local-llm-rn 1.0.0

package/cpp/ggml/src/ggml-webgpu/wgsl-shaders/soft_max.tmpl.wgsl

@@ -0,0 +1,345 @@
+#define(VARIANTS)
+[
+  {
+    "SHADER_NAME": "soft_max_f32",
+    "DECLS": ["BASE_BINDINGS", "NOT_INPLACE", "NO_MASK", "NO_SINK"]
+  },
+  {
+    "SHADER_NAME": "soft_max_f32_inplace",
+    "DECLS": ["BASE_BINDINGS_INPLACE", "INPLACE", "NO_MASK", "NO_SINK"]
+  },
+  {
+    "SHADER_NAME": "soft_max_f32_sink",
+    "DECLS": ["SINK_BINDINGS", "NOT_INPLACE", "NO_MASK", "SINK"]
+  },
+  {
+    "SHADER_NAME": "soft_max_f32_sink_inplace",
+    "DECLS": ["SINK_BINDINGS_INPLACE", "INPLACE", "NO_MASK", "SINK"]
+  },
+  {
+    "SHADER_NAME": "soft_max_f32_mask_f32",
+    "REPLS": {
+      "MASK_TYPE" : "f32",
+    },
+    "DECLS": ["MASK_BINDINGS", "NOT_INPLACE", "MASK", "NO_SINK"]
+  },
+  {
+    "SHADER_NAME": "soft_max_f32_mask_f32_inplace",
+    "REPLS": {
+      "MASK_TYPE" : "f32",
+    },
+    "DECLS": ["MASK_BINDINGS_INPLACE", "INPLACE", "MASK", "NO_SINK"]
+  },
+  {
+    "SHADER_NAME": "soft_max_f32_mask_f16",
+    "REPLS": {
+      "MASK_TYPE" : "f16",
+    },
+    "DECLS": ["MASK_BINDINGS", "NOT_INPLACE", "MASK", "NO_SINK"]
+  },
+  {
+    "SHADER_NAME": "soft_max_f32_mask_f16_inplace",
+    "REPLS": {
+      "MASK_TYPE" : "f16",
+    },
+    "DECLS": ["MASK_BINDINGS_INPLACE", "INPLACE", "MASK", "NO_SINK"]
+  },
+  {
+    "SHADER_NAME": "soft_max_f32_mask_f32_sink",
+    "REPLS": {
+      "MASK_TYPE" : "f32",
+    },
+    "DECLS": ["MASK_SINK_BINDINGS", "NOT_INPLACE", "MASK", "SINK"]
+  },
+  {
+    "SHADER_NAME": "soft_max_f32_mask_f32_sink_inplace",
+    "REPLS": {
+      "MASK_TYPE" : "f32",
+    },
+    "DECLS": ["MASK_SINK_BINDINGS_INPLACE", "INPLACE", "MASK", "SINK"]
+  },
+  {
+    "SHADER_NAME": "soft_max_f32_mask_f16_sink",
+    "REPLS": {
+      "MASK_TYPE" : "f16",
+    },
+    "DECLS": ["MASK_SINK_BINDINGS", "NOT_INPLACE", "MASK", "SINK"]
+  },
+  {
+    "SHADER_NAME": "soft_max_f32_mask_f16_sink_inplace",
+    "REPLS": {
+      "MASK_TYPE" : "f16",
+    },
+    "DECLS": ["MASK_SINK_BINDINGS_INPLACE", "INPLACE", "MASK", "SINK"]
+  }
+]
+#end(VARIANTS)
+
+#define(DECLS)
+
+#decl(BASE_BINDINGS)
+@group(0) @binding(1)
+var<storage, read_write> dst: array<f32>;
+
+@group(0) @binding(2)
+var<uniform> params: Params;
+#enddecl(BASE_BINDINGS)
+
+#decl(BASE_BINDINGS_INPLACE)
+@group(0) @binding(1)
+var<uniform> params: Params;
+#enddecl(BASE_BINDINGS_INPLACE)
+
+#decl(SINK_BINDINGS)
+@group(0) @binding(1)
+var<storage, read_write> sinks: array<f32>;
+
+@group(0) @binding(2)
+var<storage, read_write> dst: array<f32>;
+
+@group(0) @binding(3)
+var<uniform> params: Params;
+#enddecl(SINK_BINDINGS)
+
+#decl(SINK_BINDINGS_INPLACE)
+@group(0) @binding(1)
+var<storage, read_write> sinks: array<f32>;
+
+@group(0) @binding(2)
+var<uniform> params: Params;
+#enddecl(SINK_BINDINGS_INPLACE)
+
+#decl(MASK_BINDINGS)
+@group(0) @binding(1)
+var<storage, read_write> mask: array<{{MASK_TYPE}}>;
+
+@group(0) @binding(2)
+var<storage, read_write> dst: array<f32>;
+
+@group(0) @binding(3)
+var<uniform> params: Params;
+#enddecl(MASK_BINDINGS)
+
+#decl(MASK_BINDINGS_INPLACE)
+@group(0) @binding(1)
