Merge remote-tracking branch 'upstream/master'

.github/ISSUE_TEMPLATE/bug.md

@@ -1,7 +1,7 @@
 ---
 name: Bug template
 about: Used to report bugs in llama.cpp
-labels: ["bug"]
+labels: ["bug-unconfirmed"]
 assignees: ''
 
 ---

CMakeLists.txt

@@ -94,7 +94,6 @@ option(LLAMA_CLBLAST                         "llama: use CLBlast"
 option(LLAMA_METAL                           "llama: use Metal"                                 ${LLAMA_METAL_DEFAULT})
 option(LLAMA_METAL_NDEBUG                    "llama: disable Metal debugging"                   OFF)
 option(LLAMA_MPI                             "llama: use MPI"                                   OFF)
-option(LLAMA_K_QUANTS                        "llama: use k-quants"                              ON)
 option(LLAMA_QKK_64                          "llama: use super-block size of 64 for k-quants"   OFF)
 
 option(LLAMA_BUILD_TESTS                "llama: build tests"    ${LLAMA_STANDALONE})
@@ -278,13 +277,8 @@ if (LLAMA_BLAS)
     endif()
 endif()
 
-if (LLAMA_K_QUANTS)
-    set(GGML_HEADERS_EXTRA k_quants.h)
-    set(GGML_SOURCES_EXTRA k_quants.c)
-    add_compile_definitions(GGML_USE_K_QUANTS)
-    if (LLAMA_QKK_64)
-        add_compile_definitions(GGML_QKK_64)
-    endif()
+if (LLAMA_QKK_64)
+    add_compile_definitions(GGML_QKK_64)
 endif()
 
 if (LLAMA_CUBLAS)
@@ -673,6 +667,8 @@ add_library(ggml OBJECT
             ggml-alloc.h
             ggml-backend.c
             ggml-backend.h
+            ggml-quants.c
+            ggml-quants.h
             ${GGML_SOURCES_CUDA} ${GGML_HEADERS_CUDA}
             ${GGML_SOURCES_OPENCL} ${GGML_HEADERS_OPENCL}
             ${GGML_SOURCES_METAL} ${GGML_HEADERS_METAL}

Makefile

@@ -342,13 +342,9 @@ else
 	MK_CXXFLAGS += -march=rv64gcv -mabi=lp64d
 endif
 
-ifndef LLAMA_NO_K_QUANTS
-	MK_CPPFLAGS += -DGGML_USE_K_QUANTS
-	OBJS     += k_quants.o
 ifdef LLAMA_QKK_64
 	MK_CPPFLAGS += -DGGML_QKK_64
 endif
-endif
 
 ifndef LLAMA_NO_ACCELERATE
 	# Mac OS - include Accelerate framework.
@@ -365,7 +361,7 @@ ifdef LLAMA_MPI
 	MK_CPPFLAGS += -DGGML_USE_MPI
 	MK_CFLAGS   += -Wno-cast-qual
 	MK_CXXFLAGS += -Wno-cast-qual
-	OBJS     += ggml-mpi.o
+	OBJS        += ggml-mpi.o
 endif # LLAMA_MPI
 
 ifdef LLAMA_OPENBLAS
@@ -382,7 +378,7 @@ endif # LLAMA_BLIS
 ifdef LLAMA_CUBLAS
 	MK_CPPFLAGS  += -DGGML_USE_CUBLAS -I/usr/local/cuda/include -I/opt/cuda/include -I$(CUDA_PATH)/targets/x86_64-linux/include
 	MK_LDFLAGS   += -lcublas -lculibos -lcudart -lcublasLt -lpthread -ldl -lrt -L/usr/local/cuda/lib64 -L/opt/cuda/lib64 -L$(CUDA_PATH)/targets/x86_64-linux/lib
-	OBJS      += ggml-cuda.o
+	OBJS         += ggml-cuda.o
 	NVCCFLAGS = --forward-unknown-to-host-compiler -use_fast_math
 ifdef LLAMA_CUDA_NVCC
 	NVCC = $(LLAMA_CUDA_NVCC)
@@ -497,11 +493,6 @@ ggml-mpi.o: ggml-mpi.c ggml-mpi.h
 	$(CC) $(CFLAGS) -c $< -o $@
 endif # LLAMA_MPI
 
