2 + QK_K / 4 + QK_K / 8 + QK_K / 32 + 4),
+    GGMLQuantizationType.IQ2_S to (256 to 2 + QK_K / 4 + QK_K / 16),
+    GGMLQuantizationType.IQ4_XS to (256 to 2 + 2 + QK_K / 2 + QK_K / 64),
+    GGMLQuantizationType.I8 to (1 to 1),
+    GGMLQuantizationType.I16 to (1 to 2),
+    GGMLQuantizationType.I32 to (1 to 4),
+    GGMLQuantizationType.I64 to (1 to 8),
+    GGMLQuantizationType.F64 to (1 to 8),
+    GGMLQuantizationType.IQ1_M to (256 to QK_K / 8 + QK_K / 16 + QK_K / 32),
+    GGMLQuantizationType.BF16 to (1 to 2),
+    GGMLQuantizationType.TQ1_0 to (256 to 2 + 4 * 13),
+    GGMLQuantizationType.TQ2_0 to (256 to 2 + 64)
+)
+
+enum class GGUFValueType(val value: Int) {
+    UINT8(0),
+    INT8(1),
+    UINT16(2),
+    INT16(3),
+    UINT32(4),
+    INT32(5),
+    FLOAT32(6),
+    BOOL(7),
+    STRING(8),
+    ARRAY(9),
+    UINT64(10),
+    INT64(11),
+    FLOAT64(12)
+}