Merge branch 'upstream' into concedo_experimental

# Conflicts:
#	CMakeLists.txt
#	Makefile
#	README.md
#	common/common.cpp
#	requirements/requirements-convert-hf-to-gguf-update.txt
#	requirements/requirements-convert-hf-to-gguf.txt
#	requirements/requirements-convert.txt
#	tests/CMakeLists.txt
#	tests/test-json-schema-to-grammar.cpp

common/common.cpp

@@ -1,5 +1,7 @@
 #include "common.h"
 #include "build-info.h"
+// Change JSON_ASSERT from assert() to GGML_ASSERT:
+#define JSON_ASSERT GGML_ASSERT
 #include "json.hpp"
 #include "json-schema-to-grammar.h"
 #include "llama.h"
@@ -912,6 +914,10 @@ bool gpt_params_find_arg(int argc, char ** argv, const std::string & arg, gpt_pa
         params.instruct = true;
         return true;
     }
+    if (arg == "-cnv" || arg == "--conversation") {
+        params.conversation = true;
+        return true;
+    }
     if (arg == "-cml" || arg == "--chatml") {
         params.chatml = true;
         return true;
@@ -1418,6 +1424,7 @@ void gpt_print_usage(int /*argc*/, char ** argv, const gpt_params & params) {
     printf("  --version             show version and build info\n");
     printf("  -i, --interactive     run in interactive mode\n");
     printf("  --interactive-first   run in interactive mode and wait for input right away\n");
+    printf("  -cnv, --conversation  run in conversation mode (does not print special tokens and suffix/prefix)\n");
     printf("  -ins, --instruct      run in instruction mode (use with Alpaca models)\n");
     printf("  -cml, --chatml        run in chatml mode (use with ChatML-compatible models)\n");
     printf("  --multiline-input     allows you to write or paste multiple lines without ending each in '\\'\n");
@@ -1965,18 +1972,18 @@ static bool llama_download_file(const std::string & url, const std::string & pat
             try {
                 metadata_in >> metadata;
                 fprintf(stderr, "%s: previous metadata file found %s: %s\n", __func__, metadata_path.c_str(), metadata.dump().c_str());
-                if (metadata.contains("url") && metadata["url"].is_string()) {
+                if (metadata.contains("url") && metadata.at("url").is_string()) {
-                    auto previous_url = metadata["url"].get<std::string>();
+                    auto previous_url = metadata.at("url").get<std::string>();
                     if (previous_url != url) {
                         fprintf(stderr, "%s: Model URL mismatch: %s != %s\n", __func__, url.c_str(), previous_url.c_str());
                         return false;
                     }
                 }
-                if (metadata.contains("etag") && metadata["etag"].is_string()) {
+                if (metadata.contains("etag") && metadata.at("etag").is_string()) {
-                    etag = metadata["etag"];
+                    etag = metadata.at("etag");
                 }
-                if (metadata.contains("lastModified") && metadata["lastModified"].is_string()) {
+                if (metadata.contains("lastModified") && metadata.at("lastModified").is_string()) {
-                    last_modified = metadata["lastModified"];
+                    last_modified = metadata.at("lastModified");
                 }
             } catch (const nlohmann::json::exception & e) {
                 fprintf(stderr, "%s: error reading metadata file %s: %s\n", __func__, metadata_path.c_str(), e.what());

common/common.h

@@ -156,6 +156,7 @@ struct gpt_params {
     bool random_prompt     = false; // do not randomize prompt if none provided
     bool use_color         = false; // use color to distinguish generations and inputs
     bool interactive       = false; // interactive mode
+    bool conversation      = false; // conversation mode (does not print special tokens and suffix/prefix)
     bool chatml            = false; // chatml mode (used for models trained on chatml syntax)
     bool prompt_cache_all  = false; // save user input and generations to prompt cache
     bool prompt_cache_ro   = false; // open the prompt cache read-only and do not update it

common/json-schema-to-grammar.h

@@ -1,4 +1,8 @@
 #pragma once
+
+#include "ggml.h"
+// Change JSON_ASSERT from assert() to GGML_ASSERT:
+#define JSON_ASSERT GGML_ASSERT
 #include "json.hpp"
 
 std::string json_schema_to_grammar(const nlohmann::ordered_json& schema);

convert-hf-to-gguf-update.py

@@ -67,7 +67,9 @@ models = [
     {"name": "gpt-2",          "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/openai-community/gpt2", },
     {"name": "refact",         "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/smallcloudai/Refact-1_6-base", },
     {"name": "command-r",      "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/CohereForAI/c4ai-command-r-v01", },
+    {"name": "qwen2",          "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/Qwen/Qwen1.5-7B", },
     {"name": "olmo",           "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/allenai/OLMo-1.7-7B-hf", },
+    {"name": "dbrx",           "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/databricks/dbrx-base", },
 ]
 
 # make directory "models/tokenizers" if it doesn't exist

convert.py

@@ -284,6 +284,7 @@ class Params:
         n_experts      = None
         n_experts_used = None
         f_rope_freq_base = None
+        n_ff = None
 
         # hack to determine LLaMA v1 vs v2 vs CodeLlama
         if config.get("moe"):
@@ -308,6 +309,8 @@ class Params:
             n_experts_used = config["moe"]["num_experts_per_tok"]
             f_rope_freq_base = 1e6
 
+        assert n_ff is not None
+
         return Params(
             n_vocab          = model["tok_embeddings.weight"].shape[0],
             n_embd           = config["dim"],
@@ -462,7 +465,8 @@ class SentencePieceVocab(Vocab):
             # not found in alternate location either
             raise FileNotFoundError('Cannot find tokenizer.model')
 
-        self.sentencepiece_tokenizer = SentencePieceProcessor(str(fname_tokenizer))
+        self.sentencepiece_tokenizer = SentencePieceProcessor()
+        self.sentencepiece_tokenizer.LoadFromFile(str(fname_tokenizer))
         vocab_size = self.sentencepiece_tokenizer.vocab_size()
 
         new_tokens       = {id: piece for piece, id in added_tokens.items() if id >= vocab_size}
@@ -482,23 +486,23 @@ class SentencePieceVocab(Vocab):
     def sentencepiece_tokens(self) -> Iterable[tuple[bytes, float, gguf.TokenType]]:
         tokenizer = self.sentencepiece_tokenizer
         for i in range(tokenizer.vocab_size()):
-            piece = tokenizer.id_to_piece(i)
+            piece = tokenizer.IdToPiece(i)
             text         = piece.encode("utf-8")
-            score: float = tokenizer.get_score(i)
+            score: float = tokenizer.GetScore(i)
 
             toktype = gguf.TokenType.NORMAL
-            if tokenizer.is_unknown(i):
+            if tokenizer.IsUnknown(i):
                 toktype = gguf.TokenType.UNKNOWN
-            if tokenizer.is_control(i):
+            if tokenizer.IsControl(i):
                 toktype = gguf.TokenType.CONTROL
 
             # NOTE: I think added_tokens are user defined.
             # ref: https://github.com/google/sentencepiece/blob/master/src/sentencepiece_model.proto
             # if tokenizer.is_user_defined(i): toktype = gguf.TokenType.USER_DEFINED
 
-            if tokenizer.is_unused(i):
+            if tokenizer.IsUnused(i):
                 toktype = gguf.TokenType.UNUSED
-            if tokenizer.is_byte(i):
+            if tokenizer.IsByte(i):
                 toktype = gguf.TokenType.BYTE
 
             yield text, score, toktype
@@ -906,7 +910,7 @@ class LazyUnpickler(pickle.Unpickler):
     def rebuild_from_type_v2(func, new_type, args, state):
         return func(*args)
 
-    CLASSES = {
+    CLASSES: dict[tuple[str, str], type[LazyTensor] | LazyStorageKind] = {
         # getattr used here as a workaround for mypy not being smart enough to determine
         # the staticmethods have a __func__ attribute.
         ('torch._tensor', '_rebuild_from_type_v2'): getattr(rebuild_from_type_v2, '__func__'),
@@ -1508,25 +1512,27 @@ def main(args_in: list[str] | None = None) -> None:
     if args.big_endian:
         endianess = gguf.GGUFEndian.BIG
 
-    params = Params.load(model_plus)
+    params = None
-    if params.n_ctx == -1:
+    if args.pad_vocab or not args.vocab_only:
-        if args.ctx is None:
+        params = Params.load(model_plus)
-            msg = """\
+        if params.n_ctx == -1:
-                The model doesn't have a context size, and you didn't specify one with --ctx
+            if args.ctx is None:
-                Please specify one with --ctx:
+                msg = """\
-                 - LLaMA v1: --ctx 2048
+                    The model doesn't have a context size, and you didn't specify one with --ctx
-                 - LLaMA v2: --ctx 4096"""
+                    Please specify one with --ctx:
-            parser.error(textwrap.dedent(msg))
+                     - LLaMA v1: --ctx 2048
-        params.n_ctx = args.ctx
+                     - LLaMA v2: --ctx 4096"""
+                parser.error(textwrap.dedent(msg))
+            params.n_ctx = args.ctx
 
-    if args.outtype:
+        if args.outtype:
-        params.ftype = {
+            params.ftype = {
-            "f32": GGMLFileType.AllF32,
+                "f32": GGMLFileType.AllF32,
-            "f16": GGMLFileType.MostlyF16,
+                "f16": GGMLFileType.MostlyF16,
-            "q8_0": GGMLFileType.MostlyQ8_0,
+                "q8_0": GGMLFileType.MostlyQ8_0,
-        }[args.outtype]
+            }[args.outtype]
 
-    logger.info(f"params = {params}")
+        logger.info(f"params = {params}")
 
     model_parent_path = model_plus.paths[0].parent
     vocab_path = Path(args.vocab_dir or args.model or model_parent_path)
@@ -1539,6 +1545,17 @@ def main(args_in: list[str] | None = None) -> None:
         if not args.outfile:
             raise ValueError("need --outfile if using --vocab-only")
         outfile = args.outfile
+        if params is None:
+            params = Params(
+                n_vocab    = vocab.vocab_size,
+                n_embd     = 1,
+                n_layer    = 1,
+                n_ctx      = 1,
+                n_ff       = 1,
+                n_head     = 1,
+                n_head_kv  = 1,
+                f_norm_eps = 1e-5,
+            )
         OutputFile.write_vocab_only(outfile, params, vocab, special_vocab,
                                     endianess=endianess, pad_vocab=args.pad_vocab)
         logger.info(f"Wrote {outfile}")

examples/main/main.cpp

@@ -363,6 +363,9 @@ int main(int argc, char ** argv) {
         params.interactive_first = true;
         params.antiprompt.emplace_back("<|im_start|>user\n");
     }
+    else if (params.conversation) {
+        params.interactive_first = true;
+    }
 
     // enable interactive mode if interactive start is specified
     if (params.interactive_first) {
@@ -734,7 +737,7 @@ int main(int argc, char ** argv) {
         // display text
         if (input_echo && display) {
             for (auto id : embd) {
-                const std::string token_str = llama_token_to_piece(ctx, id);
+                const std::string token_str = llama_token_to_piece(ctx, id, !params.conversation);
                 printf("%s", token_str.c_str());
 
                 if (embd.size() > 1) {
@@ -817,7 +820,7 @@ int main(int argc, char ** argv) {
             if (n_past > 0 && is_interacting) {
                 LOG("waiting for user input\n");
 
-                if (params.instruct || params.chatml) {
+                if (params.conversation || params.instruct || params.chatml) {
                     printf("\n> ");
                 }
 
@@ -827,7 +830,7 @@ int main(int argc, char ** argv) {
                 }
 
                 std::string buffer;
-                if (!params.input_prefix.empty()) {
+                if (!params.input_prefix.empty() && !params.conversation) {
                     LOG("appending input prefix: '%s'\n", params.input_prefix.c_str());
                     printf("%s", params.input_prefix.c_str());
                 }
@@ -851,7 +854,7 @@ int main(int argc, char ** argv) {
                 // Entering a empty line lets the user pass control back
                 if (buffer.length() > 1) {
                     // append input suffix if any
-                    if (!params.input_suffix.empty()) {
+                    if (!params.input_suffix.empty() && !params.conversation) {
                         LOG("appending input suffix: '%s'\n", params.input_suffix.c_str());
                         printf("%s", params.input_suffix.c_str());
                     }

examples/server/README.md

@@ -331,7 +331,7 @@ Notice that each `probs` is an array of length `n_probs`.
 