-ifndef LLAMA_NO_K_QUANTS
-k_quants.o: k_quants.c k_quants.h
-	$(CC) $(CFLAGS) -c $< -o $@
-endif # LLAMA_NO_K_QUANTS
-
 # combine build flags with cmdline overrides
 override CFLAGS        := $(MK_CPPFLAGS) $(CPPFLAGS) $(MK_CFLAGS) $(CFLAGS)
 override CXXFLAGS      := $(MK_CPPFLAGS) $(CPPFLAGS) $(MK_CXXFLAGS) $(CXXFLAGS)
@@ -542,15 +533,18 @@ ggml-alloc.o: ggml-alloc.c ggml.h ggml-alloc.h
 ggml-backend.o: ggml-backend.c ggml.h ggml-backend.h
 	$(CC)  $(CFLAGS)   -c $< -o $@
 
-OBJS += ggml-alloc.o ggml-backend.o
+ggml-quants.o: ggml-quants.c ggml.h ggml-quants.h
+	$(CC) $(CFLAGS)    -c $< -o $@
+
+OBJS += ggml-alloc.o ggml-backend.o ggml-quants.o
 
 llama.o: llama.cpp ggml.h ggml-alloc.h ggml-backend.h ggml-cuda.h ggml-metal.h llama.h
 	$(CXX) $(CXXFLAGS) -c $< -o $@
 
-COMMON_H_DEPS = common/common.h common/sampling.h build-info.h common/log.h
-COMMON_DEPS   = $(COMMON_H_DEPS) common.o sampling.o grammar-parser.o
+COMMON_H_DEPS = common/common.h common/sampling.h common/log.h
+COMMON_DEPS   = common.o sampling.o grammar-parser.o
 
-common.o: common/common.cpp $(COMMON_H_DEPS)
+common.o: common/common.cpp build-info.h $(COMMON_H_DEPS)
 	$(CXX) $(CXXFLAGS) -c $< -o $@
 
 sampling.o: common/sampling.cpp $(COMMON_H_DEPS)

Package.swift

@@ -42,13 +42,12 @@ let package = Package(
                 "llama.cpp",
                 "ggml-alloc.c",
                 "ggml-backend.c",
-                "k_quants.c",
+                "ggml-quants.c",
             ] + additionalSources,
             resources: resources,
             publicHeadersPath: "spm-headers",
             cSettings: [
                 .unsafeFlags(["-Wno-shorten-64-to-32", "-O3", "-DNDEBUG"]),
-                .define("GGML_USE_K_QUANTS"),
                 .define("GGML_USE_ACCELERATE")
                 // NOTE: NEW_LAPACK will required iOS version 16.4+
                 // We should consider add this in the future when we drop support for iOS 14

build.zig

@@ -116,15 +116,10 @@ pub fn build(b: *std.build.Builder) !void {
     var make = try Maker.init(b);
     make.enable_lto = b.option(bool, "lto", "Enable LTO optimization, (default: false)") orelse false;
 
-    if (b.option(bool, "k-quants", "Enable K-quants, (default: true)") orelse true) {
-        try make.addFlag("-DGGML_USE_K_QUANTS");
-        const k_quants = make.obj("k_quants", "k_quants.c");
-        try make.objs.append(k_quants);
-    }
-
     const ggml = make.obj("ggml", "ggml.c");
     const ggml_alloc = make.obj("ggml-alloc", "ggml-alloc.c");
     const ggml_backend = make.obj("ggml-backend", "ggml-backend.c");
+    const ggml_quants = make.obj("ggml-quants", "ggml-quants.c");
     const llama = make.obj("llama", "llama.cpp");
     const common = make.obj("common", "common/common.cpp");
     const console = make.obj("console", "common/console.cpp");
@@ -133,14 +128,14 @@ pub fn build(b: *std.build.Builder) !void {
     const train = make.obj("train", "common/train.cpp");
     const clip = make.obj("clip", "examples/llava/clip.cpp");
 
