diff --git a/.gitattributes b/.gitattributes new file mode 100644 index 0000000..6313b56 --- /dev/null +++ b/.gitattributes @@ -0,0 +1 @@ +* text=auto eol=lf diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml new file mode 100644 index 0000000..15ede16 --- /dev/null +++ b/.github/workflows/ci.yml @@ -0,0 +1,167 @@ +name: ci + +permissions: + contents: read + pull-requests: read + actions: read + +env: + VERSION: "${{ github.ref_name }}" + +on: + workflow_dispatch: { } + push: + tags: + - "v*.*.*" + branches: + - main + paths-ignore: + - "docs/**" + - "**.md" + - "**.mdx" + - "**.png" + - "**.jpg" + pull_request: + branches: + - main + paths-ignore: + - "docs/**" + - "**.md" + - "**.mdx" + - "**.png" + - "**.jpg" + +concurrency: + group: ${{ github.workflow }}-${{ github.head_ref && github.ref || github.run_id }} + cancel-in-progress: true + +jobs: + darwin-metal: + if: ${{ false }} + strategy: + fail-fast: false + matrix: + # see https://github.com/actions/runner-images?tab=readme-ov-file#available-images. + os: [ macos-13, macos-14 ] + runs-on: ${{ matrix.os }} + steps: + - name: Setup XCode + if: ${{ matrix.os == 'macos-13' }} + uses: maxim-lobanov/setup-xcode@v1 + with: + xcode-version: '15.2' + - name: Checkout + uses: actions/checkout@v4 + with: + fetch-depth: 0 + submodules: 'recursive' + - name: Deps + continue-on-error: true + run: | + brew update + - name: Build + run: | + make -j LLAMA_METAL=1 + + echo "===== info =====" + file ./.dist/llama-box + - name: Release + if: ${{ startsWith(github.ref, 'refs/tags/') }} + uses: actions/upload-artifact@v4 + with: + path: ./.dist/llama-box + name: llama-box-darwin-${{ endsWith(matrix.os, '-13') && 'amd64' || 'arm64' }}-metal + + linux-hip: + strategy: + fail-fast: false + matrix: + arch: [ amd64 ] + version: [ '6.0.2' ] + runs-on: ubuntu-22.04 + steps: + - name: Maximize Space + run: | + sudo rm -rf /usr/share/dotnet + sudo rm -rf /usr/local/lib/android + sudo rm -rf /opt/ghc + sudo rm -rf /opt/hostedtoolcache/CodeQL + sudo docker image prune --all --force + - name: Clone + uses: actions/checkout@v4 + with: + fetch-depth: 0 + submodules: 'recursive' + - name: Setup QEMU + uses: docker/setup-qemu-action@v3 + with: + image: tonistiigi/binfmt:qemu-v7.0.0 + platforms: "arm64" + - name: Build + run: | + docker run \ + --rm \ + --privileged \ + --platform linux/${{ matrix.arch }} \ + --volume $(pwd):/workspace \ + --workdir /workspace \ + --env CC=/opt/rocm/llvm/bin/clang \ + --env CXX=/opt/rocm/llvm/bin/clang++ \ + --env GPU_TARGETS="gfx803 gfx900 gfx906 gfx908 gfx90a gfx1010 gfx1030 gfx1100 gfx1101 gfx1102" \ + rocm/dev-ubuntu-22.04:${{ matrix.version }}-complete \ + make -j LLAMA_HIPBLAS=1 + + echo "===== info =====" + file ./.dist/llama-box + - name: Release + if: ${{ startsWith(github.ref, 'refs/tags/') }} + uses: actions/upload-artifact@v4 + with: + path: ./.dist/llama-box + name: llama-box-linux-${{ matrix.arch }}-hip + + linux-cuda: + strategy: + fail-fast: false + matrix: + arch: [ amd64, arm64 ] + version: [ '11.7.1' ] + runs-on: ubuntu-22.04 + steps: + - name: Maximize Space + run: | + sudo rm -rf /usr/share/dotnet + sudo rm -rf /usr/local/lib/android + sudo rm -rf /opt/ghc + sudo rm -rf /opt/hostedtoolcache/CodeQL + sudo docker image prune --all --force + - name: Clone + uses: actions/checkout@v4 + with: + fetch-depth: 0 + submodules: 'recursive' + - name: Setup QEMU + uses: docker/setup-qemu-action@v3 + with: + image: tonistiigi/binfmt:qemu-v7.0.0 + platforms: "arm64" + - name: Build + run: | + docker run \ + --rm \ + --privileged \ + --platform linux/${{ matrix.arch }} \ + --volume $(pwd):/workspace \ + --workdir /workspace \ + --env CUDA_DOCKER_ARCH=all \ + nvidia/cuda:${{ matrix.version }}-devel-ubuntu22.04 \ + make -j LLAMA_CUDA=1 + + echo "===== info =====" + file ./.dist/llama-box + - name: Release + if: ${{ startsWith(github.ref, 'refs/tags/') }} + uses: actions/upload-artifact@v4 + with: + path: ./.dist/llama-box + name: llama-box-linux-${{ matrix.arch }}-cuda \ No newline at end of file diff --git a/.gitignore b/.gitignore new file mode 100644 index 0000000..76bb781 --- /dev/null +++ b/.gitignore @@ -0,0 +1,34 @@ +# Files +.DS_Store +*.o +*.a +*.so +*.gguf +*.bin +*.exe +*.exe~ +*.dll +*.dylib +*.log +*.dot +*.bat +*.tmp +*.metallib +*.out +*.swp +*.swo +.clang-tidy +version.cpp + +# Directories +.idea/ +.vscode/ +.vs/ +.build/ +.cache/ +.ccls-cache/ +.direnv/ +.sbin/ +.dist/ +out/ +tmp/ diff --git a/.gitmodules b/.gitmodules new file mode 100644 index 0000000..81bc6f3 --- /dev/null +++ b/.gitmodules @@ -0,0 +1,3 @@ +[submodule "llama.cpp"] + path = llama.cpp + url = https://github.com/ggerganov/llama.cpp diff --git a/LICENSE b/LICENSE new file mode 100644 index 0000000..c059907 --- /dev/null +++ b/LICENSE @@ -0,0 +1,21 @@ +MIT License + +Copyright (c) 2024 The llama-box authors + +Permission is hereby granted, free of charge, to any person obtaining a copy +of this software and associated documentation files (the "Software"), to deal +in the Software without restriction, including without limitation the rights +to use, copy, modify, merge, publish, distribute, sublicense, and/or sell +copies of the Software, and to permit persons to whom the Software is +furnished to do so, subject to the following conditions: + +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. diff --git a/Makefile b/Makefile new file mode 100644 index 0000000..9cec98f --- /dev/null +++ b/Makefile @@ -0,0 +1,275 @@ +# Inspired by https://github.com/ggerganov/llama.cpp/blob/61665277afde2add00c0d387acb94ed5feb95917/Makefile. + +.SILENT: +.DEFAULT_GOAL := build + +SHELL := /bin/bash + +MK_DIR := $(patsubst %/,%,$(dir $(abspath $(lastword $(MAKEFILE_LIST))))) +MK_FLAGS:= $(wordlist 3, $(words $(MAKEFLAGS)), $(MAKEFLAGS)) + +# +# System flags +# + +ifndef UNAME_S + UNAME_S := $(shell uname -s) +endif +ifndef UNAME_P + UNAME_P := $(shell uname -p) +endif +ifndef UNAME_M + UNAME_M := $(shell uname -m) +endif + +ifeq ($(origin CC),default) + CC := cc +endif +ifeq ($(origin CXX),default) + CXX := c++ +endif + +ifndef LLAMA_NO_CCACHE + CCACHE := $(shell which ccache) + ifdef CCACHE + export CCACHE_SLOPPINESS = time_macros + CC := $(CCACHE) $(CC) + CXX := $(CCACHE) $(CXX) + endif +endif + +## Mac OS + Arm can report x86_64 +## ref: https://github.com/ggerganov/whisper.cpp/issues/66#issuecomment-1282546789 +ifeq ($(UNAME_S),Darwin) + ifndef LLAMA_NO_METAL + LLAMA_METAL := 1 + endif + ifneq ($(UNAME_P),arm) + SYSCTL_M := $(shell sysctl -n hw.optional.arm64 2>/dev/null) + ifeq ($(SYSCTL_M),1) + # UNAME_P := arm + # UNAME_M := arm64 + warn := $(warning Your arch is announced as x86_64, but it seems to actually be ARM64. Not fixing that can lead to bad performance. For more info see: https://github.com/ggerganov/whisper.cpp/issues/66\#issuecomment-1282546789) + endif + endif +endif + +# +# Compile flags +# + +## standard +MK_CPPFLAGS = -I$(MK_DIR) -I$(MK_DIR)/llama.cpp -I$(MK_DIR)/llama.cpp/common +MK_CFLAGS = -std=c11 -fPIC +MK_CXXFLAGS = -std=c++11 -fPIC + +## debug or optimization +ifdef LLAMA_DEBUG + MK_CFLAGS += -O0 -g + MK_CXXFLAGS += -O0 -g + MK_LDFLAGS += -g + ifeq ($(UNAME_S),Darwin) + MK_CPPFLAGS += -D_GLIBCXX_ASSERTIONS + endif +else + MK_CPPFLAGS += -DNDEBUG + ifdef LLAMA_FAST + MK_CFLAGS += -Ofast + MK_CXXFLAGS += -Ofast + else + MK_CFLAGS += -O3 + MK_CXXFLAGS += -O3 + endif +endif + +## warning +MK_CFLAGS += -Wall -Wextra -Wpedantic -Wcast-qual -Wno-unused-function \ + -Wshadow -Wstrict-prototypes -Wpointer-arith -Wmissing-prototypes -Werror=implicit-int \ + -Werror=implicit-function-declaration +MK_CXXFLAGS += -Wall -Wextra -Wpedantic -Wcast-qual -Wno-unused-function \ + -Wmissing-declarations -Wmissing-noreturn +ifdef LLAMA_FATAL_WARNINGS + MK_CFLAGS += -Werror + MK_CXXFLAGS += -Werror +endif + +## os specific +### thread +ifneq '' '$(filter $(UNAME_S),Linux Darwin FreeBSD NetBSD OpenBSD Haiku)' + MK_CFLAGS += -pthread + MK_CXXFLAGS += -pthread +endif +### windows +ifneq ($(findstring _NT,$(UNAME_S)),) + _WIN32 := 1 + LWINSOCK2 := -lws2_32 +endif + +## arch specific +ifeq ($(UNAME_M),$(filter $(UNAME_M),x86_64 i686 amd64)) + # Use all CPU extensions that are available: + MK_CFLAGS += -march=native -mtune=native + HOST_CXXFLAGS += -march=native -mtune=native + + # Usage AVX-only + #MK_CFLAGS += -mfma -mf16c -mavx + #MK_CXXFLAGS += -mfma -mf16c -mavx + + # Usage SSSE3-only (Not is SSE3!) + #MK_CFLAGS += -mssse3 + #MK_CXXFLAGS += -mssse3 +endif +ifneq '' '$(findstring mingw,$(shell $(CC) -dumpmachine))' + # The stack is only 16-byte aligned on Windows, so don't let gcc emit aligned moves. + # https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54412 + # https://github.com/ggerganov/llama.cpp/issues/2922 + MK_CFLAGS += -Xassembler -muse-unaligned-vector-move + MK_CXXFLAGS += -Xassembler -muse-unaligned-vector-move + + # Target Windows 8 for PrefetchVirtualMemory + MK_CPPFLAGS += -D_WIN32_WINNT=0x602 +endif +ifneq ($(filter aarch64%,$(UNAME_M)),) + # Apple M1, M2, etc. + # Raspberry Pi 3, 4, Zero 2 (64-bit) + # Nvidia Jetson + MK_CFLAGS += -mcpu=native + MK_CXXFLAGS += -mcpu=native + JETSON_RELEASE_INFO = $(shell jetson_release) + ifdef JETSON_RELEASE_INFO + ifneq ($(filter TX2%,$(JETSON_RELEASE_INFO)),) + JETSON_EOL_MODULE_DETECT = 1 + CC = aarch64-unknown-linux-gnu-gcc + cxx = aarch64-unknown-linux-gnu-g++ + endif + endif +endif +ifneq ($(filter armv6%,$(UNAME_M)),) + # Raspberry Pi 1, Zero + MK_CFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access + MK_CXXFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access +endif +ifneq ($(filter armv7%,$(UNAME_M)),) + # Raspberry Pi 2 + MK_CFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access -funsafe-math-optimizations + MK_CXXFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access -funsafe-math-optimizations +endif +ifneq ($(filter armv8%,$(UNAME_M)),) + # Raspberry Pi 3, 4, Zero 2 (32-bit) + MK_CFLAGS += -mfp16-format=ieee -mno-unaligned-access + MK_CXXFLAGS += -mfp16-format=ieee -mno-unaligned-access +endif +ifneq ($(filter ppc64%,$(UNAME_M)),) + POWER9_M := $(shell grep "POWER9" /proc/cpuinfo) + ifneq (,$(findstring POWER9,$(POWER9_M))) + MK_CFLAGS += -mcpu=power9 + MK_CXXFLAGS += -mcpu=power9 + endif +endif +ifneq ($(filter ppc64le%,$(UNAME_M)),) + MK_CFLAGS += -mcpu=powerpc64le + MK_CXXFLAGS += -mcpu=powerpc64le + CUDA_POWER_ARCH = 1 +endif +ifneq ($(filter loongarch64%,$(UNAME_M)),) + MK_CFLAGS += -mlasx + MK_CXXFLAGS += -mlasx +endif +ifneq ($(filter riscv64%,$(UNAME_M)),) + MK_CFLAGS += -march=rv64gcv -mabi=lp64d + MK_CXXFLAGS += -march=rv64gcv -mabi=lp64d +endif + +## platform specific +ifdef LLAMA_METAL + MK_LDFLAGS += -framework Accelerate -framework Foundation -framework Metal -framework MetalKit -framework MetalPerformanceShaders +endif +ifdef LLAMA_HIPBLAS + ifeq ($(wildcard /opt/rocm),) + ROCM_PATH ?= /usr + else + ROCM_PATH ?= /opt/rocm + endif + MK_LDFLAGS += -L$(ROCM_PATH)/lib -Wl,-rpath=$(ROCM_PATH)/lib + MK_LDFLAGS += -L$(ROCM_PATH)/lib64 -Wl,-rpath=$(ROCM_PATH)/lib64 + MK_LDFLAGS += -lhipblas -lamdhip64 -lrocblas +endif + +## combine build flags with cmdline overrides +override CPPFLAGS := $(MK_CPPFLAGS) $(CPPFLAGS) +override CFLAGS := $(CPPFLAGS) $(MK_CFLAGS) $(CFLAGS) +override CXXFLAGS := $(MK_CXXFLAGS) $(CXXFLAGS) $(CPPFLAGS) +override LDFLAGS := $(MK_LDFLAGS) $(LDFLAGS) + +# +# Helper function +# + +## BUILD_INFO prints out the build info +define BUILD_INFO + @echo "I llama-box build info:" + @echo "I UNAME_S: $(UNAME_S)" + @echo "I UNAME_P: $(UNAME_P)" + @echo "I UNAME_M: $(UNAME_M)" + @echo "I CFLAGS: $(CFLAGS)" + @echo "I CXXFLAGS: $(CXXFLAGS)" + @echo "I LDFLAGS: $(LDFLAGS)" + @echo "I CC: $(shell $(CC) --version | head -n 1)" + @echo "I CXX: $(shell $(CXX) --version | head -n 1)" + @echo +endef + +## GET_OBJ_FILE replaces .c, .cpp, and .cu file endings with .o +define GET_OBJ_FILE + $(patsubst %.c,%.o,$(patsubst %.cpp,%.o,$(patsubst %.cu,%.o,$(1)))) +endef + +# +# Main function +# + +llama-box/version.cpp: $(wildcard .git/index) llama-box/scripts/version.sh + @sh $(MK_DIR)/llama-box/scripts/version.sh > $@.tmp + @if ! cmp -s $@ $@.tmp; then mv $@.tmp $@; else rm $@.tmp; fi + +.PHONY: clean +clean: + # first, clean llama-server. + @echo ">>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>" + @echo "I cleaning llama.cpp" + @echo ">>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>" +ifdef LLAMA_METAL + make -C $(MK_DIR)/llama.cpp -j $(MK_FLAGS) LLAMA_METAL=1 LLAMA_METAL_EMBED_LIBRARY=1 clean +else + make -C $(MK_DIR)/llama.cpp -j $(MK_FLAGS) LLAMA_NO_METAL=1 clean +endif + # then, clean llama-box. + @echo ">>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>" + @echo "I cleaning llama-box" + @echo ">>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>" + $(call BUILD_INFO) + find $(MK_DIR)/llama-box -type f -name "*.o" -delete + rm -f $(MK_DIR)/llama-box/version.cpp + +.PHONY: build +build: llama-box/version.cpp llama-box/main.cpp + # first, build llama-server. + @echo ">>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>" + @echo "I building llama.cpp" + @echo ">>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>" +ifdef LLAMA_METAL + make -C $(MK_DIR)/llama.cpp -j $(MK_FLAGS) LLAMA_METAL=1 LLAMA_METAL_EMBED_LIBRARY=1 libllama.a +else + make -C $(MK_DIR)/llama.cpp -j $(MK_FLAGS) LLAMA_NO_METAL=1 libllama.a +endif + # then, build llama-box. + @echo ">>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>" + @echo "I building llama-box" + @echo ">>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>" +ifeq ($(_WIN32),1) +SUFFIX := .exe +endif + $(call BUILD_INFO) + mkdir -p $(MK_DIR)/.dist + $(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<) + $(CXX) $(CXXFLAGS) $(MK_DIR)/llama.cpp/libllama.a $(filter-out %.h %.hpp $<,$^) $(call GET_OBJ_FILE, $<) -o $(MK_DIR)/.dist/llama-box$(SUFFIX) $(LDFLAGS) $(LWINSOCK2) diff --git a/README.md b/README.md new file mode 100644 index 0000000..431760a --- /dev/null +++ b/README.md @@ -0,0 +1,197 @@ +# llama-box + +[![ci status](https://github.com/thxcode/llama-box/actions/workflows/ci.yml/badge.svg)](https://github.com/thxcode/llama-box/actions/workflows/ci.yml)
+ +LLaMA box is a cleaning LLMs inference server rather +than [llama-server](https://github.com/ggerganov/llama.cpp/blob/master/examples/server). + +## Usage + +```shell +usage: llama-box [options] + +general: + + -h, --help, --usage print usage and exit + --version show version and build info + -m, --model FILE model path (default: models/7B/ggml-model-f16.gguf) + -a, --alias NAME model name alias (default: unknown) + -s, --seed N RNG seed (default: -1, use random seed for < 0) + -t, --threads N number of threads to use during generation (default: 8) + -tb, --threads-batch N number of threads to use during batch and prompt processing (default: same as --threads) + -lcs, --lookup-cache-static FILE + path to static lookup cache to use for lookup decoding (not updated by generation) + -lcd, --lookup-cache-dynamic FILE + path to dynamic lookup cache to use for lookup decoding (updated by generation) + -c, --ctx-size N size of the prompt context (default: 0, 0 = loaded from model) + -n, --predict N number of tokens to predict (default: -1, -1 = infinity, -2 = until context filled) + -b, --batch-size N logical maximum batch size (default: 2048) + -ub, --ubatch-size N physical maximum batch size (default: 512) + --keep N number of tokens to keep from the initial prompt (default: 0, -1 = all) + --chunks N max number of chunks to process (default: -1, -1 = all) + -fa, --flash-attn enable Flash Attention (default: disabled) + --no-escape do not process escape sequences + --samplers SAMPLERS samplers that will be used for generation in the order, separated by ';' + (default: top_k;tfs_z;typical_p;top_p;min_p;temperature) + --sampling-seq SEQUENCE simplified sequence for samplers that will be used (default: kfypmt) + --penalize-nl penalize newline tokens (default: false) + --temp N temperature (default: 0.8) + --top-k N top-k sampling (default: 40, 0 = disabled) + --top-p N top-p sampling (default: 0.9, 1.0 = disabled) + --min-p N min-p sampling (default: 0.1, 0.0 = disabled) + --tfs N tail free sampling, parameter z (default: 1.0, 1.0 = disabled) + --typical N locally typical sampling, parameter p (default: 1.0, 1.0 = disabled) + --repeat-last-n N last n tokens to consider for penalize (default: 64, 0 = disabled, -1 = ctx_size) + --repeat-penalty N penalize repeat sequence of tokens (default: 1.0, 1.0 = disabled) + --presence-penalty N repeat alpha presence penalty (default: 0.0, 0.0 = disabled) + --frequency-penalty N repeat alpha frequency penalty (default: 0.0, 0.0 = disabled) + --dynatemp-range N dynamic temperature range (default: 0.0, 0.0 = disabled) + --dynatemp-exp N dynamic temperature exponent (default: 1.0) + --mirostat N use Mirostat sampling. + Top K, Nucleus, Tail Free and Locally Typical samplers are ignored if used. + (default: 0, 0 = disabled, 1 = Mirostat, 2 = Mirostat 2.0) + --mirostat-lr N Mirostat learning rate, parameter eta (default: 0.1) + --mirostat-ent N Mirostat target entropy, parameter tau (default: 5.0) + -l --logit-bias TOKEN_ID(+/-)BIAS + modifies the likelihood of token appearing in the completion, + i.e. `--logit-bias 15043+1` to increase likelihood of token ' Hello', + or `--logit-bias 15043-1` to decrease likelihood of token ' Hello' + --grammar GRAMMAR BNF-like grammar to constrain generations (see samples in grammars/ dir) (default: '') + --grammar-file FILE file to read grammar from + -j, --json-schema SCHEMA JSON schema to constrain generations (https://json-schema.org/), e.g. `{}` for any JSON object + For schemas w/ external $refs, use --grammar + example/json_schema_to_grammar.py instead + --rope-scaling {none,linear,yarn} + RoPE frequency scaling method, defaults to linear unless specified by the model + --rope-scale N RoPE context scaling factor, expands context by a factor of N + --rope-freq-base N RoPE base frequency, used by NTK-aware scaling (default: loaded from model) + --rope-freq-scale N RoPE frequency scaling factor, expands context by a factor of 1/N + --yarn-orig-ctx N YaRN: original context size of model (default: 0 = model training context size) + --yarn-ext-factor N YaRN: extrapolation mix factor (default: -1.0, 0.0 = full interpolation) + --yarn-attn-factor N YaRN: scale sqrt(t) or attention magnitude (default: 1.0) + --yarn-beta-fast N YaRN: low correction dim or beta (default: 32.0) + --yarn-beta-slow N YaRN: high correction dim or alpha (default: 1.0) + -gan, --grp-attn-n N group-attention factor (default: 1) + -gaw, --grp-attn-w N group-attention width (default: 512.0) + -nkvo, --no-kv-offload disable KV offload + -ctk, --cache-type-k TYPE KV cache data type for K (default: f16) + -ctv, --cache-type-v TYPE KV cache data type for V (default: f16) + -dt, --defrag-thold N KV cache defragmentation threshold (default: -1.0, < 0 - disabled) + -np, --parallel N number of parallel sequences to decode (default: 1) + -cb, --cont-batching enable continuous batching (a.k.a dynamic batching) (default: enabled) + --mmproj FILE path to a multimodal projector file for LLaVA. see examples/llava/README.md + --mlock force system to keep model in RAM rather than swapping or compressing + --no-mmap do not memory-map model (slower load but may reduce pageouts if not using mlock) + --numa TYPE attempt optimizations that help on some NUMA systems + - distribute: spread execution evenly over all nodes + - isolate: only spawn threads on CPUs on the node that execution started on + - numactl: use the CPU map provided by numactl + if run without this previously, it is recommended to drop the system page cache before using this + see https://github.com/ggerganov/llama.cpp/issues/1437 + --override-kv KEY=TYPE:VALUE + advanced option to override model metadata by key. may be specified multiple times. + types: int, float, bool, str. example: --override-kv tokenizer.ggml.add_bos_token=bool:false + --lora FILE apply LoRA adapter (implies --no-mmap) + --lora-scaled FILE SCALE + apply LoRA adapter with user defined scaling S (implies --no-mmap) + --lora-base FILE optional model to use as a base for the layers modified by the LoRA adapter + --control-vector FILE add a control vector + --control-vector-scaled FILE SCALE + add a control vector with user defined scaling SCALE + --control-vector-layer-range START END + layer range to apply the control vector(s) to, start and end inclusive + -ngl, --gpu-layers N number of layers to store in VRAM + -sm, --split-mode SPLIT_MODE how to split the model across multiple GPUs, one of: + - none: use one GPU only + - layer (default): split layers and KV across GPUs + - row: split rows across GPUs + -ts, --tensor-split SPLIT fraction of the model to offload to each GPU, comma-separated list of proportions, e.g. 3,1 + -mg, --main-gpu N the GPU to use for the model (with split-mode = none), + or for intermediate results and KV (with split-mode = row) (default: 0) + +server: + + --host HOST ip address to listen (default: 127.0.0.1) + --port PORT port to listen (default: 8080) + -to --timeout N server read/write timeout in seconds (default: 600) + --threads-http N number of threads used to process HTTP requests (default: -1) + --system-prompt-file FILE + set a file to load a system prompt (initial prompt of all slots), this is useful for chat applications + --metrics enable prometheus compatible metrics endpoint (default: disabled) + --infill enable infill endpoint (default: disabled) + --embeddings enable embedding endpoint (default: disabled) + --no-slots disables slots monitoring endpoint (default: enabled) + --slot-save-path PATH path to save slot kv cache (default: disabled) + --chat-template JINJA_TEMPLATE + set custom jinja chat template (default: template taken from model's metadata) + only commonly used templates are accepted: + https://github.com/ggerganov/llama.cpp/wiki/Templates-supported-by-llama_chat_apply_template + --chat-template-file FILE + set a file to load a custom jinja chat template + -sps, --slot-prompt-similarity N + how much the prompt of a request must match the prompt of a slot in order to use that slot (default: 0.50, 0.0 = disabled) + + --conn-idle N server connection idle in seconds (default: 60) + --conn-keepalive N server connection keep-alive in seconds (default: 15) + -tps --tokens-per-second N maximum number of tokens per second (default: 0, 0 = disabled, -1 = try to detect) + +logging: + + --log-format {text,json} + log output format: json or text (default: json) +``` + +## API Endpoints + +- **GET** `/health`: Returns the current state of the llama-box. + + 503 -> `{"status": "loading model"}` if the model is still being loaded. + + 500 -> `{"status": "error"}` if the model failed to load. + + 200 -> `{"status": "ok", "slots_idle": 1, "slots_processing": 2 }` if the model is successfully loaded and the server is ready for further requests mentioned below. + + 200 -> `{"status": "no slot available", "slots_idle": 0, "slots_processing": 32}` if no slots are currently available. + + 503 -> `{"status": "no slot available", "slots_idle": 0, "slots_processing": 32}` if the query parameter `fail_on_no_slot` is provided and no slots are currently available. + +- **GET** `/metrics`: Returns the Prometheus compatible metrics of the llama-box. + + This endpoint is only available if the `--metrics` flag is enabled. + + `llamacpp:prompt_tokens_total`: (Counter) Number of prompt tokens processed. + + `llamacpp:prompt_seconds_total`: (Counter) Prompt process time. + + `llamacpp:tokens_predicted_total`: (Counter) Number of generation tokens processed. + + `llamacpp:tokens_predicted_seconds_total`: (Counter) Predict process time. + + `llamacpp:prompt_tokens_seconds`: (Gauge) Average prompt throughput in tokens/s. + + `llamacpp:predicted_tokens_seconds`: (Gauge) Average generation throughput in tokens/s. + + `llamacpp:kv_cache_usage_ratio`: (Gauge) KV-cache usage. 