     `content`: Set the text to tokenize.
 
-    Note that a special `BOS` token is never inserted.
+    `add_special`: Boolean indicating if special tokens, i.e. `BOS`, should be inserted.  Default: `false`
 
 - **POST** `/detokenize`: Convert tokens to text.
 

examples/server/server.cpp

@@ -13,6 +13,8 @@
 // increase max payload length to allow use of larger context size
 #define CPPHTTPLIB_FORM_URL_ENCODED_PAYLOAD_MAX_LENGTH 1048576
 #include "httplib.h"
+// Change JSON_ASSERT from assert() to GGML_ASSERT:
+#define JSON_ASSERT GGML_ASSERT
 #include "json.hpp"
 
 // auto generated files (update with ./deps.sh)
@@ -860,7 +862,7 @@ struct server_context {
         slot.sparams.min_keep          = json_value(data, "min_keep",          default_sparams.min_keep);
 
         // process "json_schema" and "grammar"
-        if (data.contains("json_schema") && !data["json_schema"].is_null() && data.contains("grammar") && !data["grammar"].is_null()) {
+        if (data.contains("json_schema") && !data.at("json_schema").is_null() && data.contains("grammar") && !data.at("grammar").is_null()) {
             send_error(task, "Either \"json_schema\" or \"grammar\" can be specified, but not both", ERROR_TYPE_INVALID_REQUEST);
             return false;
         } else if (data.contains("json_schema") && !data.contains("grammar")) {
@@ -1513,7 +1515,7 @@ struct server_context {
         // add subtasks
         for (int i = 0; i < prompt_count; i++) {
             json subtask_data = multiprompt_task.data;
-            subtask_data["prompt"] = subtask_data["prompt"][i];
+            subtask_data["prompt"] = subtask_data.at("prompt")[i];
 
             // subtasks inherit everything else (infill mode, embedding mode, etc.)
             request_completion(subtask_ids[i], id_multi, subtask_data, multiprompt_task.infill, multiprompt_task.embedding);
@@ -1533,7 +1535,7 @@ struct server_context {
                     }
 
                     if (task.data.contains("system_prompt")) {
-                        system_prompt_set(task.data["system_prompt"]);
+                        system_prompt_set(task.data.at("system_prompt"));
 
                         for (server_slot & slot : slots) {
                             slot.n_past    = 0;
@@ -1645,7 +1647,7 @@ struct server_context {
                 } break;
             case SERVER_TASK_TYPE_SLOT_SAVE:
                 {
-                    int id_slot = task.data["id_slot"];
+                    int id_slot = task.data.at("id_slot");
                     server_slot * slot = get_slot(id_slot);
                     if (slot == nullptr) {
                         send_error(task, "Invalid slot ID", ERROR_TYPE_INVALID_REQUEST);
@@ -1655,8 +1657,8 @@ struct server_context {
                     const size_t token_count = slot->cache_tokens.size();
                     const int64_t t_start = ggml_time_us();
 
-                    std::string filename = task.data["filename"];
+                    std::string filename = task.data.at("filename");
-                    std::string filepath = task.data["filepath"];
+                    std::string filepath = task.data.at("filepath");
 
                     const size_t nwrite = llama_state_seq_save_file(ctx, filepath.c_str(), slot->id + 1, slot->cache_tokens.data(), token_count);
 
@@ -1680,7 +1682,7 @@ struct server_context {
                 } break;
             case SERVER_TASK_TYPE_SLOT_RESTORE:
                 {
-                    int id_slot = task.data["id_slot"];
+                    int id_slot = task.data.at("id_slot");
                     server_slot * slot = get_slot(id_slot);
                     if (slot == nullptr) {
                         send_error(task, "Invalid slot ID", ERROR_TYPE_INVALID_REQUEST);
@@ -1689,8 +1691,8 @@ struct server_context {
 
                     const int64_t t_start = ggml_time_us();
 
-                    std::string filename = task.data["filename"];
+                    std::string filename = task.data.at("filename");
-                    std::string filepath = task.data["filepath"];
+                    std::string filepath = task.data.at("filepath");
 
                     slot->cache_tokens.resize(slot->n_ctx);
                     size_t token_count = 0;
@@ -1722,7 +1724,7 @@ struct server_context {
                 } break;
             case SERVER_TASK_TYPE_SLOT_ERASE:
                 {
-                    int id_slot = task.data["id_slot"];
+                    int id_slot = task.data.at("id_slot");
                     server_slot * slot = get_slot(id_slot);
                     if (slot == nullptr) {
                         send_error(task, "Invalid slot ID", ERROR_TYPE_INVALID_REQUEST);
@@ -3137,8 +3139,8 @@ int main(int argc, char ** argv) {
                     server_task_result result = ctx_server.queue_results.recv(task.id);
                     ctx_server.queue_results.remove_waiting_task_id(task.id);
 
-                    const int n_idle_slots       = result.data["idle"];
+                    const int n_idle_slots       = result.data.at("idle");
-                    const int n_processing_slots = result.data["processing"];
+                    const int n_processing_slots = result.data.at("processing");
 
                     json health = {
                         {"status",           "ok"},
@@ -3148,7 +3150,7 @@ int main(int argc, char ** argv) {
 
                     res.status = 200; // HTTP OK
                     if (sparams.slots_endpoint && req.has_param("include_slots")) {
-                        health["slots"] = result.data["slots"];
+                        health["slots"] = result.data.at("slots");
                     }
 
                     if (n_idle_slots == 0) {
@@ -3192,7 +3194,7 @@ int main(int argc, char ** argv) {
         server_task_result result = ctx_server.queue_results.recv(task.id);
         ctx_server.queue_results.remove_waiting_task_id(task.id);
 
-        res.set_content(result.data["slots"].dump(), "application/json");
+        res.set_content(result.data.at("slots").dump(), "application/json");
         res.status = 200; // HTTP OK
     };
 
@@ -3219,32 +3221,32 @@ int main(int argc, char ** argv) {
 
         json data = result.data;
 
-        const uint64_t n_prompt_tokens_processed = data["n_prompt_tokens_processed"];
+        const uint64_t n_prompt_tokens_processed = data.at("n_prompt_tokens_processed");
-        const uint64_t t_prompt_processing       = data["t_prompt_processing"];
+        const uint64_t t_prompt_processing       = data.at("t_prompt_processing");
 
-        const uint64_t n_tokens_predicted  = data["n_tokens_predicted"];
+        const uint64_t n_tokens_predicted  = data.at("n_tokens_predicted");
-        const uint64_t t_tokens_generation = data["t_tokens_generation"];
+        const uint64_t t_tokens_generation = data.at("t_tokens_generation");
 
-        const int32_t kv_cache_used_cells = data["kv_cache_used_cells"];
+        const int32_t kv_cache_used_cells = data.at("kv_cache_used_cells");
 
         // metrics definition: https://prometheus.io/docs/practices/naming/#metric-names
         json all_metrics_def = json {
             {"counter", {{
                     {"name",  "prompt_tokens_total"},
                     {"help",  "Number of prompt tokens processed."},
-                    {"value",  (uint64_t) data["n_prompt_tokens_processed_total"]}
+                    {"value",  (uint64_t) data.at("n_prompt_tokens_processed_total")}
             }, {
                     {"name",  "prompt_seconds_total"},
                     {"help",  "Prompt process time"},
-                    {"value",  (uint64_t) data["t_prompt_processing_total"] / 1.e3}
+                    {"value",  (uint64_t) data.at("t_prompt_processing_total") / 1.e3}
             }, {
                     {"name",  "tokens_predicted_total"},
                     {"help",  "Number of generation tokens processed."},
-                    {"value",  (uint64_t) data["n_tokens_predicted_total"]}
+                    {"value",  (uint64_t) data.at("n_tokens_predicted_total")}
             }, {
                     {"name",  "tokens_predicted_seconds_total"},
                     {"help",  "Predict process time"},
-                    {"value",  (uint64_t) data["t_tokens_generation_total"] / 1.e3}
+                    {"value",  (uint64_t) data.at("t_tokens_generation_total") / 1.e3}
             }}},
             {"gauge", {{
                     {"name",  "prompt_tokens_seconds"},
@@ -3261,15 +3263,15 @@ int main(int argc, char ** argv) {
             },{
                     {"name",  "kv_cache_tokens"},
                     {"help",  "KV-cache tokens."},
-                    {"value",  (uint64_t) data["kv_cache_tokens_count"]}
+                    {"value",  (uint64_t) data.at("kv_cache_tokens_count")}
             },{
                     {"name",  "requests_processing"},
                     {"help",  "Number of request processing."},
-                    {"value",  (uint64_t) data["processing"]}
+                    {"value",  (uint64_t) data.at("processing")}
             },{
                     {"name",  "requests_deferred"},
                     {"help",  "Number of request deferred."},
-                    {"value",  (uint64_t) data["deferred"]}
+                    {"value",  (uint64_t) data.at("deferred")}
             }}}
         };
 
@@ -3280,8 +3282,8 @@ int main(int argc, char ** argv) {
             const auto & metrics_def = el.value();
 
             for (const auto & metric_def : metrics_def) {
-                const std::string name = metric_def["name"];
+                const std::string name = metric_def.at("name");
-                const std::string help = metric_def["help"];
+                const std::string help = metric_def.at("help");
 
                 auto value = json_value(metric_def, "value", 0.);
                 prometheus << "# HELP llamacpp:" << name << " " << help  << "\n"
@@ -3290,7 +3292,7 @@ int main(int argc, char ** argv) {
             }
         }
 
-        const int64_t t_start = data["t_start"];
+        const int64_t t_start = data.at("t_start");
         res.set_header("Process-Start-Time-Unix", std::to_string(t_start));
 
         res.set_content(prometheus.str(), "text/plain; version=0.0.4");
@@ -3299,7 +3301,7 @@ int main(int argc, char ** argv) {
 
     const auto handle_slots_save = [&ctx_server, &res_error, &sparams](const httplib::Request & req, httplib::Response & res, int id_slot) {
         json request_data = json::parse(req.body);
-        std::string filename = request_data["filename"];
+        std::string filename = request_data.at("filename");
         if (!validate_file_name(filename)) {
             res_error(res, format_error_response("Invalid filename", ERROR_TYPE_INVALID_REQUEST));
             return;
@@ -3329,7 +3331,7 @@ int main(int argc, char ** argv) {
 
     const auto handle_slots_restore = [&ctx_server, &res_error, &sparams](const httplib::Request & req, httplib::Response & res, int id_slot) {
         json request_data = json::parse(req.body);
-        std::string filename = request_data["filename"];
+        std::string filename = request_data.at("filename");
         if (!validate_file_name(filename)) {
             res_error(res, format_error_response("Invalid filename", ERROR_TYPE_INVALID_REQUEST));
             return;
@@ -3648,7 +3650,8 @@ int main(int argc, char ** argv) {
 
         std::vector<llama_token> tokens;
         if (body.count("content") != 0) {
-            tokens = ctx_server.tokenize(body["content"], false);
+            const bool add_special = json_value(body, "add_special", false);
+            tokens = ctx_server.tokenize(body.at("content"), add_special);
         }
         const json data = format_tokenizer_response(tokens);
         return res.set_content(data.dump(), "application/json; charset=utf-8");
@@ -3660,7 +3663,7 @@ int main(int argc, char ** argv) {
 
         std::string content;
         if (body.count("tokens") != 0) {
-            const std::vector<llama_token> tokens = body["tokens"];
+            const std::vector<llama_token> tokens = body.at("tokens");
             content = tokens_to_str(ctx_server.ctx, tokens.cbegin(), tokens.cend());
         }
 