-    _ = make.exe("main", "examples/main/main.cpp", &.{ ggml, ggml_alloc, ggml_backend, llama, common, sampling, console, grammar_parser });
-    _ = make.exe("quantize", "examples/quantize/quantize.cpp", &.{ ggml, ggml_alloc, ggml_backend, llama, common });
-    _ = make.exe("perplexity", "examples/perplexity/perplexity.cpp", &.{ ggml, ggml_alloc, ggml_backend, llama, common });
-    _ = make.exe("embedding", "examples/embedding/embedding.cpp", &.{ ggml, ggml_alloc, ggml_backend, llama, common });
-    _ = make.exe("finetune", "examples/finetune/finetune.cpp", &.{ ggml, ggml_alloc, ggml_backend, llama, common, train });
-    _ = make.exe("train-text-from-scratch", "examples/train-text-from-scratch/train-text-from-scratch.cpp", &.{ ggml, ggml_alloc, ggml_backend, llama, common, train });
+    _ = make.exe("main", "examples/main/main.cpp", &.{ ggml, ggml_alloc, ggml_backend, ggml_quants, llama, common, sampling, console, grammar_parser });
+    _ = make.exe("quantize", "examples/quantize/quantize.cpp", &.{ ggml, ggml_alloc, ggml_backend, ggml_quants, llama, common });
+    _ = make.exe("perplexity", "examples/perplexity/perplexity.cpp", &.{ ggml, ggml_alloc, ggml_backend, ggml_quants, llama, common });
+    _ = make.exe("embedding", "examples/embedding/embedding.cpp", &.{ ggml, ggml_alloc, ggml_backend, ggml_quants, llama, common });
+    _ = make.exe("finetune", "examples/finetune/finetune.cpp", &.{ ggml, ggml_alloc, ggml_backend, ggml_quants, llama, common, train });
+    _ = make.exe("train-text-from-scratch", "examples/train-text-from-scratch/train-text-from-scratch.cpp", &.{ ggml, ggml_alloc, ggml_backend, ggml_quants, llama, common, train });
 
-    const server = make.exe("server", "examples/server/server.cpp", &.{ ggml, ggml_alloc, ggml_backend, llama, common, sampling, grammar_parser, clip });
+    const server = make.exe("server", "examples/server/server.cpp", &.{ ggml, ggml_alloc, ggml_backend, ggml_quants, llama, common, sampling, grammar_parser, clip });
     if (server.target.isWindows()) {
         server.linkSystemLibrary("ws2_32");
     }

common/common.cpp

@@ -224,6 +224,7 @@ bool gpt_params_parse(int argc, char ** argv, gpt_params & params) {
                 break;
             }
             sparams.temp = std::stof(argv[i]);
+            sparams.temp = std::max(sparams.temp, 0.0f);
         } else if (arg == "--tfs") {
             if (++i >= argc) {
                 invalid_param = true;
@@ -743,7 +744,7 @@ void gpt_print_usage(int /*argc*/, char ** argv, const gpt_params & params) {
 #endif // GGML_USE_CUBLAS
 #endif
     printf("  --verbose-prompt      print prompt before generation\n");
-    fprintf(stderr, "  --simple-io           use basic IO for better compatibility in subprocesses and limited consoles\n");
+    printf("  --simple-io           use basic IO for better compatibility in subprocesses and limited consoles\n");
     printf("  --lora FNAME          apply LoRA adapter (implies --no-mmap)\n");
     printf("  --lora-scaled FNAME S apply LoRA adapter with user defined scaling S (implies --no-mmap)\n");
     printf("  --lora-base FNAME     optional model to use as a base for the layers modified by the LoRA adapter\n");
@@ -888,7 +889,7 @@ std::tuple<struct llama_model *, struct llama_context *> llama_init_from_gpt_par
 