+var<storage, read_write> mask: array<{{MASK_TYPE}}>;
+
+@group(0) @binding(2)
+var<uniform> params: Params;
+#enddecl(MASK_BINDINGS_INPLACE)
+
+#decl(MASK_SINK_BINDINGS)
+@group(0) @binding(1)
+var<storage, read_write> mask: array<{{MASK_TYPE}}>;
+
+@group(0) @binding(2)
+var<storage, read_write> sinks: array<f32>;
+
+@group(0) @binding(3)
+var<storage, read_write> dst: array<f32>;
+
+@group(0) @binding(4)
+var<uniform> params: Params;
+#enddecl(MASK_SINK_BINDINGS)
+
+#decl(MASK_SINK_BINDINGS_INPLACE)
+@group(0) @binding(1)
+var<storage, read_write> mask: array<{{MASK_TYPE}}>;
+
+@group(0) @binding(2)
+var<storage, read_write> sinks: array<f32>;
+
+@group(0) @binding(3)
+var<uniform> params: Params;
+#enddecl(MASK_SINK_BINDINGS_INPLACE)
+
+#decl(NOT_INPLACE)
+fn inter_value(i: u32) -> f32 {
+    return dst[i];
+}
+
+fn update(i: u32, val: f32) {
+    dst[i] = val;
+}
+#enddecl(NOT_INPLACE)
+
+#decl(INPLACE)
+fn inter_value(i: u32) -> f32 {
+    return src[i];
+}
+
+fn update(i: u32, val: f32) {
+    src[i] = val;
+}
+#enddecl(INPLACE)
+
+#decl(NO_MASK)
+fn mask_val(i: u32) -> f32 {
+    return 0.0;
+}
+#enddecl(NO_MASK)
+
+#decl(MASK)
+fn mask_val(i: u32) -> f32 {
+    return f32(mask[i]);
+}
+#enddecl(MASK)
+
+#decl(NO_SINK)
+fn lower_max_bound(i2: u32) -> f32 {
+    return -1e30;
+}
+
+fn add_sinks(val: f32, i2: u32, max_val: f32) -> f32 {
+    return val;
+}
+#enddecl(NO_SINK)
+
+#decl(SINK)
+fn lower_max_bound(i2: u32) -> f32 {
+    return sinks[params.offset_sinks + i2];
+}
+
+fn add_sinks(val: f32, i2: u32, max_val: f32) -> f32 {
+    return val + exp(sinks[params.offset_sinks + i2] - max_val);
+}
+#enddecl(SINK)
+
+#end(DECLS)
+
+#define(SHADER)
+enable f16;
+
+struct Params {
+    offset_src0: u32,
+    offset_src1: u32,
+    offset_sinks: u32,
+    offset_dst: u32,
+
+    // Strides (in elements)
+    stride_src01: u32,
+    stride_src02: u32,
+    stride_src03: u32,
+
+    stride_src11: u32,
+    stride_src12: u32,
+    stride_src13: u32,
+
+    stride_dst1: u32,
+    stride_dst2: u32,
+    stride_dst3: u32,
+
+    // shape of src0/dst
+    ne: u32,
+    ne0: u32,
+    ne1: u32,
+    ne2: u32,
+
+    // shape of src1
+    ne12: u32,
+    ne13: u32,
+
+    scale: f32,
+    max_bias: f32,
+    n_head_log2: f32,
+    m0: f32,
+    m1: f32,
+};
+
+@group(0) @binding(0)
+var<storage, read_write> src: array<f32>;
+
+DECLS
+
+const CACHE_SIZE: u32 = 16;
+
+override wg_size: u32;
+var<workgroup> scratch: array<f32, wg_size>;
+
+@compute @workgroup_size(wg_size)
+fn main(@builtin(workgroup_id) wid: vec3<u32>,
+        @builtin(local_invocation_id) lid: vec3<u32>) {
+
+    var i = wid.x;
+    let i3 = i / (params.ne2 * params.ne1);
+    i = i % (params.ne2 * params.ne1);
+    let i2 = i / params.ne1;
+    let i1 = i % params.ne1;
+    let i_src0_row = params.offset_src0 + i3 * params.stride_src03 + i2 * params.stride_src02 + i1 * params.stride_src01;
+    let i_src1_row = params.offset_src1 + (i3 % params.ne13) * params.stride_src13 + (i2 % params.ne12) * params.stride_src12 + i1 * params.stride_src11;
+    let i_dst_row = params.offset_dst + i3 * params.stride_dst3 + i2 * params.stride_dst2 + i1 * params.stride_dst1;