@@ -3683,10 +3686,10 @@ int main(int argc, char ** argv) {
         json prompt;
         if (body.count("input") != 0) {
             is_openai = true;
-            prompt = body["input"];
+            prompt = body.at("input");
         } else if (body.count("content") != 0) {
             // with "content", we only support single prompt
-            prompt = std::vector<std::string>{body["content"]};
+            prompt = std::vector<std::string>{body.at("content")};
         } else {
             res_error(res, format_error_response("\"input\" or \"content\" must be provided", ERROR_TYPE_INVALID_REQUEST));
             return;
@@ -3705,7 +3708,7 @@ int main(int argc, char ** argv) {
             if (!result.error) {
                 if (result.data.count("results")) {
                     // result for multi-task
-                    responses = result.data["results"];
+                    responses = result.data.at("results");
                 } else {
                     // result for single task
                     responses = std::vector<json>{result.data};

examples/server/tests/features/server.feature

@@ -7,6 +7,7 @@ Feature: llama.cpp server
     And   a model file tinyllamas/stories260K.gguf from HF repo ggml-org/models
     And   a model file test-model.gguf
     And   a model alias tinyllama-2
+    And   BOS token is 1
     And   42 as server seed
       # KV Cache corresponds to the total amount of tokens
       # that can be stored across all independent sequences: #4130
@@ -91,7 +92,18 @@ Feature: llama.cpp server
     """
     What is the capital of France ?
     """
-    Then tokens can be detokenize
+    Then tokens can be detokenized
+    And  tokens do not begin with BOS
+
+  Scenario: Tokenize w/ BOS
+    Given adding special tokens
+    When  tokenizing:
+    """
+    What is the capital of Germany?
+    """
+    Then  tokens begin with BOS
+    Given first token is removed
+    Then  tokens can be detokenized
 
   Scenario: Models available
     Given available models

examples/server/tests/features/steps/steps.py

@@ -376,6 +376,11 @@ def step_seed(context, seed):
         context.seed.append(seed)
 
 
+@step('BOS token is {bos:d}')
+def step_bos_token(context, bos):
+    context.bos = bos
+
+
 @step('a prefix prompt')
 def step_prompt_prefix(context):
     context.prompt_prefix = context_text(context)
@@ -656,21 +661,29 @@ async def all_embeddings_are_generated(context):
         assert_embeddings(context.tasks_result.pop().pop())
 
 
+@step('adding special tokens')
+def step_tokenize_set_add_special(context):
+    context.tokenize_add_special = True
+
+
 @step('tokenizing')
 @async_run_until_complete
 async def step_tokenize(context):
     context.tokenized_text = context_text(context)
     async with aiohttp.ClientSession() as session:
+        tokenize_args = {
+            "content": context.tokenized_text,
+        }
+        if getattr(context, 'tokenize_add_special', None) is not None:
+            tokenize_args['add_special'] = context.tokenize_add_special
         async with session.post(f'{context.base_url}/tokenize',
-                                json={
+                                json=tokenize_args) as response:
-                                    "content": context.tokenized_text,
-                                }) as response:
             assert response.status == 200
             tokenize_json = await response.json()
             context.tokens = tokenize_json['tokens']
 
 
-@step('tokens can be detokenize')
+@step('tokens can be detokenized')
 @async_run_until_complete
 async def step_detokenize(context):
     assert len(context.tokens) > 0
@@ -685,6 +698,21 @@ async def step_detokenize(context):
             assert context.tokenized_text == detokenize_json['content'].strip()
 
 
+@step('tokens begin with BOS')
+def step_strings_for_tokenization(context):
+    assert context.tokens[0] == context.bos
+
+
+@step('tokens do not begin with BOS')
+def step_strings_for_tokenization(context):
+    assert context.tokens[0] != context.bos
+
+
+@step('first token is removed')
+def step_strings_for_tokenization(context):
+    context.tokens = context.tokens[1:]
+
+
 @step('an OPTIONS request is sent from {origin}')
 @async_run_until_complete
 async def step_options_request(context, origin):
@@ -911,7 +939,7 @@ async def oai_chat_completions(user_prompt,
                     while event_received:
                         event_received = False
                         async for line_in_bytes in response.content:
-                            line = line_in_bytes.decode('utf8')
+                            line = line_in_bytes.decode('utf-8')
                             line = line.rstrip('\n').rstrip('\r')
                             if line == '':
                                 continue

examples/server/themes/README.md

@@ -0,0 +1,5 @@
+# LLaMA.cpp Server Wild Theme
+
+Simple themes directory of sample "public" directories. To try any of these add --path to your run like `server --path=wild`.
+
+![image](wild/wild.png)

examples/server/themes/buttons-top/README.md

@@ -0,0 +1,7 @@
+# LLaMA.cpp Server Buttons Top Theme
+
+Simple tweaks to the UI. Chat buttons at the top of the page instead of bottom so you can hit Stop instead of chasing it down the page.
+
+To use simply run server with `--path=themes/buttons_top`
+
+![image](buttons_top.png)

examples/server/themes/wild/README.md

@@ -0,0 +1,5 @@
+# LLaMA.cpp Server Wild Theme
+
+Simple tweaks to the UI. To use simply run server with `--path=themes/wild`
+
+![image](wild.png)

examples/server/utils.hpp

@@ -3,6 +3,8 @@
 #include "llama.h"
 #include "common.h"
 
+// Change JSON_ASSERT from assert() to GGML_ASSERT:
+#define JSON_ASSERT GGML_ASSERT
 #include "json.hpp"
 
 #include <string>
@@ -49,18 +51,18 @@ extern bool server_log_json;
 #define LOG_WARNING(MSG, ...) server_log("WARN", __func__, __LINE__, MSG, __VA_ARGS__)
 #define LOG_INFO(   MSG, ...) server_log("INFO", __func__, __LINE__, MSG, __VA_ARGS__)
 
-static inline void server_log(const char *level, const char *function, int line, const char *message, const nlohmann::ordered_json &extra);
+static inline void server_log(const char * level, const char * function, int line, const char * message, const json & extra);
 
 template <typename T>
-static T json_value(const json &body, const std::string &key, const T &default_value) {
+static T json_value(const json & body, const std::string & key, const T & default_value) {
     // Fallback null to default value
-    if (body.contains(key) && !body.at(key).is_null()){
+    if (body.contains(key) && !body.at(key).is_null()) {
         try {
-            return body.value(key, default_value);
+            return body.at(key);
-        }
+        } catch (NLOHMANN_JSON_NAMESPACE::detail::type_error const &) {
-        catch (nlohmann::json_abi_v3_11_3::detail::type_error const&){
+            std::stringstream ss;
-            std::string message = "Wrong type supplied for parameter '" + key + "'. Expected '" + typeid(default_value).name() + "', using default value.";
+            ss << "Wrong type supplied for parameter '" << key << "'. Expected '" << json(default_value).type_name() << "', using default value.";
-            server_log("WARN", __func__, __LINE__, message.c_str(), body);
+            LOG_WARNING(ss.str().c_str(), body);
             return default_value;
         }
     } else {
@@ -68,16 +70,16 @@ static T json_value(const json &body, const std::string &key, const T &default_v
     }
 }
 
-static inline void server_log(const char *level, const char *function, int line, const char *message, const nlohmann::ordered_json &extra) {
+static inline void server_log(const char * level, const char * function, int line, const char * message, const json & extra) {
     std::stringstream ss_tid;
     ss_tid << std::this_thread::get_id();
-    json log = nlohmann::ordered_json{
+    json log = json{
         {"tid",       ss_tid.str()},
         {"timestamp", time(nullptr)},
     };
 
     if (server_log_json) {
-        log.merge_patch( {
+        log.merge_patch({
             {"level",    level},
             {"function", function},
             {"line",     line},
@@ -98,7 +100,7 @@ static inline void server_log(const char *level, const char *function, int line,
         }
         std::stringstream ss;
         ss << buf << " |";
-        for (const auto& el : log.items())
+        for (const auto & el : log.items())
         {
             const std::string value = el.value().dump(-1, ' ', false, json::error_handler_t::replace);
             ss << " " << el.key() << "=" << value;
@@ -373,11 +375,11 @@ static json oaicompat_completion_params_parse(
     llama_params["top_p"]             = json_value(body,   "top_p",             1.0);
 
     // Apply chat template to the list of messages
-    llama_params["prompt"] = format_chat(model, chat_template, body["messages"]);
+    llama_params["prompt"] = format_chat(model, chat_template, body.at("messages"));
 
     // Handle "stop" field
-    if (body.contains("stop") && body["stop"].is_string()) {
+    if (body.contains("stop") && body.at("stop").is_string()) {
-        llama_params["stop"] = json::array({body["stop"].get<std::string>()});
+        llama_params["stop"] = json::array({body.at("stop").get<std::string>()});
     } else {
         llama_params["stop"] = json_value(body, "stop", json::array());
     }

ggml-cuda.cu

@@ -1647,7 +1647,7 @@ static void ggml_cuda_op_mul_mat(
     }
 }
 
-static void ggml_cuda_mul_mat_vec_p021(ggml_backend_cuda_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst){
+static void ggml_cuda_mul_mat_vec_p021(ggml_backend_cuda_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
     GGML_ASSERT(ggml_is_permuted(src0) && ggml_is_permuted(src1));
     GGML_ASSERT(ggml_backend_buffer_is_cuda(src0->buffer));
     GGML_ASSERT(src0->nb[0] <= src0->nb[1] && src0->nb[2] <= src0->nb[3]); // 0213 permutation
@@ -1670,7 +1670,7 @@ static void ggml_cuda_mul_mat_vec_p021(ggml_backend_cuda_context & ctx, const gg
     ggml_mul_mat_p021_f16_f32_cuda(src0_ddq, src1_ddf, dst_ddf, ne00, ne01, ne02, ne12, main_stream);
 }
 
-static void ggml_cuda_mul_mat_vec_nc(ggml_backend_cuda_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst){
+static void ggml_cuda_mul_mat_vec_nc(ggml_backend_cuda_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
     GGML_ASSERT(!ggml_is_transposed(src0));
     GGML_ASSERT(!ggml_is_transposed(src1));
     GGML_ASSERT(!ggml_is_permuted(src0));
@@ -2416,32 +2416,304 @@ GGML_CALL static void ggml_backend_cuda_synchronize(ggml_backend_t backend) {
     GGML_UNUSED(backend);
 }
 
+static void set_ggml_graph_node_properties(ggml_tensor * node, ggml_graph_node_properties * graph_node_properties) {
+    graph_node_properties->node_address = node->data;
+    graph_node_properties->node_op = node->op;
+    for (int i = 0; i < GGML_MAX_DIMS; i++) {
+        graph_node_properties->ne[i] = node->ne[i];
+        graph_node_properties->nb[i] = node->nb[i];
+    }
+    for (int i = 0; i < GGML_MAX_SRC; i++) {
+        graph_node_properties->src_address[i] = node->src[i] ? node->src[i]->data : nullptr;
+    }
+}
+
+static bool ggml_graph_node_has_matching_properties(ggml_tensor * node, ggml_graph_node_properties * graph_node_properties) {
+    if (node->data != graph_node_properties->node_address &&
+          node->op != GGML_OP_CPY &&
+          node->op != GGML_OP_VIEW) {
+        return false;
+    }
+
+    if (node->op != graph_node_properties->node_op) {
+        return false;
+    }
+
+    for (int i = 0; i < GGML_MAX_DIMS; i++) {
+        if (node->ne[i] != graph_node_properties->ne[i]) {
+            return false;
+        }
+        if (node->nb[i] != graph_node_properties->nb[i]) {
+            return false;
+        }
+    }
+
+    for (int i = 0; i < GGML_MAX_SRC; i++) {
+        if (node->src[i] &&
+            node->src[i]->data != graph_node_properties->src_address[i] &&
+            node->op != GGML_OP_CPY &&
+            node->op != GGML_OP_VIEW
+        ) {
+            return false;
+        }
+    }
+    return true;
+}
+
 GGML_CALL static enum ggml_status ggml_backend_cuda_graph_compute(ggml_backend_t backend, ggml_cgraph * cgraph) {
     ggml_backend_cuda_context * cuda_ctx = (ggml_backend_cuda_context *)backend->context;
 
     ggml_cuda_set_device(cuda_ctx->device);
 
