vulkan: improve mul_mat_vec_iq1_s speed (#17874)

* q6_k faster mul mat

* 8 values

* fix comment

* switch to two at a time

* start ci for .glsl files

.github/workflows/build.yml

@@ -20,7 +20,8 @@ on:
       '**/*.swift',
       '**/*.m',
       '**/*.metal',
-      '**/*.comp'
+      '**/*.comp',
+      '**/*.glsl'
     ]
 
   pull_request:
@@ -40,7 +41,8 @@ on:
       '**/*.swift',
       '**/*.m',
       '**/*.metal',
-      '**/*.comp'
+      '**/*.comp',
+      '**/*.glsl'
     ]
 
 concurrency:

.gitignore

@@ -54,6 +54,7 @@
 /out/
 /tmp/
 /autogen-*.md
+/common/build-info.cpp
 
 # Deprecated
 

common/arg.cpp

@@ -724,7 +724,7 @@ static void add_rpc_devices(const std::string & servers) {
     }
 }
 
-bool common_params_parse(int argc, char ** argv, llama_example ex, std::map<common_arg, std::string> & out_map) {
+bool common_params_to_map(int argc, char ** argv, llama_example ex, std::map<common_arg, std::string> & out_map) {
     common_params dummy_params;
     common_params_context ctx_arg = common_params_parser_init(dummy_params, ex, nullptr);
 
@@ -733,6 +733,9 @@ bool common_params_parse(int argc, char ** argv, llama_example ex, std::map<comm
         for (const auto & arg : opt.args) {
             arg_to_options[arg] = &opt;
         }
+        for (const auto & arg : opt.args_neg) {
+            arg_to_options[arg] = &opt;
+        }
     }
 
     // TODO @ngxson : find a way to deduplicate this code

common/arg.h

@@ -115,7 +115,7 @@ bool common_params_parse(int argc, char ** argv, common_params & params, llama_e
 
 // parse input arguments from CLI into a map
 // TODO: support repeated args in the future
-bool common_params_parse(int argc, char ** argv, llama_example ex, std::map<common_arg, std::string> & out_map);
+bool common_params_to_map(int argc, char ** argv, llama_example ex, std::map<common_arg, std::string> & out_map);
 
 // initialize argument parser context - used by test-arg-parser and preset
 common_params_context common_params_parser_init(common_params & params, llama_example ex, void(*print_usage)(int, char **) = nullptr);

examples/model-conversion/scripts/causal/run-org-model.py

@@ -200,7 +200,7 @@ with torch.no_grad():
     logits = outputs.logits
 
     # Extract logits for the last token (next token prediction)
-    last_logits = logits[0, -1, :].cpu().numpy()
+    last_logits = logits[0, -1, :].float().cpu().numpy()
 
     print(f"Logits shape: {logits.shape}")
     print(f"Last token logits shape: {last_logits.shape}")

ggml/CMakeLists.txt

@@ -54,6 +54,10 @@ if (CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR)
     # TODO
 else()
     set(GGML_STANDALONE OFF)
+
+    if (NOT CMAKE_RUNTIME_OUTPUT_DIRECTORY)
+        set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/bin)
+    endif()
 endif()
 
 if (EMSCRIPTEN)

ggml/src/ggml-cpu/arch/arm/repack.cpp

@@ -24,6 +24,7 @@
 
 #define UNUSED GGML_UNUSED
 
+#if defined(__aarch64__) && defined(__ARM_NEON) && (defined(__ARM_FEATURE_MATMUL_INT8) || defined(__ARM_FEATURE_DOTPROD))
 static inline void decode_q4_Kx8_scales_mins(const uint8_t * scales_in,
                                              int16x8_t *     out_mins,
                                              int8_t *        out_scales) {
@@ -46,6 +47,7 @@ static inline void decode_q4_Kx8_scales_mins(const uint8_t * scales_in,
     scales_u32[1] = (sm[2] & kmask2) | (((sm[0] >> 6) & kmask3) << 4);
     memcpy(out_scales, scales_u32, 8);
 }
+#endif
 
 void ggml_quantize_mat_q8_0_4x4(const float * GGML_RESTRICT x, void * GGML_RESTRICT vy, int64_t k) {
     assert(QK8_0 == 32);

ggml/src/ggml-vulkan/ggml-vulkan.cpp

@@ -659,6 +659,7 @@ struct vk_device_struct {
     vk_pipeline pipeline_cos_f32;
     vk_pipeline pipeline_log[2];
     vk_pipeline pipeline_tri[2];
+    vk_pipeline pipeline_diag[2];
     vk_pipeline pipeline_clamp_f32;
     vk_pipeline pipeline_pad_f32;
     vk_pipeline pipeline_roll_f32;
@@ -722,6 +723,11 @@ struct vk_device_struct {
     vk_pipeline pipeline_soft_max_f32, pipeline_soft_max_f32_f16;
     vk_pipeline pipeline_soft_max_f32_wg512, pipeline_soft_max_f32_f16_wg512;
     vk_pipeline pipeline_soft_max_back_f32;
+
+    vk_pipeline pipeline_soft_max_large1_f32, pipeline_soft_max_large1_f32_f16;
+    vk_pipeline pipeline_soft_max_large2_f32, pipeline_soft_max_large2_f32_f16;
+    vk_pipeline pipeline_soft_max_large3_f32, pipeline_soft_max_large3_f32_f16;
+
     vk_pipeline pipeline_rope_norm_f32, pipeline_rope_norm_f16, pipeline_rope_norm_f32_f16;
     vk_pipeline pipeline_rope_neox_f32, pipeline_rope_neox_f16, pipeline_rope_neox_f32_f16;
     vk_pipeline pipeline_rope_multi_f32, pipeline_rope_multi_f16;
@@ -3732,6 +3738,7 @@ static void ggml_vk_load_shaders(vk_device& device) {
     ggml_vk_create_pipeline(device, device->pipeline_get_rows[GGML_TYPE_IQ4_XS],  "get_rows_iq4_xs",  get_rows_iq4_xs_len,  get_rows_iq4_xs_data,  "main", 3, sizeof(vk_op_binary_push_constants), {1024, 1, 1}, {}, 1);
     ggml_vk_create_pipeline(device, device->pipeline_get_rows[GGML_TYPE_IQ4_NL],  "get_rows_iq4_nl",  get_rows_iq4_nl_len,  get_rows_iq4_nl_data,  "main", 3, sizeof(vk_op_binary_push_constants), {1024, 1, 1}, {}, 1);
     ggml_vk_create_pipeline(device, device->pipeline_get_rows[GGML_TYPE_MXFP4],   "get_rows_mxfp4",   get_rows_mxfp4_len,   get_rows_mxfp4_data,   "main", 3, sizeof(vk_op_binary_push_constants), {1024, 1, 1}, {}, 1);
+    ggml_vk_create_pipeline(device, device->pipeline_get_rows[GGML_TYPE_I32],     "get_rows_i32",     get_rows_i32_len,     get_rows_i32_data,     "main", 3, sizeof(vk_op_binary_push_constants), {1024, 1, 1}, {}, 1);
 
     ggml_vk_create_pipeline(device, device->pipeline_get_rows_f32[GGML_TYPE_F32 ], "get_rows_f32_f32",  get_rows_f32_f32_len,  get_rows_f32_f32_data,  "main", 3, sizeof(vk_op_binary_push_constants), { 512, 1, 1}, {}, 1);
     ggml_vk_create_pipeline(device, device->pipeline_get_rows_f32[GGML_TYPE_F16 ], "get_rows_f16_f32",  get_rows_f16_f32_len,  get_rows_f16_f32_data,  "main", 3, sizeof(vk_op_binary_push_constants), { 512, 1, 1}, {}, 1);
@@ -3919,6 +3926,9 @@ static void ggml_vk_load_shaders(vk_device& device) {
     ggml_vk_create_pipeline(device, device->pipeline_tri[0], "tri_f32", tri_f32_len, tri_f32_data, "main", 2, sizeof(vk_op_unary_push_constants), {512, 1, 1}, {}, 1);
     ggml_vk_create_pipeline(device, device->pipeline_tri[1], "tri_f16", tri_f16_len, tri_f16_data, "main", 2, sizeof(vk_op_unary_push_constants), {512, 1, 1}, {}, 1);
 
