llama.cpp/build.zig

143 lines
6.2 KiB
Zig
Raw Normal View History

// Compatible with Zig Version 0.11.0
const std = @import("std");
const ArrayList = std.ArrayList;
const Compile = std.Build.Step.Compile;
const ConfigHeader = std.Build.Step.ConfigHeader;
const Mode = std.builtin.Mode;
const CrossTarget = std.zig.CrossTarget;
const Maker = struct {
builder: *std.build.Builder,
target: CrossTarget,
optimize: Mode,
config_header: *ConfigHeader,
enable_lto: bool,
include_dirs: ArrayList([]const u8),
cflags: ArrayList([]const u8),
cxxflags: ArrayList([]const u8),
objs: ArrayList(*Compile),
fn addInclude(m: *Maker, dir: []const u8) !void {
try m.include_dirs.append(dir);
}
fn addProjectInclude(m: *Maker, path: []const []const u8) !void {
try m.addInclude(try m.builder.build_root.join(m.builder.allocator, path));
}
fn addCFlag(m: *Maker, flag: []const u8) !void {
try m.cflags.append(flag);
}
fn addCxxFlag(m: *Maker, flag: []const u8) !void {
try m.cxxflags.append(flag);
}
fn addFlag(m: *Maker, flag: []const u8) !void {
try m.addCFlag(flag);
try m.addCxxFlag(flag);
}
fn init(builder: *std.build.Builder) !Maker {
2023-09-21 11:08:20 +02:00
const target = builder.standardTargetOptions(.{});
const zig_version = @import("builtin").zig_version_string;
const commit_hash = try std.ChildProcess.exec(
.{ .allocator = builder.allocator, .argv = &.{ "git", "rev-parse", "HEAD" } },
);
const config_header = builder.addConfigHeader(
.{ .style = .blank, .include_path = "build-info.h" },
.{
.BUILD_NUMBER = 0,
.BUILD_COMMIT = commit_hash.stdout[0 .. commit_hash.stdout.len - 1], // omit newline
.BUILD_COMPILER = builder.fmt("Zig {s}", .{zig_version}),
2023-09-21 11:08:20 +02:00
.BUILD_TARGET = try target.allocDescription(builder.allocator),
},
);
var m = Maker{
.builder = builder,
2023-09-21 11:08:20 +02:00
.target = target,
.optimize = builder.standardOptimizeOption(.{}),
.config_header = config_header,
.enable_lto = false,
.include_dirs = ArrayList([]const u8).init(builder.allocator),
.cflags = ArrayList([]const u8).init(builder.allocator),
.cxxflags = ArrayList([]const u8).init(builder.allocator),
.objs = ArrayList(*Compile).init(builder.allocator),
};
try m.addCFlag("-std=c11");
try m.addCxxFlag("-std=c++11");
try m.addProjectInclude(&.{});
2023-09-21 11:08:20 +02:00
try m.addProjectInclude(&.{"common"});
return m;
}
fn obj(m: *const Maker, name: []const u8, src: []const u8) *Compile {
const o = m.builder.addObject(.{ .name = name, .target = m.target, .optimize = m.optimize });
if (o.target.getAbi() != .msvc)
o.defineCMacro("_GNU_SOURCE", null);
o.addConfigHeader(m.config_header);
if (std.mem.endsWith(u8, src, ".c")) {
o.addCSourceFiles(&.{src}, m.cflags.items);
o.linkLibC();
} else {
o.addCSourceFiles(&.{src}, m.cxxflags.items);
if (o.target.getAbi() == .msvc) {
o.linkLibC(); // need winsdk + crt
} else {
// linkLibCpp already add (libc++ + libunwind + libc)
o.linkLibCpp();
}
}
2023-09-21 11:08:20 +02:00
o.addConfigHeader(m.config_header);
for (m.include_dirs.items) |i| o.addIncludePath(.{ .path = i });
o.want_lto = m.enable_lto;
return o;
}
fn exe(m: *const Maker, name: []const u8, src: []const u8, deps: []const *Compile) *Compile {
const e = m.builder.addExecutable(.{ .name = name, .target = m.target, .optimize = m.optimize });
e.addCSourceFiles(&.{src}, m.cxxflags.items);
for (deps) |d| e.addObject(d);
for (m.objs.items) |o| e.addObject(o);
for (m.include_dirs.items) |i| e.addIncludePath(.{ .path = i });
// https://github.com/ziglang/zig/issues/15448
if (e.target.getAbi() == .msvc) {
e.linkLibC(); // need winsdk + crt
} else {
// linkLibCpp already add (libc++ + libunwind + libc)
e.linkLibCpp();
}
e.addConfigHeader(m.config_header);
m.builder.installArtifact(e);
e.want_lto = m.enable_lto;
return e;
}
};
pub fn build(b: *std.build.Builder) !void {
var make = try Maker.init(b);
make.enable_lto = b.option(bool, "lto", "Enable LTO optimization, (default: false)") orelse false;
const ggml = make.obj("ggml", "ggml.c");
const ggml_alloc = make.obj("ggml-alloc", "ggml-alloc.c");
const ggml_backend = make.obj("ggml-backend", "ggml-backend.c");
const ggml_quants = make.obj("ggml-quants", "ggml-quants.c");
const llama = make.obj("llama", "llama.cpp");