-    for (int i = 0; i < cgraph->n_nodes; i++) {
+#ifdef USE_CUDA_GRAPH
-        ggml_tensor * node = cgraph->nodes[i];
+    static const bool disable_cuda_graphs_due_to_env = (getenv("GGML_CUDA_DISABLE_GRAPHS") != nullptr);
 
-        if (ggml_is_empty(node) || node->op == GGML_OP_RESHAPE || node->op == GGML_OP_TRANSPOSE || node->op == GGML_OP_VIEW || node->op == GGML_OP_PERMUTE || node->op == GGML_OP_NONE) {
+    // Objects required for CUDA Graph
-            continue;
+    if (cuda_ctx->cuda_graph == nullptr) {
+        cuda_ctx->cuda_graph.reset(new ggml_cuda_graph());
+    }
+
+    bool use_cuda_graph = true;
+    bool cuda_graph_update_required = false;
+    // pointer to CUDA cpy kernel, which is required to identify
+    // kernel parameters which need updated in the graph for each token
+    void * ggml_cuda_cpy_fn_ptr = nullptr;
+
+    if (cuda_ctx->cuda_graph->graph == nullptr) {
+        if (ggml_cuda_info().devices[cuda_ctx->device].cc < CC_AMPERE) {
+            cuda_ctx->cuda_graph->disable_due_to_gpu_arch = true;
+#ifndef NDEBUG
+            fprintf(stderr, "%s: disabling CUDA graphs due to GPU architecture\n", __func__);
+#endif
+        }
+    }
+
+    // Disable CUDA graphs in presence of env var, old GPU, use-case which is changing too rapidly,
+    // or previous graph capture failure.
+    // Also disable for multi-gpu for now. TO DO investigate
+    if (disable_cuda_graphs_due_to_env
+        || cuda_ctx->cuda_graph->disable_due_to_gpu_arch
+        || cuda_ctx->cuda_graph->disable_due_to_too_many_updates
+        || cuda_ctx->cuda_graph->disable_due_to_failed_graph_capture) {
+        use_cuda_graph = false;
+    }
+
+    if (use_cuda_graph) {
+        if (cuda_ctx->cuda_graph->instance == nullptr) {
+            cuda_graph_update_required = true;
         }
 
+        // Check if the graph size has changed
+        if (cuda_ctx->cuda_graph->ggml_graph_properties.size() != (size_t)cgraph->n_nodes) {
+            cuda_graph_update_required = true;
+            cuda_ctx->cuda_graph->ggml_graph_properties.resize(cgraph->n_nodes);
+        }
+
+        // Loop over nodes in GGML graph to determine if CUDA graph update is required
+        // and store properties to allow this comparison for the next token
+        for (int i = 0; i < cgraph->n_nodes; i++) {
+            bool has_matching_properties = true;
+            if (!cuda_graph_update_required) {
+                has_matching_properties = ggml_graph_node_has_matching_properties(cgraph->nodes[i], &cuda_ctx->cuda_graph->ggml_graph_properties[i]);
+            }
+            if (!has_matching_properties) {
+                cuda_graph_update_required = true;
+            }
+            set_ggml_graph_node_properties(cgraph->nodes[i], &cuda_ctx->cuda_graph->ggml_graph_properties[i]);
+        }
+
+        // Loop over nodes in GGML graph to obtain info needed for CUDA graph
+        cuda_ctx->cuda_graph->updated_kernel_arg.clear();
+        for (int i = 0; i < cgraph->n_nodes; i++) {
+            ggml_tensor * node = cgraph->nodes[i];
+
+            if (node->src[0] && ggml_backend_buffer_is_cuda_split(node->src[0]->buffer)) {
+                use_cuda_graph = false; // Split buffers are not supported by CUDA graph capture
 #ifndef NDEBUG
-        assert(node->buffer->buft == ggml_backend_cuda_buffer_type(cuda_ctx->device));
+                fprintf(stderr, "%s: disabling CUDA graphs due to split buffer\n", __func__);
-        for (int j = 0; j < GGML_MAX_SRC; j++) {
+#endif
-            if (node->src[j] != nullptr) {
+            }
-                assert(node->src[j]->buffer->buft == ggml_backend_cuda_buffer_type(cuda_ctx->device) || ggml_backend_buffer_is_cuda_split(node->src[j]->buffer));
+
+            if (node->op == GGML_OP_MUL_MAT_ID) {
+                use_cuda_graph = false; // This node type is not supported by CUDA graph capture
+#ifndef NDEBUG
+                fprintf(stderr, "%s: disabling CUDA graphs due to mul_mat_id\n", __func__);
+#endif
+            }
+
+            if (node->op == GGML_OP_ADD && node->src[1] && node->src[1]->ne[1] > 1) {
+                // disable CUDA graphs for batch size > 1 for now.
+                // Changes in batch size or context size can cause changes to the grid size of some kernels.
+                use_cuda_graph = false;
+#ifndef NDEBUG
+                fprintf(stderr, "%s: disabling CUDA graphs due to batch size > 1 [%s] [%ld %ld %ld %ld]\n", __func__, node->name, node->ne[0], node->ne[1], node->ne[2], node->ne[3]);
+#endif
+            }
+
+            if (node->op == GGML_OP_CPY) {
+                // store the copy op parameter which changes with each token.
+                cuda_ctx->cuda_graph->updated_kernel_arg.push_back((char **) &(node->src[1]->data));
+                if (ggml_cuda_cpy_fn_ptr == nullptr) {
+                    // store a pointer to the copy op CUDA kernel to identify it later
+                    ggml_cuda_cpy_fn_ptr = ggml_cuda_cpy_fn(node->src[0], node->src[1]);
+                }
+            }
+
+            if (!use_cuda_graph) {
+                break;
             }
         }
+
+        // Disable CUDA graphs (from the next token) if the use-case is demanding too many consecutive graph updates.
+        if (cuda_graph_update_required) {
+            cuda_ctx->cuda_graph->number_consecutive_updates++;
+        } else {
+            cuda_ctx->cuda_graph->number_consecutive_updates = 0;
+        }
+
+        if (cuda_ctx->cuda_graph->number_consecutive_updates >= 4) {
+            cuda_ctx->cuda_graph->disable_due_to_too_many_updates = true;
+#ifndef NDEBUG
+            fprintf(stderr, "%s: disabling CUDA graphs due to too many consecutive updates\n", __func__);
+#endif
+        }
+    }
+
+    if (use_cuda_graph && cuda_graph_update_required) { // Start CUDA graph capture
+        CUDA_CHECK(cudaStreamBeginCapture(cuda_ctx->stream(), cudaStreamCaptureModeRelaxed));
+    }
+
+#else
+    bool use_cuda_graph = false;
+    bool cuda_graph_update_required = false;
+#endif // USE_CUDA_GRAPH
+
+    bool graph_evaluated_or_captured = false;
+
+    while (!graph_evaluated_or_captured) {
+        // Only perform the graph execution if CUDA graphs are not enabled, or we are capturing the graph.
+        // With the use of CUDA graphs, the execution will be performed by the graph launch.
+        if (!use_cuda_graph || cuda_graph_update_required) {
+            for (int i = 0; i < cgraph->n_nodes; i++) {
+                ggml_tensor * node = cgraph->nodes[i];
+
+                if (ggml_is_empty(node) || node->op == GGML_OP_RESHAPE || node->op == GGML_OP_TRANSPOSE || node->op == GGML_OP_VIEW || node->op == GGML_OP_PERMUTE || node->op == GGML_OP_NONE) {
+                    continue;
+                }
+
+#ifndef NDEBUG
+                assert(node->buffer->buft == ggml_backend_cuda_buffer_type(cuda_ctx->device));
+                for (int j = 0; j < GGML_MAX_SRC; j++) {
+                    if (node->src[j] != nullptr) {
+                        assert(node->src[j]->buffer->buft == ggml_backend_cuda_buffer_type(cuda_ctx->device) || ggml_backend_buffer_is_cuda_split(node->src[j]->buffer));
+                    }
+                }
 #endif
 