+    ggml_vk_create_pipeline(device, device->pipeline_diag[0], "diag_f32", diag_f32_len, diag_f32_data, "main", 2, sizeof(vk_op_unary_push_constants), {512, 1, 1}, {}, 1);
+    ggml_vk_create_pipeline(device, device->pipeline_diag[1], "diag_f16", diag_f16_len, diag_f16_data, "main", 2, sizeof(vk_op_unary_push_constants), {512, 1, 1}, {}, 1);
+
     ggml_vk_create_pipeline(device, device->pipeline_clamp_f32, "clamp_f32", clamp_f32_len, clamp_f32_data, "main", 2, sizeof(vk_op_unary_push_constants), {512, 1, 1}, {}, 1);
 
     ggml_vk_create_pipeline(device, device->pipeline_pad_f32, "pad_f32", pad_f32_len, pad_f32_data, "main", 2, sizeof(vk_op_pad_push_constants), {512, 1, 1}, {}, 1);
@@ -3998,6 +4008,13 @@ static void ggml_vk_load_shaders(vk_device& device) {
     ggml_vk_create_pipeline(device, device->pipeline_soft_max_f32_f16_wg512, "soft_max_f32_f16_wg512", soft_max_f32_f16_len, soft_max_f32_f16_data, "main", 4, sizeof(vk_op_soft_max_push_constants), {1, 1, 1}, { 512 }, 1);
     ggml_vk_create_pipeline(device, device->pipeline_soft_max_back_f32, "soft_max_back_f32", soft_max_back_f32_len, soft_max_back_f32_data, "main", 3, sizeof(vk_op_push_constants), {1, 1, 1}, { device->subgroup_size }, 1, true);
 
+    ggml_vk_create_pipeline(device, device->pipeline_soft_max_large1_f32,     "soft_max_large1_f32",     soft_max_large1_f32_len,     soft_max_large1_f32_data,     "main", 6, sizeof(vk_op_soft_max_push_constants), {1, 1, 1}, { 128, 4 }, 1, true);
+    ggml_vk_create_pipeline(device, device->pipeline_soft_max_large2_f32,     "soft_max_large2_f32",     soft_max_large2_f32_len,     soft_max_large2_f32_data,     "main", 6, sizeof(vk_op_soft_max_push_constants), {1, 1, 1}, { 128, 4 }, 1, true);
+    ggml_vk_create_pipeline(device, device->pipeline_soft_max_large3_f32,     "soft_max_large3_f32",     soft_max_large3_f32_len,     soft_max_large3_f32_data,     "main", 6, sizeof(vk_op_soft_max_push_constants), {1, 1, 1}, { 128, 4 }, 1, true);
+    ggml_vk_create_pipeline(device, device->pipeline_soft_max_large1_f32_f16, "soft_max_large1_f32_f16", soft_max_large1_f32_f16_len, soft_max_large1_f32_f16_data, "main", 6, sizeof(vk_op_soft_max_push_constants), {1, 1, 1}, { 128, 4 }, 1, true);
+    ggml_vk_create_pipeline(device, device->pipeline_soft_max_large2_f32_f16, "soft_max_large2_f32_f16", soft_max_large2_f32_f16_len, soft_max_large2_f32_f16_data, "main", 6, sizeof(vk_op_soft_max_push_constants), {1, 1, 1}, { 128, 4 }, 1, true);
+    ggml_vk_create_pipeline(device, device->pipeline_soft_max_large3_f32_f16, "soft_max_large3_f32_f16", soft_max_large3_f32_f16_len, soft_max_large3_f32_f16_data, "main", 6, sizeof(vk_op_soft_max_push_constants), {1, 1, 1}, { 128, 4 }, 1, true);
+
     ggml_vk_create_pipeline(device, device->pipeline_rope_norm_f32, "rope_norm_f32", rope_norm_f32_len, rope_norm_f32_data, "main", 5, sizeof(vk_op_rope_push_constants), {1, 512, 1}, {}, 1);
     ggml_vk_create_pipeline(device, device->pipeline_rope_neox_f32, "rope_neox_f32", rope_neox_f32_len, rope_neox_f32_data, "main", 5, sizeof(vk_op_rope_push_constants), {1, 512, 1}, {}, 1);
     ggml_vk_create_pipeline(device, device->pipeline_rope_multi_f32, "rope_multi_f32", rope_multi_f32_len, rope_multi_f32_data, "main", 5, sizeof(vk_op_rope_push_constants), {1, 512, 1}, {}, 1);
@@ -8278,6 +8295,11 @@ static vk_pipeline ggml_vk_op_get_pipeline(ggml_backend_vk_context * ctx, const
     switch (op) {
     case GGML_OP_GET_ROWS:
         GGML_ASSERT(src1->type == GGML_TYPE_I32);
+        if (src0->type == GGML_TYPE_I32) {
+            // i32 src only supports i32 result
+            GGML_ASSERT(dst->type == GGML_TYPE_I32);
+            return ctx->device->pipeline_get_rows[src0->type];
+        }
         if (dst->type == GGML_TYPE_F16) {
             return ctx->device->pipeline_get_rows[src0->type];
         }
@@ -8404,6 +8426,12 @@ static vk_pipeline ggml_vk_op_get_pipeline(ggml_backend_vk_context * ctx, const
             return ctx->device->pipeline_tri[dst->type == GGML_TYPE_F16];
         }
         return nullptr;
+    case GGML_OP_DIAG:
+        if (src0->type == dst->type &&
+            (src0->type == GGML_TYPE_F32 || src0->type == GGML_TYPE_F16)) {
+            return ctx->device->pipeline_diag[dst->type == GGML_TYPE_F16];
+        }
+        return nullptr;
     case GGML_OP_CLAMP:
         if (src0->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32) {
             return ctx->device->pipeline_clamp_f32;
@@ -9097,6 +9125,7 @@ static void ggml_vk_op_f32(ggml_backend_vk_context * ctx, vk_context& subctx, co
     case GGML_OP_COS:
     case GGML_OP_LOG:
     case GGML_OP_TRI:
+    case GGML_OP_DIAG:
     case GGML_OP_CLAMP:
     case GGML_OP_PAD:
     case GGML_OP_ROLL:
@@ -9784,6 +9813,12 @@ static void ggml_vk_tri(ggml_backend_vk_context * ctx, vk_context& subctx, const
     ggml_vk_op_f32(ctx, subctx, src0, nullptr, nullptr, nullptr, dst, GGML_OP_TRI, std::move(p));
 }
 
+static void ggml_vk_diag(ggml_backend_vk_context * ctx, vk_context& subctx, const ggml_tensor * src0, ggml_tensor * dst) {
+    vk_op_unary_push_constants p = vk_op_unary_push_constants_init(src0, dst, ggml_nelements(dst));
+
+    ggml_vk_op_f32(ctx, subctx, src0, nullptr, nullptr, nullptr, dst, GGML_OP_DIAG, std::move(p));
+}
+
 static void ggml_vk_clamp(ggml_backend_vk_context * ctx, vk_context& subctx, const ggml_tensor * src0, ggml_tensor * dst) {
     vk_op_unary_push_constants p = vk_op_unary_push_constants_init(src0, dst);
     p.param1 = ggml_get_op_params_f32(dst, 0);
@@ -10117,7 +10152,7 @@ static void ggml_vk_soft_max(ggml_backend_vk_context * ctx, vk_context& subctx,
     const float m0 = powf(2.0f, -(max_bias       ) / n_head_log2);
     const float m1 = powf(2.0f, -(max_bias / 2.0f) / n_head_log2);
 
