Compare commits
37 Commits
| Author | SHA1 | Date | |
|---|---|---|---|
|
|
88a5d11e15 | ||
|
|
1ecb0b444b | ||
|
|
72d601370d | ||
|
|
4814b88172 | ||
|
|
cfad67a922 | ||
|
|
c28f5604ab | ||
|
|
5853b8ca8d | ||
|
|
ebfc0ce672 | ||
|
|
e62fcd152a | ||
|
|
9bd9b9ecbd | ||
|
|
17faa9c5b8 | ||
|
|
4cd445bf77 | ||
|
|
4fb35845b0 | ||
|
|
f373f1caa8 | ||
|
|
4e75531651 | ||
|
|
539c538d65 | ||
|
|
d90186db33 | ||
|
|
a2d8729ae3 | ||
|
|
edc6ac7297 | ||
|
|
e89e23621c | ||
|
|
6b9ec4c6fa | ||
|
|
ced0966ffc | ||
|
|
966b912013 | ||
|
|
dc71054e61 | ||
|
|
408f3c1a4d | ||
|
|
38b775c937 | ||
|
|
f2ec1067bf | ||
|
|
b01584c49e | ||
|
|
01e56382a3 | ||
|
|
391c067250 | ||
|
|
5b98b5c0a7 | ||
|
|
5bde0abb8d | ||
|
|
1036852924 | ||
|
|
2b10ccd507 | ||
|
|
e1df1bfc3f | ||
|
|
b41a2e7039 | ||
|
|
b63370928d |
3
.gitignore
vendored
@@ -16,3 +16,6 @@ __pycache__
|
||||
/backend-python/wkv_cuda
|
||||
*.exe
|
||||
*.old
|
||||
.DS_Store
|
||||
*.log.*
|
||||
*.log
|
||||
19
.vscode/launch.json
vendored
@@ -10,9 +10,24 @@
|
||||
"name": "Python",
|
||||
"type": "python",
|
||||
"request": "launch",
|
||||
"program": "./backend-python/main.py",
|
||||
"program": "${workspaceFolder}/backend-python/main.py",
|
||||
"console": "integratedTerminal",
|
||||
"justMyCode": false,
|
||||
"justMyCode": false
|
||||
},
|
||||
{
|
||||
"name": "Golang",
|
||||
"type": "go",
|
||||
"request": "launch",
|
||||
"mode": "exec",
|
||||
"program": "${workspaceFolder}/build/bin/testwails.exe",
|
||||
"console": "integratedTerminal",
|
||||
"preLaunchTask": "build dev"
|
||||
},
|
||||
{
|
||||
"name": "Frontend",
|
||||
"type": "node-terminal",
|
||||
"request": "launch",
|
||||
"command": "wails dev -browser"
|
||||
}
|
||||
]
|
||||
}
|
||||
40
.vscode/tasks.json
vendored
Normal file
@@ -0,0 +1,40 @@
|
||||
{
|
||||
"version": "2.0.0",
|
||||
"tasks": [
|
||||
{
|
||||
"label": "build dev",
|
||||
"type": "shell",
|
||||
"options": {
|
||||
"cwd": "${workspaceFolder}",
|
||||
"env": {
|
||||
"CGO_ENABLED": "1"
|
||||
}
|
||||
},
|
||||
"osx": {
|
||||
"options": {
|
||||
"env": {
|
||||
"CGO_CFLAGS": "-mmacosx-version-min=10.13",
|
||||
"CGO_LDFLAGS": "-framework UniformTypeIdentifiers -mmacosx-version-min=10.13"
|
||||
}
|
||||
}
|
||||
},
|
||||
"windows": {
|
||||
"options": {
|
||||
"env": {
|
||||
"CGO_ENABLED": "0"
|
||||
}
|
||||
}
|
||||
},
|
||||
"command": "go",
|
||||
"args": [
|
||||
"build",
|
||||
"-tags",
|
||||
"dev",
|
||||
"-gcflags",
|
||||
"all=-N -l",
|
||||
"-o",
|
||||
"build/bin/testwails.exe"
|
||||
]
|
||||
}
|
||||
]
|
||||
}
|
||||
11
Makefile
@@ -1,15 +1,22 @@
|
||||
ifeq ($(OS), Windows_NT)
|
||||
build: build-windows
|
||||
else
|
||||
else ifeq ($(shell uname -s), Darwin)
|
||||
build: build-macos
|
||||
else
|
||||
build: build-linux
|
||||
endif
|
||||
|
||||
build-windows:
|
||||
@echo ---- build for windows
|
||||
wails build -upx -ldflags "-s -w"
|
||||
wails build -upx -ldflags "-s -w" -platform windows/amd64
|
||||
|
||||
build-macos:
|
||||
@echo ---- build for macos
|
||||
wails build -ldflags "-s -w" -platform darwin/universal
|
||||
|
||||
build-linux:
|
||||
@echo ---- build for linux
|
||||
wails build -upx -ldflags "-s -w" -platform linux/amd64
|
||||
|
||||
dev:
|
||||
wails dev
|
||||
|
||||
@@ -73,6 +73,7 @@ body.json:
|
||||
- [x] CUDA operator int8 acceleration
|
||||
- [ ] macOS support
|
||||
- [ ] Linux support
|
||||
- [ ] Local State Cache DB
|
||||
|
||||
## Related Repositories:
|
||||
|
||||
|
||||
@@ -14,7 +14,7 @@ API兼容的接口,这意味着一切ChatGPT客户端都是RWKV客户端。
|
||||
|
||||
[English](README.md) | 简体中文
|
||||
|
||||