+    let elems = (params.ne0 + wg_size - 1) / wg_size;
+
+    let head = f32(i2);
+    let slope = select(1, select(pow(params.m1, 2 * (head - params.n_head_log2) + 1), pow(params.m0, head + 1), head < params.n_head_log2), params.max_bias > 0);
+
+    var cache: array<f32, CACHE_SIZE>;
+
+    var max_val = lower_max_bound(i2);
+    var col = lid.x;
+    for (var j: u32 = 0; j < elems; j++) {
+        if (col >= params.ne0) {
+            break;
+        }
+        let val = src[i_src0_row + col] * params.scale + slope * mask_val(i_src1_row + col);
+        max_val = max(max_val, val);
+        if (col < CACHE_SIZE) {
+            cache[col] = val;
+        }
+        col += wg_size;
+    }
+
+    scratch[lid.x] = max_val;
+    workgroupBarrier();
+    var offset = wg_size / 2;
+    while (offset > 0) {
+        if (lid.x < offset) {
+            scratch[lid.x] = max(scratch[lid.x], scratch[lid.x + offset]);
+        }
+        offset = offset / 2;
+        workgroupBarrier();
+    }
+    let row_max = scratch[0];
+    workgroupBarrier();
+
+    var sum = 0.0f;
+    col = lid.x;
+    for (var j: u32 = 0; j < elems; j++) {
+        if (col >= params.ne0) {
+            break;
+        }
+        let val = select(src[i_src0_row + col] * params.scale + slope * mask_val(i_src1_row + col),
+                         cache[col], col < CACHE_SIZE);
+        let ex = exp(val - row_max);
+        sum += ex;
+        if (col < CACHE_SIZE) {
+            cache[col] = ex;
+        } else {
+            update(i_dst_row + col, ex);
+        }
+        col += wg_size;
+    }
+
+    scratch[lid.x] = sum;
+    workgroupBarrier();
+    offset = wg_size / 2;
+    while (offset > 0) {
+        if (lid.x < offset) {
+            scratch[lid.x] += scratch[lid.x + offset];
+        }
+        offset = offset / 2;
+        workgroupBarrier();
+    }
+    let row_sum = add_sinks(scratch[0], i2, row_max);
+
+    let sum_recip = 1.0 / row_sum;
+    col = lid.x;
+    for  (var j: u32 = 0; j < elems; j++) {
+        if (col >= params.ne0) {
+            break;
+        }
+        update(i_dst_row + col, select(inter_value(i_dst_row + col), cache[col], col < CACHE_SIZE) * sum_recip);
+        col += wg_size;
+    }
+}
+#end(SHADER)

package/cpp/ggml/src/ggml-webgpu/wgsl-shaders/sum_rows.wgsl

@@ -0,0 +1,55 @@
+@group(0) @binding(0)
+var<storage, read_write> src: array<f32>;
+
+@group(0) @binding(1)
+var<storage, read_write> dst: array<f32>;
+
+struct Params {
+    offset_src: u32, // in elements
+    offset_dst: u32, // in elements
+
+    // Strides (in elements)
+    stride_src1: u32,
+    stride_src2: u32,
+    stride_src3: u32,
+
+    ne0: u32,
+    ne1: u32,
+    ne2: u32
+};
+
+@group(0) @binding(2)
+var<uniform> params: Params;
+
+var<workgroup> shared_sum: array<f32, WG_SIZE>;
+
+@compute @workgroup_size(WG_SIZE)
+fn main(@builtin(workgroup_id) wid: vec3<u32>,
+        @builtin(local_invocation_id) lid: vec3<u32>) {
+
+    var i = wid.x;
+    let i3 = i / (params.ne2 * params.ne1);
+    i = i % (params.ne2 * params.ne1);
+    let i2 = i / params.ne1;
+    let i1 = i % params.ne1;