-    ggml_vk_op_f32<vk_op_soft_max_push_constants>(ctx, subctx, src0, src1, src2, nullptr, dst, GGML_OP_SOFT_MAX, {
+    vk_op_soft_max_push_constants pc {
         ncols,
         src1 != nullptr ? nrows_y : (uint32_t)0,
         (uint32_t)src0->ne[0], (uint32_t)src0->ne[1], (uint32_t)src0->ne[2],
@@ -10128,7 +10163,55 @@ static void ggml_vk_soft_max(ggml_backend_vk_context * ctx, vk_context& subctx,
         n_head_log2,
         nrows_x,
         src2 != nullptr
-    });
+    };
+
+    if (ncols <= 16384) {
+        ggml_vk_op_f32<vk_op_soft_max_push_constants>(ctx, subctx, src0, src1, src2, nullptr, dst, GGML_OP_SOFT_MAX, std::move(pc));
+    } else {
+
+        vk_subbuffer buf_a = ggml_vk_tensor_subbuffer(ctx, src0);
+        vk_subbuffer buf_b = src1 ? ggml_vk_tensor_subbuffer(ctx, src1) : buf_a;
+        vk_subbuffer buf_c = src2 ? ggml_vk_tensor_subbuffer(ctx, src2) : buf_a;
+        vk_subbuffer buf_d = ggml_vk_tensor_subbuffer(ctx, dst);
+
+        uint32_t elems_per_wg = 128 * 4;
+        uint32_t num_wgs = CEIL_DIV(ncols, elems_per_wg);
+        size_t tmp_size = num_wgs * nrows_x * sizeof(float);
+
+        if (ctx->prealloc_size_x < tmp_size) {
+            ctx->prealloc_size_x = tmp_size;
+            ggml_vk_preallocate_buffers(ctx, subctx);
+        }
+        if (ctx->prealloc_size_y < tmp_size) {
+            ctx->prealloc_size_y = tmp_size;
+            ggml_vk_preallocate_buffers(ctx, subctx);
+        }
+        if (ctx->prealloc_x_need_sync || ctx->prealloc_y_need_sync) {
+            ggml_vk_sync_buffers(ctx, subctx);
+        }
+
+        vk_subbuffer buf_x = { ctx->prealloc_x, 0, tmp_size };
+        vk_subbuffer buf_y = { ctx->prealloc_y, 0, tmp_size };
+
+        std::array<uint32_t, 3> elements = { num_wgs, nrows_x, 1 };
+
+        vk_pipeline pipeline1 = src1 && src1->type == GGML_TYPE_F16 ? ctx->device->pipeline_soft_max_large1_f32_f16 : ctx->device->pipeline_soft_max_large1_f32;
+        vk_pipeline pipeline2 = src1 && src1->type == GGML_TYPE_F16 ? ctx->device->pipeline_soft_max_large2_f32_f16 : ctx->device->pipeline_soft_max_large2_f32;
+        vk_pipeline pipeline3 = src1 && src1->type == GGML_TYPE_F16 ? ctx->device->pipeline_soft_max_large3_f32_f16 : ctx->device->pipeline_soft_max_large3_f32;
+
+        ggml_pipeline_request_descriptor_sets(ctx, pipeline1, 1);
+        ggml_pipeline_request_descriptor_sets(ctx, pipeline2, 1);
+        ggml_pipeline_request_descriptor_sets(ctx, pipeline3, 1);
+
+        ggml_vk_dispatch_pipeline(ctx, subctx, pipeline1, { buf_a, buf_b, buf_c, buf_d, buf_x, buf_y }, pc, elements);
+        ggml_vk_sync_buffers(ctx, subctx);
+        ggml_vk_dispatch_pipeline(ctx, subctx, pipeline2, { buf_a, buf_b, buf_c, buf_d, buf_x, buf_y }, pc, elements);
+        ggml_vk_sync_buffers(ctx, subctx);
+        ggml_vk_dispatch_pipeline(ctx, subctx, pipeline3, { buf_a, buf_b, buf_c, buf_d, buf_x, buf_y }, pc, elements);
+
+        ctx->prealloc_x_need_sync = true;
+        ctx->prealloc_y_need_sync = true;
+    }
 }
 
 static void ggml_vk_soft_max_back(ggml_backend_vk_context * ctx, vk_context& subctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
@@ -11864,6 +11947,10 @@ static bool ggml_vk_build_graph(ggml_backend_vk_context * ctx, ggml_cgraph * cgr
     case GGML_OP_TRI:
         ggml_vk_tri(ctx, compute_ctx, src0, node);
 
+        break;
+    case GGML_OP_DIAG:
+        ggml_vk_diag(ctx, compute_ctx, src0, node);
+
         break;
     case GGML_OP_CLAMP:
         ggml_vk_clamp(ctx, compute_ctx, src0, node);
@@ -13883,6 +13970,7 @@ static bool ggml_backend_vk_device_supports_op(ggml_backend_dev_t dev, const ggm
                     case GGML_TYPE_IQ4_XS:
                     case GGML_TYPE_IQ4_NL:
                     case GGML_TYPE_MXFP4:
+                    case GGML_TYPE_I32:
                         return true;
                     default:
                         return false;
@@ -14007,6 +14095,7 @@ static bool ggml_backend_vk_device_supports_op(ggml_backend_dev_t dev, const ggm
             return ggml_is_contiguous(op->src[0]) && op->src[0]->type == GGML_TYPE_F32;
         case GGML_OP_LOG:
         case GGML_OP_TRI:
+        case GGML_OP_DIAG:
             return (op->src[0]->type == GGML_TYPE_F32 || op->src[0]->type == GGML_TYPE_F16) &&
                    op->type == op->src[0]->type;
         case GGML_OP_ARGSORT:
@@ -14597,6 +14686,8 @@ static void ggml_vk_check_results_0(ggml_backend_vk_context * ctx, ggml_cgraph *
             tensor_clone = ggml_log(ggml_ctx, src_clone[0]);
         } else if (tensor->op == GGML_OP_TRI) {
             tensor_clone = ggml_tri(ggml_ctx, src_clone[0], ggml_get_op_params_i32(tensor, 0));
+        } else if (tensor->op == GGML_OP_DIAG) {
+            tensor_clone = ggml_diag(ggml_ctx, src_clone[0]);
         } else if (tensor->op == GGML_OP_CLAMP) {
             const float * params = (const float *)tensor->op_params;
             tensor_clone = ggml_clamp(ggml_ctx, src_clone[0], params[0], params[1]);

ggml/src/ggml-vulkan/vulkan-shaders/diag.comp

@@ -0,0 +1,29 @@
+#version 450
+
+#include "rte.glsl"
+#include "types.glsl"
+#include "generic_unary_head.glsl"
+
+layout(local_size_x = 512, local_size_y = 1, local_size_z = 1) in;
+
+void main() {
+    const uint idx = get_idx();
+
+    if (idx >= p.ne) {
+        return;
+    }
+
+    const uint i13 = fastdiv(idx, p.ne1_012mp, p.ne1_012L);
+    const uint i13_offset = i13 * p.ne12*p.ne11*p.ne10;
+    const uint i12 = fastdiv(idx - i13_offset, p.ne1_01mp, p.ne1_01L);
+    const uint i12_offset = i12*p.ne11*p.ne10;
+    const uint i11 = fastdiv(idx - i13_offset - i12_offset, p.ne1_0mp, p.ne1_0L);
+    const uint i10 = idx - i13_offset - i12_offset - i11*p.ne10;
+
+    if (i10 == i11) {
+        const float val = float(data_a[get_aoffset() + i13*p.nb03 + i12*p.nb02 + 0*p.nb01 + i10*p.nb00]);
+        data_d[get_doffset() + dst_idx(idx)] = D_TYPE(val);
+    } else {
+        data_d[get_doffset() + dst_idx(idx)] = D_TYPE(0);
+    }
+}

ggml/src/ggml-vulkan/vulkan-shaders/get_rows.comp

@@ -26,9 +26,9 @@ void main() {
             const uint d_offset = get_doffset() + i10*p.nb21 + i11*p.nb22 + i12*p.nb23;
 
 #if defined(DATA_A_BF16)
-            FLOAT_TYPE v = FLOAT_TYPE(bf16_to_fp32(data_a[a_offset + i00]));
+            TEMP_TYPE v = TEMP_TYPE(bf16_to_fp32(data_a[a_offset + i00]));
 #else
-            FLOAT_TYPE v = FLOAT_TYPE(data_a[a_offset + i00]);
+            TEMP_TYPE v = TEMP_TYPE(data_a[a_offset + i00]);
 #endif
 #ifndef OPTIMIZATION_ERROR_WORKAROUND
             data_d[d_offset + i00] = D_TYPE(v);

ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iq1_s.comp

@@ -7,34 +7,50 @@ layout(local_size_x_id = 0, local_size_y = 1, local_size_z = 1) in;
 
 FLOAT_TYPE temp[NUM_COLS][NUM_ROWS];
 
-void calc_superblock(const uint a_offset, const uint b_offset, const uint ib32, const uint i, const uint num_blocks_per_row, const uint first_row, const uint num_rows) {
-    const uint y_idx = i * QUANT_K + 32 * ib32;
+void calc_superblock(const uint a_offset, const uint b_offset, const uint ib32, const uint i,
+                     const uint num_blocks_per_row, const uint first_row, const uint num_rows) {
+    const uint y_idx_base = i * QUANT_K + 32 * ib32;
+    [[unroll]] for (uint j = 0; j < NUM_COLS; ++j) {    
+        const uint base_b_idx = (j * p.batch_stride_b + b_offset + y_idx_base) / 4;
+        [[unroll]] for (uint l = 0; l < 4; ++l) {            
+            const vec4 b_val_0 = vec4(data_b_v4[base_b_idx + 2 * l]);
+            const vec4 b_val_1 = vec4(data_b_v4[base_b_idx + 2 * l + 1]);
 