1 means 100 percent usage. + + `llamacpp:kv_cache_tokens`: (Gauge) KV-cache tokens. + + `llamacpp:requests_processing`: (Gauge) Number of request processing. + + `llamacpp:requests_deferred`: (Gauge) Number of request deferred. + +- **GET** `/props`: Returns current server settings. + +- **POST** `/infill`: Returns the completion of the given prompt. + + This endpoint is only available if the `--infill` flag is enabled. + +- **POST** `/tokenize`: Convert text to tokens. + +- **POST** `/detokenize`: Convert tokens to text. + +- **GET** `/slots`: Returns the current slots processing state. + + This endpoint is only available if the `--no-slots` flag is no provided. + +- **POST** `/slots/:id_slot?action={save|restore|erase}`: Operate specific slot via ID. + + This endpoint is only available if the `--no-slots` flag is no provided and `--slot-save-path` is provided. + +- **POST** `/completion`: Returns the completion of the given prompt. + +- **GET** `/v1/models`: (OpenAI-compatible) Returns the list of available models, + see https://platform.openai.com/docs/api-reference/models/list. + +- **POST** `/v1/completions`: (OpenAI-compatible) Returns the completion of the given prompt, + see https://platform.openai.com/docs/api-reference/completions/create. + +- **POST** `/v1/chat/completions` (OpenAI-compatible) Returns the completion of the given prompt, + see https://platform.openai.com/docs/api-reference/chat/create. + +- **POST** `/v1/embeddings`: (OpenAI-compatible) Returns the embeddings of the given prompt, + see https://platform.openai.com/docs/api-reference/embeddings/create. + + This endpoint is only available if the `--embeddings` flag is enabled. + +## License + +MIT diff --git a/llama-box/.clang-format b/llama-box/.clang-format new file mode 100644 index 0000000..3728ce1 --- /dev/null +++ b/llama-box/.clang-format @@ -0,0 +1,14 @@ +--- +BasedOnStyle: LLVM +IndentWidth: 4 +ColumnLimit: 100 +SeparateDefinitionBlocks: Always +--- +Language: Cpp +AllowShortFunctionsOnASingleLine: None +AlignTrailingComments: true +AlignEscapedNewlines: Left +AlwaysBreakTemplateDeclarations: Yes +ConstructorInitializerAllOnOneLineOrOnePerLine: true +PackConstructorInitializers: NextLineOnly +--- diff --git a/llama-box/main.cpp b/llama-box/main.cpp new file mode 100644 index 0000000..7fe0916 --- /dev/null +++ b/llama-box/main.cpp @@ -0,0 +1,3272 @@ +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include "llama.cpp/common/common.h" +#include "llama.cpp/common/grammar-parser.h" +#include "llama.cpp/common/json-schema-to-grammar.h" +#define JSON_ASSERT GGML_ASSERT +#include "llama.cpp/common/json.hpp" +#include "llama.cpp/common/log.h" +#include "llama.cpp/ggml.h" +#include "llama.cpp/llama.h" + +#define CPPHTTPLIB_FORM_URL_ENCODED_PAYLOAD_MAX_LENGTH 10485760 +#include "llama.cpp/examples/server/httplib.h" + +#include "param.hpp" +#include "ratelimiter.hpp" +#include "utils.hpp" + +using json = nlohmann::json; + +bool server_log_json = true; + +enum stop_type { + STOP_TYPE_FULL, + STOP_TYPE_PARTIAL, +}; + +enum slot_state { + SLOT_STATE_IDLE, + SLOT_STATE_PROCESSING, +}; + +enum slot_command { + SLOT_COMMAND_NONE, + SLOT_COMMAND_LOAD_PROMPT, + SLOT_COMMAND_RELEASE, +}; + +enum server_state { + SERVER_STATE_LOADING_MODEL, // Server is starting up, model not fully + // loaded yet + SERVER_STATE_READY, // Server is ready and model is loaded + SERVER_STATE_ERROR // An error occurred, load_model failed +}; + +enum server_task_type { + SERVER_TASK_TYPE_COMPLETION, + SERVER_TASK_TYPE_CANCEL, + SERVER_TASK_TYPE_NEXT_RESPONSE, + SERVER_TASK_TYPE_METRICS, + SERVER_TASK_TYPE_SLOT_SAVE, + SERVER_TASK_TYPE_SLOT_RESTORE, + SERVER_TASK_TYPE_SLOT_ERASE, +}; + +struct server_task { + int id = -1; // to be filled by server_queue + int id_multi = -1; + int id_target = -1; + + server_task_type type; + json data; + + bool infill = false; + bool embedding = false; + + int tps = 0; +}; + +struct server_task_result { + int id = -1; + int id_multi = -1; + + json data; + + bool stop; + bool error; +}; + +struct server_task_multi { + int id = -1; + + std::set subtasks_remaining; + std::vector results; +}; + +struct slot_params { + bool stream = true; + bool cache_prompt = false; // remember the prompt to avoid reprocessing all prompt + + int32_t n_keep = 0; // number of tokens to keep from initial prompt + int32_t n_discard = 0; // number of tokens after n_keep that may be discarded when shifting + // context, 0 defaults to half + int32_t n_predict = -1; // new tokens to predict + + std::vector antiprompt; + + json input_prefix; + json input_suffix; +}; + +struct server_slot { + int id; + int id_task = -1; + int id_multi = -1; + + struct slot_params params; + + slot_state state = SLOT_STATE_IDLE; + slot_command command = SLOT_COMMAND_NONE; + + // used to determine the slot that has been used the longest + int64_t t_last_used = -1; + + // generation props + int32_t n_ctx = 0; // context size per slot + int32_t n_past = 0; + int32_t n_decoded = 0; + int32_t n_remaining = -1; + int32_t i_batch = -1; + int32_t n_predict = -1; // TODO: disambiguate from params.n_predict + + int32_t n_prompt_tokens = 0; + int32_t n_prompt_tokens_processed = 0; + + json prompt; // can be either a string, array of strings or array of token + // ids + + // when a task is submitted, we first tokenize the prompt and store it here + std::vector prompt_tokens; + + std::string generated_text; + std::vector cache_tokens; + std::vector generated_token_probs; + + bool infill = false; + bool embedding = false; + bool has_next_token = true; + bool truncated = false; + bool stopped_eos = false; + bool stopped_word = false; + bool stopped_limit = false; + + std::string stopping_word; + + bool oaicompat = false; + bool oaicompat_completion = false; + bool oaicompat_completion_chat = false; + + // sampling + llama_token sampled; + struct llama_sampling_params sparams; + llama_sampling_context *ctx_sampling = nullptr; + json json_schema; + + int32_t ga_i = 0; // group-attention state + int32_t ga_n = 1; // group-attention factor + int32_t ga_w = 512; // group-attention width + + int32_t n_past_se = 0; // self-extend + + // stats + size_t n_sent_text = 0; // number of sent text character + size_t n_sent_token_probs = 0; + + int64_t t_start_process_prompt; + int64_t t_start_generation; + + double t_prompt_processing; // ms + double t_token_generation; // ms + + token_bucket *token_bkt = nullptr; // bucket for tokens per second + + void reset() { + n_prompt_tokens = 0; + generated_text = ""; + truncated = false; + stopped_eos = false; + stopped_word = false; + stopped_limit = false; + stopping_word = ""; + n_past = 0; + n_sent_text = 0; + n_sent_token_probs = 0; + infill = false; + ga_i = 0; + n_past_se = 0; + + generated_token_probs.clear(); + + if (token_bkt != nullptr) { + delete token_bkt; + token_bkt = nullptr; + } + } + + bool has_budget(gpt_params &global_params) { + if (params.n_predict == -1 && global_params.n_predict == -1) { + return true; // limitless + } + + n_remaining = -1; + + if (params.n_predict != -1) { + n_remaining = params.n_predict - n_decoded; + } else if (global_params.n_predict != -1) { + n_remaining = global_params.n_predict - n_decoded; + } + + return n_remaining > 0; // no budget + } + + bool available() const { + return state == SLOT_STATE_IDLE && command == SLOT_COMMAND_NONE; + } + + bool is_processing() const { + return (state == SLOT_STATE_IDLE && command == SLOT_COMMAND_LOAD_PROMPT) || + state == SLOT_STATE_PROCESSING; + } + + void add_token_string(const completion_token_output &token) { + if (command == SLOT_COMMAND_RELEASE) { + return; + } + generated_token_probs.push_back(token); + } + + void release() { + if (state == SLOT_STATE_PROCESSING) { + t_token_generation = (ggml_time_us() - t_start_generation) / 1e3; + command = SLOT_COMMAND_RELEASE; + } + } + + json get_formated_timings() const { + return json{ + {"prompt_n", n_prompt_tokens_processed}, + {"prompt_ms", t_prompt_processing}, + {"prompt_per_token_ms", t_prompt_processing / n_prompt_tokens_processed}, + {"prompt_per_second", 1e3 / t_prompt_processing * n_prompt_tokens_processed}, + + {"predicted_n", n_decoded}, + {"predicted_ms", t_token_generation}, + {"predicted_per_token_ms", t_token_generation / n_decoded}, + {"predicted_per_second", 1e3 / t_token_generation * n_decoded}, + }; + } + + size_t find_stopping_strings(const std::string &text, const size_t last_token_size, + const stop_type type) { + size_t stop_pos = std::string::npos; + + for (const std::string &word : params.antiprompt) { + size_t pos; + + if (type == STOP_TYPE_FULL) { + const size_t tmp = word.size() + last_token_size; + const size_t from_pos = text.size() > tmp ? text.size() - tmp : 0; + + pos = text.find(word, from_pos); + } else { + pos = find_partial_stop_string(word, text); + } + + if (pos != std::string::npos && (stop_pos == std::string::npos || pos < stop_pos)) { + if (type == STOP_TYPE_FULL) { + stopped_word = true; + stopping_word = word; + has_next_token = false; + } + stop_pos = pos; + } + } + + return stop_pos; + } +}; + +struct server_metrics { + int64_t t_start = 0; + + uint64_t n_prompt_tokens_processed_total = 0; + uint64_t t_prompt_processing_total = 0; + uint64_t n_tokens_predicted_total = 0; + uint64_t t_tokens_generation_total = 0; + + uint64_t n_prompt_tokens_processed = 0; + uint64_t t_prompt_processing = 0; + + uint64_t n_tokens_predicted = 0; + uint64_t t_tokens_generation = 0; + + void init() { + t_start = ggml_time_us(); + } + + void on_prompt_eval(const server_slot &slot) { + n_prompt_tokens_processed_total += slot.n_prompt_tokens_processed; + n_prompt_tokens_processed += slot.n_prompt_tokens_processed; + t_prompt_processing += slot.t_prompt_processing; + t_prompt_processing_total += slot.t_prompt_processing; + } + + void on_prediction(const server_slot &slot) { + n_tokens_predicted_total += slot.n_decoded; + n_tokens_predicted += slot.n_decoded; + t_tokens_generation += slot.t_token_generation; + t_tokens_generation_total += slot.t_token_generation; + } + + void reset_bucket() { + n_prompt_tokens_processed = 0; + t_prompt_processing = 0; + n_tokens_predicted = 0; + t_tokens_generation = 0; + } +}; + +struct server_queue { + int id = 0; + bool running; + + // queues + std::vector queue_tasks; + std::vector queue_tasks_deferred; + + std::vector queue_multitasks; + + std::mutex mutex_tasks; + std::condition_variable condition_tasks; + + // callback functions + std::function callback_new_task; + std::function callback_finish_multitask; + std::function callback_update_slots; + + // Add a new task to the end of the queue + int post(server_task task) { + std::unique_lock lock(mutex_tasks); + if (task.id == -1) { + task.id = id++; + } + queue_tasks.push_back(std::move(task)); + condition_tasks.notify_one(); + return task.id; + } + + // Add a new task, but defer until one slot is available + void defer(server_task task) { + std::unique_lock lock(mutex_tasks); + queue_tasks_deferred.push_back(std::move(task)); + } + + // Get the next id for creating anew task + int get_new_id() { + std::unique_lock lock(mutex_tasks); + int new_id = id++; + return new_id; + } + + // Register function to process a new task + void on_new_task(std::function callback) { + callback_new_task = std::move(callback); + } + + // Register function to process a multitask when it is finished + void on_finish_multitask(std::function callback) { + callback_finish_multitask = std::move(callback); + } + + // Register the function to be called when all slots data is ready to be + // processed + void on_update_slots(std::function callback) { + callback_update_slots = std::move(callback); + } + + // Call when the state of one slot is changed + void notify_slot_changed() { + // move deferred tasks back to main loop + std::unique_lock lock(mutex_tasks); + for (auto &task : queue_tasks_deferred) { + queue_tasks.push_back(std::move(task)); + } + queue_tasks_deferred.clear(); + } + + // End the start_loop routine + void terminate() { + std::unique_lock lock(mutex_tasks); + running = false; + condition_tasks.notify_all(); + } + + /** + * Main loop consists of these steps: + * - Wait until a new task arrives + * - Process the task (i.e. maybe copy data into slot) + * - Check if multitask is finished + * - Update all slots + */ + void start_loop() { + running = true; + + while (true) { + while (true) { + std::unique_lock lock(mutex_tasks); + if (queue_tasks.empty()) { + lock.unlock(); + break; + } + server_task task = queue_tasks.front(); + queue_tasks.erase(queue_tasks.begin()); + lock.unlock(); + callback_new_task(task); + } + + // check if we have any finished multitasks + auto queue_iterator = queue_multitasks.begin(); + while (queue_iterator != queue_multitasks.end()) { + if (queue_iterator->subtasks_remaining.empty()) { + // all subtasks done == multitask is done + server_task_multi current_multitask = *queue_iterator; + callback_finish_multitask(current_multitask); + // remove this multitask + queue_iterator = queue_multitasks.erase(queue_iterator); + } else { + ++queue_iterator; + } + } + + // all tasks in the current loop is processed, slots data is now + // ready + callback_update_slots(); + + { + std::unique_lock lock(mutex_tasks); + if (queue_tasks.empty()) { + if (!running) { + return; + } + condition_tasks.wait(lock, [&] { return (!queue_tasks.empty() || !running); }); + } + } + } + } + + // + // functions to manage multitasks + // + + // add a multitask by specifying the id of all subtask (subtask is a + // server_task) + void add_multitask(int id_multi, std::vector &sub_ids) { + std::lock_guard lock(mutex_tasks); + server_task_multi multi; + multi.id = id_multi; + std::copy(sub_ids.begin(), sub_ids.end(), + std::inserter(multi.subtasks_remaining, multi.subtasks_remaining.end())); + queue_multitasks.push_back(multi); + } + + // updatethe remaining subtasks, while appending results to multitask + void update_multitask(int id_multi, int id_sub, server_task_result &result) { + std::lock_guard lock(mutex_tasks); + for (auto &multitask : queue_multitasks) { + if (multitask.id == id_multi) { + multitask.subtasks_remaining.erase(id_sub); + multitask.results.push_back(result); + } + } + } +}; + +struct server_response { + typedef std::function callback_multitask_t; + callback_multitask_t callback_update_multitask; + + // for keeping track of all tasks waiting for the result + std::set waiting_task_ids; + + // the main result queue + std::vector queue_results; + + std::mutex mutex_results; + std::condition_variable condition_results; + + // add the id_task to the list of tasks waiting for response + void add_waiting_task_id(int id_task) { + std::unique_lock lock(mutex_results); + waiting_task_ids.insert(id_task); + } + + // when the request is finished, we can remove task associated with it + void remove_waiting_task_id(int id_task) { + std::unique_lock lock(mutex_results); + waiting_task_ids.erase(id_task); + } + + // This function blocks the thread until there is a response for this + // id_task + server_task_result recv(int id_task) { + while (true) { + std::unique_lock lock(mutex_results); + condition_results.wait(lock, [&] { return !queue_results.empty(); }); + + for (int i = 0; i < (int)queue_results.size(); i++) { + if (queue_results[i].id == id_task) { + assert(queue_results[i].id_multi == -1); + server_task_result res = queue_results[i]; + queue_results.erase(queue_results.begin() + i); + return res; + } + } + } + + // should never reach here + } + + // Register the function to update multitask + void on_multitask_update(callback_multitask_t callback) { + callback_update_multitask = std::move(callback); + } + + // Send a new result to a waiting id_task + void send(server_task_result result) { + std::unique_lock lock(mutex_results); + for (const auto &id_task : waiting_task_ids) { + // for now, tasks that have associated parent multitasks just get + // erased once multitask picks up the result + if (result.id_multi == id_task) { + callback_update_multitask(id_task, result.id, result); + continue; + } + + if (result.id == id_task) { + queue_results.push_back(result); + condition_results.notify_all(); + return; + } + } + } +}; + +struct server_context { + llama_model *model = nullptr; + llama_context *ctx = nullptr; + + gpt_params params; + + llama_batch batch; + + bool clean_kv_cache = true; + bool add_bos_token = true; + + int32_t n_ctx; // total context for all clients / slots + int32_t n_tps; // max tokens per second + + // system prompt + bool system_need_update = false; + + std::string system_prompt; + std::vector system_tokens; + + // slots / clients + std::vector slots; + json default_generation_settings_for_props; + + server_queue queue_tasks; + server_response queue_results; + + server_metrics metrics; + + // Necessary similarity of prompt for slot selection + float slot_prompt_similarity = 0.0f; + + ~server_context() { + if (ctx) { + llama_free(ctx); + ctx = nullptr; + } + + if (model) { + llama_free_model(model); + model = nullptr; + } + + // Clear any sampling context + for (server_slot &slot : slots) { + if (slot.ctx_sampling != nullptr) { + llama_sampling_free(slot.ctx_sampling); + } + if (slot.token_bkt != nullptr) { + delete slot.token_bkt; + } + } + + llama_batch_free(batch); + } + + bool load_model(const llama_box_params &bparams) { + params = bparams.gparams; + + // dedicate one sequence to the system prompt + params.n_parallel += 1; + + std::tie(model, ctx) = llama_init_from_gpt_params(params); + params.n_parallel -= 1; // but be sneaky about it + if (model == nullptr) { + LOG_ERROR("unable to load model", {{"model", params.model}}); + return false; + } + + n_ctx = llama_n_ctx(ctx); + n_tps = bparams.n_tps; + + add_bos_token = llama_should_add_bos_token(model); + GGML_ASSERT(llama_add_eos_token(model) != 1); + + // sample tokens per second + if (n_tps < 0) { + LOG_INFO("sampling tokens per second, this will take some time...", {}); + std::vector embd = {llama_token_bos(model)}; + const int32_t n_check = std::min(n_ctx, params.n_ubatch); + llama_sampling_context *ctx_sampling = llama_sampling_init(params.sparams); + while (true) { + int32_t i = embd.size(); + if (i >= n_check) { + break; + } + if (llama_decode(ctx, llama_batch_get_one(&embd[i - 1], 1, 0, 0))) { + break; + } + const int32_t id = llama_sampling_sample(ctx_sampling, ctx, nullptr); + if (llama_token_is_eog(model, id)) { + break; + } + llama_sampling_accept(ctx_sampling, ctx, id, false); + embd.push_back(id); + } + const llama_timings timings = llama_get_timings(ctx); + n_tps = ceil(1e3 / timings.t_eval_ms * timings.n_eval); + llama_sampling_free(ctx_sampling); + llama_kv_cache_clear(ctx); + llama_synchronize(ctx); + llama_reset_timings(ctx); + LOG_INFO("sampled tokens per second", {"tps", n_tps}); + } + + return true; + } + + bool validate_model_chat_template() const { + llama_chat_message chat[] = {{"user", "test"}}; + + const int res = llama_chat_apply_template(model, nullptr, chat, 1, true, nullptr, 0); + + return res > 0; + } + + std::string load_chat_template() const { + std::vector model_template(2048, 0); // longest known template is about 1200 bytes + std::string template_key = "tokenizer.chat_template"; + int32_t res = llama_model_meta_val_str(model, template_key.c_str(), model_template.data(), + model_template.size()); + if (res < 0) { + // worst case: there is no information about template, we will use chatml by default + return "chatml"; // see llama_chat_apply_template_internal + } + return std::string(model_template.data(), res); + } + + void init() { + const int32_t n_ctx_slot = n_ctx / params.n_parallel; + + LOG_INFO("initializing slots", {{"n_slots", params.n_parallel}}); + + for (int i = 0; i < params.n_parallel; i++) { + server_slot slot; + + slot.id = i; + slot.n_ctx = n_ctx_slot; + slot.n_predict = params.n_predict; + + LOG_INFO("new slot", {{"id_slot", slot.id}, {"n_ctx_slot", slot.n_ctx}}); + + const int ga_n = params.grp_attn_n; + const int ga_w = params.grp_attn_w; + + if (ga_n != 1) { + GGML_ASSERT(ga_n > 0 && "ga_n must be positive"); // NOLINT + GGML_ASSERT(ga_w % ga_n == 0 && "ga_w must be a multiple of ga_n"); // NOLINT + // GGML_ASSERT(n_ctx_train % ga_w == 0 && "n_ctx_train must + // be a multiple of ga_w"); // NOLINT GGML_ASSERT(n_ctx >= + // n_ctx_train * ga_n && "n_ctx must be at least n_ctx_train * + // ga_n"); // NOLINT + + LOG_INFO("slot self-extend", + {{"id_slot", slot.id}, {"ga_n", ga_n}, {"ga_w", ga_w}}); + } + + slot.ga_i = 0; + slot.ga_n = ga_n; + slot.ga_w = ga_w; + + slot.reset(); + + slots.push_back(slot); + } + + default_generation_settings_for_props = get_formated_generation(slots.front()); + default_generation_settings_for_props["seed"] = -1; + + // the update_slots() logic will always