-        bool ok = ggml_cuda_compute_forward(*cuda_ctx, node);
+                bool ok = ggml_cuda_compute_forward(*cuda_ctx, node);
-        if (!ok) {
+                if (!ok) {
-            fprintf(stderr, "%s: error: op not supported %s (%s)\n", __func__, node->name, ggml_op_name(node->op));
+                    fprintf(stderr, "%s: error: op not supported %s (%s)\n", __func__, node->name, ggml_op_name(node->op));
+                }
+                GGML_ASSERT(ok);
+            }
         }
-        GGML_ASSERT(ok);
+
+#ifdef USE_CUDA_GRAPH
+        if (use_cuda_graph && cuda_graph_update_required) { // End CUDA graph capture
+            if (cuda_ctx->cuda_graph->graph != nullptr) {
+                CUDA_CHECK(cudaGraphDestroy(cuda_ctx->cuda_graph->graph));
+                cuda_ctx->cuda_graph->graph = nullptr;
+            }
+            CUDA_CHECK(cudaStreamEndCapture(cuda_ctx->stream(), &cuda_ctx->cuda_graph->graph));
+
+#if 0
+            if (disable_cuda_graphs_due_to_failed_capture) {
+                use_cuda_graph = false;
+                cuda_ctx->cuda_graph->disable_due_to_failed_graph_capture = true;
+#ifndef NDEBUG
+                fprintf(stderr, "%s: disabling CUDA graphs due to failed graph capture\n", __func__);
+#endif
+            } else {
+                graph_evaluated_or_captured = true; // CUDA graph has been captured
+            }
+#endif
+            graph_evaluated_or_captured = true; // CUDA graph has been captured
+        } else {
+            graph_evaluated_or_captured = true; // ggml graph has been directly evaluated
+        }
+    }
+
+    if (use_cuda_graph) {
+        if (cuda_ctx->cuda_graph->instance == nullptr) { // Create executable graph from captured graph.
+            CUDA_CHECK(cudaGraphInstantiate(&cuda_ctx->cuda_graph->instance, cuda_ctx->cuda_graph->graph, NULL, NULL, 0));
+        }
+
+        // Perform update to graph (if required for this token), and change copy parameter (required for every token)
+
+        if (cuda_graph_update_required) {
+            // Extract nodes from graph
+            if (cuda_ctx->cuda_graph->num_nodes == 0) {
+                // First call with null argument gets number of nodes in graph
+                CUDA_CHECK(cudaGraphGetNodes(cuda_ctx->cuda_graph->graph, nullptr, &cuda_ctx->cuda_graph->num_nodes));
+            }
+            // Subsequent call with non-null argument gets nodes
+            cuda_ctx->cuda_graph->nodes.resize(cuda_ctx->cuda_graph->num_nodes);
+            cuda_ctx->cuda_graph->params.resize(cuda_ctx->cuda_graph->num_nodes);
+            if (cuda_ctx->cuda_graph->num_nodes > 0) {
+                CUDA_CHECK(cudaGraphGetNodes(cuda_ctx->cuda_graph->graph, cuda_ctx->cuda_graph->nodes.data(), &cuda_ctx->cuda_graph->num_nodes));
+
+                // Loop over nodes, and extract kernel parameters from each node
+                for (size_t i = 0; i < cuda_ctx->cuda_graph->num_nodes; i++) {
+                    cudaGraphNodeType node_type;
+                    CUDA_CHECK(cudaGraphNodeGetType(cuda_ctx->cuda_graph->nodes[i], &node_type));
+                    if (node_type == cudaGraphNodeTypeKernel) {
+                        cudaError_t stat = cudaGraphKernelNodeGetParams(cuda_ctx->cuda_graph->nodes[i], &cuda_ctx->cuda_graph->params[i]); // Get params using runtime
+                        if (stat == cudaErrorInvalidDeviceFunction) {
+                            // Fails due to incorrect handling by CUDA runtime of CUDA BLAS node.
+                            // We don't need to update blas nodes, so clear error and move on.
+                            cudaGetLastError();
+                        } else {
+                            GGML_ASSERT(stat == cudaSuccess);
+                        }
+                    }
+                }
+            }
+        }
+
+        // One of the arguments to the copy kernel is updated for each token, hence we need to
+        // replace that argument with the updated value in the CUDA graph
+        if (!cuda_graph_update_required) { // on update steps, the live parameters will already be captured
+            int k = 0;
+            for (size_t i = 0; i < cuda_ctx->cuda_graph->num_nodes; i++) {
+                if (cuda_ctx->cuda_graph->params[i].func == ggml_cuda_cpy_fn_ptr) {
+                    char ** updated_kernel_arg_ptr = cuda_ctx->cuda_graph->updated_kernel_arg.at(k++);
+                    cuda_ctx->cuda_graph->params[i].kernelParams[1] = updated_kernel_arg_ptr;
+                    CUDA_CHECK(cudaGraphKernelNodeSetParams(cuda_ctx->cuda_graph->nodes[i], &cuda_ctx->cuda_graph->params[i]));
+                }
+            }
+        }
+
+        // Update graph executable
+        cudaGraphExecUpdateResultInfo result_info;
+        cudaError_t stat = cudaGraphExecUpdate(cuda_ctx->cuda_graph->instance, cuda_ctx->cuda_graph->graph, &result_info);
+        if (stat == cudaErrorGraphExecUpdateFailure) {
+#ifndef NDEBUG
+            fprintf(stderr, "%s: CUDA graph update failed\n", __func__);
+#endif
+            // The pre-existing graph exec cannot be updated due to violated constraints
+            // so instead clear error and re-instantiate
+            cudaGetLastError();
+            CUDA_CHECK(cudaGraphExecDestroy(cuda_ctx->cuda_graph->instance));
+            cuda_ctx->cuda_graph->instance = nullptr;
+            CUDA_CHECK(cudaGraphInstantiate(&cuda_ctx->cuda_graph->instance, cuda_ctx->cuda_graph->graph, NULL, NULL, 0));
+        } else {
+            GGML_ASSERT(stat == cudaSuccess);
+        }
+        // Launch graph
+        CUDA_CHECK(cudaGraphLaunch(cuda_ctx->cuda_graph->instance, cuda_ctx->stream()));
+#else
+        graph_evaluated_or_captured = true;
+#endif // USE_CUDA_GRAPH
     }
 
     return GGML_STATUS_SUCCESS;

ggml-cuda/clamp.cu

@@ -31,5 +31,4 @@ void ggml_cuda_op_clamp(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
     memcpy(&max, (float *) dst->op_params + 1, sizeof(float));
 
     clamp_f32_cuda(src0_d, dst_d, min, max, ggml_nelements(src0), stream);
-    CUDA_CHECK(cudaGetLastError());
 }

ggml-cuda/common.cuh

@@ -19,6 +19,7 @@
 #include <cassert>
 #include <cfloat>
 #include <string>
+#include <vector>
 
 #if defined(GGML_USE_HIPBLAS)
 #include <hip/hip_runtime.h>
@@ -526,6 +527,43 @@ struct ggml_tensor_extra_gpu {
     cudaEvent_t events[GGML_CUDA_MAX_DEVICES][GGML_CUDA_MAX_STREAMS]; // events for synchronizing multiple GPUs
 };
 
+
+#if (CUDART_VERSION >= 12000) && defined(GGML_CUDA_USE_GRAPHS)
+#define USE_CUDA_GRAPH
+#endif
+
+struct ggml_graph_node_properties {
+    void * node_address;
+    ggml_op node_op;
+    int64_t ne[GGML_MAX_DIMS];
+    size_t nb[GGML_MAX_DIMS];
+    void * src_address[GGML_MAX_SRC];
+};
+
+struct ggml_cuda_graph {
+#ifdef USE_CUDA_GRAPH
+    ~ggml_cuda_graph() {
+        if (instance != nullptr) {
+            CUDA_CHECK(cudaGraphExecDestroy(instance));
+        }
+        if (graph != nullptr) {
+            CUDA_CHECK(cudaGraphDestroy(graph));
+        }
+    }
+    cudaGraph_t graph = nullptr;
+    cudaGraphExec_t instance = nullptr;
+    size_t num_nodes = 0;
+    std::vector<cudaGraphNode_t> nodes;
+    std::vector<cudaKernelNodeParams> params;
+    bool disable_due_to_gpu_arch = false;
+    bool disable_due_to_too_many_updates = false;
+    bool disable_due_to_failed_graph_capture = false;
+    int number_consecutive_updates = 0;
+    std::vector<ggml_graph_node_properties> ggml_graph_properties;
+    std::vector<char **> updated_kernel_arg;
+#endif
+};
+
 struct ggml_backend_cuda_context {
     int device;
     std::string name;
@@ -534,6 +572,8 @@ struct ggml_backend_cuda_context {
     cudaStream_t streams[GGML_CUDA_MAX_DEVICES][GGML_CUDA_MAX_STREAMS] = { { nullptr } };
     cublasHandle_t cublas_handles[GGML_CUDA_MAX_DEVICES] = {nullptr};
 
+    std::unique_ptr<ggml_cuda_graph> cuda_graph;
+
     explicit ggml_backend_cuda_context(int device) :
         device(device),
         name(GGML_CUDA_NAME + std::to_string(device)) {

ggml-cuda/convert.cu

@@ -727,7 +727,6 @@ static void convert_unary_cuda(const void * __restrict__ vx, dst_t * __restrict_
 }
 
 to_fp16_cuda_t ggml_get_to_fp16_cuda(ggml_type type) {
-    int id;
     switch (type) {
         case GGML_TYPE_Q4_0:
             return dequantize_row_q4_0_cuda;
@@ -738,8 +737,7 @@ to_fp16_cuda_t ggml_get_to_fp16_cuda(ggml_type type) {
         case GGML_TYPE_Q5_1:
             return dequantize_block_cuda<QK5_1, QR5_1, dequantize_q5_1>;
         case GGML_TYPE_Q8_0:
-            CUDA_CHECK(cudaGetDevice(&id));
+            if (ggml_cuda_info().devices[ggml_cuda_get_device()].cc >= CC_PASCAL) {
-            if (ggml_cuda_info().devices[id].cc >= CC_PASCAL) {
                 return dequantize_block_q8_0_f16_cuda;
             }
             return dequantize_block_cuda<QK8_0, QR8_0, dequantize_q8_0>;

ggml-cuda/cpy.cu

@@ -459,3 +459,32 @@ void ggml_cuda_dup(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
     const ggml_tensor * src0 = dst->src[0];
     ggml_cuda_cpy(ctx, src0, dst);
 }
+
+void* ggml_cuda_cpy_fn(const ggml_tensor * src0, ggml_tensor * src1) {
+    if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_F32) {
+            return (void*) cpy_f32_f16<cpy_1_f32_f32>;
+    } else if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_F16) {
+            return (void*) cpy_f32_f16<cpy_1_f32_f16>;
+    } else if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_Q8_0) {
+            return (void*) cpy_f32_q<cpy_blck_f32_q8_0, QK8_0>;
+    } else if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_Q4_0) {
+            return (void*) cpy_f32_q<cpy_blck_f32_q4_0, QK4_0>;
+    } else if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_Q4_1) {
+            return (void*) cpy_f32_q<cpy_blck_f32_q4_1, QK4_1>;
+    } else if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_Q5_0) {
+            return (void*) cpy_f32_q<cpy_blck_f32_q5_0, QK5_0>;
+    } else if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_IQ4_NL) {
+            return (void*) cpy_f32_q<cpy_blck_f32_iq4_nl, QK4_NL>;
+    } else if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_Q5_1) {
+            return (void*) cpy_f32_q<cpy_blck_f32_q5_1, QK5_1>;
+    } else if (src0->type == GGML_TYPE_F16 && src1->type == GGML_TYPE_F16) {
+            return (void*) cpy_f32_f16<cpy_1_f32_f16>;
+    } else if (src0->type == GGML_TYPE_F16 && src1->type == GGML_TYPE_F32) {
+            return (void*) cpy_f32_f16<cpy_1_f16_f32>;
+    } else {
+        fprintf(stderr, "%s: unsupported type combination (%s to %s)\n", __func__,
+                ggml_type_name(src0->type), ggml_type_name(src1->type));
+        GGML_ASSERT(false);
+    }
+}
+

ggml-cuda/cpy.cuh

@@ -5,3 +5,5 @@
 void ggml_cuda_cpy(ggml_backend_cuda_context & ctx, const ggml_tensor * src0, ggml_tensor * src1);
 
 void ggml_cuda_dup(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
+
+void* ggml_cuda_cpy_fn(const ggml_tensor * src0, ggml_tensor * src1);

ggml-cuda/mmq.cu

@@ -1735,8 +1735,7 @@ static void ggml_mul_mat_q4_0_q8_1_cuda(
     const void * vx, const void * vy, float * dst, const int ncols_x, const int nrows_x,
     const int ncols_y, const int nrows_y, const int nrows_dst, cudaStream_t stream) {
 
-    int id;
+    int id = ggml_cuda_get_device();
-    CUDA_CHECK(cudaGetDevice(&id));
     const int compute_capability = ggml_cuda_info().devices[id].cc;
 
     int mmq_x, mmq_y, nwarps;
@@ -1780,8 +1779,7 @@ static void ggml_mul_mat_q4_1_q8_1_cuda(
     const void * vx, const void * vy, float * dst, const int ncols_x, const int nrows_x,
     const int ncols_y, const int nrows_y, const int nrows_dst, cudaStream_t stream) {
 
-    int id;
+    int id = ggml_cuda_get_device();
-    CUDA_CHECK(cudaGetDevice(&id));
     const int compute_capability = ggml_cuda_info().devices[id].cc;
 
     int mmq_x, mmq_y, nwarps;
@@ -1825,8 +1823,7 @@ static void ggml_mul_mat_q5_0_q8_1_cuda(
     const void * vx, const void * vy, float * dst, const int ncols_x, const int nrows_x,
     const int ncols_y, const int nrows_y, const int nrows_dst, cudaStream_t stream) {
 
-    int id;
+    int id = ggml_cuda_get_device();
-    CUDA_CHECK(cudaGetDevice(&id));
     const int compute_capability = ggml_cuda_info().devices[id].cc;
 
     int mmq_x, mmq_y, nwarps;
@@ -1870,8 +1867,7 @@ static void ggml_mul_mat_q5_1_q8_1_cuda(
     const void * vx, const void * vy, float * dst, const int ncols_x, const int nrows_x,
     const int ncols_y, const int nrows_y, const int nrows_dst, cudaStream_t stream) {
 