-    uint ibi = a_offset / QUANT_K + first_row * num_blocks_per_row + i;
-    [[unroll]] for (uint n = 0; n < num_rows; ++n) {
-        const float d = float(data_a[ibi].d);
-        const uint qh = data_a[ibi].qh[ib32];
-        const float dl = d * float(2 * bitfieldExtract(qh, 12, 3) + 1);
-        const float delta = ((qh & 0x8000) != 0) ? -IQ1S_DELTA : IQ1S_DELTA;
+            // index for data_a
+            uint ibi = a_offset / QUANT_K + first_row * num_blocks_per_row + i;
 
-        [[unroll]] for (uint l = 0; l < 4; ++l) {
-            const uint qs = data_a[ibi].qs[4 * ib32 + l];
-            const uint idxhi = bitfieldExtract(qh, 3 * int(l), 3);
-            const int16_t grid = int16_t(iq1s_grid[qs | (idxhi << 8)]);
+            [[unroll]] for (uint n = 0; n < num_rows; ++n) {                
+                const float d = float(data_a[ibi].d);
+                const uint qh = data_a[ibi].qh[ib32];
 
-            [[unroll]] for (uint j = 0; j < NUM_COLS; ++j) {
-                vec4 b0 = vec4(data_b_v4[(j*p.batch_stride_b + b_offset + y_idx) / 4 + 2*l + 0]);
-                vec4 b4 = vec4(data_b_v4[(j*p.batch_stride_b + b_offset + y_idx) / 4 + 2*l + 1]);
+                const float dl = d * float(2 * bitfieldExtract(qh, 12, 3) + 1);
+                const uint qs = data_a[ibi].qs[4 * ib32 + l];
+                const uint idxhi = bitfieldExtract(qh, 3 * int(l), 3);              
+                const uint16_t grid = uint16_t(iq1s_grid[qs | (idxhi << 8)]);
 
-                FLOAT_TYPE sum = FLOAT_TYPE(0.0);
-                [[unroll]] for (int k = 0; k < 4; ++k) {
-                    sum = fma(FLOAT_TYPE(b0[k]), bitfieldExtract(grid, 2 * k, 2) + delta,
-                          fma(FLOAT_TYPE(b4[k]), bitfieldExtract(grid, 8 + 2 * k, 2) + delta, sum));
-                }
-                temp[j][n] = fma(dl, sum, temp[j][n]);
+                const float delta_val = ((qh & 0x8000) != 0) ? -IQ1S_DELTA : IQ1S_DELTA;
+                const vec4 delta_v = vec4(delta_val);				
+                const vec4 fbits0 = vec4(
+                    float(bitfieldExtract(grid, 0, 2)),
+                    float(bitfieldExtract(grid, 2, 2)),
+                    float(bitfieldExtract(grid, 4, 2)),
+                    float(bitfieldExtract(grid, 6, 2))
+                );				
+                const vec4 fbits1 = vec4(
+                    float(bitfieldExtract(grid, 8, 2)),
+                    float(bitfieldExtract(grid, 10, 2)),
+                    float(bitfieldExtract(grid, 12, 2)),
+                    float(bitfieldExtract(grid, 14, 2))
+                );
+				
+                vec4 sum_v = fma(b_val_0, fbits0 + delta_v, vec4(0.0));
+                sum_v      = fma(b_val_1, fbits1 + delta_v, sum_v);
+				FLOAT_TYPE sum = dot(sum_v, vec4(1.0));
+				
+                temp[j][n] = fma(dl, sum, temp[j][n]);				
+                ibi += num_blocks_per_row;
             }
         }
-        ibi += num_blocks_per_row;
     }
 }
 

ggml/src/ggml-vulkan/vulkan-shaders/mul_mm_funcs.glsl

@@ -244,17 +244,20 @@ void load_a_to_shmem(const uint pos_a, const uint row, const uint col, const uin
             const uint iqs = idx % 128;                 // 0..127
 
             const uint n = iqs / 64;                    // 0,1
-            const uint b = (iqs % 64) / 32;             // 0,1
+            const uint b = ((iqs % 64) / 32) * 4;       // 0,4
             const uint is_b = (iqs % 16) / 8;           // 0,1
             const uint qhshift = ((iqs % 64) / 16) * 2; // 0,2,4,6
             const uint is = 8 * n + qhshift + is_b;     // 0..15
-            const uint qsi = n * 64 + (iqs % 32) * 2;   // 0,2,4..126
-            const uint qhi = n * 32 + (iqs % 16) * 2;   // 0,2,4..62
+            const uint qsi = n * 32 + (iqs % 32);       // 0..63
+            const uint qhi = n * 16 + (iqs % 16);       // 0..31
 
             const float dscale = float(data_a[ib].d) * float(data_a[ib].scales[is]);
 
-            buf_a[buf_idx] = FLOAT_TYPE_VEC2(dscale * float(int8_t(((data_a[ib].ql[qsi    ] >> (b * 4)) & 0xF) | (((data_a[ib].qh[qhi    ] >> qhshift) & 3) << 4)) - 32),
-                                             dscale * float(int8_t(((data_a[ib].ql[qsi + 1] >> (b * 4)) & 0xF) | (((data_a[ib].qh[qhi + 1] >> qhshift) & 3) << 4)) - 32));
+            const uint ql = (uint(data_a_packed16[ib].ql[qsi]) >> b) & 0x0F0F;
+            const uint qh = (uint(data_a_packed16[ib].qh[qhi]) >> qhshift) & 0x0303;
+            const vec2 q = (vec2(unpack8(ql | (qh << 4)).xy) - 32) * dscale;
+
+            buf_a[buf_idx] = FLOAT_TYPE_VEC2(q.x, q.y);
 #elif defined(DATA_A_IQ1_S)
             const uint idx = pos_a + col * p.stride_a / LOAD_VEC_A + row;
             const uint buf_idx = col * SHMEM_STRIDE + row * LOAD_VEC_A / 2;

ggml/src/ggml-vulkan/vulkan-shaders/soft_max_large1.comp

@@ -0,0 +1,62 @@
+#version 450
+
+#include "soft_max_large_common.glsl"
+
+void main() {
+    const uint tid = gl_LocalInvocationID.x;
+    const uint rowx = gl_WorkGroupID.y;
+    const uint wg_start = gl_WorkGroupID.x * BLOCK_SIZE * num_iters;
+
+    const uint32_t i03 = rowx / (p.ne01 * p.ne02);
+    const uint32_t i02 = (rowx - i03 * p.ne01 * p.ne02) / p.ne01;
+    const uint32_t i01 = rowx % p.ne01;
+
+    uint rowy_start = 0;
+    if (p.KY > 0) {
+        rowy_start = i01 * p.nb11 + (i02 % p.ne12) * p.nb12 + (i03 % p.ne13) * p.nb13;
+    }
+
+    if (rowx >= p.nrows_x) {
+        return;
+    }
+
+    float slope = get_slope(rowx);
+
+    // Find max
+    FLOAT_TYPE max_val = p.has_sinks == 0 ? uintBitsToFloat(0xFF800000) : data_c[i02];
+
+    [[unroll]] for (uint col0 = wg_start, idx = 0; idx < num_iters; col0 += BLOCK_SIZE, ++idx) {
+        const uint col = col0 + tid;
+
+        FLOAT_TYPE a = FLOAT_TYPE(0);
+        if (col < p.KX) {
+            a = data_a[rowx * p.KX + col];
+        }
+
+        FLOAT_TYPE b = FLOAT_TYPE(0);
+        if (p.KY > 0 && col < p.KX) {
+            b = data_b[rowy_start + col];
+        }
+
+        FLOAT_TYPE v = a * p.scale + slope * b;
+
+        if (col < p.KX) {
+            max_val = max(max_val, v);
+        }
+    }
+
+    // reduce across the workgroup
+    vals[tid] = max_val;
+    barrier();
+    [[unroll]] for (uint s = BLOCK_SIZE / 2; s > 0; s >>= 1) {
+        if (tid < s) {
+            vals[tid] = max(vals[tid], vals[tid + s]);
+        }
+        barrier();
+    }
+
+    if (tid == 0) {
+        max_val = vals[0];
+        data_m[rowx * gl_NumWorkGroups.x + gl_WorkGroupID.x] = max_val;
+    }
+}