submit a maximum of n_batch + // tokens note that n_batch can be > n_ctx (e.g. for non-causal + // attention models such as BERT where the KV cache is not used) + { + const int32_t n_batch = llama_n_batch(ctx); + + // only a single seq_id per token is needed + batch = llama_batch_init(n_batch, 0, 1); + } + + metrics.init(); + } + + std::vector tokenize(const json &json_prompt, bool add_special) const { + // TODO: currently, we tokenize using special tokens by default + // this is not always correct (see + // https://github.com/ggerganov/llama.cpp/pull/4160#issuecomment-1824826216) + // but it's better compared to completely ignoring ChatML and + // other chat templates + const bool TMP_FORCE_SPECIAL = true; + + // If `add_bos` is true, we only add BOS, when json_prompt is a string, + // or the first element of the json_prompt array is a string. + std::vector prompt_tokens; + + if (json_prompt.is_array()) { + bool first = true; + for (const auto &p : json_prompt) { + if (p.is_string()) { + auto s = p.template get(); + + std::vector p; + if (first) { + p = ::llama_tokenize(ctx, s, add_special, TMP_FORCE_SPECIAL); + first = false; + } else { + p = ::llama_tokenize(ctx, s, false, TMP_FORCE_SPECIAL); + } + + prompt_tokens.insert(prompt_tokens.end(), p.begin(), p.end()); + } else { + if (first) { + first = false; + } + + prompt_tokens.push_back(p.template get()); + } + } + } else { + auto s = json_prompt.template get(); + prompt_tokens = ::llama_tokenize(ctx, s, add_special, TMP_FORCE_SPECIAL); + } + + return prompt_tokens; + } + + server_slot *get_slot_by_id(int id) { + for (server_slot &slot : slots) { + if (slot.id == id) { + return &slot; + } + } + + return nullptr; + } + + server_slot *get_available_slot(const std::string &prompt) { + server_slot *ret = nullptr; + + // find the slot that has at least n% prompt similarity + if (ret == nullptr && slot_prompt_similarity != 0.0f && !prompt.empty()) { + int max_lcp_len = 0; + float similarity = 0; + + for (server_slot &slot : slots) { + // skip the slot if it is not available + if (!slot.available()) { + continue; + } + + // skip the slot if it does not contains prompt + if (!slot.prompt.is_string()) { + continue; + } + + // current slot's prompt + std::string slot_prompt = slot.prompt.get(); + + // length of the current slot's prompt + int slot_prompt_len = slot_prompt.size(); + + // length of the Longest Common Prefix between the current + // slot's prompt and the input prompt + int lcp_len = common_part(slot_prompt, prompt); + + // fraction of the common substring length compared to the + // current slot's prompt length + similarity = static_cast(lcp_len) / slot_prompt_len; + + // select the current slot if the criteria match + if (lcp_len > max_lcp_len && similarity > slot_prompt_similarity) { + max_lcp_len = lcp_len; + ret = &slot; + } + } + } + + // find the slot that has been least recently used + if (ret == nullptr) { + int64_t t_last = ggml_time_us(); + for (server_slot &slot : slots) { + // skip the slot if it is not available + if (!slot.available()) { + continue; + } + + // select the current slot if the criteria match + if (slot.t_last_used < t_last) { + t_last = slot.t_last_used; + ret = &slot; + } + } + } + + return ret; + } + + bool launch_slot_with_task(server_slot &slot, const server_task &task) { + llama_sampling_params sparams = params.sparams; + auto &data = task.data; + + slot.oaicompat = json_value(data, "__oaicompat", false); + slot.oaicompat_completion = json_value(data, "__oaicompat_completion", false); + slot.oaicompat_completion_chat = json_value(data, "__oaicompat_completion_chat", false); + + slot.params.stream = json_value(data, "stream", false); + slot.params.cache_prompt = json_value(data, "cache_prompt", false); + slot.params.n_keep = json_value(data, "n_keep", params.n_keep); + slot.params.n_predict = json_value(data, "n_predict", params.n_predict); + slot.params.n_discard = json_value(data, "n_discard", 0); + slot.params.input_prefix = json_value(data, "input_prefix", params.input_prefix); + slot.params.input_suffix = json_value(data, "input_suffix", params.input_suffix); + + slot.sparams.top_k = json_value(data, "top_k", sparams.top_k); + slot.sparams.top_p = json_value(data, "top_p", sparams.top_p); + slot.sparams.min_p = json_value(data, "min_p", sparams.min_p); + slot.sparams.tfs_z = json_value(data, "tfs_z", sparams.tfs_z); + slot.sparams.typical_p = json_value(data, "typical_p", sparams.typical_p); + slot.sparams.temp = json_value(data, "temperature", sparams.temp); + slot.sparams.dynatemp_range = json_value(data, "dynatemp_range", sparams.dynatemp_range); + slot.sparams.dynatemp_exponent = + json_value(data, "dynatemp_exponent", sparams.dynatemp_exponent); + slot.sparams.penalty_last_n = json_value(data, "repeat_last_n", sparams.penalty_last_n); + slot.sparams.penalty_repeat = json_value(data, "repeat_penalty", sparams.penalty_repeat); + slot.sparams.penalty_freq = json_value(data, "frequency_penalty", sparams.penalty_freq); + slot.sparams.penalty_present = + json_value(data, "presence_penalty", sparams.penalty_present); + slot.sparams.mirostat = json_value(data, "mirostat", sparams.mirostat); + slot.sparams.mirostat_tau = json_value(data, "mirostat_tau", sparams.mirostat_tau); + slot.sparams.mirostat_eta = json_value(data, "mirostat_eta", sparams.mirostat_eta); + slot.sparams.penalize_nl = json_value(data, "penalize_nl", sparams.penalize_nl); + slot.sparams.seed = json_value(data, "seed", sparams.seed); + slot.sparams.n_probs = json_value(data, "n_probs", sparams.n_probs); + slot.sparams.min_keep = json_value(data, "min_keep", sparams.min_keep); + + // process "json_schema" and "grammar" + 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")) { + try { + auto schema = json_value(data, "json_schema", json::object()); + slot.sparams.grammar = json_schema_to_grammar(schema); + } catch (const std::exception &e) { + send_error(task, std::string("\"json_schema\": ") + e.what(), + ERROR_TYPE_INVALID_REQUEST); + return false; + } + } else { + slot.sparams.grammar = json_value(data, "grammar", sparams.grammar); + } + + if (slot.params.cache_prompt && slot.ga_n != 1) { + LOG_WARNING("cache_prompt is not supported with group-attention", {}); + slot.params.cache_prompt = false; + } + + if (slot.n_predict > 0 && slot.params.n_predict > slot.n_predict) { + // Might be better to reject the request with a 400 ? + LOG_WARNING("Max tokens to predict exceeds server configuration", + { + {"params.n_predict", slot.params.n_predict}, + {"slot.n_predict", slot.n_predict}, + }); + slot.params.n_predict = slot.n_predict; + } + + // get prompt + if (!task.infill) { + const auto &prompt = data.find("prompt"); + if (prompt == data.end()) { + send_error(task, "\"prompt\" must be provided", ERROR_TYPE_INVALID_REQUEST); + return false; + } + + if ((prompt->is_string()) || + (prompt->is_array() && prompt->size() == 1 && prompt->at(0).is_string()) || + (prompt->is_array() && !prompt->empty() && prompt->at(0).is_number_integer())) { + slot.prompt = *prompt; + } else { + send_error(task, "\"prompt\" must be a string or an array of integers", + ERROR_TYPE_INVALID_REQUEST); + return false; + } + } + + // penalize user-provided tokens + { + slot.sparams.penalty_prompt_tokens.clear(); + slot.sparams.use_penalty_prompt_tokens = false; + + const auto &penalty_prompt = data.find("penalty_prompt"); + + if (penalty_prompt != data.end()) { + if (penalty_prompt->is_string()) { + const auto penalty_prompt_string = penalty_prompt->get(); + slot.sparams.penalty_prompt_tokens = + llama_tokenize(model, penalty_prompt_string, false); + + if (slot.params.n_predict > 0) { + slot.sparams.penalty_prompt_tokens.reserve( + slot.sparams.penalty_prompt_tokens.size() + slot.params.n_predict); + } + slot.sparams.use_penalty_prompt_tokens = true; + } else if (penalty_prompt->is_array()) { + const auto n_tokens = penalty_prompt->size(); + slot.sparams.penalty_prompt_tokens.reserve(n_tokens + + std::max(0, slot.params.n_predict)); + + const int n_vocab = llama_n_vocab(model); + for (const auto &penalty_token : *penalty_prompt) { + if (penalty_token.is_number_integer()) { + const auto tok = penalty_token.get(); + if (tok >= 0 && tok < n_vocab) { + slot.sparams.penalty_prompt_tokens.push_back(tok); + } + } + } + slot.sparams.use_penalty_prompt_tokens = true; + } + } + } + + { + slot.sparams.logit_bias.clear(); + + if (json_value(data, "ignore_eos", false)) { + slot.sparams.logit_bias[llama_token_eos(model)] = -INFINITY; + } + + const auto &logit_bias = data.find("logit_bias"); + if (logit_bias != data.end() && logit_bias->is_array()) { + const int n_vocab = llama_n_vocab(model); + for (const auto &el : *logit_bias) { + // TODO: we may want to throw errors here, in case "el" is + // incorrect + if (el.is_array() && el.size() == 2) { + float bias; + if (el[1].is_number()) { + bias = el[1].get(); + } else if (el[1].is_boolean() && !el[1].get()) { + bias = -INFINITY; + } else { + continue; + } + + if (el[0].is_number_integer()) { + llama_token tok = el[0].get(); + if (tok >= 0 && tok < n_vocab) { + slot.sparams.logit_bias[tok] = bias; + } + } else if (el[0].is_string()) { + auto toks = llama_tokenize(model, el[0].get(), false); + for (auto tok : toks) { + slot.sparams.logit_bias[tok] = bias; + } + } + } + } + } + } + + { + slot.params.antiprompt.clear(); + + const auto &stop = data.find("stop"); + if (stop != data.end() && stop->is_array()) { + for (const auto &word : *stop) { + if (!word.empty()) { + slot.params.antiprompt.push_back(word); + } + } + } + } + + { + const auto &samplers_sequence = data.find("samplers"); + if (samplers_sequence != data.end() && samplers_sequence->is_array()) { + std::vector sampler_names; + for (const auto &sampler_name : *samplers_sequence) { + if (sampler_name.is_string()) { + sampler_names.emplace_back(sampler_name); + } + } + slot.sparams.samplers_sequence = + llama_sampling_types_from_names(sampler_names, false); + } else { + slot.sparams.samplers_sequence = sparams.samplers_sequence; + } + } + + { + if (slot.ctx_sampling != nullptr) { + llama_sampling_free(slot.ctx_sampling); + } + slot.ctx_sampling = llama_sampling_init(slot.sparams); + if (slot.ctx_sampling == nullptr) { + // for now, the only error that may happen here is invalid + // grammar + send_error(task, "Failed to parse grammar", ERROR_TYPE_INVALID_REQUEST); + return false; + } + } + + { + if (slot.token_bkt != nullptr) { + delete slot.token_bkt; + slot.token_bkt = nullptr; + } + int tps = task.tps; +#ifndef NDEBUG + tps = json_value(data, "tps", + task.tps); // allow overriding tps for debugging + if (tps > n_tps) { + tps = n_tps; + } +#endif + if (tps > 0) { + slot.token_bkt = new token_bucket(tps, tps); + if (slot.token_bkt == nullptr) { + send_error(task, "Failed to create token bucket", ERROR_TYPE_SERVER); + return false; + } + } + } + + slot.command = SLOT_COMMAND_LOAD_PROMPT; + slot.prompt_tokens.clear(); + + LOG_INFO("slot is processing task", { + {"id_slot", slot.id}, + {"id_task", slot.id_task}, + }); + + return true; + } + + void kv_cache_clear() { + // clear the entire KV cache + llama_kv_cache_clear(ctx); + clean_kv_cache = false; + } + + void system_prompt_update() { + kv_cache_clear(); + system_tokens.clear(); + + if (!system_prompt.empty()) { + system_tokens = ::llama_tokenize(ctx, system_prompt, true); + + llama_batch_clear(batch); + + for (int i = 0; i < (int)system_tokens.size(); ++i) { + llama_batch_add(batch, system_tokens[i], i, {0}, false); + } + + const int32_t n_batch = llama_n_batch(ctx); + + for (int32_t i = 0; i < batch.n_tokens; i += n_batch) { + const int32_t n_tokens = std::min(params.n_batch, batch.n_tokens - i); + llama_batch batch_view = { + n_tokens, + batch.token + i, + nullptr, + batch.pos + i, + batch.n_seq_id + i, + batch.seq_id + i, + batch.logits + i, + 0, + 0, + 0, // unused + }; + + if (llama_decode(ctx, batch_view) != 0) { + LOG_ERROR("llama_decode() failed", {}); + return; + } + } + + // assign the system KV cache to all parallel sequences + for (int32_t i = 1; i <= params.n_parallel; ++i) { + llama_kv_cache_seq_cp(ctx, 0, i, -1, -1); + } + } + + system_need_update = false; + } + + bool system_prompt_set(const std::string &sys_prompt) { + system_prompt = sys_prompt; + + // release all slots + for (server_slot &slot : slots) { + slot.release(); + } + + system_need_update = true; + return true; + } + + bool process_token(completion_token_output &result, server_slot &slot) { + // remember which tokens were sampled - used for repetition penalties + // during sampling + const std::string token_str = llama_token_to_piece(ctx, result.tok, false); + slot.sampled = result.tok; + + // search stop word and delete it + slot.generated_text += token_str; + slot.has_next_token = true; + + if (slot.ctx_sampling->params.use_penalty_prompt_tokens && result.tok != -1) { + // we can change penalty_prompt_tokens because it is always created + // from scratch each request + slot.ctx_sampling->params.penalty_prompt_tokens.push_back(result.tok); + } + + // check if there is incomplete UTF-8 character at the end + bool incomplete = false; + for (unsigned i = 1; i < 5 && i <= slot.generated_text.size(); ++i) { + unsigned char c = slot.generated_text[slot.generated_text.size() - i]; + if ((c & 0xC0) == 0x80) { + // continuation byte: 10xxxxxx + continue; + } + if ((c & 0xE0) == 0xC0) { + // 2-byte character: 110xxxxx ... + incomplete = i < 2; + } else if ((c & 0xF0) == 0xE0) { + // 3-byte character: 1110xxxx ... + incomplete = i < 3; + } else if ((c & 0xF8) == 0xF0) { + // 4-byte character: 11110xxx ... + incomplete = i < 4; + } + // else 1-byte character or invalid byte + break; + } + + if (!incomplete) { + size_t pos = std::min(slot.n_sent_text, slot.generated_text.size()); + + const std::string str_test = slot.generated_text.substr(pos); + bool is_stop_full = false; + + size_t stop_pos = + slot.find_stopping_strings(str_test, token_str.size(), STOP_TYPE_FULL); + if (stop_pos != std::string::npos) { + is_stop_full = true; + slot.generated_text.erase(slot.generated_text.begin() + pos + stop_pos, + slot.generated_text.end()); + pos = std::min(slot.n_sent_text, slot.generated_text.size()); + } else { + is_stop_full = false; + stop_pos = + slot.find_stopping_strings(str_test, token_str.size(), STOP_TYPE_PARTIAL); + } + + // check if there is any token to predict + if (stop_pos == std::string::npos || + (!slot.has_next_token && !is_stop_full && stop_pos > 0)) { + // no send the stop word in the response + result.text_to_send = slot.generated_text.substr(pos, std::string::npos); + slot.n_sent_text += result.text_to_send.size(); + // add the token to slot queue and cache + } + + slot.add_token_string(result); + if (slot.params.stream) { + send_partial_response(slot, result); + } + } + + if (incomplete) { + slot.has_next_token = true; + } + + // check the limits + if (slot.n_decoded > 0 && slot.has_next_token && !slot.has_budget(params)) { + slot.stopped_limit = true; + slot.has_next_token = false; + } + + if (llama_token_is_eog(model, result.tok)) { + slot.stopped_eos = true; + slot.has_next_token = false; + } + + auto n_ctx_train = llama_n_ctx_train(model); + if (slot.params.n_predict < 1 && slot.n_predict < 1 && slot.ga_n == 1 && + slot.n_prompt_tokens + slot.n_decoded >= n_ctx_train) { + LOG_WARNING("n_predict is not set and self-context extend is disabled." + " Limiting generated tokens to n_ctx_train to avoid EOS-less " + "generation " + "infinite loop", + { + {"id_slot", slot.id}, + {"params.n_predict", slot.params.n_predict}, + {"slot.n_prompt_tokens", slot.n_prompt_tokens}, + {"slot.n_decoded", slot.n_decoded}, + {"slot.n_predict", slot.n_predict}, + {"n_slots", params.n_parallel}, + {"slot.n_ctx", slot.n_ctx}, + {"n_ctx", n_ctx}, + {"n_ctx_train", n_ctx_train}, + {"ga_n", slot.ga_n}, + }); + slot.truncated = true; + slot.stopped_limit = true; + slot.has_next_token = false; // stop prediction + } + + return slot.has_next_token; // continue + } + + json get_formated_generation(const server_slot &slot) const { + const auto eos_bias = slot.sparams.logit_bias.find(llama_token_eos(model)); + const bool ignore_eos = eos_bias != slot.sparams.logit_bias.end() && + eos_bias->second < 0.0f && std::isinf(eos_bias->second); + + std::vector samplers_sequence; + samplers_sequence.reserve(slot.sparams.samplers_sequence.size()); + for (const auto &sampler_type : slot.sparams.samplers_sequence) { + samplers_sequence.emplace_back(llama_sampling_type_to_str(sampler_type)); + } + + return json{{"n_ctx", slot.n_ctx}, + {"n_predict", slot.n_predict}, + {"model", params.model_alias}, + {"seed", slot.sparams.seed}, + {"temperature", slot.sparams.temp}, + {"dynatemp_range", slot.sparams.dynatemp_range}, + {"dynatemp_exponent", slot.sparams.dynatemp_exponent}, + {"top_k", slot.sparams.top_k}, + {"top_p", slot.sparams.top_p}, + {"min_p", slot.sparams.min_p}, + {"tfs_z", slot.sparams.tfs_z}, + {"typical_p", slot.sparams.typical_p}, + {"repeat_last_n", slot.sparams.penalty_last_n}, + {"repeat_penalty", slot.sparams.penalty_repeat}, + {"presence_penalty", slot.sparams.penalty_present}, + {"frequency_penalty", slot.sparams.penalty_freq}, + {"penalty_prompt_tokens", slot.sparams.penalty_prompt_tokens}, + {"use_penalty_prompt_tokens", slot.sparams.use_penalty_prompt_tokens}, + {"mirostat", slot.sparams.mirostat}, + {"mirostat_tau", slot.sparams.mirostat_tau}, + {"mirostat_eta", slot.sparams.mirostat_eta}, + {"penalize_nl", slot.sparams.penalize_nl}, + {"stop", slot.params.antiprompt}, + {"n_predict", slot.params.n_predict}, // TODO: fix duplicate key n_predict + {"n_keep", slot.params.n_keep}, + {"n_discard", slot.params.n_discard}, + {"ignore_eos", ignore_eos}, + {"stream", slot.params.stream}, + {"logit_bias", slot.sparams.logit_bias}, + {"n_probs", slot.sparams.n_probs}, + {"min_keep", slot.sparams.min_keep}, + {"grammar", slot.sparams.grammar}, + {"samplers", samplers_sequence}}; + } + + void send_error(const server_task &task, const std::string &error, + const enum error_type type = ERROR_TYPE_SERVER) { + send_error(task.id, task.id_multi, error, type); + } + + void send_error(const server_slot &slot, const std::string &error, + const enum error_type type = ERROR_TYPE_SERVER) { + send_error(slot.id_task, slot.id_multi, error, type); + } + + void send_error(const int id_task, const int id_multi, const std::string &error, + const enum error_type type = ERROR_TYPE_SERVER) { + LOG_ERROR("task error", { + {"id_multi", id_multi}, + {"id_task", id_task}, + {"error", error}, + }); + + server_task_result res; + res.id = id_task; + res.id_multi = id_multi; + res.stop = false; + res.error = true; + res.data = format_error_response(error, type); + + queue_results.send(res); + } + + void send_partial_response(server_slot &slot, completion_token_output tkn) { + server_task_result res; + res.id = slot.id_task; + res.id_multi = slot.id_multi; + res.error = false; + res.stop = false; + res.data = json{{"content", tkn.text_to_send}, + {"stop", false}, + {"id_slot", slot.id}, + {"multimodal", false}}; + + if (slot.sparams.n_probs > 0) { + const std::vector to_send_toks = + llama_tokenize(ctx, tkn.text_to_send, false); + const size_t probs_pos = + std::min(slot.n_sent_token_probs, slot.generated_token_probs.size()); + const size_t probs_stop_pos = std::min(slot.n_sent_token_probs + to_send_toks.size(), + slot.generated_token_probs.size()); + + std::vector probs_output; + if (probs_pos < probs_stop_pos) { + probs_output = std::vector( + slot.generated_token_probs.begin() + probs_pos, + slot.generated_token_probs.begin() + probs_stop_pos); + } + slot.n_sent_token_probs = probs_stop_pos; + + res.data["completion_probabilities"] = probs_vector_to_json( + ctx, probs_output, slot.oaicompat_completion, slot.oaicompat_completion_chat); + } + + queue_results.send(res); + } + + void send_final_response(const server_slot &slot) { + server_task_result res; + res.id = slot.id_task; + res.id_multi = slot.id_multi; + res.error = false; + res.stop = true; + res.data = json{{"content", !slot.params.stream ? slot.generated_text : ""}, + {"id_slot", slot.id}, + {"stop", true}, + {"model", params.model_alias}, + {"tokens_predicted", slot.n_decoded}, + {"tokens_evaluated", slot.n_prompt_tokens}, + {"generation_settings", get_formated_generation(slot)}, + {"prompt", slot.prompt}, + {"truncated", slot.truncated}, + {"stopped_eos", slot.stopped_eos}, + {"stopped_word", slot.stopped_word}, + {"stopped_limit", slot.stopped_limit}, + {"stopping_word", slot.stopping_word}, + {"tokens_cached", slot.n_past}, + {"timings", slot.get_formated_timings()}}; + + if (slot.sparams.n_probs > 0) { + std::vector probs; + if (!slot.params.stream && slot.stopped_word) { + const std::vector stop_word_toks = + llama_tokenize(ctx, slot.stopping_word, false); + + size_t safe_offset = + std::min(slot.generated_token_probs.size(), stop_word_toks.size()); + probs = std::vector(slot.generated_token_probs.begin(), + slot.generated_token_probs.end() - + safe_offset); + } else { + probs = std::vector(slot.generated_token_probs.begin(), + slot.generated_token_probs.end()); + } + + res.data["completion_probabilities"] = probs_vector_to_json( + ctx, probs, slot.oaicompat_completion, slot.oaicompat_completion_chat); + } + + queue_results.send(res); + } + + void send_embedding(const server_slot &slot, const llama_batch &batch) { + server_task_result res; + res.id = slot.id_task; + res.id_multi = slot.id_multi; + res.error = false; + res.stop = true; + + const int n_embd = llama_n_embd(model); + + std::vector embd_res(n_embd, 0.0f); + + for (int i = 0; i < batch.n_tokens; ++i) { + if (!batch.logits[i] || batch.seq_id[i][0] != slot.id + 1) { + continue; + } + + const float *embd = llama_get_embeddings_seq(ctx, batch.seq_id[i][0]); + if (embd == NULL) { + embd = llama_get_embeddings_ith(ctx, i); + } + + if (embd == NULL) { + LOG_ERROR("failed to get embeddings", + {{"token", batch.token[i]}, {"seq_id", batch.seq_id[i][0]}}); + + res.data = json{ + {"embedding", std::vector(n_embd, 0.0f)}, + }; + + continue; + } + + llama_embd_normalize(embd, embd_res.data(), n_embd); + + res.data = json{ + {"embedding", embd_res}, + }; + } + + res.data["tokens_evaluated"] = slot.n_prompt_tokens; + queue_results.send(res); + } + + void request_completion(int id_task, int id_multi, json data, bool infill, bool embedding, + int tps = 0) { + server_task task; + task.id = id_task; + task.id_multi = id_multi; + task.id_target = 0; + task.data = std::move(data); + task.infill = infill; + task.embedding = embedding; + task.type = SERVER_TASK_TYPE_COMPLETION; + task.tps = tps; + + // when a completion task's prompt array is not a singleton, we split it + // into multiple requests otherwise, it's a single-prompt task, we + // actually queue it if there's numbers in the prompt array it will be + // treated as an array of tokens + if (task.data.count("prompt") != 0 && task.data.at("prompt").size() > 1) { + bool numbers = false; + for (const auto &e : task.data.at("prompt")) { + if (e.is_number()) { + numbers = true; + break; + } + } + + // NOTE: split_multiprompt_task() does not handle a mix of strings + // and numbers, it will completely stall the server. I don't know + // where the bug for this is. + // + // if there are numbers, it needs to be treated like a single + // prompt, queue_tasks handles a mix of strings and numbers just + // fine. + if (numbers) { + queue_tasks.post(task); + } else { + split_multiprompt_task(id_task, task); + } + } else { + queue_tasks.post(task); + } + } + + void request_cancel(int id_task) { + server_task task; + task.type = SERVER_TASK_TYPE_CANCEL; + task.id_target = id_task; + + queue_tasks.post(task); + } + + void split_multiprompt_task(int id_multi, const server_task &multiprompt_task) { + const int prompt_count = multiprompt_task.data.at("prompt").size(); + if (prompt_count <= 1) { + send_error(multiprompt_task, "error while handling multiple prompts"); + return; + } + + // generate all the ID for subtask + std::vector subtask_ids(prompt_count); + for (int i = 0; i < prompt_count; i++) { + subtask_ids[i] = queue_tasks.get_new_id(); + } + + // queue up the multitask so we can track its subtask progression + queue_tasks.add_multitask(id_multi, subtask_ids); + + // add subtasks + for (int i = 0; i < prompt_count; i++) { + json subtask_data = multiprompt_task.data; + 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, multiprompt_task.tps); + } + } + + void process_single_task(const server_task &task) { + switch (task.type) { + case SERVER_TASK_TYPE_COMPLETION: { + const int id_slot = json_value(task.data, "id_slot", -1); + + server_slot *slot; + + if (id_slot != -1) { + slot = get_slot_by_id(id_slot); + } else { + std::string prompt; + if (task.data.contains("prompt") && task.data.at("prompt").is_string()) { + prompt = json_value(task.data, "prompt", std::string()); + } + + slot = get_available_slot(prompt); + } + + if (slot == nullptr) { + // if no slot is available, we defer this task for + // processing later + queue_tasks.defer(task); + break; + } + if (!slot->available()) { + // if requested slot is unavailable, we defer this task for + // processing later + queue_tasks.defer(task); + break; + } + + if (task.data.contains("system_prompt")) { + std::string sys_prompt = json_value(task.data, "system_prompt", std::string()); + system_prompt_set(sys_prompt); + + for (server_slot &slot : slots) { + slot.n_past = 0; + slot.n_past_se = 0; + } + } + + slot->reset(); + + slot->id_task = task.id; + slot->id_multi = task.id_multi; + slot->infill = task.infill; + slot->embedding = task.embedding; + + if (!launch_slot_with_task(*slot, task)) { + LOG_ERROR("error while launching slot", task.data); + break; + } + } break; + case SERVER_TASK_TYPE_CANCEL: { + // release slot linked with the task id + for (auto &slot : slots) { + if (slot.id_task == task.id_target) { + slot.release(); + break; + } + } + } break; + case SERVER_TASK_TYPE_NEXT_RESPONSE: { + // do nothing + } break; + case SERVER_TASK_TYPE_METRICS: { + json slots_data = json::array(); + + int n_idle_slots = 0; + int n_processing_slots = 0; + + for (server_slot &slot : slots) { + json slot_data = get_formated_generation(slot); + slot_data["id"] = slot.id; + slot_data["id_task"] = slot.id_task; + slot_data["state"] = slot.state; + slot_data["prompt"] = slot.prompt; + slot_data["next_token"] = { + {"has_next_token", slot.has_next_token}, {"n_remain", slot.n_remaining}, + {"n_decoded", slot.n_decoded}, {"stopped_eos", slot.stopped_eos}, + {"stopped_word", slot.stopped_word}, {"stopped_limit", slot.stopped_limit}, + {"stopping_word", slot.stopping_word}, + }; + + if (slot_data["state"] == SLOT_STATE_IDLE) { + n_idle_slots++; + } else { + n_processing_slots++; + } + + slots_data.push_back(slot_data); + } + LOG_INFO("slot data", {{"id_task", task.id}, + {"n_idle_slots", n_idle_slots}, + {"n_processing_slots", n_processing_slots}}); + + server_task_result res; + res.id = task.id; + res.id_multi = task.id_multi; + res.stop = true; + res.error = false; + res.data = { + {"idle", n_idle_slots}, + {"processing", n_processing_slots}, + {"deferred", queue_tasks.queue_tasks_deferred.size()}, + {"t_start", metrics.t_start}, + + {"n_prompt_tokens_processed_total", metrics.n_prompt_tokens_processed_total}, + {"t_tokens_generation_total", metrics.t_tokens_generation_total}, + {"n_tokens_predicted_total", metrics.n_tokens_predicted_total}, + {"t_prompt_processing_total", metrics.t_prompt_processing_total}, + + {"n_prompt_tokens_processed", metrics.n_prompt_tokens_processed}, + {"t_prompt_processing", metrics.t_prompt_processing}, + {"n_tokens_predicted", metrics.n_tokens_predicted}, + {"t_tokens_generation", metrics.t_tokens_generation}, + + {"kv_cache_tokens_count", llama_get_kv_cache_token_count(ctx)}, + {"kv_cache_used_cells", llama_get_kv_cache_used_cells(ctx)}, + + {"slots", slots_data}, + }; + + if (json_value(task.data, "reset_bucket", false)) { + metrics.reset_bucket(); + } + queue_results.send(res); + } break; + case SERVER_TASK_TYPE_SLOT_SAVE: { + int id_slot = task.data.at("id_slot"); + server_slot *slot = get_slot_by_id(id_slot); + if (slot == nullptr) { + send_error(task, "Invalid slot ID", ERROR_TYPE_INVALID_REQUEST); + break; + } + if (!slot->available()) { + // if requested slot is unavailable, we defer this task for + // processing later + queue_tasks.defer(task); + break; + } + + const size_t token_count = slot->cache_tokens.size(); + const int64_t t_start = ggml_time_us(); + + std::string filename = task.data.at("filename"); + 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); + + const int64_t t_end = ggml_time_us(); + const double t_save_ms = (t_end - t_start) / 1000.0; + + server_task_result result; + result.id = task.id; + result.stop = true; + result.error = false; + result.data = json{{"id_slot", id_slot}, + {"filename", filename}, + {"n_saved", token_count}, // tokens saved + {"n_written", nwrite}, // bytes written + {"timings", {{"save_ms", t_save_ms}}}}; + queue_results.send(result); + } break; + case SERVER_TASK_TYPE_SLOT_RESTORE: { + int id_slot = task.data.at("id_slot"); + server_slot *slot = get_slot_by_id(id_slot); + if (slot == nullptr) { + send_error(task, "Invalid slot ID", ERROR_TYPE_INVALID_REQUEST); + break; + } + if (!slot->available()) { + // if requested slot is unavailable, we defer this task for + // processing later + queue_tasks.defer(task); + break; + } + + const int64_t t_start = ggml_time_us(); + + std::string filename = task.data.at("filename"); + std::string filepath = task.data.at("filepath"); + + slot->cache_tokens.resize(slot->n_ctx); + size_t token_count = 0; + size_t nread = llama_state_seq_load_file(ctx, filepath.c_str(), slot->id + 1, + slot->cache_tokens.data(), + slot->cache_tokens.size(), &token_count); + if (nread == 0) { + slot->cache_tokens.resize(0); + send_error(task, + "Unable to restore slot, no available space in KV " + "cache or invalid slot " + "save file", + ERROR_TYPE_INVALID_REQUEST); + break; + } + slot->cache_tokens.resize(token_count); + + const int64_t t_end = ggml_time_us(); + const double t_restore_ms = (t_end - t_start) / 1000.0; + + server_task_result result; + result.id = task.id; + result.stop = true; + result.error = false; + result.data = json{{"id_slot", id_slot}, + {"filename", filename}, + {"n_restored", token_count}, // tokens restored + {"n_read", nread}, // bytes read + {"timings", {{"restore_ms", t_restore_ms}}}}; + queue_results.send(result); + } break; + case SERVER_TASK_TYPE_SLOT_ERASE: { + int id_slot = task.data.at("id_slot"); + server_slot *slot = get_slot_by_id(id_slot); + if (slot == nullptr) { + send_error(task, "Invalid slot ID", ERROR_TYPE_INVALID_REQUEST); + break; + } + if (!slot->available()) { + // if requested slot is unavailable, we defer this task for + // processing later + queue_tasks.defer(task); + break; + } + + // Erase token cache + const size_t n_erased = slot->cache_tokens.size(); + llama_kv_cache_seq_rm(ctx, slot->id + 1, -1, -1); + slot->cache_tokens.clear(); + + server_task_result result; + result.id = task.id; + result.stop = true; + result.error = false; + result.data = json{{"id_slot", id_slot}, {"n_erased", n_erased}}; + queue_results.send(result); + } break; + } + } + + void on_finish_multitask(const server_task_multi &multitask) { + // all subtasks done == multitask is done + server_task_result result; + result.id = multitask.id; + result.stop = true; + result.error = false; + + // collect json results into one json result + std::vector result_jsons; + for (const auto &subres : multitask.results) { + result_jsons.push_back(subres.data); + result.error = result.error && subres.error; + } + result.data = json{{"results", result_jsons}}; + + queue_results.send(result); + } + + void update_slots() { + if (system_need_update) { + system_prompt_update(); + } + + // release slots + for (auto &slot : slots) { + if (slot.command == SLOT_COMMAND_RELEASE) { + slot.state = SLOT_STATE_IDLE; + slot.command = SLOT_COMMAND_NONE; + slot.t_last_used = ggml_time_us(); + + LOG_INFO("slot released", {{"id_slot", slot.id}, + {"id_task", slot.id_task}, + {"n_ctx", n_ctx}, + {"n_past", slot.n_past}, + {"n_system_tokens", system_tokens.size()}, + {"n_cache_tokens", slot.cache_tokens.size()}, + {"truncated", slot.truncated}}); + + queue_tasks.notify_slot_changed(); + } + } + + // check if all slots are idle + { + bool all_idle = true; + + for (auto &slot : slots) { + if (slot.state != SLOT_STATE_IDLE || slot.command != SLOT_COMMAND_NONE) { + all_idle = false; + break; + } + } + + if (all_idle) { + LOG_INFO("all slots are idle", {}); + if (system_prompt.empty() && clean_kv_cache) { + kv_cache_clear(); + } + + return; + } + } + + { + server_task task; + task.type = SERVER_TASK_TYPE_NEXT_RESPONSE; + task.id_target = -1; + + queue_tasks.post(task); + } + + // apply context-shift if needed + // TODO: simplify and improve + for (server_slot &slot : slots) { + if (slot.ga_n == 1) { + if (slot.is_processing() && + (int)system_tokens.size() + slot.n_past >= slot.n_ctx - 1) { + // Shift context + const int n_keep = slot.params.n_keep + add_bos_token; + const int n_left = (int)system_tokens.size() + slot.n_past - n_keep; + const int n_discard = + slot.params.n_discard ? slot.params.n_discard : (n_left / 2); + + LOG_INFO("slot context shift", {{"id_slot", slot.id}, + {"id_task", slot.id_task}, + {"n_keep", n_keep}, + {"n_left", n_left}, + {"n_discard", n_discard}, + {"n_ctx", n_ctx}, + {"n_past", slot.n_past}, + {"n_system_tokens", system_tokens.size()}, + {"n_cache_tokens", slot.cache_tokens.size()}}); + + llama_kv_cache_seq_rm(ctx, slot.id + 1, n_keep, n_keep + n_discard); + llama_kv_cache_seq_add(ctx, slot.id + 1, n_keep + n_discard, + system_tokens.size() + slot.n_past, -n_discard); + + if (slot.params.cache_prompt) { + for (size_t i = n_keep + n_discard; i < slot.cache_tokens.size(); i++) { + slot.cache_tokens[i - n_discard] = slot.cache_tokens[i]; + } + + slot.cache_tokens.resize(slot.cache_tokens.size() - n_discard); + } + + slot.n_past -= n_discard; + + slot.truncated = true; + } + } + } + + // start populating the batch for this iteration + llama_batch_clear(batch); + + // frist, add sampled tokens from any ongoing sequences + for (auto &slot : slots) { + if (slot.state == SLOT_STATE_IDLE) { + continue; + } + + if (slot.token_bkt && !slot.token_bkt->acquire()) { + continue; + } + + slot.i_batch = batch.n_tokens; + + const int32_t slot_npast = slot.n_past_se > 0 ? slot.n_past_se : slot.n_past; + + // TODO: we always have to take into account the "system_tokens" + // this is not great and needs to be improved somehow + llama_batch_add(batch, slot.sampled, system_tokens.size() + slot_npast, {slot.id + 1}, + true); + + slot.n_past += 1; + + if (slot.params.cache_prompt) { + slot.cache_tokens.push_back(slot.sampled); + } + } + + // process in chunks of params.n_batch + int32_t n_batch = llama_n_batch(ctx); + int32_t n_ubatch = llama_n_ubatch(ctx); + + // next, batch any pending prompts without exceeding n_batch + if (params.cont_batching || batch.n_tokens == 0) { + for (auto &slot : slots) { + // this slot still has a prompt to be processed + if (slot.state == SLOT_STATE_IDLE && slot.command == SLOT_COMMAND_LOAD_PROMPT) { + auto &prompt_tokens = slot.prompt_tokens; + + // we haven't tokenized the prompt yet - do it now: + if (prompt_tokens.empty()) { + slot.t_start_process_prompt = ggml_time_us(); + slot.t_start_generation = 0; + + if (slot.infill) { + bool suff_rm_leading_spc = true; + if (params.input_suffix.find_first_of(' ') == 0 && + params.input_suffix.size() > 1) { + params.input_suffix.erase(0, 1); + suff_rm_leading_spc = false; + } + + auto prefix_tokens = tokenize(slot.params.input_prefix, false); + auto suffix_tokens = tokenize(slot.params.input_suffix, false); + + const int space_token = 29871; // TODO: this should not be hardcoded + if (suff_rm_leading_spc && !suffix_tokens.empty() && + suffix_tokens[0] == space_token) { + suffix_tokens.erase(suffix_tokens.begin()); + } + + prefix_tokens.insert(prefix_tokens.begin(), llama_token_prefix(model)); + prefix_tokens.insert(prefix_tokens.begin(), + llama_token_bos(model)); // always add BOS + prefix_tokens.insert(prefix_tokens.end(), llama_token_suffix(model)); + prefix_tokens.insert(prefix_tokens.end(), suffix_tokens.begin(), + suffix_tokens.end()); + + const llama_token middle_token = llama_token_middle(model); + if (middle_token >= 0) { + prefix_tokens.push_back(middle_token); + } + + prompt_tokens = prefix_tokens; + } else { + prompt_tokens = tokenize(slot.prompt, + system_prompt.empty()); // add BOS if there + // isn't system prompt + } + + slot.n_past = 0; + slot.n_prompt_tokens = prompt_tokens.size(); + + // empty prompt passed -> release the slot and send + // empty response + if (prompt_tokens.empty()) { + LOG_INFO("empty prompt - releasing slot", + {{"id_slot", slot.id}, {"id_task", slot.id_task}}); + + slot.state = SLOT_STATE_PROCESSING; + slot.command = SLOT_COMMAND_NONE; + slot.release(); + send_final_response(slot); + continue; + } + + if (slot.embedding) { + // this prompt is too large to process - discard it + if (slot.n_prompt_tokens > n_ubatch) { + slot.state = SLOT_STATE_PROCESSING; + slot.command = SLOT_COMMAND_NONE; + slot.release(); + send_error(slot, + "input is too large to process. " + "increase the physical " + "batch size", + ERROR_TYPE_SERVER); + continue; + } + } else { + if (slot.params.n_keep < 0) { + slot.params.n_keep = slot.n_prompt_tokens; + } + slot.params.n_keep = std::min(slot.n_ctx - 4, slot.params.n_keep); + + // if input prompt is too big, truncate it (if group + // attention self-extend is disabled) + if (slot.ga_n == 1 && slot.n_prompt_tokens >= slot.n_ctx) { + const int n_left = slot.n_ctx - slot.params.n_keep; + + const int n_block_size = n_left / 2; + const int erased_blocks = + (slot.n_prompt_tokens - slot.params.n_keep - n_block_size) / + n_block_size; + + std::vector new_tokens(prompt_tokens.begin(), + prompt_tokens.begin() + + slot.params.n_keep); + + new_tokens.insert(new_tokens.end(), + prompt_tokens.begin() + slot.params.n_keep + + erased_blocks * n_block_size, + prompt_tokens.end()); + + prompt_tokens = std::move(new_tokens); + + slot.truncated = true; + slot.n_prompt_tokens = prompt_tokens.size(); + + GGML_ASSERT(slot.n_prompt_tokens < slot.n_ctx); + } + + llama_sampling_reset(slot.ctx_sampling); + + if (!slot.params.cache_prompt) { + slot.n_past_se = 0; + slot.ga_i = 0; + } else { + GGML_ASSERT(slot.ga_n == 1); + + // reuse any previously computed tokens that are + // common with the new prompt + slot.n_past = common_part(slot.cache_tokens, prompt_tokens); + + // push the prompt into the sampling context (do + // not apply grammar) + for (int i = 0; i < slot.n_past; ++i) { + llama_sampling_accept(slot.ctx_sampling, ctx, + slot.cache_tokens[i], false); + } + } + } + + if (slot.n_past == slot.n_prompt_tokens && slot.n_past > 0) { + // we have to evaluate at least 1 token to generate + // logits. + LOG_INFO("we have to evaluate at least 1 token to " + "generate logits", + {{"id_slot", slot.id}, {"id_task", slot.id_task}}); + + slot.n_past--; + if (slot.ga_i > 0) { + slot.n_past_se--; + } + } + + slot.n_prompt_tokens_processed = 0; + } + + if (slot.embedding) { + // cannot fit the prompt in the current batch - will try + // next iter + if (batch.n_tokens + slot.n_prompt_tokens > n_batch) { + continue; + } + } + + // keep only the common part + int p0 = (int)system_tokens.size() + slot.n_past; + if (!llama_kv_cache_seq_rm(ctx, slot.id + 1, p0, -1)) { + // could not partially delete (likely using a + // non-Transformer model) + llama_kv_cache_seq_rm(ctx, slot.id + 1, -1, -1); + + p0 = (int)system_tokens.size(); + if (p0 != 0) { + // copy over the system prompt when there is one + llama_kv_cache_seq_cp(ctx, 0, slot.id + 1, -1, -1); + } + + // there is no common part left (except for the system + // prompt) + slot.n_past = 0; + slot.n_past_se = 0; + slot.ga_i = 0; + // TODO: is the system prompt ever in the sampling + // context? + llama_sampling_reset(slot.ctx_sampling); + } + + // remove the non-common part from the cache + slot.cache_tokens.resize(slot.n_past); + + LOG_INFO("kv cache rm [p0, end)", + {{"id_slot", slot.id}, {"id_task", slot.id_task}, {"p0", p0}}); + + int32_t slot_npast = slot.n_past_se > 0 ? slot.n_past_se : slot.n_past; + + int32_t ga_i = slot.ga_i; + int32_t ga_n = slot.ga_n; + int32_t ga_w = slot.ga_w; + + // add prompt tokens for processing in the current batch + // TODO: the self-extend stuff here is a mess - simplify + // and/or abstract it somehow + for (; slot.n_past < slot.n_prompt_tokens && batch.n_tokens < n_batch; + ++slot.n_past) { + if (slot.ga_n != 1) { + while (slot_npast >= ga_i + ga_w) { + const int bd = (ga_w / ga_n) * (ga_n - 1); + slot_npast -= bd; + ga_i += ga_w / ga_n; + } + } + + llama_batch_add(batch, prompt_tokens[slot.n_past], + system_tokens.size() + slot_npast, {slot.id + 1}, false); + + if (slot.params.cache_prompt) { + slot.cache_tokens.push_back(prompt_tokens[slot.n_past]); + } + + slot.n_prompt_tokens_processed++; + slot_npast++; + } + + // entire prompt has been processed - start decoding new + // tokens + if (slot.n_past == slot.n_prompt_tokens) { + slot.state = SLOT_STATE_PROCESSING; + slot.command = SLOT_COMMAND_NONE; + + GGML_ASSERT(batch.n_tokens > 0); + + // extract the logits only for the last token + batch.logits[batch.n_tokens - 1] = true; + + slot.n_decoded = 0; + slot.i_batch = batch.n_tokens - 1; + } + } + + if (batch.n_tokens >= n_batch) { + break; + } + } + } + + if (batch.n_tokens == 0) { + return; + } + + // process the created batch of tokens + for (int32_t i = 0; i < batch.n_tokens; i += n_batch) { + const int32_t n_tokens = std::min(n_batch, batch.n_tokens - i); + + for (auto &slot : slots) { + if (slot.ga_n != 1) { + // context extension via Self-Extend + // TODO: simplify and/or abstract this + while (slot.n_past_se >= slot.ga_i + slot.ga_w) { + const int ib = (slot.ga_n * slot.ga_i) / slot.ga_w; + const int bd = (slot.ga_w / slot.ga_n) * (slot.ga_n - 1); + const int dd = (slot.ga_w / slot.ga_n) - ib * bd - slot.ga_w; + + LOG_TEE("\n"); + LOG_TEE("shift: [%6d, %6d] + %6d -> [%6d, %6d]\n", slot.ga_i, + slot.n_past_se, ib * bd, slot.ga_i + ib * bd, + slot.n_past_se + ib * bd); + LOG_TEE("div: [%6d, %6d] / %6d -> [%6d, %6d]\n", slot.ga_i + ib * bd, + slot.ga_i + ib * bd + slot.ga_w, slot.ga_n, + (slot.ga_i + ib * bd) / slot.ga_n, + (slot.ga_i + ib * bd + slot.ga_w) / slot.ga_n); + LOG_TEE("shift: [%6d, %6d] + %6d -> [%6d, %6d]\n", + slot.ga_i + ib * bd + slot.ga_w, slot.n_past_se + ib * bd, dd, + slot.ga_i + ib * bd + slot.ga_w + dd, + slot.n_past_se + ib * bd + dd); + + llama_kv_cache_seq_add(ctx, slot.id + 1, slot.ga_i, slot.n_past_se, + ib * bd); + llama_kv_cache_seq_div(ctx, slot.id + 1, slot.ga_i + ib * bd, + slot.ga_i + ib * bd + slot.ga_w, slot.ga_n); + llama_kv_cache_seq_add(ctx, slot.id + 1, slot.ga_i + ib * bd + slot.ga_w, + slot.n_past_se + ib * bd, dd); + + slot.n_past_se -= bd; + + slot.ga_i += slot.ga_w / slot.ga_n; + + LOG_TEE("\nn_past_old = %d, n_past = %d, ga_i = %d\n\n", + slot.n_past_se + bd, slot.n_past_se, slot.ga_i); + } + + slot.n_past_se += n_tokens; + } + } + + llama_batch batch_view = { + n_tokens, + batch.token + i, + nullptr, + batch.pos + i, + batch.n_seq_id + i, + batch.seq_id + i, + batch.logits + i, + 0, + 0, + 0, // unused + }; + + const int ret = llama_decode(ctx, batch_view); + + if (ret != 0) { + if (n_batch == 1 || ret < 0) { + // if you get here, it means the KV cache is full - try + // increasing it via the context size + LOG_ERROR("failed to decode the batch: KV cache is full - " + "try increasing it " + "via the context size", + { + {"i", i}, + {"n_batch", ret}, + {"ret", ret}, + }); + for (auto &slot : slots) { + slot.state = SLOT_STATE_PROCESSING; + slot.command = SLOT_COMMAND_NONE; + slot.release(); + send_error(slot, "Input prompt is too big compared to " + "KV size. Please try " + "increasing KV size."); + } + break; // break loop of n_batch + } + + // retry with half the batch size to try to find a free slot in + // the KV cache + n_batch /= 2; + i -= n_batch; + + LOG_WARNING("failed to find free space in the KV cache, " + "retrying with smaller batch size - " + "try increasing it via the context size or enable " + "defragmentation", + { + {"i", i}, + {"n_batch", n_batch}, + {"ret", ret}, + }); + + continue; // continue loop of n_batch + } + + for (auto &slot : slots) { + if (slot.state != SLOT_STATE_PROCESSING || slot.i_batch < (int)i || + slot.i_batch >= (int)(i + n_tokens)) { + continue; // continue loop of slots + } + + // prompt evaluated for embedding + if (slot.embedding) { + send_embedding(slot, batch_view); + slot.release(); + slot.i_batch = -1; + continue; // continue loop of slots + } + + completion_token_output result; + const llama_token id = + llama_sampling_sample(slot.ctx_sampling, ctx, NULL, slot.i_batch - i); + + llama_sampling_accept(slot.ctx_sampling, ctx, id, true); + + slot.n_decoded += 1; + if (slot.n_decoded == 1) { + slot.t_start_generation = ggml_time_us(); + slot.t_prompt_processing = + (slot.t_start_generation - slot.t_start_process_prompt) / 1e3; + metrics.on_prompt_eval(slot); + } + + llama_token_data_array cur_p = {slot.ctx_sampling->cur.data(), + slot.ctx_sampling->cur.size(), false}; + result.tok = id; + + const size_t n_probs = std::min(cur_p.size, (size_t)slot.sparams.n_probs); + if (n_probs > 0) { + const size_t n_valid = slot.ctx_sampling->n_valid; + + // Make sure at least n_probs top tokens are at the front of + // the vector: + if (slot.sparams.temp == 0.0f && n_probs > n_valid) { + llama_sample_top_k(ctx, &cur_p, n_probs, 0); + } + + if (slot.sparams.temp == 0.0f) { + // With greedy sampling the probabilities have possibly + // not been calculated. + for (size_t i = 0; i < n_probs; ++i) { + result.probs.push_back({cur_p.data[i].id, i == 0 ? 