@fugood/llama.node 0.2.2 → 0.3.0

package/src/llama.cpp/examples/quantize-stats/quantize-stats.cpp

@@ -154,7 +154,7 @@ static void test_roundtrip_on_chunk(
     }
 
     if (use_reference) {
-        qfns.from_float_reference(input_scratch, quantized_scratch, chunk_size);
+        qfns.from_float_ref(input_scratch, quantized_scratch, chunk_size);
     } else {
         qfns.from_float(input_scratch, quantized_scratch, chunk_size);
     }

package/src/llama.cpp/examples/retrieval/CMakeLists.txt

@@ -1,4 +1,4 @@
-set(TARGET retrieval)
+set(TARGET llama-retrieval)
 add_executable(${TARGET} retrieval.cpp)
 install(TARGETS ${TARGET} RUNTIME)
 target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})

package/src/llama.cpp/examples/retrieval/retrieval.cpp

@@ -4,72 +4,12 @@
 #include <algorithm>
 #include <fstream>
 
-struct retrieval_params {
-    std::vector<std::string> context_files; // context files to embed
-    int32_t chunk_size            = 64;     // chunk size for context embedding
-    std::string chunk_separator   = "\n";   // chunk separator for context embedding
-};
-
-static void retrieval_params_print_usage(int argc, char ** argv, gpt_params & gpt_params, retrieval_params & params) {
-    gpt_params_print_usage(argc, argv, gpt_params);
-    printf("retrieval options:\n");
-    printf("  --context-file FNAME  file containing context to embed.\n");
-    printf("                        specify multiple files by providing --context-file option multiple times.\n");
-    printf("  --chunk-size N        minimum length of embedded text chunk (default:%d)\n", params.chunk_size);
-    printf("  --chunk-separator STRING\n");
-    printf("                        string to separate chunks (default: \"\\n\")\n");
-    printf("\n");
-}
+static void print_usage(int argc, char ** argv, const gpt_params & params) {
+    gpt_params_print_usage(argc, argv, params);
 
-static void retrieval_params_parse(int argc, char ** argv, gpt_params & gpt_params, retrieval_params & retrieval_params) {
-    int i = 1;
-    std::string arg;
-    while (i < argc) {
-        arg = argv[i];
-        bool invalid_gpt_param = false;
-        if(gpt_params_find_arg(argc, argv, argv[i], gpt_params, i, invalid_gpt_param)) {
-            if (invalid_gpt_param) {
-                fprintf(stderr, "error: invalid argument: %s\n", arg.c_str());
-                retrieval_params_print_usage(argc, argv, gpt_params, retrieval_params);
-                exit(1);
-            }
-            // option was parsed by gpt_params_find_arg
-        } else if (arg == "--context-file") {
-            if (++i >= argc) {
-                fprintf(stderr, "error: missing argument for --context-file\n");
-                retrieval_params_print_usage(argc, argv, gpt_params, retrieval_params);
-                exit(1);
-            }
-            std::ifstream file(argv[i]);
-            if (!file) {
-                fprintf(stderr, "error: failed to open file '%s'\n", argv[i]);
-                retrieval_params_print_usage(argc, argv, gpt_params, retrieval_params);
-                exit(1);
-            }
-            // store the external file name in params
-            retrieval_params.context_files.push_back(argv[i]);
-        } else if (arg == "--chunk-size") {
-            if (++i >= argc) {
-                fprintf(stderr, "error: missing argument for --chunk-size\n");
-                retrieval_params_print_usage(argc, argv, gpt_params, retrieval_params);
-                exit(1);
-            }
-            retrieval_params.chunk_size = std::stoi(argv[i]);
-        } else if (arg == "--chunk-separator") {
-            if (++i >= argc) {
-                fprintf(stderr, "error: missing argument for --chunk-separator\n");
-                retrieval_params_print_usage(argc, argv, gpt_params, retrieval_params);
-                exit(1);
-            }
-            retrieval_params.chunk_separator = argv[i];
-        } else {
-            // unknown argument
-            fprintf(stderr, "error: unknown argument: %s\n", arg.c_str());
-            retrieval_params_print_usage(argc, argv, gpt_params, retrieval_params);
-            exit(1);
-        }
-        i++;
-    }
+    LOG_TEE("\nexample usage:\n");
+    LOG_TEE("\n    %s --model ./models/bge-base-en-v1.5-f16.gguf --top-k 3 --context-file README.md --context-file License --chunk-size 100 --chunk-separator .\n", argv[0]);
+    LOG_TEE("\n");
 }
 
 struct chunk {
@@ -133,9 +73,10 @@ static std::vector<chunk> chunk_file(const std::string & filename, int chunk_siz
     return chunks;
 }
 
-static void batch_add_seq(llama_batch & batch, const std::vector<int32_t> & tokens, int seq_id) {
-    for (size_t i = 0; i < tokens.size(); i++) {
-        llama_batch_add(batch, tokens[i], i, { seq_id }, i == tokens.size() - 1);
+static void batch_add_seq(llama_batch & batch, const std::vector<int32_t> & tokens, llama_seq_id seq_id) {
+    size_t n_tokens = tokens.size();
+    for (size_t i = 0; i < n_tokens; i++) {
+        llama_batch_add(batch, tokens[i], i, { seq_id }, true);
     }
 }
 