+    let i_src_row = params.offset_src + i3 * params.stride_src3 + i2 * params.stride_src2 + i1 * params.stride_src1;
+    var local_sum: f32 = 0.0;
+    for (var col = lid.x; col < params.ne0; col += WG_SIZE) {
+        local_sum += src[i_src_row + col];
+    }
+    shared_sum[lid.x] = local_sum;
+    workgroupBarrier();
+    // reduce within workgroup
+    var offset: u32 = WG_SIZE >> 1;
+    while (offset > 0) {
+        if (lid.x < offset) {
+            shared_sum[lid.x] = shared_sum[lid.x] + shared_sum[lid.x + offset];
+        }
+        workgroupBarrier();
+        offset >>= 1;
+    }
+
+    if (lid.x == 0) {
+        dst[params.offset_dst + wid.x] = shared_sum[0];
+    }
+}

package/cpp/ggml/src/ggml-webgpu/wgsl-shaders/unary.wgsl

@@ -0,0 +1,193 @@
+#ifdef TYPE_F16
+enable f16;
+#define TYPE f16
+#else
+#define TYPE f32
+#endif
+
+
+@group(0) @binding(0)
+var<storage, read_write> src: array<TYPE>;
+
+#ifndef INPLACE
+@group(0) @binding(1)
+var<storage, read_write> dst: array<TYPE>;
+#define PARAMS_BINDING 2
+#else
+#define PARAMS_BINDING 1
+#endif
+
+struct Params {
+    ne: u32,            // total number of elements
+    offset_src: u32,    // in elements
+    offset_dst: u32,    // in elements
+
+    // Strides (in elements)
+    stride_src0: u32,
+    stride_src1: u32,
+    stride_src2: u32,
+    stride_src3: u32,
+
+    // Logical shapes
+    ne0: u32,
+    ne1: u32,
+    ne2: u32,
+#ifdef CLAMP
+    clamp_min: f32,
+    clamp_max: f32,
+#endif
+#ifdef FILL
+    fill_val: f32,
+#endif
+#ifdef XIELU
+    alpha_n: f32,
+    alpha_p: f32,
+    beta: f32,
+    eps: f32,
+#endif
+
+};
+
+@group(0) @binding(PARAMS_BINDING)
+var<uniform> params: Params;
+
+@compute @workgroup_size(WG_SIZE)
+fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
+    if (gid.x >= params.ne) {
+      return;
+    }
+    var i = gid.x;
+    let i3 = i / (params.ne2 * params.ne1 * params.ne0);
+    i = i % (params.ne2 * params.ne1 * params.ne0);
+    let i2 = i / (params.ne1 * params.ne0);
+    i = i % (params.ne1 * params.ne0);
+    let i1 = i / params.ne0;
+    let i0 = i % params.ne0;
+
+    let src_idx = i0 * params.stride_src0 + i1 * params.stride_src1 +
+                  i2 * params.stride_src2 + i3 * params.stride_src3;
+
+#ifdef ABS
+    let res = abs(src[params.offset_src + src_idx]);
+#endif
+#ifdef SGN
+    let res = select(TYPE(select(0.0, -1.0, src[params.offset_src + src_idx] < 0.0)), TYPE(1.0),
+                     src[params.offset_src + src_idx] > 0.0);
+#endif
+#ifdef NEG
+    let res = -src[params.offset_src + src_idx];
+#endif
+#ifdef STEP
+    let res = TYPE(select(0.0, 1.0, src[params.offset_src + src_idx] > 0.0));
+#endif
+#ifdef TANH
+    let res = tanh(clamp(src[params.offset_src + src_idx], -9.010913, 9.010913));
+#endif
+#ifdef RELU
+    let res = select(0.0, src[params.offset_src + src_idx], src[params.offset_src + src_idx] > 0.0);
+#endif
+#ifdef ELU
+    let res = select(exp(src[params.offset_src + src_idx]) - 1.0, src[params.offset_src + src_idx],
+                     src[params.offset_src + src_idx] > 0.0);
+#endif
+#ifdef HARDSIGMOID
+    let res = min(1.0, max(0.0, (src[params.offset_src + src_idx] + 3.0) / 6.0));
+#endif
+#ifdef SIGMOID