-    int id;
+    int id = ggml_cuda_get_device();
-    CUDA_CHECK(cudaGetDevice(&id));
     const int compute_capability = ggml_cuda_info().devices[id].cc;
 
     int mmq_x, mmq_y, nwarps;
@@ -1915,8 +1911,7 @@ static void ggml_mul_mat_q8_0_q8_1_cuda(
     const void * vx, const void * vy, float * dst, const int ncols_x, const int nrows_x,
     const int ncols_y, const int nrows_y, const int nrows_dst, cudaStream_t stream) {
 
-    int id;
+    int id = ggml_cuda_get_device();
-    CUDA_CHECK(cudaGetDevice(&id));
     const int compute_capability = ggml_cuda_info().devices[id].cc;
 
     int mmq_x, mmq_y, nwarps;
@@ -1960,8 +1955,7 @@ static void ggml_mul_mat_q2_K_q8_1_cuda(
     const void * vx, const void * vy, float * dst, const int ncols_x, const int nrows_x,
     const int ncols_y, const int nrows_y, const int nrows_dst, cudaStream_t stream) {
 
-    int id;
+    int id = ggml_cuda_get_device();
-    CUDA_CHECK(cudaGetDevice(&id));
     const int compute_capability = ggml_cuda_info().devices[id].cc;
 
     int mmq_x, mmq_y, nwarps;
@@ -2007,8 +2001,7 @@ static void ggml_mul_mat_q3_K_q8_1_cuda(
 
 #if QK_K == 256
 
-    int id;
+    int id = ggml_cuda_get_device();
-    CUDA_CHECK(cudaGetDevice(&id));
     const int compute_capability = ggml_cuda_info().devices[id].cc;
 
     int mmq_x, mmq_y, nwarps;
@@ -2053,8 +2046,7 @@ static void ggml_mul_mat_q4_K_q8_1_cuda(
     const void * vx, const void * vy, float * dst, const int ncols_x, const int nrows_x,
     const int ncols_y, const int nrows_y, const int nrows_dst, cudaStream_t stream) {
 
-    int id;
+    int id = ggml_cuda_get_device();
-    CUDA_CHECK(cudaGetDevice(&id));
     const int compute_capability = ggml_cuda_info().devices[id].cc;
 
     int mmq_x, mmq_y, nwarps;
@@ -2098,8 +2090,7 @@ static void ggml_mul_mat_q5_K_q8_1_cuda(
     const void * vx, const void * vy, float * dst, const int ncols_x, const int nrows_x,
     const int ncols_y, const int nrows_y, const int nrows_dst, cudaStream_t stream) {
 
-    int id;
+    int id = ggml_cuda_get_device();
-    CUDA_CHECK(cudaGetDevice(&id));
     const int compute_capability = ggml_cuda_info().devices[id].cc;
 
     int mmq_x, mmq_y, nwarps;
@@ -2143,8 +2134,7 @@ static void ggml_mul_mat_q6_K_q8_1_cuda(
     const void * vx, const void * vy, float * dst, const int ncols_x, const int nrows_x,
     const int ncols_y, const int nrows_y, const int nrows_dst, cudaStream_t stream) {
 
-    int id;
+    int id = ggml_cuda_get_device();
-    CUDA_CHECK(cudaGetDevice(&id));
     const int compute_capability = ggml_cuda_info().devices[id].cc;
 
     int mmq_x, mmq_y, nwarps;

ggml-cuda/mmvq.cu

@@ -89,8 +89,7 @@ static void mul_mat_vec_q_cuda(
     GGML_ASSERT(ncols_x % qk == 0);
     GGML_ASSERT(ncols_y <= MMVQ_MAX_BATCH_SIZE);
 
-    int id;
+    int id = ggml_cuda_get_device();
-    CUDA_CHECK(cudaGetDevice(&id));
 
     int64_t nwarps = 1;
     int64_t rows_per_cuda_block = 1;
@@ -328,8 +327,7 @@ void ggml_cuda_op_mul_mat_vec_q(
 
     const int64_t ne0 = dst->ne[0];
 
-    int id;
+    int id = ggml_cuda_get_device();
-    CUDA_CHECK(cudaGetDevice(&id));
 
     // the main device has a larger memory buffer to hold the results from all GPUs
     // nrows_dst == nrows of the matrix that the kernel writes into

ggml-cuda/scale.cu

@@ -28,5 +28,4 @@ void ggml_cuda_op_scale(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
     memcpy(&scale, dst->op_params, sizeof(float));
 
     scale_f32_cuda(src0_d, dst_d, scale, ggml_nelements(src0), stream);
-    CUDA_CHECK(cudaGetLastError());
 }

ggml-metal.m

@@ -265,11 +265,20 @@ static void ggml_metal_log(enum ggml_log_level level, const char * format, ...){
 
 static void * ggml_metal_host_malloc(size_t n) {
     void * data = NULL;
+
+#if TARGET_OS_OSX
+    kern_return_t err = vm_allocate((vm_map_t) mach_task_self(), (void *) &data, n, VM_FLAGS_ANYWHERE);
+    if (err != KERN_SUCCESS) {
+        GGML_METAL_LOG_ERROR("%s: error: vm_allocate failed\n", __func__);
+        return NULL;
+    }
+#else
     const int result = posix_memalign((void **) &data, sysconf(_SC_PAGESIZE), n);
     if (result != 0) {
         GGML_METAL_LOG_ERROR("%s: error: posix_memalign failed\n", __func__);
         return NULL;
     }
+#endif
 
     return data;
 }
@@ -2840,7 +2849,11 @@ GGML_CALL static void ggml_backend_metal_buffer_free_buffer(ggml_backend_buffer_
     ggml_backend_metal_free_device();
 
     if (ctx->owned) {
+#if TARGET_OS_OSX
+        vm_deallocate((vm_map_t)mach_task_self(), (vm_address_t)ctx->all_data, ctx->all_size);
+#else
         free(ctx->all_data);
+#endif
     }
 
     free(ctx);
@@ -2944,14 +2957,16 @@ GGML_CALL static ggml_backend_buffer_t ggml_backend_metal_buffer_type_alloc_buff
     ctx->owned = true;
     ctx->n_buffers = 1;
 
-    ctx->buffers[0].data = ctx->all_data;
+    if (ctx->all_data != NULL) {
-    ctx->buffers[0].size = size;
+        ctx->buffers[0].data = ctx->all_data;
-    ctx->buffers[0].metal = [device newBufferWithBytesNoCopy:ctx->all_data
+        ctx->buffers[0].size = size;
-                    length:size_aligned
+        ctx->buffers[0].metal = [device newBufferWithBytesNoCopy:ctx->all_data
-                    options:MTLResourceStorageModeShared
+                        length:size_aligned
-                    deallocator:nil];
+                        options:MTLResourceStorageModeShared
+                        deallocator:nil];
+    }
 
-    if (ctx->buffers[0].metal == nil) {
+    if (ctx->all_data == NULL || ctx->buffers[0].metal == nil) {
         GGML_METAL_LOG_ERROR("%s: error: failed to allocate buffer, size = %8.2f MiB\n", __func__, size_aligned / 1024.0 / 1024.0);
         free(ctx);
         ggml_backend_metal_free_device();

gguf-py/gguf/constants.py

@@ -860,7 +860,7 @@ class GGUFValueType(IntEnum):
 # Note: Does not support GGML_QKK_64
 QK_K = 256
 # Items here are (block size, type size)
-GGML_QUANT_SIZES = {
+GGML_QUANT_SIZES: dict[GGMLQuantizationType, tuple[int, int]] = {
     GGMLQuantizationType.F32:     (1, 4),
     GGMLQuantizationType.F16:     (1, 2),
     GGMLQuantizationType.Q4_0:    (32, 2 + 16),

gguf-py/gguf/gguf_reader.py

@@ -65,7 +65,7 @@ class ReaderTensor(NamedTuple):
 
 class GGUFReader:
     # I - same as host, S - swapped
-    byte_order: Literal['I' | 'S'] = 'I'
+    byte_order: Literal['I'] | Literal['S'] = 'I'
     alignment: int = GGUF_DEFAULT_ALIGNMENT
 
     # Note: Internal helper, API may change.
@@ -83,7 +83,7 @@ class GGUFReader:
         GGUFValueType.BOOL:    np.bool_,
     }
 
-    def __init__(self, path: os.PathLike[str] | str, mode: Literal['r' | 'r+' | 'c'] = 'r'):
+    def __init__(self, path: os.PathLike[str] | str, mode: Literal['r'] | Literal['r+'] | Literal['c'] = 'r'):
         self.data = np.memmap(path, mode = mode)
         offs = 0
         if self._get(offs, np.uint32, override_order = '<')[0] != GGUF_MAGIC:
@@ -128,7 +128,7 @@ class GGUFReader:
         return self.tensors[idx]
 
     def _get(
-        self, offset: int, dtype: npt.DTypeLike, count: int = 1, override_order: None | Literal['I' | 'S' | '<'] = None,
+        self, offset: int, dtype: npt.DTypeLike, count: int = 1, override_order: None | Literal['I'] | Literal['S'] | Literal['<'] = None,
     ) -> npt.NDArray[Any]:
         count = int(count)
         itemsize = int(np.empty([], dtype = dtype).itemsize)
@@ -250,7 +250,7 @@ class GGUFReader:
                 raise ValueError(f'Found duplicated tensor with name {tensor_name}')
             tensor_names.add(tensor_name)
             ggml_type = GGMLQuantizationType(raw_dtype[0])
-            n_elems = np.prod(dims)
+            n_elems = int(np.prod(dims))
             block_size, type_size = GGML_QUANT_SIZES[ggml_type]
             n_bytes = n_elems * type_size // block_size
             data_offs = int(start_offs + offset_tensor[0])

gguf-py/gguf/gguf_writer.py

@@ -7,7 +7,7 @@ import struct
 import tempfile
 from enum import Enum, auto
 from io import BufferedWriter
-from typing import IO, Any, Sequence, Mapping
+from typing import IO, Any, Callable, Sequence, Mapping
 from string import ascii_letters, digits
 
 import numpy as np
@@ -28,6 +28,47 @@ from .constants import (
 logger = logging.getLogger(__name__)
 
 
+class LazyTensor:
+    data: Callable[[], np.ndarray[Any, Any]]
+    # to avoid too deep recursion
+    functions: list[Callable[[np.ndarray[Any, Any]], np.ndarray[Any, Any]]]
+    dtype: np.dtype[Any]
+    shape: tuple[int, ...]
+
+    def __init__(self, data: Callable[[], np.ndarray[Any, Any]], *, dtype: type, shape: tuple[int, ...]):
+        self.data = data
+        self.functions = []
+        self.dtype = np.dtype(dtype)
+        self.shape = shape
+
+    def astype(self, dtype: type, **kwargs) -> LazyTensor:
+        self.functions.append(lambda n: n.astype(dtype, **kwargs))
+        self.dtype = np.dtype(dtype)
+        return self
+
+    @property
+    def nbytes(self) -> int:
+        size = 1
+        for n in self.shape:
+            size *= n
+        return size * self.dtype.itemsize
+
+    def tofile(self, *args, **kwargs) -> None:
+        data = self.data()
+        for f in self.functions:
+            data = f(data)
+        assert data.shape == self.shape
+        assert data.dtype == self.dtype
+        assert data.nbytes == self.nbytes
+        self.functions = []
+        self.data = lambda: data
+        data.tofile(*args, **kwargs)
+
+    def byteswap(self, *args, **kwargs) -> LazyTensor:
+        self.functions.append(lambda n: n.byteswap(*args, **kwargs))
+        return self
+
+
 class WriterState(Enum):
     EMPTY   = auto()
     HEADER  = auto()
@@ -38,7 +79,7 @@ class WriterState(Enum):
 class GGUFWriter:
     fout: BufferedWriter
     temp_file: tempfile.SpooledTemporaryFile[bytes] | None
-    tensors: list[np.ndarray[Any, Any]]
+    tensors: list[np.ndarray[Any, Any] | LazyTensor]
     _simple_value_packing = {
         GGUFValueType.UINT8:   "B",
         GGUFValueType.INT8:    "b",
@@ -176,7 +217,7 @@ class GGUFWriter:
         if pack_fmt is not None:
             self.kv_data += self._pack(pack_fmt, val, skip_pack_prefix = vtype == GGUFValueType.BOOL)
         elif vtype == GGUFValueType.STRING:
-            encoded_val = val.encode("utf8") if isinstance(val, str) else val
+            encoded_val = val.encode("utf-8") if isinstance(val, str) else val
             self.kv_data += self._pack("Q", len(encoded_val))
             self.kv_data += encoded_val
         elif vtype == GGUFValueType.ARRAY and isinstance(val, Sequence) and val:
@@ -205,7 +246,7 @@ class GGUFWriter:
             raise ValueError(f'Duplicated tensor name {name}')
         self.ti_names.add(name)
 