[视频演示](https://www.bilibili.com/video/BV1hM4y1v76R) | [疑难解答](https://www.bilibili.com/read/cv23921171) | [预览](#Preview) | [下载][download-url]
|
||||
[视频演示](https://www.bilibili.com/video/BV1hM4y1v76R) | [疑难解答](https://www.bilibili.com/read/cv23921171) | [预览](#Preview) | [下载][download-url] | [懒人包](https://pan.baidu.com/s/1wchIUHgne3gncIiLIeKBEQ?pwd=1111)
|
||||
|
||||
[license-image]: http://img.shields.io/badge/license-MIT-blue.svg
|
||||
|
||||
@@ -73,6 +73,7 @@ body.json:
|
||||
- [x] CUDA算子int8提速
|
||||
- [ ] macOS支持
|
||||
- [ ] linux支持
|
||||
- [ ] 本地状态缓存数据库
|
||||
|
||||
## 相关仓库:
|
||||
|
||||
|
||||
@@ -5,6 +5,7 @@ import (
|
||||
"net/http"
|
||||
"os"
|
||||
"os/exec"
|
||||
"path/filepath"
|
||||
"runtime"
|
||||
|
||||
"github.com/minio/selfupdate"
|
||||
@@ -13,7 +14,9 @@ import (
|
||||
|
||||
// App struct
|
||||
type App struct {
|
||||
ctx context.Context
|
||||
ctx context.Context
|
||||
exDir string
|
||||
cmdPrefix string
|
||||
}
|
||||
|
||||
// NewApp creates a new App application struct
|
||||
@@ -25,6 +28,14 @@ func NewApp() *App {
|
||||
// so we can call the runtime methods
|
||||
func (a *App) OnStartup(ctx context.Context) {
|
||||
a.ctx = ctx
|
||||
a.exDir = ""
|
||||
a.cmdPrefix = ""
|
||||
|
||||
if runtime.GOOS == "darwin" {
|
||||
ex, _ := os.Executable()
|
||||
a.exDir = filepath.Dir(ex) + "/../../../"
|
||||
a.cmdPrefix = "cd " + a.exDir + " && "
|
||||
}
|
||||
|
||||
a.downloadLoop()
|
||||
}
|
||||
|
||||
@@ -1,6 +1,7 @@
|
||||
package backend_golang
|
||||
|
||||
import (
|
||||
"context"
|
||||
"path/filepath"
|
||||
"time"
|
||||
|
||||
@@ -9,7 +10,7 @@ import (
|
||||
)
|
||||
|
||||
func (a *App) DownloadFile(path string, url string) error {
|
||||
_, err := grab.Get(path, url)
|
||||
_, err := grab.Get(a.exDir+path, url)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
@@ -18,6 +19,7 @@ func (a *App) DownloadFile(path string, url string) error {
|
||||
|
||||
type DownloadStatus struct {
|
||||
resp *grab.Response
|
||||
cancel context.CancelFunc
|
||||
Name string `json:"name"`
|
||||
Path string `json:"path"`
|
||||
Url string `json:"url"`
|
||||
@@ -29,7 +31,7 @@ type DownloadStatus struct {
|
||||
Done bool `json:"done"`
|
||||
}
|
||||
|
||||
var downloadList []DownloadStatus
|
||||
var downloadList []*DownloadStatus
|
||||
|
||||
func existsInDownloadList(url string) bool {
|
||||
for _, ds := range downloadList {
|
||||
@@ -41,49 +43,58 @@ func existsInDownloadList(url string) bool {
|
||||
}
|
||||
|
||||
func (a *App) PauseDownload(url string) {
|
||||
for i, ds := range downloadList {
|
||||
for _, ds := range downloadList {
|
||||
if ds.Url == url {
|
||||
if ds.resp != nil {
|
||||
ds.resp.Cancel()
|
||||
}
|
||||
|
||||
downloadList[i] = DownloadStatus{
|
||||
resp: ds.resp,
|
||||
Name: ds.Name,
|
||||
Path: ds.Path,
|
||||
Url: ds.Url,
|
||||
Downloading: false,
|
||||
if ds.cancel != nil {
|
||||
ds.cancel()
|
||||
}
|
||||
ds.resp = nil
|
||||
ds.Downloading = false
|
||||
ds.Speed = 0
|
||||
break
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func (a *App) ContinueDownload(url string) {
|
||||
for i, ds := range downloadList {
|
||||
for _, ds := range downloadList {
|
||||
if ds.Url == url {
|
||||
client := grab.NewClient()
|
||||
req, _ := grab.NewRequest(ds.Path, ds.Url)
|
||||
resp := client.Do(req)
|
||||
if !ds.Downloading && ds.resp == nil && !ds.Done {
|
||||
ds.Downloading = true
|
||||
|
||||
downloadList[i] = DownloadStatus{
|
||||
resp: resp,
|
||||
Name: ds.Name,
|
||||
Path: ds.Path,
|
||||
Url: ds.Url,
|
||||
Downloading: true,
|
||||
req, err := grab.NewRequest(ds.Path, ds.Url)