         std::vector<llama_token> tmp = { llama_token_bos(model), llama_token_eos(model), };
         llama_decode(lctx, llama_batch_get_one(tmp.data(), std::min(tmp.size(), (size_t) params.n_batch), 0, 0));
-        llama_kv_cache_tokens_rm(lctx, -1, -1);
+        llama_kv_cache_clear(lctx);
         llama_reset_timings(lctx);
     }
 

common/sampling.cpp

@@ -167,8 +167,12 @@ llama_token llama_sampling_sample(
         llama_sample_grammar(ctx_main, &cur_p, ctx_sampling->grammar);
     }
 
-    if (temp <= 0) {
-        // greedy sampling
+    if (temp < 0.0) {
+        // greedy sampling, with probs
+        llama_sample_softmax(ctx_main, &cur_p);
+        id = cur_p.data[0].id;
+    } else if (temp == 0.0) {
+        // greedy sampling, no probs
         id = llama_sample_token_greedy(ctx_main, &cur_p);
     } else {
         if (mirostat == 1) {

convert.py

@@ -366,16 +366,19 @@ def __init__(self, fname_tokenizer: Path, fname_added_tokens: Path | None) -> No
             added_tokens = {}
 
         vocab_size: int = self.sentencepiece_tokenizer.vocab_size()
-        expected_ids = list(range(vocab_size, vocab_size + len(added_tokens)))
-        actual_ids   = sorted(added_tokens.values())
-        if expected_ids != actual_ids:
-            raise Exception(f"Expected added token IDs to be sequential and start at {vocab_size}; got {actual_ids}")
 
-        items = sorted(added_tokens.items(), key=lambda text_idx: text_idx[1])
-        self.added_tokens_list = [text for (text, idx) in items]
-        self.vocab_size_base: int = vocab_size
-        self.vocab_size: int = self.vocab_size_base + len(self.added_tokens_list)
-        self.fname_tokenizer = fname_tokenizer
+        new_tokens       = {id: piece for piece, id in added_tokens.items() if id >= vocab_size}
+        expected_new_ids = list(range(vocab_size, vocab_size + len(new_tokens)))
+        actual_new_ids   = sorted(new_tokens.keys())
+
+        if expected_new_ids != actual_new_ids:
+            raise ValueError(f"Expected new token IDs {expected_new_ids} to be sequential; got {actual_new_ids}")
+
+        # Token pieces that were added to the base vocabulary.
+        self.added_tokens_list  = [new_tokens[id] for id in actual_new_ids]
+        self.vocab_size_base    = vocab_size
+        self.vocab_size         = self.vocab_size_base + len(self.added_tokens_list)
+        self.fname_tokenizer    = fname_tokenizer
         self.fname_added_tokens = fname_added_tokens
 
     def sentencepiece_tokens(self) -> Iterable[tuple[bytes, float, gguf.TokenType]]:

examples/batched-bench/batched-bench.cpp

@@ -185,7 +185,7 @@ int main(int argc, char ** argv) {
 
                 const auto t_pp_start = ggml_time_us();
 
-                llama_kv_cache_tokens_rm(ctx, -1, -1);
+                llama_kv_cache_clear(ctx);
 
                 if (!decode_helper(ctx, batch, ctx_params.n_batch)) {
                     LOG_TEE("%s: llama_decode() failed\n", __func__);

examples/llama-bench/llama-bench.cpp

@@ -1037,7 +1037,7 @@ int main(int argc, char ** argv) {
 
         test t(inst, lmodel, ctx);
 
-        llama_kv_cache_tokens_rm(ctx, -1, -1);
+        llama_kv_cache_clear(ctx);
 