2023-09-21 11:08:20 +02:00
const common = make.obj("common", "common/common.cpp");
const console = make.obj("console", "common/console.cpp");
const sampling = make.obj("sampling", "common/sampling.cpp");
2023-09-21 11:08:20 +02:00
const grammar_parser = make.obj("grammar-parser", "common/grammar-parser.cpp");
const train = make.obj("train", "common/train.cpp");
server : parallel decoding and multimodal (#3677) * implementing parallel decoding in server example * crash fixed * save dev progress * refactored sampling function * completion endpoint working * multiple client support * grammar + no stream completion * cached prompt support * chat.mjs support cached prompt + some fixes * server ui now support multiple clients * unused change reverted * fixed timings per slot * add context swap * add changes to README.md * llava multimodal integration * fixed tokens probs * add multimodal input - alfa * refactor code + remove unused comments + improved README.md * fix compilation errors with llvm * notify the user from server ui that multimodality is unavialable * some ci fixes * fix ci make build undefined ref errors * fix long prompt than ctx proposed in #3639 * fixed premature end due stop word * context shift fixed * fix llava implementation * sync README.md changes * readme change * update api like OpenAI * multimodal support enabled by default * fix make bui;d errors * fix multiple clients * fix zig build * new sampling API * latest changes of sampling API * server : coding-style normalization * server : coding-style normalization (part 2) * server : remove beam-search functionality * server : bug fix in ingest_images n_tokens is incremented internally by llama_batch_add * server : use refs + use llama_batch_clear() * server : snake case * server : minor sync * added thread safe pipeline * server : bach has to be allocated for n_parallel sequences * server : no need for atomic int - already using mutex * server : logs + minor code style * server : fix multibyte handle in partial response (#3706) * fix image load + view image in chat * make : silence stb warnings * clip : link to ggml, not to llama * server : fix switch fallthrough * server : fix crash in Debug on macOS (I have no idea why this fixes it!?) * server : refactor ctx_sampling init + n_ctx + names * server : bug fix for prompt caching * Do not save/load image_data to localStorage * editorconfig : new line in index.html * server : completion requests remember slot_id * Update readme to document multimodal in server * server : minor style * Update readme to document multimodal in server * server : hide ctx_sampling->prev behind API (#3696) * server : apply fix from #3722 * server : fix slot reuse * server : add comment about changing slot_state to bool --------- Co-authored-by: FSSRepo <go778sgt@gmail.com> Co-authored-by: Damian Stewart <d@damianstewart.com> Co-authored-by: Steward Garcia <57494570+FSSRepo@users.noreply.github.com> Co-authored-by: Jhen-Jie Hong <iainst0409@gmail.com> Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com>
2023-10-22 21:53:08 +02:00
const clip = make.obj("clip", "examples/llava/clip.cpp");
_ = make.exe("main", "examples/main/main.cpp", &.{ ggml, ggml_alloc, ggml_backend, ggml_quants, llama, common, sampling, console, grammar_parser });
_ = make.exe("quantize", "examples/quantize/quantize.cpp", &.{ ggml, ggml_alloc, ggml_backend, ggml_quants, llama, common });
_ = make.exe("perplexity", "examples/perplexity/perplexity.cpp", &.{ ggml, ggml_alloc, ggml_backend, ggml_quants, llama, common });
_ = make.exe("embedding", "examples/embedding/embedding.cpp", &.{ ggml, ggml_alloc, ggml_backend, ggml_quants, llama, common });
_ = make.exe("finetune", "examples/finetune/finetune.cpp", &.{ ggml, ggml_alloc, ggml_backend, ggml_quants, llama, common, train });
_ = make.exe("train-text-from-scratch", "examples/train-text-from-scratch/train-text-from-scratch.cpp", &.{ ggml, ggml_alloc, ggml_backend, ggml_quants, llama, common, train });
const server = make.exe("server", "examples/server/server.cpp", &.{ ggml, ggml_alloc, ggml_backend, ggml_quants, llama, common, sampling, grammar_parser, clip });
if (server.target.isWindows()) {
server.linkSystemLibrary("ws2_32");
}
}