ggml/src/ggml-vulkan/vulkan-shaders/soft_max_large2.comp

@@ -0,0 +1,79 @@
+#version 450
+
+#include "soft_max_large_common.glsl"
+
+void main() {
+    const uint tid = gl_LocalInvocationID.x;
+    const uint rowx = gl_WorkGroupID.y;
+    const uint wg_start = gl_WorkGroupID.x * BLOCK_SIZE * num_iters;
+
+    const uint32_t i03 = rowx / (p.ne01 * p.ne02);
+    const uint32_t i02 = (rowx - i03 * p.ne01 * p.ne02) / p.ne01;
+    const uint32_t i01 = rowx % p.ne01;
+
+    uint rowy_start = 0;
+    if (p.KY > 0) {
+        rowy_start = i01 * p.nb11 + (i02 % p.ne12) * p.nb12 + (i03 % p.ne13) * p.nb13;
+    }
+
+    if (rowx >= p.nrows_x) {
+        return;
+    }
+
+    float slope = get_slope(rowx);
+
+    // Find max
+    FLOAT_TYPE max_val = p.has_sinks == 0 ? uintBitsToFloat(0xFF800000) : data_c[i02];
+
+    [[unroll]] for (uint i = 0; i < gl_NumWorkGroups.x; i += BLOCK_SIZE) {
+        if (i + tid < gl_NumWorkGroups.x) {
+            max_val = max(max_val, data_m[rowx * gl_NumWorkGroups.x + i + tid]);
+        }
+    }
+
+    // reduce across the workgroup
+    vals[tid] = max_val;
+    barrier();
+    [[unroll]] for (uint s = BLOCK_SIZE / 2; s > 0; s >>= 1) {
+        if (tid < s) {
+            vals[tid] = max(max_val, vals[tid + s]);
+        }
+        barrier();
+    }
+
+    max_val = vals[0];
+    barrier();
+
+    FLOAT_TYPE sum = FLOAT_TYPE(0.0f);
+
+    // Compute sum{exp(x - max)}
+    [[unroll]] for (uint col0 = wg_start, idx = 0; idx < num_iters; col0 += BLOCK_SIZE, ++idx) {
+        const uint col = col0 + tid;
+
+        if (col >= p.KX) {
+            break;
+        }
+
+        // compute exp(a*scale+b*slope), add it to sum
+        const uint i = rowx * p.KX + col;
+        FLOAT_TYPE val;
+        val = exp(FLOAT_TYPE(data_a[i]) * p.scale + (p.KY > 0 ? slope * FLOAT_TYPE(data_b[rowy_start + col]) : FLOAT_TYPE(0.0f)) - max_val);
+        sum += val;
+        data_d[i] = D_TYPE(val);
+    }
+
+    // reduce across the workgroup
+    vals[tid] = sum;
+    barrier();
+    [[unroll]] for (uint s = BLOCK_SIZE / 2; s > 0; s >>= 1) {
+        if (tid < s) {
+            vals[tid] += vals[tid + s];
+        }
+        barrier();
+    }
+
+    if (tid == 0) {
+        sum = vals[0];
+        data_s[rowx * gl_NumWorkGroups.x + gl_WorkGroupID.x] = sum;
+    }
+}

ggml/src/ggml-vulkan/vulkan-shaders/soft_max_large3.comp

@@ -0,0 +1,65 @@
+#version 450
+
+#include "soft_max_large_common.glsl"
+
+shared FLOAT_TYPE sumsh[BLOCK_SIZE];
+
+void main() {
+    const uint tid = gl_LocalInvocationID.x;
+    const uint rowx = gl_WorkGroupID.y;
+    const uint wg_start = gl_WorkGroupID.x * BLOCK_SIZE * num_iters;
+
+    const uint32_t i03 = rowx / (p.ne01 * p.ne02);
+    const uint32_t i02 = (rowx - i03 * p.ne01 * p.ne02) / p.ne01;
+    const uint32_t i01 = rowx % p.ne01;
+
+    uint rowy_start = 0;
+    if (p.KY > 0) {
+        rowy_start = i01 * p.nb11 + (i02 % p.ne12) * p.nb12 + (i03 % p.ne13) * p.nb13;
+    }
+
+    if (rowx >= p.nrows_x) {
+        return;
+    }
+
+    FLOAT_TYPE max_val = p.has_sinks == 0 ? uintBitsToFloat(0xFF800000) : data_c[i02];
+    FLOAT_TYPE sum = FLOAT_TYPE(0.0f);
+
+    [[unroll]] for (uint i = 0; i < gl_NumWorkGroups.x; i += BLOCK_SIZE) {
+        if (i + tid < gl_NumWorkGroups.x) {
+            max_val = max(max_val, data_m[rowx * gl_NumWorkGroups.x + i + tid]);
+            sum += data_s[rowx * gl_NumWorkGroups.x + i + tid];
+        }
+    }
+
+    // reduce across the workgroup
+    vals[tid] = max_val;
+    sumsh[tid] = sum;
+    barrier();
+    [[unroll]] for (uint s = BLOCK_SIZE / 2; s > 0; s >>= 1) {
+        if (tid < s) {
+            vals[tid] = max(max_val, vals[tid + s]);
+            sumsh[tid] += sumsh[tid + s];
+        }
+        barrier();
+    }
+
+    max_val = vals[0];
+    sum = sumsh[0];
+
+    if (p.has_sinks != 0) {
+        sum += FLOAT_TYPE(exp(FLOAT_TYPE(data_c[i02]) - max_val));
+    }
+
+    FLOAT_TYPE rcpdivisor = 1.0/sum;
+
+    [[unroll]] for (uint col0 = wg_start, idx = 0; idx < num_iters; col0 += BLOCK_SIZE, ++idx) {
+        const uint col = col0 + tid;
+
+        if (col >= p.KX) {
+            continue;
+        }
+
+        data_d[rowx*p.KX + col] *= D_TYPE(rcpdivisor);
+    }
+}

ggml/src/ggml-vulkan/vulkan-shaders/soft_max_large_common.glsl

@@ -0,0 +1,53 @@
+#extension GL_EXT_control_flow_attributes : enable
+
+layout (push_constant) uniform parameter
+{
+    uint KX;
+    uint KY;
+    uint ne00;
+    uint ne01;
+    uint ne02;
+    uint ne12;
+    uint ne13;
+    uint nb11;
+    uint nb12;
+    uint nb13;
+    float scale;
+    float max_bias;
+    float m0;
+    float m1;
+    uint n_head_log2;
+    uint nrows_x;
+    uint has_sinks;
+} p;
+
+#include "types.glsl"
+
+layout(constant_id = 0) const uint BLOCK_SIZE = 128;
+layout(local_size_x_id = 0, local_size_y = 1, local_size_z = 1) in;
+layout(constant_id = 1) const uint num_iters = 4;
+
+layout (binding = 0) readonly buffer X {A_TYPE data_a[];};
+layout (binding = 1) readonly buffer Y {B_TYPE data_b[];};
+layout (binding = 2) readonly buffer Z {float data_c[];};
+layout (binding = 3) buffer D {D_TYPE data_d[];};
+layout (binding = 4) buffer M {float data_m[];};
+layout (binding = 5) buffer S {float data_s[];};
+
+shared FLOAT_TYPE vals[BLOCK_SIZE];
+
+float get_slope(uint rowx) {
+    float slope = 1.0f;
+
+    // ALiBi
+    if (p.max_bias > 0.0f) {
+        const uint h = (rowx / p.ne01) % p.ne02; // head index
+
+        const float base = h < p.n_head_log2 ? p.m0 : p.m1;
+        const uint   exp = h < p.n_head_log2 ? h + 1 : 2*(h - p.n_head_log2) + 1;
+
+        slope = pow(base, exp);
+    }
+
+    return slope;
+}

ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp

@@ -704,13 +704,15 @@ void process_shaders() {
         shader = (tname == "f32" || tname == "f16" || tname == "bf16") ? "get_rows.comp" : "get_rows_quant.comp";
 
         if (tname == "f16") {
-            string_to_spv("get_rows_" + tname, shader, merge_maps(base_dict, {{data_a_key, "1"}, {"B_TYPE", "int"}, {"D_TYPE", "float16_t"}, {"OPTIMIZATION_ERROR_WORKAROUND", "1"}}));
+            string_to_spv("get_rows_" + tname, shader, merge_maps(base_dict, {{"TEMP_TYPE", "FLOAT_TYPE"}, {data_a_key, "1"}, {"B_TYPE", "int"}, {"D_TYPE", "float16_t"}, {"OPTIMIZATION_ERROR_WORKAROUND", "1"}}));
         } else {
-            string_to_spv("get_rows_" + tname, shader, merge_maps(base_dict, {{data_a_key, "1"}, {"B_TYPE", "int"}, {"D_TYPE", "float16_t"}}));
+            string_to_spv("get_rows_" + tname, shader, merge_maps(base_dict, {{"TEMP_TYPE", "FLOAT_TYPE"}, {data_a_key, "1"}, {"B_TYPE", "int"}, {"D_TYPE", "float16_t"}}));
         }
-        string_to_spv("get_rows_" + tname + "_f32", shader, merge_maps(base_dict, {{data_a_key, "1"}, {"B_TYPE", "int"}, {"D_TYPE", "float"}}));
+        string_to_spv("get_rows_" + tname + "_f32", shader, merge_maps(base_dict, {{"TEMP_TYPE", "FLOAT_TYPE"}, {data_a_key, "1"}, {"B_TYPE", "int"}, {"D_TYPE", "float"}}));
     }
 
