LocalAI/main.go

103 lines
2.6 KiB
Go
Raw Normal View History

2023-03-18 22:59:06 +00:00
package main
import (
"fmt"
2023-03-18 22:59:06 +00:00
"os"
"path/filepath"
2023-03-18 22:59:06 +00:00
api "github.com/go-skynet/LocalAI/api"
model "github.com/go-skynet/LocalAI/pkg/model"
2023-04-20 22:55:19 +00:00
"github.com/jaypipes/ghw"
2023-04-20 16:33:02 +00:00
"github.com/rs/zerolog"
"github.com/rs/zerolog/log"
2023-03-18 22:59:06 +00:00
"github.com/urfave/cli/v2"
)
func main() {
2023-04-20 16:33:02 +00:00
log.Logger = log.Output(zerolog.ConsoleWriter{Out: os.Stderr})
path, err := os.Getwd()
2023-03-18 22:59:06 +00:00
if err != nil {
2023-04-20 16:33:02 +00:00
log.Error().Msgf("error: %s", err.Error())
os.Exit(1)
2023-03-18 22:59:06 +00:00
}
2023-04-20 22:55:19 +00:00
threads := 4
cpu, err := ghw.CPU()
if err == nil {
threads = int(cpu.TotalCores)
}
2023-03-18 22:59:06 +00:00
app := &cli.App{
Name: "LocalAI",
Usage: "OpenAI compatible API for running LLaMA/GPT models locally on CPU with consumer grade hardware.",
Flags: []cli.Flag{
&cli.BoolFlag{
Name: "f16",
EnvVars: []string{"F16"},
},
2023-04-20 16:33:02 +00:00
&cli.BoolFlag{
Name: "debug",
EnvVars: []string{"DEBUG"},
},
&cli.IntFlag{
Name: "threads",
DefaultText: "Number of threads used for parallel computation. Usage of the number of physical cores in the system is suggested.",
EnvVars: []string{"THREADS"},
2023-04-20 22:55:19 +00:00
Value: threads,
2023-03-18 22:59:06 +00:00
},
&cli.StringFlag{
Name: "models-path",
DefaultText: "Path containing models used for inferencing",
EnvVars: []string{"MODELS_PATH"},
Value: filepath.Join(path, "models"),
2023-03-18 22:59:06 +00:00
},
&cli.StringFlag{
Name: "config-file",
DefaultText: "Config file",
EnvVars: []string{"CONFIG_FILE"},
},
2023-03-18 22:59:06 +00:00
&cli.StringFlag{
Name: "address",
DefaultText: "Bind address for the API server.",
EnvVars: []string{"ADDRESS"},
Value: ":8080",
},
&cli.IntFlag{
Name: "context-size",
DefaultText: "Default context size of the model",
EnvVars: []string{"CONTEXT_SIZE"},
Value: 512,
},
},
Description: `
LocalAI is a drop-in replacement OpenAI API which runs inference locally.
2023-03-18 22:59:06 +00:00
Some of the models compatible are:
- Vicuna
- Koala
- GPT4ALL
- GPT4ALL-J
- Cerebras
- Alpaca
- StableLM (ggml quantized)
2023-03-18 22:59:06 +00:00
It uses llama.cpp, ggml and gpt4all as backend with golang c bindings.
2023-03-18 22:59:06 +00:00
`,
UsageText: `local-ai [options]`,
2023-03-18 22:59:06 +00:00
Copyright: "go-skynet authors",
Action: func(ctx *cli.Context) error {
fmt.Printf("Starting LocalAI using %d threads, with models path: %s\n", ctx.Int("threads"), ctx.String("models-path"))
return api.App(ctx.String("config-file"), model.NewModelLoader(ctx.String("models-path")), ctx.Int("threads"), ctx.Int("context-size"), ctx.Bool("f16"), ctx.Bool("debug"), false).Listen(ctx.String("address"))
2023-03-18 22:59:06 +00:00
},
}
err = app.Run(os.Args)
2023-03-18 22:59:06 +00:00
if err != nil {
2023-04-20 16:33:02 +00:00
log.Error().Msgf("error: %s", err.Error())
2023-03-18 22:59:06 +00:00
os.Exit(1)
}
}