+    let res = 1.0 / (1.0 + exp(-src[params.offset_src + src_idx]));
+#endif
+#ifdef SILU
+    let res = src[params.offset_src + src_idx] / (1.0 + exp(-src[params.offset_src + src_idx]));
+#endif
+#ifdef EXP
+    let res = exp(src[params.offset_src + src_idx]);
+#endif
+#ifdef LOG
+    let res = TYPE(log(f32(src[params.offset_src + src_idx])));
+#endif
+#ifdef CLAMP
+    let res = clamp(src[params.offset_src + src_idx], TYPE(params.clamp_min), TYPE(params.clamp_max));
+#endif
+#ifdef FILL
+    let res = TYPE(params.fill_val);
+#endif
+#ifdef HARDSWISH
+    let res = src[params.offset_src + src_idx] *
+              min(1.0, max(0.0, (src[params.offset_src + src_idx] + 3.0) / 6.0));
+#endif
+#ifdef GELU
+    let res = 0.5 * src[params.offset_src + src_idx] *
+              (1.0 + tanh(clamp(sqrt(2.0 / 3.14159265) *
+                               (src[params.offset_src + src_idx] +
+                                0.044715 * pow(src[params.offset_src + src_idx], 3.0)),
+                               -9.010913, 9.010913)));
+#endif
+#ifdef GELU_QUICK
+    let res = src[params.offset_src + src_idx] * 0.5 *
+              (1.0 + tanh(clamp(0.79788456 *
+                               (src[params.offset_src + src_idx] +
+                                0.044715 * src[params.offset_src + src_idx] *
+                                    src[params.offset_src + src_idx] * src[params.offset_src + src_idx]),
+                               -9.010913, 9.010913)));
+#endif
+#ifdef GELU_ERF
+    let res = 0.5 * src[params.offset_src + src_idx] *
+              (1.0 + tanh(clamp(0.79788456 *
+                               (src[params.offset_src + src_idx] +
+                                0.044715 * src[params.offset_src + src_idx] *
+                                    src[params.offset_src + src_idx] * src[params.offset_src + src_idx]),
+                               -9.010913, 9.010913)));
+#endif
+#ifdef XIELU
+    let res =
+        select(((exp(min(src[params.offset_src + src_idx], TYPE(params.eps))) - 1.0) -
+                src[params.offset_src + src_idx]) *
+                   TYPE(params.alpha_n) +
+               TYPE(params.beta) * src[params.offset_src + src_idx],
+               TYPE(params.alpha_p) * src[params.offset_src + src_idx] *
+                   src[params.offset_src + src_idx] +
+                   TYPE(params.beta) * src[params.offset_src + src_idx],
+               src[params.offset_src + src_idx] > 0.0);
+#endif
+#ifdef SOFTPLUS
+    let src_f32 = f32(src[params.offset_src + src_idx]);
+    let res = TYPE(select(log(1.0 + exp(src_f32)), src_f32, src_f32 > 20.0));
+#endif
+#ifdef EXPM1
+    let res = exp(src[params.offset_src + src_idx]) - 1.0;
+#endif
+#ifdef FLOOR
+    let res = floor(src[params.offset_src + src_idx]);
+#endif
+#ifdef CEIL
+    let res = ceil(src[params.offset_src + src_idx]);
+#endif
+#ifdef ROUND
+    let src_f32 = f32(src[params.offset_src + src_idx]);
+    let result = select(ceil(src_f32 - 0.5), floor(src_f32 + 0.5), src_f32 >= 0.0);
+    let res = TYPE(result);
+#endif
+#ifdef TRUNC
+    let res = trunc(src[params.offset_src + src_idx]);
+#endif
+#ifdef SQR