1.0f : 0.0f}); + } + } else { + for (size_t i = 0; i < n_probs; ++i) { + result.probs.push_back({ + cur_p.data[i].id, + i >= n_valid ? 0.0f + : cur_p.data[i].p // Tokens filtered out due to e.g. + // top_k have 0 probability. + }); + } + } + } + + if (!process_token(result, slot)) { + slot.release(); + send_final_response(slot); + metrics.on_prediction(slot); + } + + slot.i_batch = -1; + } + } + } + + json model_meta() const { + return json{ + {"vocab_type", llama_vocab_type(model)}, {"n_vocab", llama_n_vocab(model)}, + {"n_ctx_train", llama_n_ctx_train(model)}, {"n_embd", llama_n_embd(model)}, + {"n_params", llama_model_n_params(model)}, {"size", llama_model_size(model)}, + }; + } +}; + +static void log_server_request(const httplib::Request &req, const httplib::Response &res) { + // skip GH copilot requests when using default port + if (req.path == "/v1/health" || req.path == "/v1/completions") { + return; + } + + LOG_INFO("request", { + {"remote_addr", req.remote_addr}, + {"remote_port", req.remote_port}, + {"status", res.status}, + {"method", req.method}, + {"path", req.path}, + {"params", req.params}, + }); +} + +std::function shutdown_handler; +std::atomic_flag is_terminating = ATOMIC_FLAG_INIT; + +inline void signal_handler(int signal) { + if (is_terminating.test_and_set()) { + // in case it hangs, we can force terminate the server by hitting Ctrl+C + // twice this is for better developer experience, we can remove when the + // server is stable enough + fprintf(stderr, "Received second interrupt, terminating immediately.\n"); + exit(1); + } + + shutdown_handler(signal); +} + +int main(int argc, char **argv) { + log_set_target(stderr); + + llama_box_params bparams; + if (!llama_box_params_parse(argc, argv, bparams)) { + llama_box_params_print_usage(argc, argv, bparams); + return 1; + } + gpt_params ¶ms = bparams.gparams; + + server_log_json = params.log_json; + + server_context ctx_server; + if (!params.system_prompt.empty()) { + ctx_server.system_prompt_set(params.system_prompt); + } + + if (params.model_alias == "unknown") { + params.model_alias = params.model; + } + + llama_backend_init(); + llama_numa_init(params.numa); + + LOG_INFO("build info", {{"build", LLAMA_BUILD_NUMBER}, {"commit", LLAMA_COMMIT}}); + LOG_INFO("system info", { + {"n_threads", params.n_threads}, + {"n_threads_batch", params.n_threads_batch}, + {"total_threads", std::thread::hardware_concurrency()}, + {"system_info", llama_print_system_info()}, + }); + + httplib::Server svr; + std::atomic state{SERVER_STATE_LOADING_MODEL}; + + // default headers + svr.set_default_headers({{"Server", "llama-box"}}); + + // CORS preflight + svr.Options(R"(.*)", [](const httplib::Request &req, httplib::Response &res) { + res.set_header("Access-Control-Allow-Origin", req.get_header_value("Origin")); + res.set_header("Access-Control-Allow-Methods", "POST"); + res.set_header("Access-Control-Allow-Headers", "*"); + return res.set_content("", "application/json; charset=utf-8"); + }); + + // logger + svr.set_logger(log_server_request); + + // error handlers + auto res_error = [](httplib::Response &res, json error_data) { + json final_response{{"error", error_data}}; + res.set_content(final_response.dump(), "application/json; charset=utf-8"); + res.status = json_value(error_data, "code", httplib::StatusCode::InternalServerError_500); + }; + svr.set_exception_handler( + [&res_error](const httplib::Request &, httplib::Response &res, std::exception_ptr ep) { + std::string message; + try { + std::rethrow_exception(std::move(ep)); + } catch (std::exception &e) { + message = e.what(); + } catch (...) { + message = "Unknown Exception"; + } + + json formatted_error = format_error_response(message, ERROR_TYPE_SERVER); + LOG_ERROR("Got exception", formatted_error); + res_error(res, formatted_error); + }); + svr.set_error_handler([&res_error](const httplib::Request &, httplib::Response &res) { + if (res.status == 404) { + res_error(res, format_error_response("Not Found", ERROR_TYPE_NOT_FOUND)); + } + // for other error codes, we skip processing here because it's + // already done by res_error() + }); + + // configure and bind + svr.set_read_timeout(params.timeout_read); + svr.set_write_timeout(params.timeout_write); + svr.set_payload_max_length(1024 * 1024 * 10); + svr.set_idle_interval(bparams.conn_idle); + svr.set_keep_alive_timeout(bparams.conn_keepalive); + if (!svr.bind_to_port(params.hostname, params.port)) { + LOG_ERROR("couldn't bind to server socket", + {{"hostname", params.hostname}, {"port", params.port}}); + return 1; + } + + std::unordered_map log_data; + + log_data["hostname"] = params.hostname; + log_data["port"] = std::to_string(params.port); + + // necessary similarity of prompt for slot selection + ctx_server.slot_prompt_similarity = params.slot_prompt_similarity; + + // load the model + if (!ctx_server.load_model(bparams)) { + return 1; + } + ctx_server.init(); + + LOG_INFO("model loaded", {}); + + // if a custom chat template is not supplied, we will use the one that comes + // with the model (if any) + if (params.chat_template.empty()) { + params.chat_template = ctx_server.load_chat_template(); + } + LOG_INFO("chat template", {{"template", params.chat_template}}); + + // + // Handlers + // + + const auto handle_health = [&](const httplib::Request &req, httplib::Response &res) { + server_state current_state = state.load(); + switch (current_state) { + case SERVER_STATE_READY: { + // request slots data using task queue + server_task task; + task.id_target = -1; + task.type = SERVER_TASK_TYPE_METRICS; + + // post the task + task.id = ctx_server.queue_tasks.post(task); + ctx_server.queue_results.add_waiting_task_id(task.id); + + // get the result + 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.at("idle"); + const int n_processing_slots = result.data.at("processing"); + + json health = {{"status", "ok"}, + {"slots_idle", n_idle_slots}, + {"slots_processing", n_processing_slots}}; + + if (params.endpoint_slots && req.has_param("include_slots")) { + health["slots"] = result.data.at("slots"); + } + + if (n_idle_slots == 0) { + health["status"] = "no slot available"; + if (req.has_param("fail_on_no_slot")) { + res.status = httplib::StatusCode::ServiceUnavailable_503; + } + } + + res.set_content(health.dump(), "application/json"); + break; + } + case SERVER_STATE_LOADING_MODEL: + res_error(res, format_error_response("Loading model", ERROR_TYPE_UNAVAILABLE)); + break; + case SERVER_STATE_ERROR: + res_error(res, format_error_response("Model failed to load", ERROR_TYPE_SERVER)); + break; + } + }; + + const auto handle_metrics = [&](const httplib::Request &, httplib::Response &res) { + // request slots data using task queue + server_task task; + task.id_multi = -1; + task.id_target = -1; + task.type = SERVER_TASK_TYPE_METRICS; + task.data.push_back({{"reset_bucket", true}}); + + // post the task + task.id = ctx_server.queue_tasks.post(task); + ctx_server.queue_results.add_waiting_task_id(task.id); + + // get the result + server_task_result result = ctx_server.queue_results.recv(task.id); + ctx_server.queue_results.remove_waiting_task_id(task.id); + + std::stringstream metrics; + { + json data = result.data; + uint64_t n_prompt_tokens_processed_total = data.at("n_prompt_tokens_processed_total"); + uint64_t t_prompt_processing_total = data.at("t_prompt_processing_total"); + uint64_t n_tokens_predicted_total = data.at("n_tokens_predicted_total"); + uint64_t t_tokens_generation_total = data.at("t_tokens_generation_total"); + uint64_t n_prompt_tokens_processed = data.at("n_prompt_tokens_processed"); + uint64_t t_prompt_processing = data.at("t_prompt_processing"); + uint64_t n_tokens_predicted = data.at("n_tokens_predicted"); + uint64_t t_tokens_generation = data.at("t_tokens_generation"); + int32_t kv_cache_used_cells = data.at("kv_cache_used_cells"); + uint64_t kv_cache_tokens_count = data.at("kv_cache_tokens_count"); + uint64_t processing = data.at("processing"); + uint64_t deferred = data.at("deferred"); + + // 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", n_prompt_tokens_processed_total}}, + {{"name", "prompt_seconds_total"}, + {"help", "Prompt process time"}, + {"value", t_prompt_processing_total / 1.e3}}, + {{"name", "tokens_predicted_total"}, + {"help", "Number of generation tokens processed."}, + {"value", n_tokens_predicted_total}}, + {{"name", "tokens_predicted_seconds_total"}, + {"help", "Predict process time"}, + {"value", t_tokens_generation_total / 1.e3}}}}, + {"gauge", + {{{"name", "prompt_tokens_seconds"}, + {"help", "Average prompt throughput in tokens/s."}, + {"value", n_prompt_tokens_processed + ? 1.e3 / t_prompt_processing * n_prompt_tokens_processed + : 0.}}, + {{"name", "predicted_tokens_seconds"}, + {"help", "Average generation throughput in tokens/s."}, + {"value", + n_tokens_predicted ? 1.e3 / t_tokens_generation * n_tokens_predicted : 0.}}, + {{"name", "kv_cache_usage_ratio"}, + {"help", "KV-cache usage. 1 means 100 percent usage."}, + {"value", 1. * kv_cache_used_cells / params.n_ctx}}, + {{"name", "kv_cache_tokens"}, + {"help", "KV-cache tokens."}, + {"value", kv_cache_tokens_count}}, + {{"name", "requests_processing"}, + {"help", "Number of request processing."}, + {"value", processing}}, + {{"name", "requests_deferred"}, + {"help", "Number of request deferred."}, + {"value", deferred}}}}}; + + for (const auto &el : all_metrics_def.items()) { + const auto &type = el.key(); + const auto &metrics_def = el.value(); + + for (const auto &metric_def : metrics_def) { + const std::string name = metric_def.at("name"); + const std::string help = metric_def.at("help"); + + auto value = json_value(metric_def, "value", 0.); + metrics << "# HELP llamacpp:" << name << " " << help << "\n" + << "# TYPE llamacpp:" << name << " " << type << "\n" + << "llamacpp:" << name << " " << value << "\n"; + } + } + } + res.set_content(metrics.str(), "text/plain; version=0.0.4"); + }; + + const auto handle_props = [&ctx_server](const httplib::Request &, httplib::Response &res) { + json props = { + {"system_prompt", ctx_server.system_prompt.c_str()}, + {"default_generation_settings", ctx_server.default_generation_settings_for_props}, + {"total_slots", ctx_server.params.n_parallel}}; + + res.set_content(props.dump(), "application/json; charset=utf-8"); + }; + + const auto handle_infill = [&ctx_server, &res_error](const httplib::Request &req, + httplib::Response &res) { + int tps = 0; + { + const std::string tps_s = req.get_header_value("X-Request-Tokens-Per-Second"); + if (!tps_s.empty()) { + try { + tps = std::stoi(tps_s); + } catch (const std::exception &) { + tps = ctx_server.n_tps; + } + } + if (tps > ctx_server.n_tps) { + // if the request exceeds the maximum tokens per second, return + // 410 Gone + if (ctx_server.n_tps > 0) { + res.status = httplib::StatusCode::Gone_410; + res.set_content("This request exceeds the maximum tokens per second", + "text/plain; charset=utf-8"); + return; + } + // if the server is not limited by tokens per second, set tps to + // 0 + tps = 0; + } + } + + const json request = json::parse(req.body); + + // post the task + const int id_task = ctx_server.queue_tasks.get_new_id(); + ctx_server.queue_results.add_waiting_task_id(id_task); + ctx_server.request_completion(id_task, -1, request, true, false, tps); + + // process non-streaming requests + if (!json_value(request, "stream", false)) { + server_task_result result = ctx_server.queue_results.recv(id_task); + if (result.error || !result.stop) { + res_error(res, result.data); + } else { + res.set_header("X-Response-Tokens-Per-Second", + std::to_string(json_value(result.data.at("timings"), + "predicted_per_second", tps))); + const std::string infill = + result.data.dump(-1, ' ', false, json::error_handler_t::replace); + res.set_content(infill, "application/json; charset=utf-8"); + } + + ctx_server.queue_results.remove_waiting_task_id(id_task); + return; + } + + // process streaming requests + const auto on_chunk = [id_task, &ctx_server, tps](size_t, httplib::DataSink &sink) { + while (true) { + server_task_result result = ctx_server.queue_results.recv(id_task); + if (result.error) { + const std::string str = "error: failed to infill\n\n"; + sink.write(str.c_str(), str.size()); + sink.done(); + return false; + } + + const std::string infill = + "data: " + result.data.dump(-1, ' ', false, json::error_handler_t::replace) + + "\n\n"; + if (!sink.write(infill.c_str(), infill.size())) { + sink.done(); + return false; + } + + if (!result.stop) { + continue; + } + + sink.done_with_trailer({{"X-Response-Tokens-Per-Second", + std::to_string(json_value(result.data.at("timings"), + "predicted_per_second", tps))}}); + return true; + } + }; + const auto on_complete = [id_task, &ctx_server](bool) { + ctx_server.request_cancel(id_task); + ctx_server.queue_results.remove_waiting_task_id(id_task); + }; + + res.set_header("Trailer", "X-Response-Tokens-Per-Second"); + res.set_chunked_content_provider("text/event-stream", on_chunk, on_complete); + }; + + const auto handle_tokenize = [&ctx_server](const httplib::Request &req, + httplib::Response &res) { + const json request = json::parse(req.body); + + std::vector tokens; + if (request.count("content") != 0) { + const bool add_special = json_value(request, "add_special", false); + tokens = ctx_server.tokenize(request.at("content"), add_special); + } + + const json response = json{{"tokens", tokens}}; + return res.set_content(response.dump(), "application/json; charset=utf-8"); + }; + + const auto handle_detokenize = [&ctx_server](const httplib::Request &req, + httplib::Response &res) { + const json request = json::parse(req.body); + + std::string content; + if (request.count("tokens") != 0) { + const std::vector tokens = request.at("tokens"); + content = llama_detokenize_bpe(ctx_server.ctx, tokens); + } + + const json response = json{{"content", content}}; + return res.set_content(response.dump(), "application/json; charset=utf-8"); + }; + + const auto handle_slots = [&](const httplib::Request &, httplib::Response &res) { + // request slots data using task queue + server_task task; + task.id_multi = -1; + task.id_target = -1; + task.type = SERVER_TASK_TYPE_METRICS; + + // post the task + task.id = ctx_server.queue_tasks.post(task); + ctx_server.queue_results.add_waiting_task_id(task.id); + + // get the result + 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.at("slots").dump(), "application/json"); + }; + + const auto handle_slots_save = [&ctx_server, &res_error, ¶ms](const httplib::Request &req, + httplib::Response &res, + int id_slot) { + json request = json::parse(req.body); + + std::string filename = request.at("filename"); + if (!fs_validate_filename(filename)) { + res_error(res, format_error_response("Invalid filename", ERROR_TYPE_INVALID_REQUEST)); + return; + } + std::string filepath = params.slot_save_path + filename; + + server_task task; + task.type = SERVER_TASK_TYPE_SLOT_SAVE; + task.data = {{"id_slot", id_slot}, {"filename", filename}, {"filepath", filepath}}; + + // post the task + task.id = ctx_server.queue_tasks.post(task); + ctx_server.queue_results.add_waiting_task_id(task.id); + + // get the result + server_task_result result = ctx_server.queue_results.recv(task.id); + ctx_server.queue_results.remove_waiting_task_id(task.id); + + if (result.error) { + res_error(res, result.data); + return; + } + + res.set_content(result.data.dump(), "application/json"); + }; + + const auto handle_slots_restore = [&ctx_server, &res_error, + ¶ms](const httplib::Request &req, httplib::Response &res, + int id_slot) { + json request = json::parse(req.body); + + std::string filename = request.at("filename"); + if (!fs_validate_filename(filename)) { + res_error(res, format_error_response("Invalid filename", ERROR_TYPE_INVALID_REQUEST)); + return; + } + std::string filepath = params.slot_save_path + filename; + + server_task task; + task.type = SERVER_TASK_TYPE_SLOT_RESTORE; + task.data = {{"id_slot", id_slot}, {"filename", filename}, {"filepath", filepath}}; + + // post the task + task.id = ctx_server.queue_tasks.post(task); + ctx_server.queue_results.add_waiting_task_id(task.id); + + // get the result + server_task_result result = ctx_server.queue_results.recv(task.id); + ctx_server.queue_results.remove_waiting_task_id(task.id); + + if (result.error) { + res_error(res, result.data); + return; + } + + res.set_content(result.data.dump(), "application/json"); + }; + + const auto handle_slots_erase = [&ctx_server, &res_error](const httplib::Request &, + httplib::Response &res, int id_slot) { + server_task task; + task.type = SERVER_TASK_TYPE_SLOT_ERASE; + task.data = {{"id_slot", id_slot}}; + + // post the task + task.id = ctx_server.queue_tasks.post(task); + ctx_server.queue_results.add_waiting_task_id(task.id); + + // get the result + server_task_result result = ctx_server.queue_results.recv(task.id); + ctx_server.queue_results.remove_waiting_task_id(task.id); + + if (result.error) { + res_error(res, result.data); + return; + } + + res.set_content(result.data.dump(), "application/json"); + }; + + const auto handle_slots_action = [&res_error, &handle_slots_save, &handle_slots_restore, + &handle_slots_erase](const httplib::Request &req, + httplib::Response &res) { + int id_slot = -1; + { + const std::string id_slot_str = req.path_params.at("id_slot"); + if (!id_slot_str.empty()) { + try { + id_slot = std::stoi(id_slot_str); + } catch (const std::exception &) { + id_slot = -1; + } + } + if (id_slot < 0) { + res_error(res, + format_error_response("Invalid slot ID", ERROR_TYPE_INVALID_REQUEST)); + return; + } + } + + // forward + const std::string action = req.get_param_value("action"); + if (action == "save") { + handle_slots_save(req, res, id_slot); + } else if (action == "restore") { + handle_slots_restore(req, res, id_slot); + } else if (action == "erase") { + handle_slots_erase(req, res, id_slot); + } else { + res_error(res, format_error_response("Invalid action", ERROR_TYPE_INVALID_REQUEST)); + } + }; + + const auto handle_completions = [&ctx_server, &res_error](const httplib::Request &req, + httplib::Response &res) { + int tps = 0; + { + const std::string tps_s = req.get_header_value("X-Request-Tokens-Per-Second"); + if (!tps_s.empty()) { + try { + tps = std::stoi(tps_s); + } catch (const std::exception &) { + tps = ctx_server.n_tps; + } + } + if (tps > ctx_server.n_tps) { + // if the request exceeds the maximum tokens per second, return + // 410 Gone + if (ctx_server.n_tps > 0) { + res.status = httplib::StatusCode::Gone_410; + res.set_content("This request exceeds the maximum tokens per second", + "text/plain; charset=utf-8"); + return; + } + // if the server is not limited by tokens per second, set tps to + // 0 + tps = 0; + } + } + + bool oaicompat = req.path.compare("/v1/completions") == 0; + json request = json::parse(req.body); + if (!request.contains("prompt")) { + res_error(res, format_error_response("\"prompt\" must be provided", + ERROR_TYPE_INVALID_REQUEST)); + return; + } + if (oaicompat) { + request = oaicompat_completion_request(ctx_server.model, request, std::string()); + } + + // post the task + const int id_task = ctx_server.queue_tasks.get_new_id(); + ctx_server.queue_results.add_waiting_task_id(id_task); + ctx_server.request_completion(id_task, -1, request, false, false, tps); + + const std::string completion_id = gen_cmplid(); + + // process non-streaming requests + if (!json_value(request, "stream", false)) { + server_task_result result = ctx_server.queue_results.recv(id_task); + if (result.error || !result.stop) { + res_error(res, result.data); + } else { + res.set_header("X-Response-Tokens-Per-Second", + std::to_string(json_value(result.data.at("timings"), + "predicted_per_second", tps))); + + json completions_json = result.data; + if (req.path.compare("/v1/completions") == 0) { + completions_json = + oaicompat_completion_response(request, completions_json, completion_id); + } + const std::string completions = + completions_json.dump(-1, ' ', false, json::error_handler_t::replace); + res.set_content(completions, "application/json; charset=utf-8"); + } + + ctx_server.queue_results.remove_waiting_task_id(id_task); + return; + } + + // process streaming requests + const auto on_chunk = [id_task, &ctx_server, completion_id, oaicompat, request, + tps](size_t, httplib::DataSink &sink) { + while (true) { + server_task_result result = ctx_server.queue_results.recv(id_task); + if (result.error) { + const std::string str = "error: failed to complete\n\n"; + sink.write(str.c_str(), str.size()); + sink.done(); + return false; + } + + json completions_json = result.data; + if (oaicompat) { + completions_json = oaicompat_completion_response(request, completions_json, + completion_id, true); + } + const std::string completions = + "data: " + + completions_json.dump(-1, ' ', false, json::error_handler_t::replace) + "\n\n"; + if (!sink.write(completions.c_str(), completions.size())) { + sink.done(); + return false; + } + + if (!result.stop) { + continue; + } + + sink.done_with_trailer({{"X-Response-Tokens-Per-Second", + std::to_string(json_value(result.data.at("timings"), + "predicted_per_second", tps))}}); + return