+    string_to_spv("get_rows_i32", "get_rows.comp", {{"TEMP_TYPE", "uint"}, {"A_TYPE", "uint"}, {"B_TYPE", "int"}, {"D_TYPE", "uint"}});
+
     string_to_spv("mul_mat_vec_p021_f16_f32_subgroup_add", "mul_mat_vec_p021.comp", {{"A_TYPE", "float16_t"}, {"A_TYPE_VEC4", "f16vec4"}, {"B_TYPE", "float"}, {"B_TYPE_VEC4", "vec4"}, {"D_TYPE", "float"}, {"USE_SUBGROUP_ADD", "1"}});
     string_to_spv("mul_mat_vec_p021_f16_f32",              "mul_mat_vec_p021.comp", {{"A_TYPE", "float16_t"}, {"A_TYPE_VEC4", "f16vec4"}, {"B_TYPE", "float"}, {"B_TYPE_VEC4", "vec4"}, {"D_TYPE", "float"}});
     string_to_spv("mul_mat_vec_nc_f16_f32", "mul_mat_vec_nc.comp", {{"A_TYPE", "float16_t"}, {"A_TYPE_VEC4", "f16vec4"}, {"B_TYPE", "float"}, {"B_TYPE_VEC4", "vec4"}, {"D_TYPE", "float"}});
@@ -854,6 +856,8 @@ void process_shaders() {
 
     string_to_spv("tri_f16",        "tri.comp",         {{"A_TYPE", "float16_t"},   {"D_TYPE", "float16_t"}});
     string_to_spv("tri_f32",        "tri.comp",         {{"A_TYPE", "float"},       {"D_TYPE", "float"}});
+    string_to_spv("diag_f16",       "diag.comp",        {{"A_TYPE", "float16_t"},   {"D_TYPE", "float16_t"}});
+    string_to_spv("diag_f32",       "diag.comp",        {{"A_TYPE", "float"},       {"D_TYPE", "float"}});
 
     string_to_spv("softplus_f16",   "softplus.comp",    {{"A_TYPE", "float16_t"},   {"D_TYPE", "float16_t"}});
     string_to_spv("softplus_f32",   "softplus.comp",    {{"A_TYPE", "float"},       {"D_TYPE", "float"}});
@@ -899,6 +903,13 @@ void process_shaders() {
     string_to_spv("soft_max_f32_f16", "soft_max.comp", merge_maps(base_dict, {{"A_TYPE", "float"}, {"B_TYPE", "float16_t"}, {"D_TYPE", "float"}}));
     string_to_spv("soft_max_back_f32", "soft_max_back.comp", merge_maps(base_dict, {{"A_TYPE", "float"}, {"B_TYPE", "float"}, {"D_TYPE", "float"}}));
 
+    string_to_spv("soft_max_large1_f32", "soft_max_large1.comp", merge_maps(base_dict, {{"A_TYPE", "float"}, {"B_TYPE", "float"}, {"D_TYPE", "float"}}));
+    string_to_spv("soft_max_large2_f32", "soft_max_large2.comp", merge_maps(base_dict, {{"A_TYPE", "float"}, {"B_TYPE", "float"}, {"D_TYPE", "float"}}));
+    string_to_spv("soft_max_large3_f32", "soft_max_large3.comp", merge_maps(base_dict, {{"A_TYPE", "float"}, {"B_TYPE", "float"}, {"D_TYPE", "float"}}));
+    string_to_spv("soft_max_large1_f32_f16", "soft_max_large1.comp", merge_maps(base_dict, {{"A_TYPE", "float"}, {"B_TYPE", "float16_t"}, {"D_TYPE", "float"}}));
+    string_to_spv("soft_max_large2_f32_f16", "soft_max_large2.comp", merge_maps(base_dict, {{"A_TYPE", "float"}, {"B_TYPE", "float16_t"}, {"D_TYPE", "float"}}));
+    string_to_spv("soft_max_large3_f32_f16", "soft_max_large3.comp", merge_maps(base_dict, {{"A_TYPE", "float"}, {"B_TYPE", "float16_t"}, {"D_TYPE", "float"}}));
+
     string_to_spv("rope_norm_f32", "rope_norm.comp", {{"A_TYPE", "float"}, {"ROPE_D_TYPE", "float"}});
     string_to_spv("rope_norm_f16", "rope_norm.comp", {{"A_TYPE", "float16_t"}, {"ROPE_D_TYPE", "float16_t"}});
     string_to_spv("rope_norm_f16_rte", "rope_norm.comp", {{"A_TYPE", "float16_t"}, {"ROPE_D_TYPE", "float16_t"}, {"RTE16", "1"}});