-        encoded_name = name.encode("utf8")
+        encoded_name = name.encode("utf-8")
         self.ti_data += self._pack("Q", len(encoded_name))
         self.ti_data += encoded_name
         n_dims = len(tensor_shape)
@@ -237,7 +278,7 @@ class GGUFWriter:
         self.ti_data_count += 1
 
     def add_tensor(
-        self, name: str, tensor: np.ndarray[Any, Any], raw_shape: Sequence[int] | None = None,
+        self, name: str, tensor: np.ndarray[Any, Any] | LazyTensor, raw_shape: Sequence[int] | None = None,
         raw_dtype: GGMLQuantizationType | None = None,
     ) -> None:
         if self.endianess == GGUFEndian.BIG:
@@ -262,7 +303,7 @@ class GGUFWriter:
         if pad != 0:
             fp.write(bytes([0] * pad))
 
-    def write_tensor_data(self, tensor: np.ndarray[Any, Any]) -> None:
+    def write_tensor_data(self, tensor: np.ndarray[Any, Any] | LazyTensor) -> None:
         if self.state is not WriterState.TI_DATA:
             raise ValueError(f'Expected output file to contain tensor info, got {self.state}')
 
@@ -272,15 +313,33 @@ class GGUFWriter:
         tensor.tofile(self.fout)
         self.write_padding(self.fout, tensor.nbytes)
 
-    def write_tensors_to_file(self) -> None:
+    def write_tensors_to_file(self, *, progress: bool = False) -> None:
         self.write_ti_data_to_file()
 
         self.write_padding(self.fout, self.fout.tell())
 
         if self.temp_file is None:
+            self.tensors.reverse()  # to pop from the "beginning" in constant time
+
+            if progress:
+                from tqdm import tqdm
+
+                total_bytes = sum(t.nbytes for t in self.tensors)
+
+                bar = tqdm(desc="Writing", total=total_bytes, unit="byte", unit_scale=True)
+
+                while True:
+                    try:
+                        tensor = self.tensors.pop()
+                    except IndexError:
+                        break
+                    tensor.tofile(self.fout)
+                    bar.update(tensor.nbytes)
+                    self.write_padding(self.fout, tensor.nbytes)
+                return
             while True:
                 try:
-                    tensor = self.tensors.pop(0)
+                    tensor = self.tensors.pop()
                 except IndexError:
                     break
                 tensor.tofile(self.fout)
@@ -479,7 +538,7 @@ class GGUFWriter:
         self.add_bool(Keys.Tokenizer.ADD_PREFIX, value)
 
     def add_chat_template(self, value: str | Sequence[Mapping[str, str]]) -> None:
-        if isinstance(value, list):
+        if not isinstance(value, str):
             template_default = None
             template_names = set()
 

gguf-py/gguf/vocab.py

@@ -4,7 +4,7 @@ import logging
 import json
 import os
 from pathlib import Path
-from typing import Any, Callable
+from typing import Any, Callable, Sequence, Mapping, Iterable
 
 from .gguf_writer import GGUFWriter
 
@@ -15,11 +15,11 @@ class SpecialVocab:
     merges: list[str]
     add_special_token: dict[str, bool]
     special_token_ids: dict[str, int]
-    chat_template: str | None
+    chat_template: str | Sequence[Mapping[str, str]] | None
 
     def __init__(
         self, path: str | os.PathLike[str], load_merges: bool = False,
-        special_token_types: tuple[str, ...] | None = None,
+        special_token_types: Iterable[str] | None = None,
         n_vocab: int | None = None,
     ):
         self.special_token_ids = {}

gguf-py/pyproject.toml

@@ -21,6 +21,7 @@ classifiers = [
 [tool.poetry.dependencies]
 python = ">=3.8"
 numpy = ">=1.17"
+tqdm = ">=4.27"
 
 [tool.poetry.dev-dependencies]
 pytest = "^5.2"

gguf-py/scripts/gguf-dump.py

@@ -47,7 +47,7 @@ def dump_metadata(reader: GGUFReader, args: argparse.Namespace) -> None:
         if len(field.types) == 1:
             curr_type = field.types[0]
             if curr_type == GGUFValueType.STRING:
-                log_message += ' = {0}'.format(repr(str(bytes(field.parts[-1]), encoding='utf8')[:60]))
+                log_message += ' = {0}'.format(repr(str(bytes(field.parts[-1]), encoding='utf-8')[:60]))
             elif field.types[0] in reader.gguf_scalar_to_np:
                 log_message += ' = {0}'.format(field.parts[-1][0])
         print(log_message)  # noqa: NP100

gguf-py/scripts/gguf-new-metadata.py

@@ -7,7 +7,7 @@ import json
 from pathlib import Path
 
 import numpy as np
-from typing import Any, Mapping, Sequence
+from typing import Any, Sequence
 
 # Necessary to load the local gguf package
 if "NO_LOCAL_GGUF" not in os.environ and (Path(__file__).parent.parent.parent / 'gguf-py').exists():
@@ -34,7 +34,7 @@ def get_byteorder(reader: gguf.GGUFReader) -> gguf.GGUFEndian:
         return host_endian
 
 
-def decode_field(field: gguf.ReaderField) -> Any:
+def decode_field(field: gguf.ReaderField | None) -> Any:
     if field and field.types:
         main_type = field.types[0]
 
@@ -42,11 +42,11 @@ def decode_field(field: gguf.ReaderField) -> Any:
             sub_type = field.types[-1]
 
             if sub_type == gguf.GGUFValueType.STRING:
-                return [str(bytes(field.parts[idx]), encoding='utf8') for idx in field.data]
+                return [str(bytes(field.parts[idx]), encoding='utf-8') for idx in field.data]
             else:
                 return [pv for idx in field.data for pv in field.parts[idx].tolist()]
         if main_type == gguf.GGUFValueType.STRING:
-            return str(bytes(field.parts[-1]), encoding='utf8')
+            return str(bytes(field.parts[-1]), encoding='utf-8')
         else:
             return field.parts[-1][0]
 
@@ -59,7 +59,7 @@ def get_field_data(reader: gguf.GGUFReader, key: str) -> Any:
     return decode_field(field)
 
 
-def copy_with_new_metadata(reader: gguf.GGUFReader, writer: gguf.GGUFWriter, new_metadata: Mapping[str, str], remove_metadata: Sequence[str]) -> None:
+def copy_with_new_metadata(reader: gguf.GGUFReader, writer: gguf.GGUFWriter, new_metadata: dict[str, str], remove_metadata: Sequence[str]) -> None:
     for field in reader.fields.values():
         # Suppress virtual fields and fields written by GGUFWriter
         if field.name == gguf.Keys.General.ARCHITECTURE or field.name.startswith('GGUF.'):
@@ -101,7 +101,7 @@ def copy_with_new_metadata(reader: gguf.GGUFReader, writer: gguf.GGUFWriter, new
 
     for tensor in reader.tensors:
         # Dimensions are written in reverse order, so flip them first
-        shape = np.flipud(tensor.shape)
+        shape = np.flipud(tensor.shape).tolist()
         writer.add_tensor_info(tensor.name, shape, tensor.data.dtype, tensor.data.nbytes, tensor.tensor_type)
 
     writer.write_header_to_file()

llama.cpp

@@ -4445,9 +4445,15 @@ static void llm_load_vocab(
             } else if (
                 tokenizer_pre == "command-r") {
                 vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_COMMAND_R;
+            } else if (
+                tokenizer_pre == "qwen2") {
+                vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_QWEN2;
             } else if (
                 tokenizer_pre == "olmo") {
                 vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_OLMO;
+            } else if (
+                tokenizer_pre == "dbrx") {
+                vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DBRX;
             } else {
                 throw std::runtime_error(format("unknown pre-tokenizer type: '%s'", tokenizer_pre.c_str()));
             }
@@ -12491,6 +12497,7 @@ struct llm_tokenizer_bpe {
             case LLAMA_VOCAB_TYPE_BPE:
                 switch (vocab.type_pre) {
                     case LLAMA_VOCAB_PRE_TYPE_LLAMA3:
+                    case LLAMA_VOCAB_PRE_TYPE_DBRX:
                         word_collection = unicode_regex_split(text, {
                             // original regex from tokenizer.json
                             //"(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?\\p{L}+|\\p{N}{1,3}| ?[^\\s\\p{L}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
@@ -12550,6 +12557,13 @@ struct llm_tokenizer_bpe {
                             "'s|'t|'re|'ve|'m|'ll|'d| ?\\p{L}+| ?\\p{N}+| ?[^\\s\\p{L}\\p{N}]+|\\s+(?!\\S)",
                         });
                         break;
+                    case LLAMA_VOCAB_PRE_TYPE_QWEN2:
+                        word_collection = unicode_regex_split(text, {
+                            // original regex from tokenizer.json
+                            // "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?\\p{L}+|\\p{N}| ?[^\\s\\p{L}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+"
+                            "(?:'[sS]|'[tT]|'[rR][eE]|'[vV][eE]|'[mM]|'[lL][lL]|'[dD])|[^\\r\\n\\p{L}\\p{N}]?\\p{L}+|\\p{N}| ?[^\\s\\p{L}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+                        });
+                        break;
                     default:
                         // default regex for BPE tokenization pre-processing
                         word_collection = unicode_regex_split(text, {

llama.h

@@ -81,7 +81,9 @@ extern "C" {
         LLAMA_VOCAB_PRE_TYPE_GPT2           = 7,
         LLAMA_VOCAB_PRE_TYPE_REFACT         = 8,
         LLAMA_VOCAB_PRE_TYPE_COMMAND_R      = 9,
-        LLAMA_VOCAB_PRE_TYPE_OLMO           = 10,
+        LLAMA_VOCAB_PRE_TYPE_QWEN2          = 10,
+        LLAMA_VOCAB_PRE_TYPE_OLMO           = 11,
+        LLAMA_VOCAB_PRE_TYPE_DBRX           = 12,
     };
 