true; + } + }; + const auto on_complete = [id_task, &ctx_server](bool) { + ctx_server.request_cancel(id_task); + ctx_server.queue_results.remove_waiting_task_id(id_task); + }; + + res.set_header("Trailer", "X-Response-Tokens-Per-Second"); + res.set_chunked_content_provider("text/event-stream", on_chunk, on_complete); + }; + + const auto handle_models = [&ctx_server, ¶ms](const httplib::Request &, + httplib::Response &res) { + json models = {{"object", "list"}, + {"data", + { + {{"id", params.model_alias}, + {"object", "model"}, + {"created", std::time(0)}, + {"owned_by", "llama-box"}, + {"meta", ctx_server.model_meta()}}, + }}}; + + res.set_content(models.dump(), "application/json; charset=utf-8"); + }; + + const auto handle_chat_completions = [&ctx_server, ¶ms, &res_error]( + const httplib::Request &req, httplib::Response &res) { + int tps = 0; + { + const std::string tps_s = req.get_header_value("X-Request-Tokens-Per-Second"); + if (!tps_s.empty()) { + try { + tps = std::stoi(tps_s); + } catch (const std::exception &) { + tps = ctx_server.n_tps; + } + } + if (tps > ctx_server.n_tps) { + // if the request exceeds the maximum tokens per second, return + // 410 Gone + if (ctx_server.n_tps > 0) { + res.status = httplib::StatusCode::Gone_410; + res.set_content("This request exceeds the maximum tokens per second", + "text/plain; charset=utf-8"); + return; + } + // if the server is not limited by tokens per second, set tps to + // 0 + tps = 0; + } + } + + json request = json::parse(req.body); + if (!request.contains("messages")) { + res_error(res, format_error_response("\"messages\" must be provided", + ERROR_TYPE_INVALID_REQUEST)); + return; + } + request = oaicompat_completion_request(ctx_server.model, request, params.chat_template); + + // post the task + const int id_task = ctx_server.queue_tasks.get_new_id(); + ctx_server.queue_results.add_waiting_task_id(id_task); + ctx_server.request_completion(id_task, -1, request, false, false, tps); + + const std::string completion_id = gen_chatcmplid(); + + // process non-streaming requests + if (!json_value(request, "stream", false)) { + server_task_result result = ctx_server.queue_results.recv(id_task); + if (result.error || !result.stop) { + res_error(res, result.data); + } else { + res.set_header("X-Response-Tokens-Per-Second", + std::to_string(json_value(result.data.at("timings"), + "predicted_per_second", tps))); + + const json chats_completion_json = + oaicompat_completion_response(request, result.data, completion_id); + const std::string chats_completion = + chats_completion_json.dump(-1, ' ', false, json::error_handler_t::replace); + res.set_content(chats_completion, "application/json; charset=utf-8"); + } + + ctx_server.queue_results.remove_waiting_task_id(id_task); + return; + } + + // process streaming requests + const auto on_chunk = [id_task, &ctx_server, completion_id, request, + tps](size_t, httplib::DataSink &sink) { + bool first = true; + while (true) { + server_task_result result = ctx_server.queue_results.recv(id_task); + if (result.error) { + const std::string str = "error: failed to chat\n\n"; + sink.write(str.c_str(), str.size()); + sink.done(); + return false; + } + + if (first) { + first = false; + json chat_completions_json = oaicompat_completion_response( + request, result.data, completion_id, true, true); + const std::string chat_completions = + "data: " + + chat_completions_json.dump(-1, ' ', false, json::error_handler_t::replace) + + "\n\n"; + if (!sink.write(chat_completions.c_str(), chat_completions.size())) { + sink.done(); + return false; + } + } + + json chat_completions_json = + oaicompat_completion_response(request, result.data, completion_id, true); + const std::string chat_completions = + "data: " + + chat_completions_json.dump(-1, ' ', false, json::error_handler_t::replace) + + "\n\n"; + if (!sink.write(chat_completions.c_str(), chat_completions.size())) { + sink.done(); + return false; + } + + if (!result.stop) { + continue; + } + + sink.done_with_trailer({{"X-Response-Tokens-Per-Second", + std::to_string(json_value(result.data.at("timings"), + "predicted_per_second", tps))}}); + return true; + } + }; + auto on_complete = [id_task, &ctx_server](bool) { + ctx_server.request_cancel(id_task); + ctx_server.queue_results.remove_waiting_task_id(id_task); + }; + + res.set_header("Trailer", "X-Response-Tokens-Per-Second"); + res.set_chunked_content_provider("text/event-stream", on_chunk, on_complete); + }; + + const auto handle_embeddings = [&ctx_server, &res_error](const httplib::Request &req, + httplib::Response &res) { + json request = json::parse(req.body); + if (!request.contains("input")) { + res_error(res, format_error_response("\"input\" must be provided", + ERROR_TYPE_INVALID_REQUEST)); + return; + } + request = oaicompat_embedding_request(ctx_server.params, request); + + // post the task + const int id_task = ctx_server.queue_tasks.get_new_id(); + ctx_server.queue_results.add_waiting_task_id(id_task); + ctx_server.request_completion(id_task, -1, request, false, true); + + // get the result + server_task_result result = ctx_server.queue_results.recv(id_task); + if (result.error || !result.stop) { + res_error(res, result.data); + } else { + const json embeddings_json = oaicompat_embedding_response(request, result.data); + + const std::string embeddings = embeddings_json.dump(); + return res.set_content(embeddings, "application/json; charset=utf-8"); + } + + ctx_server.queue_results.remove_waiting_task_id(id_task); + }; + + // + // Router + // + + svr.Get("/health", handle_health); + if (params.endpoint_metrics) { + svr.Get("/metrics", handle_metrics); + } + svr.Get("/props", handle_props); + if (params.infill) { + svr.Post("/infill", handle_infill); + } + svr.Post("/tokenize", handle_tokenize); + svr.Post("/detokenize", handle_detokenize); + if (params.endpoint_slots) { + svr.Get("/slots", handle_slots); + if (!params.slot_save_path.empty()) { + // only enable slot endpoints if slot_save_path is set + svr.Post("/slots/:id_slot", handle_slots_action); + } + } + svr.Post("/completion", handle_completions); + svr.Get("/v1/models", handle_models); + svr.Post("/v1/completions", handle_completions); + svr.Post("/v1/chat/completions", handle_chat_completions); + if (params.embedding) { + svr.Post("/v1/embeddings", handle_embeddings); + } + + // + // Middlewares + // + + svr.set_post_routing_handler([](const httplib::Request &req, httplib::Response &res) { + if (req.method == "POST") { + res.set_header("Access-Control-Allow-Origin", req.get_header_value("Origin")); + } + return httplib::Server::HandlerResponse::Handled; + }); + + // + // Start + // + + if (params.n_threads_http < 1) { + // +2 threads for monitoring endpoints: /metrics and /slots + params.n_threads_http = + std::max(params.n_parallel + 2, (int32_t)std::thread::hardware_concurrency() - 1); + } + log_data["n_threads_http"] = std::to_string(params.n_threads_http); + svr.new_task_queue = [¶ms] { return new httplib::ThreadPool(params.n_threads_http); }; + + LOG_INFO("HTTP server listening", log_data); + // run the HTTP server in a thread - see comment below + std::thread t([&]() { + if (!svr.listen_after_bind()) { + state.store(SERVER_STATE_ERROR); + return 1; + } + + return 0; + }); + + ctx_server.queue_tasks.on_new_task( + std::bind(&server_context::process_single_task, &ctx_server, std::placeholders::_1)); + ctx_server.queue_tasks.on_finish_multitask( + std::bind(&server_context::on_finish_multitask, &ctx_server, std::placeholders::_1)); + ctx_server.queue_tasks.on_update_slots(std::bind(&server_context::update_slots, &ctx_server)); + ctx_server.queue_results.on_multitask_update( + std::bind(&server_queue::update_multitask, &ctx_server.queue_tasks, std::placeholders::_1, + std::placeholders::_2, std::placeholders::_3)); + + shutdown_handler = [&](int) { ctx_server.queue_tasks.terminate(); }; + +#if defined(__unix__) || (defined(__APPLE__) && defined(__MACH__)) + struct sigaction sigint_action; + sigint_action.sa_handler = signal_handler; + sigemptyset(&sigint_action.sa_mask); + sigint_action.sa_flags = 0; + sigaction(SIGINT, &sigint_action, NULL); + sigaction(SIGTERM, &sigint_action, NULL); +#elif defined(_WIN32) + auto console_ctrl_handler = +[](DWORD ctrl_type) -> BOOL { + return (ctrl_type == CTRL_C_EVENT) ? (signal_handler(SIGINT), true) : false; + }; + SetConsoleCtrlHandler(reinterpret_cast(console_ctrl_handler), true); +#endif + + ctx_server.queue_tasks.start_loop(); + svr.stop(); + t.join(); + + llama_backend_free(); + return 0; +} diff --git a/llama-box/param.hpp b/llama-box/param.hpp new file mode 100644 index 0000000..bdb97e7 --- /dev/null +++ b/llama-box/param.hpp @@ -0,0 +1,1144 @@ +#pragma once + +#include + +#include "llama.cpp/common/common.h" +#include "llama.cpp/common/grammar-parser.h" +#include "llama.cpp/common/json-schema-to-grammar.h" +#define JSON_ASSERT GGML_ASSERT +#include "llama.cpp/common/json.hpp" +#include "llama.cpp/ggml.h" +#include "llama.cpp/llama.h" + +// version +extern const char *LLAMA_BOX_BUILD_DATE; +extern const char *LLAMA_BOX_GIT_TREE_STATE; +extern const char *LLAMA_BOX_GIT_VERSION; +extern const char *LLAMA_BOX_GIT_COMMIT; + +using json = nlohmann::json; + +struct llama_box_params { + gpt_params gparams; + + int32_t conn_idle = 60; // connection idle in seconds + int32_t conn_keepalive = 15; // connection keep-alive in seconds + int32_t n_tps = 0; // maximum number of tokens per seconds +}; + +static int unknown(const char *flag) { + throw std::invalid_argument("Unknown argument: " + std::string(flag)); + return 1; +} + +static int missing(const char *flag) { + throw std::invalid_argument("Missing argument: " + std::string(flag)); + return 1; +} + +static int invalid(const char *flag) { + throw std::invalid_argument("Invalid argument: " + std::string(flag)); + return 1; +} + +#ifdef __GNUC__ +#ifdef __MINGW32__ +#define LLAMA_COMMON_ATTRIBUTE_FORMAT(...) __attribute__((format(gnu_printf, __VA_ARGS__))) +#else +#define LLAMA_COMMON_ATTRIBUTE_FORMAT(...) __attribute__((format(printf, __VA_ARGS__))) +#endif +#else +#define LLAMA_COMMON_ATTRIBUTE_FORMAT(...) +#endif + +void llama_box_params_print_usage(int, char **argv, const llama_box_params &bparams) { + struct opt { + LLAMA_COMMON_ATTRIBUTE_FORMAT(4, 5) + + opt(const std::string &tags, const char *args, const char *desc, ...) + : tags(tags), args(args), desc(desc) { + va_list args_list; + va_start(args_list, desc); + char buffer[1024]; + vsnprintf(buffer, sizeof(buffer), desc, args_list); + va_end(args_list); + this->desc = buffer; + } + + opt(const std::string &grp) + : grp(grp) { + } + + std::string tags; + std::string args; + std::string desc; + std::string grp; + }; + + const auto ¶ms = bparams.gparams; + const auto &sparams = params.sparams; + std::string sampler_type_chars; + std::string sampler_type_names; + for (const auto sampler_type : sparams.samplers_sequence) { + sampler_type_chars += static_cast(sampler_type); + sampler_type_names += llama_sampling_type_to_str(sampler_type) + ";"; + } + sampler_type_names.pop_back(); + + std::vector opts; + // clang-format off + opts.push_back({ "general" }); + opts.push_back({ "*", "-h, --help, --usage", "print usage and exit" }); + opts.push_back({ "*", " --version", "show version and build info" }); + opts.push_back({ "*", "-m, --model FILE", "model path (default: %s)", DEFAULT_MODEL_PATH }); + opts.push_back({ "*", "-a, --alias NAME", "model name alias (default: %s)", params.model_alias.c_str() }); + opts.push_back({ "*", "-s, --seed N", "RNG seed (default: %d, use random seed for < 0)", params.seed }); + opts.push_back({ "*", "-t, --threads N", "number of threads to use during generation (default: %d)", params.n_threads }); + opts.push_back({ "*", "-tb, --threads-batch N", "number of threads to use during batch and prompt processing (default: same as --threads)" }); + opts.push_back({ "*", "-lcs, --lookup-cache-static FILE", + "path to static lookup cache to use for lookup decoding (not updated by generation)" }); + opts.push_back({ "*", "-lcd, --lookup-cache-dynamic FILE", + "path to dynamic lookup cache to use for lookup decoding (updated by generation)" }); + opts.push_back({ "*", "-c, --ctx-size N", "size of the prompt context (default: %d, 0 = loaded from model)", params.n_ctx }); + opts.push_back({ "*", "-n, --predict N", "number of tokens to predict (default: %d, -1 = infinity, -2 = until context filled)", params.n_predict }); + opts.push_back({ "*", "-b, --batch-size N", "logical maximum batch size (default: %d)", params.n_batch }); + opts.push_back({ "*", "-ub, --ubatch-size N", "physical maximum batch size (default: %d)", params.n_ubatch }); + opts.push_back({ "*", " --keep N", "number of tokens to keep from the initial prompt (default: %d, -1 = all)", params.n_keep }); + opts.push_back({ "*", " --chunks N", "max number of chunks to process (default: %d, -1 = all)", params.n_chunks }); + opts.push_back({ "*", "-fa, --flash-attn", "enable Flash Attention (default: %s)", params.flash_attn ? "enabled" : "disabled" }); + opts.push_back({ "*", " --no-escape", "do not process escape sequences" }); + opts.push_back({ "*", " --samplers SAMPLERS", "samplers that will be used for generation in the order, separated by \';\'\n" + "(default: %s)", sampler_type_names.c_str() }); + opts.push_back({ "*", " --sampling-seq SEQUENCE", + "simplified sequence for samplers that will be used (default: %s)", sampler_type_chars.c_str() }); + opts.push_back({ "*", " --penalize-nl", "penalize newline tokens (default: %s)", sparams.penalize_nl ? "true" : "false" }); + opts.push_back({ "*", " --temp N", "temperature (default: %.1f)", (double)sparams.temp }); + opts.push_back({ "*", " --top-k N", "top-k sampling (default: %d, 0 = disabled)", sparams.top_k }); + opts.push_back({ "*", " --top-p N", "top-p sampling (default: %.1f, 1.0 = disabled)", (double)sparams.top_p }); + opts.push_back({ "*", " --min-p N", "min-p sampling (default: %.1f, 0.0 = disabled)", (double)sparams.min_p }); + opts.push_back({ "*", " --tfs N", "tail free sampling, parameter z (default: %.1f, 1.0 = disabled)", (double)sparams.tfs_z }); + opts.push_back({ "*", " --typical N", "locally typical sampling, parameter p (default: %.1f, 1.0 = disabled)", (double)sparams.typical_p }); + opts.push_back({ "*", " --repeat-last-n N", "last n tokens to consider for penalize (default: %d, 0 = disabled, -1 = ctx_size)", sparams.penalty_last_n }); + opts.push_back({ "*", " --repeat-penalty N", "penalize repeat sequence of tokens (default: %.1f, 1.0 = disabled)", (double)sparams.penalty_repeat }); + opts.push_back({ "*", " --presence-penalty N", "repeat alpha presence penalty (default: %.1f, 0.0 = disabled)", (double)sparams.penalty_present }); + opts.push_back({ "*", " --frequency-penalty N", "repeat alpha frequency penalty (default: %.1f, 0.0 = disabled)", (double)sparams.penalty_freq }); + opts.push_back({ "*", " --dynatemp-range N", "dynamic temperature range (default: %.1f, 0.0 = disabled)", (double)sparams.dynatemp_range }); + opts.push_back({ "*", " --dynatemp-exp N", "dynamic temperature exponent (default: %.1f)", (double)sparams.dynatemp_exponent }); + opts.push_back({ "*", " --mirostat N", "use Mirostat sampling.\n" + "Top K, Nucleus, Tail Free and Locally Typical samplers are ignored if used.\n" + "(default: %d, 0 = disabled, 1 = Mirostat, 2 = Mirostat 2.0)", sparams.mirostat }); + opts.push_back({ "*", " --mirostat-lr N", "Mirostat learning rate, parameter eta (default: %.1f)", (double)sparams.mirostat_eta }); + opts.push_back({ "*", " --mirostat-ent N", "Mirostat target entropy, parameter tau (default: %.1f)", (double)sparams.mirostat_tau }); + opts.push_back({ "*", "-l --logit-bias TOKEN_ID(+/-)BIAS", + "modifies the likelihood of token appearing in the completion,\n" + "i.e. `--logit-bias 15043+1` to increase likelihood of token ' Hello',\n" + "or `--logit-bias 15043-1` to decrease likelihood of token ' Hello'" }); + opts.push_back({ "*", " --grammar GRAMMAR", "BNF-like grammar to constrain generations (see samples in grammars/ dir) (default: '%s')", sparams.grammar.c_str() }); + opts.push_back({ "*", " --grammar-file FILE", "file to read grammar from" }); + opts.push_back({ "*", "-j, --json-schema SCHEMA", + "JSON schema to constrain generations (https://json-schema.org/), e.g. `{}` for any JSON object\n" + "For schemas w/ external $refs, use --grammar + example/json_schema_to_grammar.py instead" }); + opts.push_back({ "*", " --rope-scaling {none,linear,yarn}", + "RoPE frequency scaling method, defaults to linear unless specified by the model" }); + opts.push_back({ "*", " --rope-scale N", "RoPE context scaling factor, expands context by a factor of N" }); + opts.push_back({ "*", " --rope-freq-base N", "RoPE base frequency, used by NTK-aware scaling (default: loaded from model)" }); + opts.push_back({ "*", " --rope-freq-scale N", "RoPE frequency scaling factor, expands context by a factor of 1/N" }); + opts.push_back({ "*", " --yarn-orig-ctx N", "YaRN: original context size of model (default: %d = model training context size)", params.yarn_orig_ctx }); + opts.push_back({ "*", " --yarn-ext-factor N", "YaRN: extrapolation mix factor (default: %.1f, 0.0 = full interpolation)", (double)params.yarn_ext_factor }); + opts.push_back({ "*", " --yarn-attn-factor N", "YaRN: scale sqrt(t) or attention magnitude (default: %.1f)", (double)params.yarn_attn_factor }); + opts.push_back({ "*", " --yarn-beta-fast N", "YaRN: low correction dim or beta (default: %.1f)", (double)params.yarn_beta_fast }); + opts.push_back({ "*", " --yarn-beta-slow N", "YaRN: high correction dim or alpha (default: %.1f)", (double)params.yarn_beta_slow }); + opts.push_back({ "*", "-gan, --grp-attn-n N", "group-attention factor (default: %d)", params.grp_attn_n }); + opts.push_back({ "*", "-gaw, --grp-attn-w N", "group-attention width (default: %.1f)", (double)params.grp_attn_w }); + opts.push_back({ "*", "-nkvo, --no-kv-offload", "disable KV offload" }); + opts.push_back({ "*", "-ctk, --cache-type-k TYPE", "KV cache data type for K (default: %s)", params.cache_type_k.c_str() }); + opts.push_back({ "*", "-ctv, --cache-type-v TYPE", "KV cache data type for V (default: %s)", params.cache_type_v.c_str() }); + opts.push_back({ "*", "-dt, --defrag-thold N", "KV cache defragmentation threshold (default: %.1f, < 0 - disabled)", (double)params.defrag_thold }); + opts.push_back({ "*", "-np, --parallel N", "number of parallel sequences to decode (default: %d)", params.n_parallel }); + opts.push_back({ "*", "-cb, --cont-batching", "enable continuous batching (a.k.a dynamic batching) (default: %s)", params.cont_batching ? "enabled" : "disabled" }); + opts.push_back({ "*", " --mmproj FILE", "path to a multimodal projector file for LLaVA. see examples/llava/README.md" }); + if (llama_supports_mlock()) { + opts.push_back({ "*", " --mlock", "force system to keep model in RAM rather than swapping or compressing" }); + } + if (llama_supports_mmap()) { + opts.push_back({ "*", " --no-mmap", "do not memory-map model (slower load but may reduce pageouts if not using mlock)" }); + } + opts.push_back({ "*", " --numa TYPE", "attempt optimizations that help on some NUMA systems\n" + " - distribute: spread execution evenly over all nodes\n" + " - isolate: only spawn threads on CPUs on the node that execution started on\n" + " - numactl: use the CPU map provided by numactl\n" + "if run without this previously, it is recommended to drop the system page cache before using this\n" + "see https://github.com/ggerganov/llama.cpp/issues/1437" }); + opts.push_back({ "*", " --override-kv KEY=TYPE:VALUE", + "advanced option to override model metadata by key. may be specified multiple times.