         // warmup run
         if (t.n_prompt > 0) {
@@ -1048,7 +1048,7 @@ int main(int argc, char ** argv) {
         }
 
         for (int i = 0; i < params.reps; i++) {
-            llama_kv_cache_tokens_rm(ctx, -1, -1);
+            llama_kv_cache_clear(ctx);
 
             uint64_t t_start = get_time_ns();
             if (t.n_prompt > 0) {

examples/main/main.cpp

@@ -298,7 +298,7 @@ int main(int argc, char ** argv) {
         }
 
         // remove any "future" tokens that we might have inherited from the previous session
-        llama_kv_cache_tokens_rm(ctx, n_matching_session_tokens, -1);
+        llama_kv_cache_seq_rm(ctx, -1, n_matching_session_tokens, -1);
     }
 
     LOGLN(

examples/perplexity/perplexity.cpp

@@ -210,7 +210,7 @@ static results_perplexity perplexity_v2(llama_context * ctx, const gpt_params &
         const auto t_start = std::chrono::high_resolution_clock::now();
 
         // clear the KV cache
-        llama_kv_cache_tokens_rm(ctx, -1, -1);
+        llama_kv_cache_clear(ctx);
 
         for (int j = 0; j < num_batches; ++j) {
             const int batch_start = start + j * n_batch;
@@ -339,7 +339,7 @@ static results_perplexity perplexity(llama_context * ctx, const gpt_params & par
         const auto t_start = std::chrono::high_resolution_clock::now();
 
         // clear the KV cache
-        llama_kv_cache_tokens_rm(ctx, -1, -1);
+        llama_kv_cache_clear(ctx);
 
         for (int j = 0; j < num_batches; ++j) {
             const int batch_start = start + j * n_batch;
@@ -573,7 +573,7 @@ static void hellaswag_score(llama_context * ctx, const gpt_params & params) {
         }
 
         // clear the KV cache
-        llama_kv_cache_tokens_rm(ctx, -1, -1);
+        llama_kv_cache_clear(ctx);
 
         auto logits = hellaswag_evaluate_tokens(ctx, query_embd, 0, params.n_batch, n_vocab);
         if (logits.empty()) {

examples/quantize/quantize.cpp

@@ -18,7 +18,6 @@ static const std::vector<struct quant_option> QUANT_OPTIONS = {
     { "Q4_1",   LLAMA_FTYPE_MOSTLY_Q4_1,   " 3.90G, +0.1585 ppl @ LLaMA-v1-7B", },
     { "Q5_0",   LLAMA_FTYPE_MOSTLY_Q5_0,   " 4.33G, +0.0683 ppl @ LLaMA-v1-7B", },
     { "Q5_1",   LLAMA_FTYPE_MOSTLY_Q5_1,   " 4.70G, +0.0349 ppl @ LLaMA-v1-7B", },
-#ifdef GGML_USE_K_QUANTS
     { "Q2_K",   LLAMA_FTYPE_MOSTLY_Q2_K,   " 2.63G, +0.6717 ppl @ LLaMA-v1-7B", },
     { "Q3_K",   LLAMA_FTYPE_MOSTLY_Q3_K_M, "alias for Q3_K_M" },
     { "Q3_K_S", LLAMA_FTYPE_MOSTLY_Q3_K_S, " 2.75G, +0.5551 ppl @ LLaMA-v1-7B", },
@@ -31,7 +30,6 @@ static const std::vector<struct quant_option> QUANT_OPTIONS = {
     { "Q5_K_S", LLAMA_FTYPE_MOSTLY_Q5_K_S, " 4.33G, +0.0400 ppl @ LLaMA-v1-7B", },
     { "Q5_K_M", LLAMA_FTYPE_MOSTLY_Q5_K_M, " 4.45G, +0.0122 ppl @ LLaMA-v1-7B", },
     { "Q6_K",   LLAMA_FTYPE_MOSTLY_Q6_K,   " 5.15G, -0.0008 ppl @ LLaMA-v1-7B", },
-#endif
     { "Q8_0",   LLAMA_FTYPE_MOSTLY_Q8_0,   " 6.70G, +0.0004 ppl @ LLaMA-v1-7B", },
     { "F16",    LLAMA_FTYPE_MOSTLY_F16,    "13.00G              @ 7B", },
     { "F32",    LLAMA_FTYPE_ALL_F32,       "26.00G              @ 7B", },
@@ -70,13 +68,14 @@ static bool try_parse_ftype(const std::string & ftype_str_in, llama_ftype & ftyp
 }
 