@@ -171,33 +112,35 @@ static void batch_decode(llama_context * ctx, llama_batch & batch, float * outpu
 
 int main(int argc, char ** argv) {
     gpt_params params;
-    retrieval_params retrieval_params;
 
-    retrieval_params_parse(argc, argv, params, retrieval_params);
+    if (!gpt_params_parse(argc, argv, params)) {
+        print_usage(argc, argv, params);
+        return 1;
+    }
 
     // For BERT models, batch size must be equal to ubatch size
     params.n_ubatch = params.n_batch;
+    params.embedding = true;
 
-    if (retrieval_params.chunk_size <= 0) {
+    if (params.chunk_size <= 0) {
         fprintf(stderr, "chunk_size must be positive\n");
         return 1;
     }
-    if (retrieval_params.context_files.empty()) {
+    if (params.context_files.empty()) {
         fprintf(stderr, "context_files must be specified\n");
         return 1;
     }
-    params.embedding = true;
 
     print_build_info();
 
     printf("processing files:\n");
-    for (auto & context_file : retrieval_params.context_files) {
+    for (auto & context_file : params.context_files) {
         printf("%s\n", context_file.c_str());
     }
 
     std::vector<chunk> chunks;
-    for (auto & context_file : retrieval_params.context_files) {
-        std::vector<chunk> file_chunk = chunk_file(context_file, retrieval_params.chunk_size, retrieval_params.chunk_separator);
+    for (auto & context_file : params.context_files) {
+        std::vector<chunk> file_chunk = chunk_file(context_file, params.chunk_size, params.chunk_separator);
         chunks.insert(chunks.end(), file_chunk.begin(), file_chunk.end());
     }
     printf("Number of chunks: %ld\n", chunks.size());
@@ -218,6 +161,12 @@ int main(int argc, char ** argv) {
     const int n_ctx_train = llama_n_ctx_train(model);
     const int n_ctx = llama_n_ctx(ctx);
 
+    const enum llama_pooling_type pooling_type = llama_pooling_type(ctx);
+    if (pooling_type == LLAMA_POOLING_TYPE_NONE) {
+        fprintf(stderr, "%s: error: pooling type NONE not supported\n", __func__);
+        return 1;
+    }
+
     if (n_ctx > n_ctx_train) {
         fprintf(stderr, "%s: warning: model was trained on only %d context tokens (%d specified)\n",
                 __func__, n_ctx_train, n_ctx);
@@ -242,7 +191,7 @@ int main(int argc, char ** argv) {
             return 1;
         }
         // add eos if not present
-        if (inp.empty() || inp.back() != llama_token_eos(model)) {
+        if (llama_token_eos(model) >= 0 && (inp.empty() || inp.back() != llama_token_eos(model))) {
             inp.push_back(llama_token_eos(model));
         }
         chunk.tokens = inp;

package/src/llama.cpp/examples/save-load-state/CMakeLists.txt

@@ -1,4 +1,4 @@
-set(TARGET save-load-state)
+set(TARGET llama-save-load-state)
 add_executable(${TARGET} save-load-state.cpp)
 install(TARGETS ${TARGET} RUNTIME)
 target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})

package/src/llama.cpp/examples/save-load-state/save-load-state.cpp

@@ -11,6 +11,7 @@ int main(int argc, char ** argv) {
     params.prompt = "The quick brown fox";
 
     if (!gpt_params_parse(argc, argv, params)) {
+        gpt_params_print_usage(argc, argv, params);
         return 1;
     }
 
@@ -46,7 +47,7 @@ int main(int argc, char ** argv) {
     // save state (rng, logits, embedding and kv_cache) to file
     {
         std::vector<uint8_t> state_mem(llama_state_get_size(ctx));
-        const size_t written = llama_state_get_data(ctx, state_mem.data());
+        const size_t written = llama_state_get_data(ctx, state_mem.data(), state_mem.size());
 
         FILE *fp_write = fopen("dump_state.bin", "wb");
         fwrite(state_mem.data(), 1, written, fp_write);
@@ -98,13 +99,16 @@ int main(int argc, char ** argv) {
 
     // load state (rng, logits, embedding and kv_cache) from file
     {
-        std::vector<uint8_t> state_mem(llama_state_get_size(ctx2));
+        std::vector<uint8_t> state_mem;
 
         FILE * fp_read = fopen("dump_state.bin", "rb");
+        fseek(fp_read, 0, SEEK_END);
+        state_mem.resize(ftell(fp_read));
+        fseek(fp_read, 0, SEEK_SET);
         const size_t read = fread(state_mem.data(), 1, state_mem.size(), fp_read);
         fclose(fp_read);
 
-        if (read != llama_state_set_data(ctx2, state_mem.data())) {
+        if (read != llama_state_set_data(ctx2, state_mem.data(), state_mem.size())) {
             fprintf(stderr, "\n%s : failed to read state\n", __func__);
             llama_free(ctx2);
             llama_free_model(model);
@@ -158,13 +162,16 @@ int main(int argc, char ** argv) {
 