\n" + "types: int, float, bool, str. example: --override-kv tokenizer.ggml.add_bos_token=bool:false" }); + opts.push_back({ "*", " --lora FILE", "apply LoRA adapter (implies --no-mmap)" }); + opts.push_back({ "*", " --lora-scaled FILE SCALE", + "apply LoRA adapter with user defined scaling S (implies --no-mmap)" }); + opts.push_back({ "*", " --lora-base FILE", "optional model to use as a base for the layers modified by the LoRA adapter" }); + opts.push_back({ "*", " --control-vector FILE", "add a control vector" }); + opts.push_back({ "*", " --control-vector-scaled FILE SCALE", + "add a control vector with user defined scaling SCALE" }); + opts.push_back({ "*", " --control-vector-layer-range START END", + "layer range to apply the control vector(s) to, start and end inclusive" }); + if (llama_supports_gpu_offload()) { + opts.push_back({ "*", "-ngl, --gpu-layers N", "number of layers to store in VRAM" }); + opts.push_back({ "*", "-sm, --split-mode SPLIT_MODE", + "how to split the model across multiple GPUs, one of:\n" + " - none: use one GPU only\n" + " - layer (default): split layers and KV across GPUs\n" + " - row: split rows across GPUs" }); + opts.push_back({ "*", "-ts, --tensor-split SPLIT", + "fraction of the model to offload to each GPU, comma-separated list of proportions, e.g. 3,1" }); + opts.push_back({ "*", "-mg, --main-gpu N", "the GPU to use for the model (with split-mode = none),\n" + "or for intermediate results and KV (with split-mode = row) (default: %d)", params.main_gpu }); + } + + opts.push_back({ "server" }); + opts.push_back({ "server", " --host HOST", "ip address to listen (default: %s)", params.hostname.c_str() }); + opts.push_back({ "server", " --port PORT", "port to listen (default: %d)", params.port }); + opts.push_back({ "server", "-to --timeout N", "server read/write timeout in seconds (default: %d)", params.timeout_read }); + opts.push_back({ "server", " --threads-http N", "number of threads used to process HTTP requests (default: %d)", params.n_threads_http }); + opts.push_back({ "server", " --system-prompt-file FILE", + "set a file to load a system prompt (initial prompt of all slots), this is useful for chat applications" }); + opts.push_back({ "server", " --metrics", "enable prometheus compatible metrics endpoint (default: %s)", params.endpoint_metrics ? "enabled" : "disabled" }); + opts.push_back({ "server", " --infill", "enable infill endpoint (default: %s)", params.infill? "enabled" : "disabled" }); + opts.push_back({ "server", " --embeddings", "enable embedding endpoint (default: %s)", params.embedding ? "enabled" : "disabled" }); + opts.push_back({ "server", " --no-slots", "disables slots monitoring endpoint (default: %s)", params.endpoint_slots ? "enabled" : "disabled" }); + opts.push_back({ "server", " --slot-save-path PATH", "path to save slot kv cache (default: disabled)" }); + opts.push_back({ "server", " --chat-template JINJA_TEMPLATE", + "set custom jinja chat template (default: template taken from model's metadata)\n" + "only commonly used templates are accepted:\n" + "https://github.com/ggerganov/llama.cpp/wiki/Templates-supported-by-llama_chat_apply_template" }); + opts.push_back({ "server", " --chat-template-file FILE", + "set a file to load a custom jinja chat template" }); + opts.push_back({ "server", "-sps, --slot-prompt-similarity N", + "how much the prompt of a request must match the prompt of a slot in order to use that slot (default: %.2f, 0.0 = disabled)\n", params.slot_prompt_similarity }); + opts.push_back({ "server", " --conn-idle N", "server connection idle in seconds (default: %d)", bparams.conn_idle }); + opts.push_back({ "server", " --conn-keepalive N", "server connection keep-alive in seconds (default: %d)", bparams.conn_keepalive }); + opts.push_back({ "server", "-tps --tokens-per-second N", "maximum number of tokens per second (default: %d, 0 = disabled, -1 = try to detect)", bparams.n_tps }); + + opts.push_back({ "logging" }); + opts.push_back({ "logging", " --log-format {text,json}", + "log output format: json or text (default: json)" }); + // clang-format on + + printf("usage: %s [options]\n", argv[0]); + + for (const auto &o : opts) { + if (!o.grp.empty()) { + printf("\n%s:\n\n", o.grp.c_str()); + continue; + } + printf(" %-32s", o.args.c_str()); + if (o.args.length() > 30) { + printf("\n%34s", ""); + } + + const auto desc = o.desc; + size_t start = 0; + size_t end = desc.find('\n'); + while (end != std::string::npos) { + printf("%s\n%34s", desc.substr(start, end - start).c_str(), ""); + start = end + 1; + end = desc.find('\n', start); + } + + printf("%s\n", desc.substr(start).c_str()); + } + printf("\n"); +} + +bool llama_box_params_parse(int argc, char **argv, llama_box_params &bparams) { + try { + for (int i = 1; i < argc;) { + const char *flag = argv[i++]; + + if (*flag != '-') { + unknown(flag); + } + + // general flags + + if (!strcmp(flag, "-h") || !strcmp(flag, "--help") || !strcmp(flag, "--usage")) { + llama_box_params_print_usage(argc, argv, bparams); + exit(0); + } + + if (!strcmp(flag, "--version")) { + fprintf(stderr, "version: %s (%s)\n", LLAMA_BOX_GIT_VERSION, LLAMA_BOX_GIT_COMMIT); + fprintf(stderr, "llama.cpp version: %d (%s)\n", LLAMA_BUILD_NUMBER, LLAMA_COMMIT); + fprintf(stderr, "built with %s for %s\n", LLAMA_COMPILER, LLAMA_BUILD_TARGET); + exit(0); + } + + if (!strcmp(flag, "-m") || !strcmp(flag, "--model")) { + if (i == argc) { + missing("--model"); + } + char *arg = argv[i++]; + bparams.gparams.model = std::string(arg); + continue; + } + + if (!strcmp(flag, "-a") || !strcmp(flag, "--alias")) { + if (i == argc) { + missing("--alias"); + } + char *arg = argv[i++]; + bparams.gparams.model_alias = std::string(arg); + continue; + } + + if (!strcmp(flag, "-s") || !strcmp(flag, "--seed")) { + if (i == argc) { + missing("--seed"); + } + char *arg = argv[i++]; + bparams.gparams.seed = std::stoul(std::string(arg)); + bparams.gparams.sparams.seed = bparams.gparams.seed; + continue; + } + + if (!strcmp(flag, "-t") || !strcmp(flag, "--threads")) { + if (i == argc) { + missing("--threads"); + } + char *arg = argv[i++]; + bparams.gparams.n_threads = std::stoi(std::string(arg)); + if (bparams.gparams.n_threads <= 0) { + bparams.gparams.n_threads = std::thread::hardware_concurrency(); + } + continue; + } + + if (!strcmp(flag, "-tb") || !strcmp(flag, "--threads-batch")) { + if (i == argc) { + missing("--threads-batch"); + } + char *arg = argv[i++]; + bparams.gparams.n_threads_batch = std::stoi(std::string(arg)); + if (bparams.gparams.n_threads_batch <= 0) { + bparams.gparams.n_threads_batch = std::thread::hardware_concurrency(); + } + continue; + } + + if (!strcmp(flag, "-lcs") || !strcmp(flag, "--lookup-cache-static")) { + if (i == argc) { + missing("--lookup-cache-static"); + } + char *arg = argv[i++]; + bparams.gparams.lookup_cache_static = std::string(arg); + continue; + } + + if (!strcmp(flag, "-lcd") || !strcmp(flag, "--lookup-cache-dynamic")) { + if (i == argc) { + missing("--lookup-cache-dynamic"); + } + char *arg = argv[i++]; + bparams.gparams.lookup_cache_dynamic = std::string(arg); + continue; + } + + if (!strcmp(flag, "-c") || !strcmp(flag, "--ctx-size")) { + if (i == argc) { + missing("--ctx-size"); + } + char *arg = argv[i++]; + bparams.gparams.n_ctx = std::stoi(std::string(arg)); + continue; + } + + if (!strcmp(flag, "-n") || !strcmp(flag, "--predict")) { + if (i == argc) { + missing("--predict"); + } + char *arg = argv[i++]; + bparams.gparams.n_predict = std::stoi(std::string(arg)); + continue; + } + + if (!strcmp(flag, "-b") || !strcmp(flag, "--batch-size")) { + if (i == argc) { + missing("--batch-size"); + } + char *arg = argv[i++]; + bparams.gparams.n_batch = std::stoi(std::string(arg)); + continue; + } + + if (!strcmp(flag, "-ub") || !strcmp(flag, "--ubatch-size")) { + if (i == argc) { + missing("--ubatch-size"); + } + char *arg = argv[i++]; + bparams.gparams.n_ubatch = std::stoi(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--keep")) { + if (i == argc) { + missing("--keep"); + } + char *arg = argv[i++]; + bparams.gparams.n_keep = std::stoi(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--chunks")) { + if (i == argc) { + missing("--chunks"); + } + char *arg = argv[i++]; + bparams.gparams.n_chunks = std::stoi(std::string(arg)); + continue; + } + + if (!strcmp(flag, "-fa") || !strcmp(flag, "--flash-attn")) { + bparams.gparams.flash_attn = true; + continue; + } + + if (!strcmp(flag, "--no-escape")) { + bparams.gparams.escape = false; + continue; + } + + if (!strcmp(flag, "--samplers")) { + if (i == argc) { + missing("--samplers"); + } + char *arg = argv[i++]; + const auto sampler_names = string_split(arg, ';'); + bparams.gparams.sparams.samplers_sequence = + llama_sampling_types_from_names(sampler_names, true); + continue; + } + + if (!strcmp(flag, "--sampling-seq")) { + if (i == argc) { + missing("--sampling-seq"); + } + char *arg = argv[i++]; + bparams.gparams.sparams.samplers_sequence = llama_sampling_types_from_chars(arg); + continue; + } + + if (!strcmp(flag, "--penalize-nl")) { + bparams.gparams.sparams.penalize_nl = true; + continue; + } + + if (!strcmp(flag, "--temp")) { + if (i == argc) { + missing("--temp"); + } + char *arg = argv[i++]; + bparams.gparams.sparams.temp = std::stof(std::string(arg)); + bparams.gparams.sparams.temp = std::max(bparams.gparams.sparams.temp, 0.0f); + continue; + } + + if (!strcmp(flag, "--top-k")) { + if (i == argc) { + missing("--top-k"); + } + char *arg = argv[i++]; + bparams.gparams.sparams.top_k = std::stoi(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--top-p")) { + if (i == argc) { + missing("--top-p"); + } + char *arg = argv[i++]; + bparams.gparams.sparams.top_p = std::stof(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--min-p")) { + if (i == argc) { + missing("--min-p"); + } + char *arg = argv[i++]; + bparams.gparams.sparams.min_p = std::stof(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--tfs")) { + if (i == argc) { + missing("--tfs"); + } + char *arg = argv[i++]; + bparams.gparams.sparams.tfs_z = std::stof(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--typical")) { + if (i == argc) { + missing("--typical"); + } + char *arg = argv[i++]; + bparams.gparams.sparams.typical_p = std::stof(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--repeat-last-n")) { + if (i == argc) { + missing("--repeat-last-n"); + } + char *arg = argv[i++]; + bparams.gparams.sparams.penalty_last_n = std::stoi(std::string(arg)); + bparams.gparams.sparams.n_prev = std::max(bparams.gparams.sparams.n_prev, + bparams.gparams.sparams.penalty_last_n); + continue; + } + + if (!strcmp(flag, "--repeat-penalty")) { + if (i == argc) { + missing("--repeat-penalty"); + } + char *arg = argv[i++]; + bparams.gparams.sparams.penalty_repeat = std::stof(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--presence-penalty")) { + if (i == argc) { + missing("--presence-penalty"); + } + char *arg = argv[i++]; + bparams.gparams.sparams.penalty_present = std::stof(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--frequency-penalty")) { + if (i == argc) { + missing("--frequency-penalty"); + } + char *arg = argv[i++]; + bparams.gparams.sparams.penalty_freq = std::stof(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--dynatemp-range")) { + if (i == argc) { + missing("--dynatemp-range"); + } + char *arg = argv[i++]; + bparams.gparams.sparams.dynatemp_range = std::stof(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--dynatemp-exp")) { + if (i == argc) { + missing("--dynatemp-exp"); + } + char *arg = argv[i++]; + bparams.gparams.sparams.dynatemp_exponent = std::stof(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--mirostat")) { + if (i == argc) { + missing("--mirostat"); + } + char *arg = argv[i++]; + bparams.gparams.sparams.mirostat = std::stoi(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--mirostat-lr")) { + if (i == argc) { + missing("--mirostat-lr"); + } + char *arg = argv[i++]; + bparams.gparams.sparams.mirostat_eta = std::stof(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--mirostat-ent")) { + if (i == argc) { + missing("--mirostat-ent"); + } + char *arg = argv[i++]; + bparams.gparams.sparams.mirostat_tau = std::stof(std::string(arg)); + continue; + } + + if (!strcmp(flag, "-l") || !strcmp(flag, "--logit-bias")) { + if (i == argc) { + missing("--logit-bias"); + } + char *arg = argv[i++]; + std::stringstream ss(arg); + llama_token key; + char sign; + std::string value_str; + if (ss >> key && ss >> sign && std::getline(ss, value_str) && + (sign == '+' || sign == '-')) { + bparams.gparams.sparams.logit_bias[key] = + std::stof(value_str) * ((sign == '-') ? -1.0f : 1.0f); + } else { + invalid("--logit-bias"); + } + continue; + } + + if (!strcmp(flag, "--grammar")) { + if (i == argc) { + missing("--grammar"); + } + char *arg = argv[i++]; + bparams.gparams.sparams.grammar = std::string(arg); + continue; + } + + if (!strcmp(flag, "--grammar-file")) { + if (i == argc) { + missing("--grammar-file"); + } + char *arg = argv[i++]; + std::ifstream file(arg); + if (!file) { + invalid("--grammar-file"); + } + std::copy(std::istreambuf_iterator(file), std::istreambuf_iterator(), + std::back_inserter(bparams.gparams.sparams.grammar)); + continue; + } + + if (!strcmp(flag, "-j") || !strcmp(flag, "--json-schema")) { + if (i == argc) { + missing("--json-schema"); + } + char *arg = argv[i++]; + bparams.gparams.sparams.grammar = + json_schema_to_grammar(json::parse(std::string(arg))); + continue; + } + + if (!strcmp(flag, "--rope-scaling")) { + if (i == argc) { + missing("--rope-scaling"); + } + char *arg = argv[i++]; + std::string value(arg); + if (value == "none") { + bparams.gparams.rope_scaling_type = LLAMA_ROPE_SCALING_TYPE_NONE; + } else if (value == "linear") { + bparams.gparams.rope_scaling_type = LLAMA_ROPE_SCALING_TYPE_LINEAR; + } else if (value == "yarn") { + bparams.gparams.rope_scaling_type = LLAMA_ROPE_SCALING_TYPE_YARN; + } else { + invalid("--rope-scaling"); + } + continue; + } + + if (!strcmp(flag, "--rope-scale")) { + if (i == argc) { + missing("--rope-scale"); + } + char *arg = argv[i++]; + bparams.gparams.rope_freq_scale = 1.0f / std::stof(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--rope-freq-base")) { + if (i == argc) { + missing("--rope-freq-base"); + } + char *arg = argv[i++]; + bparams.gparams.rope_freq_base = std::stof(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--rope-freq-scale")) { + if (i == argc) { + missing("--rope-freq-scale"); + } + char *arg = argv[i++]; + bparams.gparams.rope_freq_scale = std::stof(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--yarn-orig-ctx")) { + if (i == argc) { + missing("--yarn-orig-ctx"); + } + char *arg = argv[i++]; + bparams.gparams.yarn_orig_ctx = std::stoi(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--yarn-ext-factor")) { + if (i == argc) { + missing("--yarn-ext-factor"); + } + char *arg = argv[i++]; + bparams.gparams.yarn_ext_factor = std::stof(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--yarn-attn-factor")) { + if (i == argc) { + missing("--yarn-attn-factor"); + } + char *arg = argv[i++]; + bparams.gparams.yarn_attn_factor = std::stof(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--yarn-beta-fast")) { + if (i == argc) { + missing("--yarn-beta-fast"); + } + char *arg = argv[i++]; + bparams.gparams.yarn_beta_fast = std::stof(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--yarn-beta-slow")) { + if (i == argc) { + missing("--yarn-beta-slow"); + } + char *arg = argv[i++]; + bparams.gparams.yarn_beta_slow = std::stof(std::string(arg)); + continue; + } + + if (!strcmp(flag, "-gan") || !strcmp(flag, "--grp-attn-n")) { + if (i == argc) { + missing("--grp-attn-n"); + } + char *arg = argv[i++]; + bparams.gparams.grp_attn_n = std::stoi(std::string(arg)); + continue; + } + + if (!strcmp(flag, "-gaw") || !strcmp(flag, "--grp-attn-w")) { + if (i == argc) { + missing("--grp-attn-w"); + } + char *arg = argv[i++]; + bparams.gparams.grp_attn_w = std::stoi(std::string(arg)); + continue; + } + + if (!strcmp(flag, "-nkvo") || !strcmp(flag, "--no-kv-offload")) { + bparams.gparams.no_kv_offload = true; + continue; + } + + if (!strcmp(flag, "-ctk") || !strcmp(flag, "--cache-type-k")) { + if (i == argc) { + missing("--cache-type-k"); + } + char *arg = argv[i++]; + bparams.gparams.cache_type_k = std::string(arg); + continue; + } + + if (!strcmp(flag, "-ctv") || !strcmp(flag, "--cache-type-v")) { + if (i == argc) { + missing("--cache-type-v"); + } + char *arg = argv[i++]; + bparams.gparams.cache_type_v = std::string(arg); + continue; + } + + if (!strcmp(flag, "-dt") || !strcmp(flag, "--defrag-thold")) { + if (i == argc) { + missing("--defrag-thold"); + } + char *arg = argv[i++]; + bparams.gparams.defrag_thold = std::stof(std::string(arg)); + continue; + } + + if (!strcmp(flag, "-np") || !strcmp(flag, "--parallel")) { + if (i == argc) { + missing("--parallel"); + } + char *arg = argv[i++]; + bparams.gparams.n_parallel = std::stoi(std::string(arg)); + continue; + } + + if (!strcmp(flag, "-cb") || !strcmp(flag, "--cont-batching")) { + bparams.gparams.cont_batching = true; + continue; + } + + if (!strcmp(flag, "--mmproj")) { + if (i == argc) { + missing("--mmproj"); + } + char *arg = argv[i++]; + bparams.gparams.mmproj = std::string(arg); + continue; + } + + if (llama_supports_mlock()) { + if (!strcmp(flag, "--mlock")) { + bparams.gparams.use_mlock = true; + continue; + } + } + + if (llama_supports_mmap()) { + if (!strcmp(flag, "--no-mmap")) { + bparams.gparams.use_mmap = false; + continue; + } + } + + if (!strcmp(flag, "--numa")) { + if (i == argc) { + missing("--numa"); + } + char *arg = argv[i++]; + std::string value(arg); + if (value == "distribute") { + bparams.gparams.numa = GGML_NUMA_STRATEGY_DISTRIBUTE; + } else if (value == "isolate") { + bparams.gparams.numa = GGML_NUMA_STRATEGY_ISOLATE; + } else if (value == "numactl") { + bparams.gparams.numa = GGML_NUMA_STRATEGY_NUMACTL; + } else { + invalid("--numa"); + } + } + + if (!strcmp(flag, "--override-kv")) { + if (i == argc) { + missing("--override-kv"); + } + char *arg = argv[i++]; + if (!string_parse_kv_override(arg, bparams.gparams.kv_overrides)) { + invalid("--override-kv"); + } + continue; + } + + if (!strcmp(flag, "--lora")) { + if (i == argc) { + missing("--lora"); + } + char *arg = argv[i++]; + bparams.gparams.lora_adapter.emplace_back(std::string(arg), 1.0f); + bparams.gparams.use_mmap = false; + continue; + } + + if (!strcmp(flag, "--lora-scaled")) { + if (i == argc) { + missing("--lora-scaled"); + } + char *n = argv[i++]; + if (i == argc) { + invalid("--lora-scaled"); + } + char *s = argv[i++]; + bparams.gparams.lora_adapter.emplace_back(std::string(n), + std::stof(std::string(s))); + bparams.gparams.use_mmap = false; + continue; + } + + if (!strcmp(flag, "--lora-base")) { + if (i == argc) { + missing("--lora-base"); + } + char *arg = argv[i++]; + bparams.gparams.lora_base = std::string(arg); + continue; + } + + if (!strcmp(flag, "--control-vector")) { + if (i == argc) { + missing("--control-vector"); + } + char *arg = argv[i++]; + bparams.gparams.control_vectors.push_back({1.0f, std::string(arg)}); + continue; + } + + if (!strcmp(flag, "--control-vector-scaled")) { + if (i == argc) { + missing("--control-vector-scaled"); + } + char *n = argv[i++]; + if (i == argc) { + invalid("--control-vector-scaled"); + } + char *s = argv[i++]; + bparams.gparams.control_vectors.push_back( + {std::stof(std::string(s)), std::string(n)}); + continue; + } + + if (!strcmp(flag, "--control-vector-layer-range")) { + if (i == argc) { + missing("--control-vector-layer-range"); + } + char *s = argv[i++]; + if (i == argc) { + invalid("--control-vector-layer-range"); + } + char *e = argv[i++]; + bparams.gparams.control_vector_layer_start = std::stoi(std::string(s)); + bparams.gparams.control_vector_layer_end = std::stoi(std::string(e)); + continue; + } + + if (llama_supports_gpu_offload()) { + if (!strcmp(flag, "-ngl") || !strcmp(flag, "--gpu-layers")) { + if (i == argc) { + missing("--gpu-layers"); + } + char *arg = argv[i++]; + bparams.gparams.n_gpu_layers = std::stoi(arg); + continue; + } + + if (!strcmp(flag, "-sm") || !strcmp(flag, "--split-mode")) { + if (i == argc) { + missing("--split-mode"); + } + char *arg = argv[i++]; + if (!strcmp(arg, "none")) { + bparams.gparams.split_mode = LLAMA_SPLIT_MODE_NONE; + } else if (!strcmp(arg, "layer")) { + bparams.gparams.split_mode = LLAMA_SPLIT_MODE_LAYER; + } else if (!strcmp(arg, "row")) { + bparams.gparams.split_mode = LLAMA_SPLIT_MODE_ROW; + } else { + invalid("--split-mode"); + } + continue; + } + + if (!strcmp(flag, "-ts") || !strcmp(flag, "--tensor-split")) { + if (i == argc) { + missing("--tensor-split"); + } + char *arg = argv[i++]; + const std::regex regex{R"([,/]+)"}; + std::string arg_s{arg}; + std::sregex_token_iterator it{arg_s.begin(), arg_s.end(), regex, -1}; + std::vector split_arg{it, {}}; + if (split_arg.size() >= llama_max_devices()) { + invalid("--tensor-split"); + } + for (size_t i = 0; i < llama_max_devices(); ++i) { + if (i < split_arg.size()) { + bparams.gparams.tensor_split[i] = std::stof(split_arg[i]); + } else { + bparams.gparams.tensor_split[i] = 0.0f; + } + } + continue; + } + + if (!strcmp(flag, "-mg") || !strcmp(flag, "--main-gpu")) { + if (i == argc) { + missing("--main-gpu"); + } + char *arg = argv[i++]; + bparams.gparams.main_gpu = std::stoi(std::string(arg)); + continue; + } + } + + // server flags + + if (!strcmp(flag, "--host")) { + if (i == argc) { + missing("--host"); + } + char *arg = argv[i++]; + bparams.gparams.hostname = std::string(arg); + continue; + } + + if (!strcmp(flag, "--port")) { + if (i == argc) { + missing("--port"); + } + char *arg = argv[i++]; + bparams.gparams.port = std::stoi(std::string(arg)); + continue; + } + + if (!strcmp(flag, "-to") || !strcmp(flag, "--timeout")) { + if (i == argc) { + missing("--timeout"); + } + char *arg = argv[i++]; + bparams.gparams.timeout_read = std::stoi(std::string(arg)); + bparams.gparams.timeout_write = bparams.gparams.timeout_read; + continue; + } + + if (!strcmp(flag, "--threads-http")) { + if (i == argc) { + missing("--threads-http"); + } + char *arg = argv[i++]; + bparams.gparams.n_threads_http = std::stoi(std::string(arg)); + continue; + } + + if (!strcmp(flag, "-spf") || !strcmp(flag, "--system-prompt-file")) { + if (i == argc) { + missing("--system-prompt-file"); + } + char *arg = argv[i++]; + std::ifstream file(arg); + if (!file) { + invalid("--system-prompt-file"); + } + std::copy(std::istreambuf_iterator(file), std::istreambuf_iterator(), + std::back_inserter(bparams.gparams.system_prompt)); + continue; + } + + if (!strcmp(flag, "--metrics")) { + bparams.gparams.endpoint_metrics = true; + continue; + } + + if (!strcmp(flag, "--infill")) { + bparams.gparams.infill = true; + continue; + } + + if (!strcmp(flag, "--embedding") || !strcmp(flag, "--embeddings")) { + bparams.gparams.embedding = true; + continue; + } + + if (!strcmp(flag, "--no-slots")) { + bparams.gparams.endpoint_slots = false; + continue; + } + + if (!strcmp(flag, "--slot-save-path")) { + if (i == argc) { + missing("--slot-save-path"); + } + char *arg = argv[i++]; + if (arg[0] == '\0') { + invalid("--slot-save-path"); + } + std::string p(arg); + if (p[p.size() - 1] != DIRECTORY_SEPARATOR) { + p += DIRECTORY_SEPARATOR; + } + bparams.gparams.slot_save_path = p; + continue; + } + + if (!strcmp(flag, "--chat-template")) { + if (i == argc) { + missing("--chat-template"); + } + char *arg = argv[i++]; + if (arg[0] == '\0') { + invalid("--chat-template"); + } + std::string t(arg); + if (!llama_chat_verify_template(t)) { + invalid("--chat-template"); + } + bparams.gparams.chat_template = t; + continue; + } + + if (!strcmp(flag, "--chat-template-file")) { + if (i == argc) { + missing("--chat-template-file"); + } + char *arg = argv[i++]; + std::ifstream file(arg); + if (!file) { + invalid("--chat-template-file"); + } + std::copy(std::istreambuf_iterator(file), std::istreambuf_iterator(), + std::back_inserter(bparams.gparams.chat_template)); + continue; + } + + if (!strcmp(flag, "-sps") || !strcmp(flag, "--slot-prompt-similarity")) { + if (i == argc) { + missing("--slot-prompt-similarity"); + } + char *arg = argv[i++]; + bparams.gparams.slot_prompt_similarity = std::stof(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--conn-idle")) { // extend + if (i == argc) { + missing("--conn-idle"); + } + char *arg = argv[i++]; + bparams.conn_idle = std::stoi(std::string(arg)); + continue; + } + + if (!strcmp(flag, "--conn-keepalive")) { // extend + if (i == argc) { + missing("--conn-keepalive"); + } + char *arg = argv[i++]; + bparams.conn_keepalive = std::stoi(std::string(arg)); + continue; + } + + if (!strcmp(flag, "-tps") || !strcmp(flag, "--tokens-per-second")) { // extend + if (i == argc) { + missing("--tokens-per-second"); + } + char *arg = argv[i++]; + bparams.n_tps = std::stoi(std::string(arg)); + continue; + } + + // logging flags + + if (!strcmp(flag, "--log-format")) { + if (i == argc) { + missing("--log-format"); + } + char *arg = argv[i++]; + if (!strcmp(arg, "json")) { + bparams.gparams.log_json = true; + } else if (!strcmp(arg, "text")) { + bparams.gparams.log_json = false; + } else { + unknown("--log-format"); + } + continue; + } + + unknown(flag); + } + } catch (const std::invalid_argument &ex) { + fprintf(stderr, "%s\n", ex.what()); + return false; + } + + if (!bparams.gparams.kv_overrides.empty()) { + bparams.gparams.kv_overrides.emplace_back(); + bparams.gparams.kv_overrides.back().key[0] = 0; + } + + return true; +} \ No newline at end of file diff --git a/llama-box/ratelimiter.hpp b/llama-box/ratelimiter.hpp new file mode 100644 index 0000000..571f9f0 --- /dev/null +++ b/llama-box/ratelimiter.hpp @@ -0,0 +1,39 @@ +#include +#include + +// lockless token bucket rate limiter +class token_bucket { + +private: + int capacity; + int rate; + int tokens; + std::chrono::steady_clock::time_point last_time; + + void refill() { + auto const now = std::chrono::steady_clock::now(); + auto const elapsed = std::chrono::duration_cast(now - last_time).count(); + int new_tokens = elapsed * rate / 1000; + if (new_tokens > 0) { + tokens = std::min(capacity, tokens + new_tokens); + last_time = now; + } + } + +public: + token_bucket(int capacity, int rate) : capacity(capacity), rate(rate) { + tokens = capacity; + last_time = std::chrono::steady_clock::now(); + } + + bool acquire(int tokens = 1) { + if (this->tokens < tokens) { + refill(); + if (this->tokens < tokens) { + return false; + } + } + this->tokens -= tokens; + return true; + } +}; \ No newline at end of file diff --git a/llama-box/scripts/version.sh b/llama-box/scripts/version.sh new file mode 100755 index 0000000..556339d --- /dev/null +++ b/llama-box/scripts/version.sh @@ -0,0 +1,60 @@ +#!