 // usage:
-//  ./quantize [--allow-requantize] [--leave-output-tensor] models/llama/ggml-model.gguf [models/llama/ggml-model-quant.gguf] type [nthreads]
+//  ./quantize [--allow-requantize] [--leave-output-tensor] [--pure] models/llama/ggml-model.gguf [models/llama/ggml-model-quant.gguf] type [nthreads]
 //
 [[noreturn]]
 static void usage(const char * executable) {
-    printf("usage: %s [--help] [--allow-requantize] [--leave-output-tensor] model-f32.gguf [model-quant.gguf] type [nthreads]\n\n", executable);
+    printf("usage: %s [--help] [--allow-requantize] [--leave-output-tensor] [--pure] model-f32.gguf [model-quant.gguf] type [nthreads]\n\n", executable);
     printf("  --allow-requantize: Allows requantizing tensors that have already been quantized. Warning: This can severely reduce quality compared to quantizing from 16bit or 32bit\n");
     printf("  --leave-output-tensor: Will leave output.weight un(re)quantized. Increases model size but may also increase quality, especially when requantizing\n");
+    printf("  --pure: Disable k-quant mixtures and quantize all tensors to the same type\n");
     printf("\nAllowed quantization types:\n");
     for (auto & it : QUANT_OPTIONS) {
         if (it.name != "COPY") {
@@ -103,6 +102,8 @@ int main(int argc, char ** argv) {
             params.quantize_output_tensor = false;
         } else if (strcmp(argv[arg_idx], "--allow-requantize") == 0) {
             params.allow_requantize = true;
+        } else if (strcmp(argv[arg_idx], "--pure") == 0) {
+            params.pure = true;
         } else {
             usage(argv[0]);
         }

examples/server/server.cpp

@@ -857,7 +857,7 @@ struct llama_server_context
 
     void kv_cache_clear() {
         // clear the entire KV cache
-        llama_kv_cache_tokens_rm(ctx, -1, -1);
+        llama_kv_cache_clear(ctx);
         clean_kv_cache = false;
     }
 

examples/speculative/speculative.cpp

@@ -148,7 +148,7 @@ int main(int argc, char ** argv) {
     std::vector<seq_draft> drafts(n_seq_dft);
 
     params.sparams.grammar.clear(); // the draft samplers will copy the target sampler's grammar
-    params.sparams.temp = std::max(0.01f, params.sparams.temp);
+    params.sparams.temp = -1.0f;    // force greedy sampling with probs for the draft model
 
     for (int s = 0; s < n_seq_dft; ++s) {
         drafts[s].ctx_sampling = llama_sampling_init(params.sparams);

flake.nix

@@ -51,6 +51,9 @@
         };
         llama-python =
           pkgs.python3.withPackages (ps: with ps; [ numpy sentencepiece ]);
+        # TODO(Green-Sky): find a better way to opt-into the heavy ml python runtime
+        llama-python-extra =
+          pkgs.python3.withPackages (ps: with ps; [ numpy sentencepiece torchWithoutCuda transformers ]);
         postPatch = ''
           substituteInPlace ./ggml-metal.m \
             --replace '[bundle pathForResource:@"ggml-metal" ofType:@"metal"];' "@\"$out/bin/ggml-metal.metal\";"
@@ -126,5 +129,9 @@
           buildInputs = [ llama-python ];
           packages = nativeBuildInputs ++ osSpecific;
         };
+        devShells.extra = pkgs.mkShell {
+          buildInputs = [ llama-python-extra ];
+          packages = nativeBuildInputs ++ osSpecific;
+        };
       });
 }