     // load state (rng, logits, embedding and kv_cache) from file
     {
-        std::vector<uint8_t> state_mem(llama_state_get_size(ctx3));
+        std::vector<uint8_t> state_mem;
 
         FILE * fp_read = fopen("dump_state.bin", "rb");
+        fseek(fp_read, 0, SEEK_END);
+        state_mem.resize(ftell(fp_read));
+        fseek(fp_read, 0, SEEK_SET);
         const size_t read = fread(state_mem.data(), 1, state_mem.size(), fp_read);
         fclose(fp_read);
 
-        if (read != llama_state_set_data(ctx3, state_mem.data())) {
+        if (read != llama_state_set_data(ctx3, state_mem.data(), state_mem.size())) {
             fprintf(stderr, "\n%s : failed to read state\n", __func__);
             llama_free(ctx3);
             llama_free_model(model);
@@ -181,7 +188,7 @@ int main(int argc, char ** argv) {
     {
         // save kv of seq 0
         std::vector<uint8_t> seq_store(llama_state_seq_get_size(ctx3, 0));
-        const size_t ncopy = llama_state_seq_get_data(ctx3, seq_store.data(), 0);
+        const size_t ncopy = llama_state_seq_get_data(ctx3, seq_store.data(), seq_store.size(), 0);
         if (ncopy != seq_store.size()) {
             fprintf(stderr, "\n%s : seq copy data length %zd does not match expected length %zd\n", __func__, ncopy, seq_store.size());
             llama_free(ctx3);
@@ -195,7 +202,7 @@ int main(int argc, char ** argv) {
         fprintf(stderr, "%s : kv cache cleared\n", __func__);
 
         // restore kv into seq 1
-        const size_t nset = llama_state_seq_set_data(ctx3, seq_store.data(), 1);
+        const size_t nset = llama_state_seq_set_data(ctx3, seq_store.data(), seq_store.size(), 1);
         if (nset != seq_store.size()) {
             fprintf(stderr, "\n%s : seq set data length %zd does not match expected length %zd\n", __func__, nset, seq_store.size());
             llama_free(ctx3);

package/src/llama.cpp/examples/server/CMakeLists.txt

@@ -1,18 +1,37 @@
-set(TARGET server)
+set(TARGET llama-server)
 option(LLAMA_SERVER_VERBOSE "Build verbose logging option for Server" ON)
-option(LLAMA_SERVER_SSL "Build SSL support for the server" OFF)
+option(LLAMA_SERVER_SSL     "Build SSL support for the server"        OFF)
+
 include_directories(${CMAKE_CURRENT_SOURCE_DIR} ${CMAKE_CURRENT_BINARY_DIR})
+
+if (MINGW)
+    # fix: https://github.com/ggerganov/llama.cpp/actions/runs/9651004652/job/26617901362?pr=8006
+    add_compile_definitions(_WIN32_WINNT=${GGML_WIN_VER})
+endif()
+
 set(TARGET_SRCS
     server.cpp
     utils.hpp
     httplib.h
 )
 set(PUBLIC_ASSETS
+    colorthemes.css
+    style.css
+    theme-beeninorder.css
+    theme-ketivah.css
+    theme-mangotango.css
+    theme-playground.css
+    theme-polarnight.css
+    theme-snowstorm.css
     index.html
+    index-new.html
     index.js
     completion.js
+    system-prompts.js
+    prompt-formats.js
     json-schema-to-grammar.mjs
 )
+
 foreach(asset ${PUBLIC_ASSETS})
     set(input "${CMAKE_CURRENT_SOURCE_DIR}/public/${asset}")
     set(output "${CMAKE_CURRENT_BINARY_DIR}/${asset}.hpp")
@@ -23,18 +42,23 @@ foreach(asset ${PUBLIC_ASSETS})
         COMMAND "${CMAKE_COMMAND}" "-DINPUT=${input}" "-DOUTPUT=${output}" -P "${PROJECT_SOURCE_DIR}/scripts/xxd.cmake"
     )
 endforeach()
+
 add_executable(${TARGET} ${TARGET_SRCS})
 install(TARGETS ${TARGET} RUNTIME)
 target_compile_definitions(${TARGET} PRIVATE
     SERVER_VERBOSE=$<BOOL:${LLAMA_SERVER_VERBOSE}>
 )
+
 target_link_libraries(${TARGET} PRIVATE common ${CMAKE_THREAD_LIBS_INIT})
+
 if (LLAMA_SERVER_SSL)
     find_package(OpenSSL REQUIRED)
     target_link_libraries(${TARGET} PRIVATE OpenSSL::SSL OpenSSL::Crypto)
     target_compile_definitions(${TARGET} PRIVATE CPPHTTPLIB_OPENSSL_SUPPORT)
 endif()
+
 if (WIN32)
     TARGET_LINK_LIBRARIES(${TARGET} PRIVATE ws2_32)
 endif()
+
 target_compile_features(${TARGET} PRIVATE cxx_std_11)