     // note: these values should be synchronized with ggml_rope

models/ggml-vocab-qwen2.gguf.inp

@@ -0,0 +1,106 @@
+ied 4 ½ months
+__ggml_vocab_test__
+Führer
+__ggml_vocab_test__
+
+__ggml_vocab_test__
+ 
+__ggml_vocab_test__
+  
+__ggml_vocab_test__
+   
+__ggml_vocab_test__
+	
+__ggml_vocab_test__
+
+
+__ggml_vocab_test__
+
+
+
+__ggml_vocab_test__
+
+
+
+
+__ggml_vocab_test__
+	
+
+__ggml_vocab_test__
+Hello world
+__ggml_vocab_test__
+ Hello world
+__ggml_vocab_test__
+Hello World
+__ggml_vocab_test__
+ Hello World
+__ggml_vocab_test__
+ Hello World!
+__ggml_vocab_test__
+Hello, world!
+__ggml_vocab_test__
+ Hello, world!
+__ggml_vocab_test__
+ this is 🦙.cpp
+__ggml_vocab_test__
+w048 7tuijk dsdfhu
+__ggml_vocab_test__
+нещо на Български
+__ggml_vocab_test__
+កាន់តែពិសេសអាចខលចេញ
+__ggml_vocab_test__
+🚀 (normal) 😶‍🌫️ (multiple emojis concatenated) ✅ (only emoji that has its own token)
+__ggml_vocab_test__
+Hello
+__ggml_vocab_test__
+ Hello
+__ggml_vocab_test__
+  Hello
+__ggml_vocab_test__
+   Hello
+__ggml_vocab_test__
+    Hello
+__ggml_vocab_test__
+    Hello
+    Hello
+__ggml_vocab_test__
+ (
+__ggml_vocab_test__
+
+ =
+__ggml_vocab_test__
+' era
+__ggml_vocab_test__
+Hello, y'all! How are you 😁 ?我想在apple工作1314151天～
+__ggml_vocab_test__
+3
+__ggml_vocab_test__
+33
+__ggml_vocab_test__
+333
+__ggml_vocab_test__
+3333
+__ggml_vocab_test__
+33333
+__ggml_vocab_test__
+333333
+__ggml_vocab_test__
+3333333
+__ggml_vocab_test__
+33333333
+__ggml_vocab_test__
+333333333
+__ggml_vocab_test__
+
+ 
+
+ 
+
+
+ 	 		 	
+  
+   
+    
+     
+🚀 (normal) 😶‍🌫️ (multiple emojis concatenated) ✅ 🦙🦙 3 33 333 3333 33333 333333 3333333 33333333 3.3 3..3 3...3 កាន់តែពិសេសអាច😁 ?我想在apple工作1314151天～ ------======= нещо на Български ''''''```````""""......!!!!!!?????? I've been 'told he's there, 'RE you sure? 'M not sure I'll make it, 'D you like some tea? We'Ve a'lL
+__ggml_vocab_test__

models/ggml-vocab-qwen2.gguf.out

@@ -0,0 +1,43 @@
+ 1122 220 19 220 26062 3951
+ 37 50753 261
+
+ 220
+ 256
+ 262
+ 197
+ 198
+ 271
+ 1406
+ 1572
+ 9707 1879
+ 21927 1879
+ 9707 4337
+ 21927 4337
+ 21927 4337 0
+ 9707 11 1879 0
+ 21927 11 1879 0
+ 419 374 11162 99 247 13 10821
+ 86 15 19 23 220 22 83 1963 41808 11472 2940 16739
+ 78762 14144 1456 13073 63471 33594 3038 133178 79012
+ 146394 97529 241 44258 233 146568 44258 224 147603 20879 115 146280 44258 223 146280 147272 97529 227 147805 148301 147270 44258 223 146848
+ 145836 320 8252 8 26525 114 378 235 149921 30543 320 35673 99066 97534 8 25521 227 320 3243 42365 429 702 1181 1828 3950 8
+ 9707
+ 21927
+ 220 21927
+ 256 21927
+ 262 21927
+ 262 21927 198 262 21927
+ 320
+ 198 284
+ 6 11385
+ 9707 11 379 64848 0 2585 525 498 26525 223 937 104100 18493 22377 99257 16 18 16 19 16 20 16 35727 21216
+ 18
+ 18 18
+ 18 18 18
+ 18 18 18 18
+ 18 18 18 18 18
+ 18 18 18 18 18 18
+ 18 18 18 18 18 18 18
+ 18 18 18 18 18 18 18 18
+ 18 18 18 18 18 18 18 18 18
+ 198 4710 14731 65497 7847 1572 2303 78672 10947 145836 320 8252 8 26525 114 378 235 149921 30543 320 35673 99066 97534 8 25521 227 11162 99 247 149955 220 18 220 18 18 220 18 18 18 220 18 18 18 18 220 18 18 18 18 18 220 18 18 18 18 18 18 220 18 18 18 18 18 18 18 220 18 18 18 18 18 18 18 18 220 18 13 18 220 18 496 18 220 18 1112 18 220 146394 97529 241 44258 233 146568 44258 224 147603 20879 115 146280 44258 223 146280 147272 97529 227 144534 937 104100 18493 22377 99257 16 18 16 19 16 20 16 35727 21216 55460 53237 18658 14144 1456 13073 63471 33594 3038 133178 79012 3355 4605 4605 13874 13874 73594 3014 3014 28149 17085 2928 26610 7646 358 3003 1012 364 83 813 566 594 1052 11 364 787 498 2704 30 364 44 537 2704 358 3278 1281 432 11 364 35 498 1075 1045 15243 30 1205 6 42612 264 63866 43

pyrightconfig.json

@@ -0,0 +1,3 @@
+{
+  "extraPaths": ["gguf-py"],
+}

sgemm.cpp

@@ -1,6 +1,3 @@
-// -*- mode:c++;indent-tabs-mode:nil;c-basic-offset:4;coding:utf-8 -*-
-// vi: set et ft=c++ ts=4 sts=4 sw=4 fenc=utf-8 :vi
-//
 // Copyright 2024 Mozilla Foundation
 //
 // Permission is hereby granted, free of charge, to any person obtaining
@@ -585,15 +582,15 @@ class tinyBLAS_Q0_ARM {
 };
 #endif // __ARM_FEATURE_DOTPROD
 
-#if defined(__AVX2__) || defined(__AVX512F__)
+#if defined(__AVX2__) || defined(__AVX512F__) || defined(__AVX__)
 template <typename TA, typename TB, typename TC>
-class tinyBLAS_Q0_AVX2 {
+class tinyBLAS_Q0_AVX {
   public:
-    tinyBLAS_Q0_AVX2(int64_t k,
+    tinyBLAS_Q0_AVX(int64_t k,
-                     const TA *A, int64_t lda,
+                    const TA *A, int64_t lda,
-                     const TB *B, int64_t ldb,
+                    const TB *B, int64_t ldb,
-                     TC *C, int64_t ldc,
+                    TC *C, int64_t ldc,
-                     int ith, int nth)
+                    int ith, int nth)
         : A(A), B(B), C(C), k(k), lda(lda), ldb(ldb), ldc(ldc), ith(ith), nth(nth) {
     }
 
@@ -728,14 +725,34 @@ class tinyBLAS_Q0_AVX2 {
             __m256 Cv[RN][RM] = {};
             for (int64_t l = 0; l < k; ++l)
                 for (int64_t j = 0; j < RN; ++j)
-                    for (int64_t i = 0; i < RM; ++i)
+                    for (int64_t i = 0; i < RM; ++i) {
+#if defined(__AVX2__)
+                        __m256 udTmp = updot(_mm256_sign_epi8(load(A + lda * (ii + i) + l),
+                                                              load(A + lda * (ii + i) + l)),
+                                             _mm256_sign_epi8(load(B + ldb * (jj + j) + l),
+                                                              load(A + lda * (ii + i) + l)));
+#else
+                        __m128i ali0 = load0(A + lda * (ii + i) + l);
+                        __m128i ali1 = load1(A + lda * (ii + i) + l);
+                        __m128i blj0 = load0(B + ldb * (jj + j) + l);
+                        __m128i blj1 = load1(B + ldb * (jj + j) + l);
+
+                        __m128i sepAA0 = _mm_sign_epi8(ali0, ali0);
+                        __m128i sepAA1 = _mm_sign_epi8(ali1, ali1);
+                        __m128i sepBA0 = _mm_sign_epi8(blj0, ali0);
+                        __m128i sepBA1 = _mm_sign_epi8(blj1, ali1);
+
+                        // updot
+                        const __m128i oneFill = _mm_set1_epi16(1);
+                        __m128i mad0 = _mm_maddubs_epi16(sepAA0, sepBA0);
+                        __m128i mad1 = _mm_maddubs_epi16(sepAA1, sepBA1);
+                        __m256 udTmp = _mm256_cvtepi32_ps(MM256_SET_M128I(_mm_madd_epi16(oneFill, mad1), _mm_madd_epi16(oneFill, mad0)));
+#endif
                         Cv[j][i] = madd(_mm256_set1_ps(unhalf(A[lda * (ii + i) + l].d) *
                                                        unhalf(B[ldb * (jj + j) + l].d)),
-                                        updot(_mm256_sign_epi8(load(A + lda * (ii + i) + l),
+                                                       udTmp,
-                                                               load(A + lda * (ii + i) + l)),
+                                                       Cv[j][i]);
-                                              _mm256_sign_epi8(load(B + ldb * (jj + j) + l),
+                    }
-                                                               load(A + lda * (ii + i) + l))),
-                                        Cv[j][i]);
             for (int64_t j = 0; j < RN; ++j)
                 for (int64_t i = 0; i < RM; ++i)
                     C[ldc * (jj + j) + (ii + i)] = hsum(Cv[j][i]);
@@ -746,10 +763,28 @@ class tinyBLAS_Q0_AVX2 {
         return _mm256_loadu_si256((const __m256i *)b->qs);
     }
 
+    inline __m128i load0(const block_q8_0 *b) {
+        return _mm_loadu_si128((const __m128i *)b->qs);
+    }
+
+    inline __m128i load1(const block_q8_0 *b) {
+        return _mm_loadu_si128(((const __m128i *)b->qs) + 1);
+    }
+
     inline __m256i load(const block_q4_0 *b) {
         return _mm256_sub_epi8(denibble(b->qs), _mm256_set1_epi8(8));
     }
 
+    inline __m128i load0(const block_q4_0 *b) {
+        const __m128i x = _mm_loadu_si128((const __m128i *)(b->qs));
+        return _mm_sub_epi8(_mm_and_si128(_mm_set1_epi8(15), x), _mm_set1_epi8(8));
+    }
+
+    inline __m128i load1(const block_q4_0 *b) {
+        const __m128i x = _mm_loadu_si128((const __m128i *)(b->qs));
+        return _mm_sub_epi8(_mm_and_si128(_mm_set1_epi8(15), _mm_srli_epi16(x, 4)), _mm_set1_epi8(8));
+    }
+
     inline __m256 updot(__m256i u, __m256i s) {
         __m256i res;
 #if defined(__AVXVNNI__) || (defined(__AVX512VNNI__) && defined(__AVX512VL__))
@@ -777,7 +812,7 @@ class tinyBLAS_Q0_AVX2 {
     const int ith;
     const int nth;
 };
-#endif // __AVX2__
+#endif // __AVX__
 
 } // namespace
 
@@ -928,8 +963,8 @@ bool llamafile_sgemm(int64_t m, int64_t n, int64_t k, const void *A, int64_t lda
     case GGML_TYPE_Q8_0: {
         if (Btype != GGML_TYPE_Q8_0)
            return false;
-#if defined(__AVX2__) || defined(__AVX512F__)
+#if defined(__AVX2__) || defined(__AVX512F__) || defined(__AVX__)
-        tinyBLAS_Q0_AVX2<block_q8_0, block_q8_0, float> tb{
+        tinyBLAS_Q0_AVX<block_q8_0, block_q8_0, float> tb{
             k, (const block_q8_0 *)A, lda,
             (const block_q8_0 *)B, ldb,
             (float *)C, ldc,
@@ -952,8 +987,8 @@ bool llamafile_sgemm(int64_t m, int64_t n, int64_t k, const void *A, int64_t lda
     case GGML_TYPE_Q4_0: {
         if (Btype != GGML_TYPE_Q8_0)
             return false;
-#if defined(__AVX2__) || defined(__AVX512F__)
+#if defined(__AVX2__) || defined(__AVX512F__) || defined(__AVX__)
-        tinyBLAS_Q0_AVX2<block_q4_0, block_q8_0, float> tb{
+        tinyBLAS_Q0_AVX<block_q4_0, block_q8_0, float> tb{
             k, (const block_q4_0 *)A, lda,
             (const block_q8_0 *)B, ldb,
             (float *)C, ldc,