scripts/compare-logprobs.py

@@ -0,0 +1,281 @@
+import argparse
+import requests
+import json
+from pathlib import Path
+import logging
+
+logger = logging.getLogger("compare-logprobs")
+logging.basicConfig(level=logging.INFO)
+
+
+DESCRIPTION = """
+Compare logits between llama.cpp and another inference engine using OpenAI-compatible server endpoints.
+
+Unlike compare-logits.py, it allows dumping logits from a hosted API endpoint. Useful when it's not possible to run both models locally.
+
+Example usage:
+    Step 1: Dump logits from two different servers
+        python scripts/compare-logprobs.py dump logits_llama.log http://localhost:8080/v1/completions
+        python scripts/compare-logprobs.py dump logits_other.log http://other-engine:8000/v1/completions
+
+        (optionally, you can add --api-key <key> if the endpoint requires authentication)
+
+    Step 2: Compare the dumped logits
+        python scripts/compare-logprobs.py compare logits_llama.log logits_other.log report.md
+"""
+
+
+def generate_input_prompt(length: int) -> list[str]:
+    CORPUS = """
+    You are an advanced AI assistant capable of using tools to gather information, perform calculations, or execute tasks. Always think step by step before responding. If a user's query requires external data, computation, or actions beyond your internal knowledge, use the appropriate tools via function calls.
+
+    ### Tool Call Format:
+    When you need to use a tool, output the call in this exact XML format. Include the opening and closing tags. Do not escape arguments; they will be parsed as plain text.
+
+    You can make multiple calls in one go by placing them one after another.
+    """
+    words = [w.strip() for w in CORPUS.strip().split(" ")]
+    words = [w for w in words if len(w) > 0]  # filter out empty strings
+    while len(words) < length:
+        words += words
+    return words[:length]
+
+
+def dump_logits(
+    endpoint: str,
+    output_path: Path,
+    input_words: list[str],
+    pattern: list[tuple[bool, int]],
+    api_key=None,
+):
+    logger.info(f"Dumping logits to {output_path} from endpoint {endpoint}...")
+    words = input_words
+    curr_text = ""
+    n_total = sum(n for get, n in pattern if get)
+    n_done = 0
+    i_cur = 0
+    i_total = len(words)
+    with output_path.open("w") as f:
+        for get, n in pattern:
+            if not get:
+                # skip n words
+                for i in range(n):
+                    curr_text += words.pop(0) + " "
+                    i_cur += 1
+                continue
+            # get n words
+            for i in range(n):
+                curr_text += words.pop(0) + " "
+                payload = {
+                    "prompt": curr_text.strip(),
+                    "temperature": 0.0,
+                    "top_k": 1,
+                    "max_tokens": 1,
+                    "logprobs": 1,
+                    "stream": False,
+                }
+                response = requests.post(
+                    endpoint,
+                    json=payload,
+                    headers={"Authorization": f"Bearer {api_key}"} if api_key else {},
+                )
+                response.raise_for_status()
+                data = response.json()
+                data["__index"] = i_cur  # add index for easier debugging later
+                data = json.dumps(data)
+                f.write(f"{data}\n")
+                n_done += 1
+                i_cur += 1
+                logger.info(
+                    f"\n\n{data}\n\n[Step: {n_done}/{n_total} | Word: {i_cur}/{i_total}]"
+                )
+    logger.info(f"Logits dumped to {output_path}")
+
+
+def get_token_logprobs(data: dict):
+    logprobs = data["choices"][0]["logprobs"]
+    if "content" in logprobs:
+        # llama.cpp case
+        top = logprobs["content"][0]["top_logprobs"][0]
+        return top["token"], top["logprob"]
+    else:
+        # vllm case
+        tokens = logprobs["tokens"]
+        token_logprobs = logprobs["token_logprobs"]
+        return tokens[0], token_logprobs[0]
+
+
+def clean_text(text: str) -> str:
+    return (
+        "'"
+        + text.replace("\n", "\\n")
+        .replace("\t", "\\t")
+        .replace("\r", "\\r")
+        .replace("|", "\\|")
+        + "'"
+    )
+
+
+def compare_logits(input1: Path, input2: Path, output_path: Path):
+    with input1.open("r") as f1, input2.open("r") as f2, output_path.open("w") as fout:
+        lines1 = f1.readlines()
+        lines2 = f2.readlines()
+
+        tab_header = [
+            "idx",
+            input1.name,
+            "logprob_1",
+            input2.name,
+            "logprob_2",
+            "diff (abs)",
+        ]
+        tab_entries = []
+        tab_max_widths = [len(h) for h in tab_header]
+
+        assert len(lines1) == len(
+            lines2
+        ), "Input files must have the same number of lines."
+
+        fout.write("# Logits Comparison Report\n\n")
+        for i, (line1, line2) in enumerate(zip(lines1, lines2)):
+            if not line1.strip() or not line2.strip():
+                continue  # skip empty lines
+
+            data1 = json.loads(line1)
+            data2 = json.loads(line2)
+
+            idx1 = data1.get("__index", -1)
+            idx2 = data2.get("__index", -1)
+            if idx1 != idx2:
+                logger.warning(
+                    f"Warning: Mismatched indices at line {i}: {idx1} vs {idx2}"
+                )
+
+            token1, logprob1 = get_token_logprobs(data1)
+            token2, logprob2 = get_token_logprobs(data2)
+
+            token1 = clean_text(token1)
+            token2 = clean_text(token2)
+            abs_diff = abs(logprob1 - logprob2)
+
+            tab_entries.append(
+                (
+                    str(idx1 + 1),
+                    token1,
+                    f"{logprob1:.4f}",
+                    token2,
+                    f"{logprob2:.4f}",
+                    f"{(abs_diff):.4f}",
+                )
+            )
+
+        for i in range(len(tab_entries)):
+            for j in range(len(tab_header)):
+                tab_max_widths[j] = max(tab_max_widths[j], len(tab_entries[i][j]))
+
+        output = ""
+        for j in range(len(tab_header)):
+            output += f"| {tab_header[j]:<{tab_max_widths[j]}} "
+        output += "|\n"
+        for j in range(len(tab_header)):
+            output += f"|{'-' * (tab_max_widths[j] + 2)}"
+        output += "|\n"
+        for entry in tab_entries:
+            for j in range(len(tab_header)):
+                output += f"| {entry[j]:<{tab_max_widths[j]}} "
+            output += "|\n"
+
+        logger.info("\n" + output)
+        fout.write(output)
+        logger.info(f"Report written to {output_path}")
+
+
+def parse_pattern(pattern: str) -> list[tuple[bool, int]]:
+    parts = pattern.split(",")
+    result = []
+    for i, part in enumerate(parts):
+        n = int(part)
+        if i % 2 == 0:
+            result.append((True, n))  # get n words
+        else:
+            result.append((False, n))  # skip n words
+    return result
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(
+        description=DESCRIPTION, formatter_class=argparse.RawTextHelpFormatter
+    )
+    subparsers = parser.add_subparsers(
+        dest="verb", required=True, help="action to perform"
+    )
+
+    # dump subcommand
+    parser_dump = subparsers.add_parser("dump", help="dump logits from an endpoint")
+    parser_dump.add_argument(
+        "output", type=Path, help="output path for dumped logits (.log)"
+    )
+    parser_dump.add_argument(
+        "endpoint", type=str, help="OAI-compat /completions endpoint"
+    )
+    parser_dump.add_argument(
+        "--api-key",
+        type=str,
+        default=None,
+        help="API key for authentication (if required)",
+    )
+    parser_dump.add_argument(
+        "--file",
+        type=Path,
+        default=None,
+        help="File containing prompt to use instead of the default",
+    )
+    parser_dump.add_argument(
+        "--pattern",
+        type=str,
+        default="10,1000,10,4000,10",
+        help="Pattern n_get,n_skip,... where n_get is number of words to get and n_skip is number of words to skip (num of words, NOT num of tokens)",
+    )
+
+    # compare subcommand
+    parser_compare = subparsers.add_parser(
+        "compare", help="compare two dumped logits files"
+    )
+    parser_compare.add_argument("input1", type=Path, help="first input file (.log)")
+    parser_compare.add_argument("input2", type=Path, help="second input file (.log)")
+    parser_compare.add_argument(
+        "output", type=Path, help="output path for comparison report (.md)"
+    )
+
+    try:
+        return parser.parse_args()
+    except Exception as e:
+        parser.print_help()
+        raise e
+
+
+def main():
+    args = parse_args()
+
+    if args.verb == "dump":
+        pattern = parse_pattern(args.pattern)
+        input_length = sum(n for _, n in pattern)
+        input_words = generate_input_prompt(input_length)
+        if args.file is not None:
+            with args.file.open("r") as f:
+                input_words = f.read().strip().split(" ")
+                if input_length < sum(n for _, n in pattern):
+                    raise ValueError(
+                        f"Input file has only {input_length} words, but pattern requires at least {input_length} words."
+                    )
+                input_length = len(input_words)
+        logger.info(f"Using {input_length} words")
+        dump_logits(args.endpoint, args.output, input_words, pattern, args.api_key)
+    elif args.verb == "compare":
+        compare_logits(args.input1, args.input2, args.output)
+    else:
+        raise ValueError(f"Unknown verb: {args.verb}")
+
+
+if __name__ == "__main__":
+    main()

scripts/sync-ggml.last

@@ -1 +1 @@
-55bc9320a4aae82af18e23eefd5de319a755d7b9
+130bc125a88bb57664b88932c48c38a1cb316fac

src/llama-context.cpp

@@ -9,6 +9,7 @@
 #include "llama-model.h"
 
 #include <cinttypes>
+#include <cmath>
 #include <cstring>
 #include <limits>
 #include <stdexcept>
@@ -72,6 +73,43 @@ llama_context::llama_context(
         cparams.yarn_ext_factor = rope_scaling_type == LLAMA_ROPE_SCALING_TYPE_YARN ? 1.0f : 0.0f;
     }
 
+    if (cparams.yarn_ext_factor != 0) {
+        static auto get_mscale = [](float scale, float mscale) {
+            return scale <= 1.0f ? 1.0f : (0.1f * mscale * logf(scale) + 1.0f);
+        };
+
+        const float factor = 1.0f / cparams.rope_freq_scale;
+
+        // ref: https://github.com/huggingface/transformers/blob/6d00f6b0a5679c36510f203e4226e36f517c3032/src/transformers/modeling_rope_utils.py#L336-L348
+        if (hparams.rope_yarn_log_mul != 0.0f) {
+            // note: here we assume `mscale == 1.0f`
+            // TODO: start reading the actual value of mscale and handle the case where it is not 1.0f
+                  float mscale          = 1.0f;
+            const float mscale_all_dims = hparams.rope_yarn_log_mul;
+
+            // [TAG_DEEPSEEK2_YARN_LOG_MUL_FIX]
+            // special-case DEEPSEEK v2:
+            // https://huggingface.co/deepseek-ai/DeepSeek-V2-Lite-Chat/blob/main/config.json#L42-L43
+            if (model.arch == LLM_ARCH_DEEPSEEK2 && mscale_all_dims != 1.0f) {
+                mscale = mscale_all_dims;
+            }
+
+            cparams.yarn_attn_factor = get_mscale(factor, mscale) / get_mscale(factor, mscale_all_dims);
+
+            LLAMA_LOG_WARN("%s: setting new yarn_attn_factor = %.4f (mscale == %.1f, mscale_all_dim = %.1f)\n",
+                    __func__, cparams.yarn_attn_factor, mscale, mscale_all_dims);
+        } else {
+            cparams.yarn_attn_factor = get_mscale(factor, 1.0f);
+        }
+
+        // when YARN is applied with yarn_ext_factor != 0.0f, we need to cancel this factor:
+        // https://github.com/ggml-org/llama.cpp/blob/a81a569577cc38b32558958b048228150be63eae/ggml/src/ggml-cpu/ops.cpp#L5541-L5544
+        //
+        // ref: https://github.com/ggml-org/llama.cpp/discussions/7416
+        //      https://github.com/ggml-org/llama.cpp/pull/17945
+        cparams.yarn_attn_factor *= 1.0f / (1.0f + 0.1f * logf(factor));
+    }
+
     cparams.yarn_attn_factor *= hparams.rope_attn_factor;
 