/bin/sh + +## +# Inspired by github.com/kubernetes/kubernetes/hack/lib/version.sh +## + +# ----------------------------------------------------------------------------- +# Version management helpers. These functions help to set the +# following variables: +# +# GIT_TREE_STATE - "clean" indicates no changes since the git commit id. +# "dirty" indicates source code changes after the git commit id. +# "unknown" indicates cannot find out the git tree. +# GIT_COMMIT - The git commit id corresponding to this +# source code. +# GIT_VERSION - "vX.Y" used to indicate the last release version, +# it can be specified via "VERSION". +# BUILD_DATE - The build date of the version. + +BUILD_DATE=$(date -u '+%Y-%m-%dT%H:%M:%SZ') +GIT_TREE_STATE="unknown" +GIT_COMMIT="unknown" +GIT_VERSION="unknown" + +# return directly if not found git client. +if [ -z "$(command -v git)" ]; then + # respect specified version. + GIT_VERSION=${VERSION:-${GIT_VERSION}} + return +fi + +# find out git info via git client. +if GIT_COMMIT=$(git rev-parse "HEAD^{commit}" 2>/dev/null); then + # specify as dirty if the tree is not clean. + if git_status=$(git status --porcelain 2>/dev/null) && [ -n "${git_status}" ]; then + GIT_TREE_STATE="dirty" + else + GIT_TREE_STATE="clean" + fi + + # specify with the tag if the head is tagged. + if GIT_VERSION="$(git rev-parse --abbrev-ref HEAD 2>/dev/null)"; then + if git_tag=$(git tag -l --contains HEAD 2>/dev/null | head -n 1 2>/dev/null) && [ -n "${git_tag}" ]; then + GIT_VERSION="${git_tag}" + fi + fi + + # specify to dev if the tree is dirty. + if [ "${GIT_TREE_STATE:-dirty}" = "dirty" ]; then + GIT_VERSION="dev" + fi + + # respect specified version + GIT_VERSION=${VERSION:-${GIT_VERSION}} +fi + +echo "char const *LLAMA_BOX_BUILD_DATE = \"${BUILD_DATE:-0}\";" +echo "char const *LLAMA_BOX_GIT_TREE_STATE = \"${GIT_TREE_STATE}\";" +echo "char const *LLAMA_BOX_GIT_COMMIT = \"${GIT_COMMIT}\";" +echo "char const *LLAMA_BOX_GIT_VERSION = \"${GIT_VERSION}\";" diff --git a/llama-box/utils.hpp b/llama-box/utils.hpp new file mode 100644 index 0000000..9e7f6ee --- /dev/null +++ b/llama-box/utils.hpp @@ -0,0 +1,645 @@ +#pragma once + +#include +#include +#include +#include + +#include "llama.cpp/common/common.h" +#define JSON_ASSERT GGML_ASSERT +#include "llama.cpp/common/json.hpp" +#include "llama.cpp/llama.h" + +#define DEFAULT_OAICOMPAT_MODEL "gpt-3.5-turbo-0613" + +using json = nlohmann::json; + +// https://community.openai.com/t/openai-chat-list-of-error-codes-and-types/357791/11 +enum error_type { + ERROR_TYPE_INVALID_REQUEST, + ERROR_TYPE_AUTHENTICATION, + ERROR_TYPE_SERVER, + ERROR_TYPE_NOT_FOUND, + ERROR_TYPE_PERMISSION, + ERROR_TYPE_UNAVAILABLE, // custom error + ERROR_TYPE_NOT_SUPPORTED, // custom error +}; + +#define LOG_ERROR(MSG, ...) server_log("ERR", __func__, __LINE__, MSG, __VA_ARGS__) +#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 json &extra); + +template +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()) { + try { + return body.at(key); + } catch (NLOHMANN_JSON_NAMESPACE::detail::type_error const &) { + std::stringstream ss; + ss << "Wrong type supplied for parameter '" << key << "'. Expected '" + << json(default_value).type_name() << "', using default value."; + LOG_WARNING(ss.str().c_str(), body); + return default_value; + } + } else { + return default_value; + } +} + +extern bool server_log_json; + +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 = json{ + {"tid", ss_tid.str()}, + {"timestamp", time(nullptr)}, + }; + + if (server_log_json) { + log.merge_patch({ + {"level", level}, + {"function", function}, + {"line", line}, + {"msg", message}, + }); + + if (!extra.empty()) { + log.merge_patch(extra); + } + + printf("%s\n", log.dump(-1, ' ', false, json::error_handler_t::replace).c_str()); + fflush(stdout); + return; + } + + char buf[1024]; + snprintf(buf, 1024, "%4s [%24s] %s", level, function, message); + + if (!extra.empty()) { + log.merge_patch(extra); + } + std::stringstream ss; + ss << buf << " |"; + for (const auto &el : log.items()) { + const std::string value = el.value().dump(-1, ' ', false, json::error_handler_t::replace); + ss << " " << el.key() << "=" << value; + } + + const std::string str = ss.str(); + printf("%.*s\n", (int)str.size(), str.data()); + fflush(stdout); +} + +// +// chat template utils +// + +// Format given chat. If tmpl is empty, we take the template from model metadata +inline std::string format_chat(const struct llama_model *model, const std::string &tmpl, + const std::vector &messages) { + size_t alloc_size = 0; + // vector holding all allocated string to be passed to llama_chat_apply_template + std::vector str(messages.size() * 2); + std::vector chat(messages.size()); + + for (size_t i = 0; i < messages.size(); ++i) { + const auto &curr_msg = messages[i]; + str[i * 2 + 0] = json_value(curr_msg, "role", std::string("")); + str[i * 2 + 1] = json_value(curr_msg, "content", std::string("")); + alloc_size += str[i * 2 + 1].length(); + chat[i].role = str[i * 2 + 0].c_str(); + chat[i].content = str[i * 2 + 1].c_str(); + } + + const char *ptr_tmpl = tmpl.empty() ? nullptr : tmpl.c_str(); + std::vector buf(alloc_size * 2); + + // run the first time to get the total output length + int32_t res = llama_chat_apply_template(model, ptr_tmpl, chat.data(), chat.size(), true, + buf.data(), buf.size()); + + // if it turns out that our buffer is too small, we resize it + if ((size_t)res > buf.size()) { + buf.resize(res); + res = llama_chat_apply_template(model, ptr_tmpl, chat.data(), chat.size(), true, buf.data(), + buf.size()); + } + + const std::string formatted_chat(buf.data(), res); + return formatted_chat; +} + +// +// base64 utils (TODO: move to common in the future) +// + +static const std::string base64_chars = "ABCDEFGHIJKLMNOPQRSTUVWXYZ" + "abcdefghijklmnopqrstuvwxyz" + "0123456789+/"; + +static inline bool is_base64(uint8_t c) { + return (isalnum(c) || (c == '+') || (c == '/')); +} + +static inline std::vector base64_decode(const std::string &encoded_string) { + int i = 0; + int j = 0; + int in_ = 0; + + int in_len = encoded_string.size(); + + uint8_t char_array_4[4]; + uint8_t char_array_3[3]; + + std::vector ret; + + while (in_len-- && (encoded_string[in_] != '=') && is_base64(encoded_string[in_])) { + char_array_4[i++] = encoded_string[in_]; + in_++; + if (i == 4) { + for (i = 0; i < 4; i++) { + char_array_4[i] = base64_chars.find(char_array_4[i]); + } + + char_array_3[0] = ((char_array_4[0]) << 2) + ((char_array_4[1] & 0x30) >> 4); + char_array_3[1] = ((char_array_4[1] & 0xf) << 4) + ((char_array_4[2] & 0x3c) >> 2); + char_array_3[2] = ((char_array_4[2] & 0x3) << 6) + char_array_4[3]; + + for (i = 0; (i < 3); i++) { + ret.push_back(char_array_3[i]); + } + + i = 0; + } + } + + if (i) { + for (j = i; j < 4; j++) { + char_array_4[j] = 0; + } + + for (j = 0; j < 4; j++) { + char_array_4[j] = base64_chars.find(char_array_4[j]); + } + + char_array_3[0] = ((char_array_4[0]) << 2) + ((char_array_4[1] & 0x30) >> 4); + char_array_3[1] = ((char_array_4[1] & 0xf) << 4) + ((char_array_4[2] & 0x3c) >> 2); + char_array_3[2] = ((char_array_4[2] & 0x3) << 6) + char_array_4[3]; + + for (j = 0; j < i - 1; j++) { + ret.push_back(char_array_3[j]); + } + } + + return ret; +} + +// +// random string / id +// + +static std::string random_string() { + static const std::string str("0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz"); + + std::random_device rd; + std::mt19937 generator(rd()); + + std::string result(32, ' '); + + for (int i = 0; i < 32; ++i) { + result[i] = str[generator() % str.size()]; + } + + return result; +} + +static std::string gen_chatcmplid() { + std::stringstream chatcmplid; + chatcmplid << "chatcmpl-" << random_string(); + + return chatcmplid.str(); +} + +static std::string gen_cmplid() { + std::stringstream cmplid; + cmplid << "cmpl-" << random_string(); + + return cmplid.str(); +} + +// +// other common utils +// + +static size_t common_part(const std::vector &a, const std::vector &b) { + size_t i; + for (i = 0; i < a.size() && i < b.size() && a[i] == b[i]; i++) { + } + + return i; +} + +static size_t common_part(const std::string &a, const std::string &b) { + size_t i; + for (i = 0; i < a.size() && i < b.size() && a[i] == b[i]; i++) { + } + + return i; +} + +static bool ends_with(const std::string & str, const std::string & suffix) { + return str.size() >= suffix.size() && 0 == str.compare(str.size() - suffix.size(), suffix.size(), suffix); +} + +static size_t find_partial_stop_string(const std::string &stop, const std::string &text) { + if (!text.empty() && !stop.empty()) { + const char text_last_char = text.back(); + for (int64_t char_index = stop.size() - 1; char_index >= 0; char_index--) { + if (stop[char_index] == text_last_char) { + const std::string current_partial = stop.substr(0, char_index + 1); + if (ends_with(text, current_partial)) { + return text.size() - char_index - 1; + } + } + } + } + + return std::string::npos; +} + +// format incomplete utf-8 multibyte character for output +static std::string tokens_to_output_formatted_string(const llama_context * ctx, const llama_token token) { + std::string out = token == -1 ? "" : llama_token_to_piece(ctx, token); + + // if the size is 1 and first bit is 1, meaning it's a partial character + // (size > 1 meaning it's already a known token) + if (out.size() == 1 && (out[0] & 0x80) == 0x80) { + std::stringstream ss; + ss << std::hex << (out[0] & 0xff); + std::string res(ss.str()); + out = "byte: \\x" + res; + } + + return out; +} + +struct completion_token_output { + llama_token tok; + std::string text_to_send; + + struct token_prob { + llama_token tok; + float prob; + }; + + std::vector probs; +}; + +// convert a vector of completion_token_output to json +static json probs_vector_to_json(const llama_context *ctx, + const std::vector &probs, + const bool oaicompat_completion = false, + const bool oaicompat_completion_chat = false) { + if (oaicompat_completion) { + if (oaicompat_completion_chat) { + json content = json::array(); + + for (const auto &prob : probs) { + const std::string token = tokens_to_output_formatted_string(ctx, prob.tok); + float token_logprob = 1.0f; + std::vector token_bytes(token.begin(), token.end()); + json token_top_logprobs = json::array(); + for (const auto &p : prob.probs) { + const std::string p_token = tokens_to_output_formatted_string(ctx, p.tok); + float p_token_logprob = p.prob; + std::vector p_token_bytes(p_token.begin(), p_token.end()); + token_top_logprobs.push_back(json{ + {"token", p_token}, + {"logprob", p_token_logprob}, + {"bytes", p_token_bytes}, + }); + if (p.tok == prob.tok) { + token_logprob = p_token_logprob; + } + } + + content.push_back(json{ + {"token", token}, + {"logprob", token_logprob}, + {"bytes", token_bytes}, + {"top_logprobs", token_top_logprobs}, + }); + } + + return json{{"content", content}}; + } else { + json token_logprobs = json::array(); + json tokens = json::array(); + json top_logprobs = json::array(); + + for (const auto &prob : probs) { + const std::string token = tokens_to_output_formatted_string(ctx, prob.tok); + float token_logprob = 1.0f; + json token_top_logprobs; + for (const auto &p : prob.probs) { + const std::string p_token = tokens_to_output_formatted_string(ctx, p.tok); + float p_token_logprob = p.prob; + token_top_logprobs[p_token] = p_token_logprob; + if (p.tok == prob.tok) { + token_logprob = p_token_logprob; + } + } + + tokens.push_back(token); + token_logprobs.push_back(token_logprob); + top_logprobs.push_back(token_top_logprobs); + } + + return json{{"tokens", tokens}, + {"token_logprobs", token_logprobs}, + {"top_logprobs", top_logprobs}}; + } + } + + json out = json::array(); + + for (const auto &prob : probs) { + json probs_for_token = json::array(); + + for (const auto &p : prob.probs) { + const std::string tok_str = tokens_to_output_formatted_string(ctx, p.tok); + probs_for_token.push_back(json{ + {"tok_str", tok_str}, + {"prob", p.prob}, + }); + } + + const std::string tok_str = tokens_to_output_formatted_string(ctx, prob.tok); + out.push_back(json{ + {"content", tok_str}, + {"probs", probs_for_token}, + }); + } + + return out; +} + +// +// OAI utils +// + +static json oaicompat_completion_request(const struct llama_model *model, const json &body, + const std::string &chat_template) { + bool chat = !chat_template.empty(); + json llama_params; + + // Annotations for OAI compatibility + llama_params["__oaicompat"] = true; + llama_params["__oaicompat_completion"] = true; + llama_params["__oaicompat_completion_chat"] = chat; + + // Handle default field + llama_params["model"] = json_value(body, "model", std::string(DEFAULT_OAICOMPAT_MODEL)); + llama_params["frequency_penalty"] = json_value(body, "frequency_penalty", 0.0f); + llama_params["temperature"] = json_value(body, "temperature", 1.0f); + llama_params["top_p"] = json_value(body, "top_p", 1.0f); + + // Handle "max_tokens" field + llama_params["n_predict"] = json_value(body, "max_tokens", -1); + + // Apply chat template to the list of messages + if (!chat_template.empty()) { + llama_params["prompt"] = format_chat(model, chat_template, body.at("messages")); + } else { + llama_params["prompt"] = json_value(body, "prompt", std::string()); + } + + // Handle "stop" field + if (body.contains("stop") && body.at("stop").is_string()) { + llama_params["stop"] = json::array({body.at("stop").get()}); + } else { + llama_params["stop"] = json_value(body, "stop", json::array()); + } + + // Handle "response_format" field + if (body.contains("response_format")) { + json response_format = json_value(body, "response_format", json::object()); + std::string response_type = json_value(response_format, "type", std::string()); + if (response_type == "json_object") { + llama_params["json_schema"] = json_value(response_format, "schema", json::object()); + } else if (!response_type.empty() && response_type != "text") { + throw std::runtime_error( + "\"response_format\" must be one of \"text\" or \"json_object\", but got: " + + response_type); + } + } + + // Handle "n" field + int n_choices = json_value(body, "n", 1); + if (n_choices != 1) { + throw std::runtime_error("Only one completion choice is allowed"); + } + + // Handle "logprobs" field + if (body.contains("logprobs")) { + if (chat) { + llama_params["n_probs"] = std::min(json_value(body, "top_logprobs", 2), 20); + } else { + llama_params["n_probs"] = std::min(json_value(body, "logprobs", 2), 5); + } + } else if (body.contains("top_logprobs")) { + throw std::runtime_error("\"top_logprobs\" requires \"logprobs\" to be set"); + } + + // Params supported by OAI but unsupported by llama.cpp + static const std::vector unsupported_params{"tools", "tool_choice"}; + for (auto ¶m : unsupported_params) { + if (body.contains(param)) { + throw std::runtime_error("Unsupported param: " + param); + } + } + + // Copy remaining properties to llama_params + // This allows user to use llama.cpp-specific params like "mirostat", "tfs_z",... via OAI + // endpoint. See "launch_slot_with_task()" for a complete list of params supported by llama.cpp + for (const auto &item : body.items()) { + // Exception: if "n_predict" is present, we overwrite the value specified earlier by + // "max_tokens" + if (!llama_params.contains(item.key()) || item.key() == "n_predict") { + llama_params[item.key()] = item.value(); + } + } + + return llama_params; +} + +static json oaicompat_completion_response(const json &request, const json result, + const std::string &completion_id, bool streaming = false, + bool first = false) { + bool stopped_word = json_value(result, "stopped_word", false); + bool stopped_eos = json_value(result, "stopped_eos", false); + bool stopped_limit = json_value(result, "stopped_limit", false); + std::string content = json_value(result, "content", std::string("")); + + std::string finish_reason; + if (stopped_word || stopped_eos) { + finish_reason = "stop"; + } + if (stopped_limit) { + finish_reason = "length"; + } + + json res = json{ + {"id", completion_id}, + {"created", std::time(0)}, + {"model", json_value(request, "model", std::string(DEFAULT_OAICOMPAT_MODEL))}, + }; + + bool chat = json_value(request, "__oaicompat_completion_chat", false); + bool finish = !finish_reason.empty(); + json choice; + if (chat) { + // chat completion + if (streaming) { + res["object"] = "chat.completion.chunk"; + if (!finish && first) { + choice = json{{"finish_reason", nullptr}, + {"index", 0}, + {"delta", json{{"role", "assistant"}}}}; + } else if (!finish) { + choice = json{{"finish_reason", nullptr}, + {"index", 0}, + {"delta", json{{"content", content}}}}; + } else { + // finished + choice = + json{{"finish_reason", finish_reason}, {"index", 0}, {"delta", json::object()}}; + } + } else { + res["object"] = "chat.completion"; + if (!finish) { + choice = json{{"finish_reason", nullptr}, + {"index", 0}, + {"message", json{{"content", content}, {"role", "assistant"}}}}; + } else { + choice = json{{"finish_reason", finish_reason}, + {"index", 0}, + {"message", json{{"content", content}, {"role", "assistant"}}}}; + } + } + + } else { + // completion + res["object"] = "text_completion"; + if (!finish) { + choice = json{{"finish_reason", nullptr}, {"index", 0}, {"text", content}}; + } else { + choice = json{{"finish_reason", finish_reason}, {"index", 0}, {"text", content}}; + } + } + bool logprobs = result.contains("completion_probabilities"); + if (!logprobs) { + choice["logprobs"] = nullptr; + } else { + choice["logprobs"] = result.at("completion_probabilities"); + } + res["choices"] = json::array({choice}); + + // Add usage information + if (!streaming) { + int completion_tokens = json_value(result, "tokens_predicted", 0); + int prompt_tokens = json_value(result, "tokens_evaluated", 0); + json timings = json_value(result, "timings", json::object()); + int ttft = json_value(timings, "prompt_ms", 0); // time to first token in milliseconds. + int tpot = json_value(timings, "predicted_per_token_ms", + 0); // time per output token in milliseconds. + res["usage"] = json{{"completion_tokens", completion_tokens}, + {"prompt_tokens", prompt_tokens}, + {"total_tokens", completion_tokens + prompt_tokens}, + {"time_to_first_token_ms", ttft}, + {"time_per_output_token_ms", tpot}}; + } + + return res; +} + +static json oaicompat_embedding_request(const struct gpt_params ¶ms, const json &body) { + json llama_params; + + // Annotations for OAI compatibility + llama_params["__oaicompat"] = true; + llama_params["__oaicompat_embedding"] = true; + + // Handle "model" field + llama_params["model"] = json_value(body, "model", params.model_alias); + + // Handle "input" field + llama_params["prompt"] = json_value(body, "input", std::string()); + + // Handle "encoding_format" field + llama_params["encoding_format"] = json_value(body, "encoding_format", std::string("float")); + + return llama_params; +} + +static json oaicompat_embedding_response(const json &request, const json &result) { + json data = json::array(); + data.push_back(json{{"embedding", json_value(result, "embedding", json::array())}, + {"index", 0}, + {"object", "embedding"}}); + + int num_prompt_tokens = json_value(result, "tokens_evaluated", 0); + json res = json{ + {"model", json_value(request, "model", std::string(DEFAULT_OAICOMPAT_MODEL))}, + {"object", "list"}, + {"usage", json{{"prompt_tokens", num_prompt_tokens}, {"total_tokens", num_prompt_tokens}}}, + {"data", data}}; + + return res; +} + +static json format_error_response(const std::string &message, const enum error_type type) { + std::string type_str; + int code = 500; + switch (type) { + case ERROR_TYPE_INVALID_REQUEST: + type_str = "invalid_request_error"; + code = 400; + break; + case ERROR_TYPE_AUTHENTICATION: + type_str = "authentication_error"; + code = 401; + break; + case ERROR_TYPE_NOT_FOUND: + type_str = "not_found_error"; + code = 404; + break; + case ERROR_TYPE_SERVER: + type_str = "server_error"; + code = 500; + break; + case ERROR_TYPE_PERMISSION: + type_str = "permission_error"; + code = 403; + break; + case ERROR_TYPE_NOT_SUPPORTED: + type_str = "not_supported_error"; + code = 501; + break; + case ERROR_TYPE_UNAVAILABLE: + type_str = "unavailable_error"; + code = 503; + break; + } + return json{ + {"code", code}, + {"message", message}, + {"type", type_str}, + }; +} diff --git a/llama.cpp b/llama.cpp new file mode 160000 index 0000000..d50f889 --- /dev/null +++ b/llama.cpp @@ -0,0 +1 @@ +Subproject commit d50f8897a797a5a03f31228d1b5a7b8130ee1bc2