+    let res = src[params.offset_src + src_idx] * src[params.offset_src + src_idx];
+#endif
+#ifdef SQRT
+    let res = sqrt(src[params.offset_src + src_idx]);
+#endif
+#ifdef SIN
+    let res_f32 = sin(f32(src[params.offset_src + src_idx]));
+    let res = TYPE(res_f32);
+#endif
+#ifdef COS
+    let res_f32 = cos(f32(src[params.offset_src + src_idx]));
+    let res = TYPE(res_f32);
+#endif
+
+#ifdef INPLACE
+    src[params.offset_src + src_idx] = res;
+#else
+    dst[params.offset_dst + gid.x] = res;
+#endif
+}

package/cpp/ggml/src/ggml-zdnn/CMakeLists.txt

@@ -0,0 +1,36 @@
+if (DEFINED ZDNN_ROOT)
+    message(STATUS "zdnn: using ZDNN_ROOT override: ${ZDNN_ROOT}")
+    set(ZDNN_HINT "${ZDNN_ROOT}")
+else()
+    set(ZDNN_HINT "")
+endif()
+
+find_path(ZDNN_INCLUDE
+            NAMES zdnn.h
+            HINTS ${ZDNN_HINT} /usr /usr/local
+            PATH_SUFFIXES include)
+if (ZDNN_INCLUDE)
+    message(STATUS "zdnn: found include: ${ZDNN_INCLUDE}")
+else()
+    message(FATAL_ERROR "zdnn: include directory not found, please set ZDNN_ROOT to the proper path if necessary")
+endif()
+
+find_library(ZDNN_LIB
+                NAMES zdnn
+                HINTS ${ZDNN_HINT} /usr /usr/local
+                PATH_SUFFIXES lib lib64)
+if (ZDNN_LIB)
+    message(STATUS "zdnn: found library: ${ZDNN_LIB}")
+else()
+    message(FATAL_ERROR "zdnn: library not found, please set ZDNN_ROOT to the proper path if necessary")
+endif()
+
+file(GLOB GGML_SOURCES_ZDNN "*.c" "*.cpp")
+file(GLOB GGML_HEADERS_ZDNN "*.h" "*.hpp")
+
+ggml_add_backend_library(ggml-zdnn ${GGML_HEADERS_ZDNN} ${GGML_SOURCES_ZDNN})
+target_link_libraries(ggml-zdnn PRIVATE ${ZDNN_LIB})
+target_include_directories(ggml-zdnn PRIVATE ${ZDNN_INCLUDE})
+target_link_directories(ggml-zdnn PRIVATE ${ZDNN_LIB})
+
+target_compile_definitions(ggml-zdnn PRIVATE GGML_USE_ZDNN)

package/cpp/ggml/src/ggml-zdnn/common.hpp

@@ -0,0 +1,59 @@
+#ifndef GGML_ZDNN_COMMON_HPP
+#define GGML_ZDNN_COMMON_HPP
+
+#include "ggml.h"
+#include "ggml-impl.h"
+
+#include "zdnn.h"
+
+#include <vector>
+#include <memory>
+
+#define GGML_ZDNN_NAME    "zDNN"
+#define GGML_ZDNN_VERSION ZDNN_VERNUM
+
+#define ZDNN_CHECK(stmt)                \
+    do {                                \
+        zdnn_status status = (stmt);    \
+        GGML_ASSERT(status == ZDNN_OK); \
+    } while (0);
+
+struct ggml_backend_zdnn_device_context {
+    int zdnn_device;
+    int zdnn_device_ref_count;
+
+    bool has_parmblkformat_0;
+    bool has_parmblkformat_1;  // checks for z17
+
+    size_t max_size;
+
+    char name[128];
+};
+
+struct ggml_backend_zdnn_context {
+    int device;
+    ggml_cgraph * gf;
+};
+
+struct ggml_backend_zdnn_buffer {
+    void * data;
+    ggml_backend_zdnn_buffer * extra;  // for bias, etc.
+    size_t size;
+
+    zdnn_tensor_desc pre_tfm_desc;
+    zdnn_tensor_desc tfm_desc;
+    zdnn_ztensor     ztensor;
+
+    char name[GGML_MAX_NAME];
+};
+
+struct ggml_backend_zdnn_buffer_context {
+    void * all_data;
+    size_t all_size;
+    bool owned;
+
+    int n_buffers;
+    std::vector<std::unique_ptr<ggml_backend_zdnn_buffer>> buffers;
+};
+
+#endif  // GGML_ZDNN_COMMON_HPP