     if (cparams.pooling_type == LLAMA_POOLING_TYPE_UNSPECIFIED) {
@@ -1318,6 +1356,7 @@ uint32_t llama_context::output_reserve(int32_t n_outputs) {
             // This doesn't happen often, but may be annoying in some cases (like the HellaSwag benchmark)
             LLAMA_LOG_INFO("%s: reallocating output buffer from size %.02f MiB to %.02f MiB\n", __func__, prev_size / 1024.0 / 1024.0, new_size / 1024.0 / 1024.0);
 #endif
+            synchronize();
             buf_output = nullptr;
             logits = nullptr;
             embd = nullptr;

src/llama-graph.cpp

@@ -574,7 +574,7 @@ llm_graph_context::llm_graph_context(const llm_graph_params & params) :
     freq_base        (cparams.rope_freq_base),
     freq_scale       (cparams.rope_freq_scale),
     ext_factor       (cparams.yarn_ext_factor),
-    attn_factor      (llama_hparams::yarn_attn_factor_adjust(cparams.yarn_attn_factor, cparams.rope_freq_scale, cparams.yarn_ext_factor)),
+    attn_factor      (cparams.yarn_attn_factor),
     beta_fast        (cparams.yarn_beta_fast),
     beta_slow        (cparams.yarn_beta_slow),
     norm_eps         (hparams.f_norm_eps),

src/llama-hparams.cpp

@@ -3,7 +3,6 @@
 #include "ggml.h"
 
 #include <cassert>
-#include <cmath>
 
 void llama_hparams::set_swa_pattern(uint32_t n_pattern, bool dense_first) {
     if (dense_first) {
@@ -231,13 +230,3 @@ bool llama_hparams::is_masked_swa(uint32_t n_swa, llama_swa_type swa_type, llama
 
     return false;
 }
-
-float llama_hparams::yarn_attn_factor_adjust(float attn_factor, float freq_scale, float ext_factor) {
-    GGML_ASSERT(ext_factor >= 0.0f);
-
-    if (ext_factor != 0.0f) {
-        attn_factor *= 1.0f / (1.0f + 0.1f * logf(1.0f / freq_scale));
-    }
-
-    return attn_factor;
-}

src/llama-hparams.h

@@ -268,13 +268,6 @@ struct llama_hparams {
     // TODO: think of a better place for this function
     // TODO: pack the SWA params in a struct?
     static bool is_masked_swa(uint32_t n_swa, llama_swa_type swa_type, llama_pos p0, llama_pos p1);
-
-    // when YARN is applied with yarn_ext_factor != 0.0f, we need to cancel this factor:
-    // https://github.com/ggml-org/llama.cpp/blob/a81a569577cc38b32558958b048228150be63eae/ggml/src/ggml-cpu/ops.cpp#L5541-L5544
-    //
-    // ref: https://github.com/ggml-org/llama.cpp/discussions/7416
-    //      https://github.com/ggml-org/llama.cpp/pull/17945
-    static float yarn_attn_factor_adjust(float attn_factor, float freq_scale, float ext_factor);
 };
 
 static_assert(std::is_trivially_copyable<llama_hparams>::value, "llama_hparams must be trivially copyable");

src/llama-kv-cache.cpp

@@ -1372,7 +1372,7 @@ ggml_tensor * llama_kv_cache::build_rope_shift(
     const auto & yarn_ext_factor  = cparams.yarn_ext_factor;
     const auto & yarn_beta_fast   = cparams.yarn_beta_fast;
     const auto & yarn_beta_slow   = cparams.yarn_beta_slow;
-    const auto & yarn_attn_factor = llama_hparams::yarn_attn_factor_adjust(cparams.yarn_attn_factor, cparams.rope_freq_scale, cparams.yarn_ext_factor);
+    const auto & yarn_attn_factor = cparams.yarn_attn_factor;
 
     const auto & n_rot     = hparams.n_rot;
     const auto & rope_type = hparams.rope_type == LLAMA_ROPE_TYPE_MROPE || hparams.rope_type == LLAMA_ROPE_TYPE_IMROPE

src/llama-model.cpp

@@ -2294,32 +2294,6 @@ void llama_model::load_hparams(llama_model_loader & ml) {
         default: throw std::runtime_error("unsupported model architecture");
     }
 
-    // ref: https://github.com/huggingface/transformers/blob/6d00f6b0a5679c36510f203e4226e36f517c3032/src/transformers/modeling_rope_utils.py#L336-L348
-    if (hparams.rope_yarn_log_mul != 0.0f) {
-        const float factor = 1.0f / hparams.rope_freq_scale_train;
-
-        // note: here we assume `mscale == 1.0f`
-        // TODO: start reading the actual value of mscale and handle the case where it is not 1.0f
-              float mscale          = 1.0f;
-        const float mscale_all_dims = hparams.rope_yarn_log_mul;
-
-        // [TAG_DEEPSEEK2_YARN_LOG_MUL_FIX]
-        // special-case DEEPSEEK v2:
-        // https://huggingface.co/deepseek-ai/DeepSeek-V2-Lite-Chat/blob/main/config.json#L42-L43
-        if (arch == LLM_ARCH_DEEPSEEK2 && mscale_all_dims != 1.0f) {
-            mscale = mscale_all_dims;
-        }
-
-        static auto get_mscale = [](float scale, float mscale) {
-            return scale <= 1.0f ? 1.0f : (0.1f * mscale * logf(scale) + 1.0f);
-        };
-
-        hparams.yarn_attn_factor = get_mscale(factor, mscale) / get_mscale(factor, mscale_all_dims);
-
-        LLAMA_LOG_WARN("%s: setting new yarn_attn_factor = %.4f (mscale == %.1f, mscale_all_dim = %.1f)\n",
-                __func__, hparams.yarn_attn_factor, mscale, mscale_all_dims);
-    }
-
     pimpl->n_bytes = ml.n_bytes;
 
     pimpl->desc_str = arch_name() + " " + type_name() + " " + ml.ftype_name();

tests/test-arg-parser.cpp

@@ -72,6 +72,10 @@ int main(void) {
     argv = {"binary_name", "--draft", "123"};
     assert(false == common_params_parse(argv.size(), list_str_to_char(argv).data(), params, LLAMA_EXAMPLE_EMBEDDING));
 
+    // negated arg
+    argv = {"binary_name", "--no-mmap"};
+    assert(false == common_params_parse(argv.size(), list_str_to_char(argv).data(), params, LLAMA_EXAMPLE_COMMON));
+
 
     printf("test-arg-parser: test valid usage\n\n");
 

tests/test-backend-ops.cpp

@@ -7652,6 +7652,9 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_eval() {
     test_cases.emplace_back(new test_soft_max(GGML_TYPE_F32, {32, 2, 32, 1}, true,  true,  GGML_TYPE_F32, {1, 1}, 0.1f, 8.0f));
     test_cases.emplace_back(new test_soft_max(GGML_TYPE_F32, {32, 2, 32, 1}, true,  true,  GGML_TYPE_F16, {1, 1}, 0.1f, 8.0f));
 
+    test_cases.emplace_back(new test_soft_max(GGML_TYPE_F32, {200001, 2, 3, 1}, true,  true,  GGML_TYPE_F32, {1, 1}, 0.1f, 8.0f));
+    test_cases.emplace_back(new test_soft_max(GGML_TYPE_F32, {200001, 2, 3, 1}, true,  true,  GGML_TYPE_F16, {1, 1}, 0.1f, 8.0f));
+
     for (float max_bias : {0.0f, 8.0f}) {
         for (float scale : {1.0f, 0.1f}) {
             for (int64_t ne0 : {16, 1024}) {

tools/server/server-models.cpp

@@ -16,6 +16,7 @@
 #include <atomic>
 #include <chrono>
 #include <queue>
+#include <filesystem>
 
 #ifdef _WIN32
 #include <winsock2.h>
@@ -171,7 +172,7 @@ server_presets::server_presets(int argc, char ** argv, common_params & base_para
     }
 
     // read base args from router's argv
-    common_params_parse(argc, argv, LLAMA_EXAMPLE_SERVER, base_args);
+    common_params_to_map(argc, argv, LLAMA_EXAMPLE_SERVER, base_args);
 
     // remove any router-controlled args from base_args
     for (const auto & cargs : control_args) {

vendor/cpp-httplib/CMakeLists.txt

@@ -11,8 +11,9 @@ endif()
 target_link_libraries  (${TARGET} PRIVATE Threads::Threads)
 
 if (WIN32 AND NOT MSVC)
-    target_link_libraries(${TARGET} PUBLIC ws2_32)
+    target_link_libraries(${TARGET} PRIVATE ws2_32)
 endif()
+
 target_compile_features(${TARGET} PRIVATE cxx_std_17)
 
 target_compile_definitions(${TARGET} PRIVATE