|
||||
if err != nil {
|
||||
ds.Downloading = false
|
||||
break
|
||||
}
|
||||
// if PauseDownload() is called before the request finished, ds.Downloading will be false
|
||||
// if the user keeps clicking pause and resume, it may result in multiple requests being successfully downloaded at the same time
|
||||
// so we have to create a context and cancel it when PauseDownload() is called
|
||||
ctx, cancel := context.WithCancel(context.Background())
|
||||
ds.cancel = cancel
|
||||
req = req.WithContext(ctx)
|
||||
resp := grab.DefaultClient.Do(req)
|
||||
|
||||
if resp != nil && resp.HTTPResponse != nil &&
|
||||
resp.HTTPResponse.StatusCode >= 200 && resp.HTTPResponse.StatusCode < 300 {
|
||||
ds.resp = resp
|
||||
} else {
|
||||
ds.Downloading = false
|
||||
}
|
||||
}
|
||||
break
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func (a *App) AddToDownloadList(path string, url string) {
|
||||
if !existsInDownloadList(url) {
|
||||
downloadList = append(downloadList, DownloadStatus{
|
||||
downloadList = append(downloadList, &DownloadStatus{
|
||||
resp: nil,
|
||||
Name: filepath.Base(path),
|
||||
Path: path,
|
||||
Path: a.exDir + path,
|
||||
Url: url,
|
||||
Downloading: true,
|
||||
Downloading: false,
|
||||
})
|
||||
a.ContinueDownload(url)
|
||||
} else {
|
||||
@@ -96,32 +107,17 @@ func (a *App) downloadLoop() {
|
||||
go func() {
|
||||
for {
|
||||
<-ticker.C
|
||||
for i, ds := range downloadList {
|
||||
transferred := int64(0)
|
||||
size := int64(0)
|
||||
speed := float64(0)
|
||||
progress := float64(0)
|
||||
downloading := ds.Downloading
|
||||
done := false
|
||||
for _, ds := range downloadList {
|
||||
if ds.resp != nil {
|
||||
transferred = ds.resp.BytesComplete()
|
||||
size = ds.resp.Size()
|
||||
speed = ds.resp.BytesPerSecond()
|
||||
progress = 100 * ds.resp.Progress()
|
||||
downloading = !ds.resp.IsComplete()
|
||||
done = ds.resp.Progress() == 1
|
||||
}
|
||||
downloadList[i] = DownloadStatus{
|
||||
resp: ds.resp,
|
||||
Name: ds.Name,
|
||||
Path: ds.Path,
|
||||
Url: ds.Url,
|
||||
Transferred: transferred,
|
||||
Size: size,
|
||||
Speed: speed,
|
||||
Progress: progress,
|
||||
Downloading: downloading,
|
||||
Done: done,
|
||||
ds.Transferred = ds.resp.BytesComplete()
|
||||
ds.Size = ds.resp.Size()
|
||||
ds.Speed = ds.resp.BytesPerSecond()
|
||||
ds.Progress = 100 * ds.resp.Progress()
|
||||
ds.Downloading = !ds.resp.IsComplete()
|
||||
ds.Done = ds.resp.Progress() == 1
|
||||
if !ds.Downloading {
|
||||
ds.resp = nil
|
||||
}
|
||||
}
|
||||
}
|
||||
runtime.EventsEmit(a.ctx, "downloadList", downloadList)
|
||||
|
||||
@@ -2,7 +2,7 @@ package backend_golang
|
||||
|
||||
import (
|
||||
"encoding/json"
|
||||
"fmt"
|
||||
"errors"
|
||||
"io"
|
||||
"os"
|
||||
"os/exec"
|
||||
@@ -18,14 +18,14 @@ func (a *App) SaveJson(fileName string, jsonData any) error {
|
||||
return err
|
||||
}
|
||||
|
||||
if err := os.WriteFile(fileName, text, 0644); err != nil {
|
||||
if err := os.WriteFile(a.exDir+fileName, text, 0644); err != nil {
|
||||
return err
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
func (a *App) ReadJson(fileName string) (any, error) {
|
||||
file, err := os.ReadFile(fileName)
|
||||
file, err := os.ReadFile(a.exDir + fileName)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
@@ -40,7 +40,7 @@ func (a *App) ReadJson(fileName string) (any, error) {
|
||||
}
|
||||
|
||||
func (a *App) FileExists(fileName string) bool {
|
||||
_, err := os.Stat(fileName)
|
||||
_, err := os.Stat(a.exDir + fileName)
|
||||
return err == nil
|
||||
}
|
||||
|
||||
@@ -52,7 +52,7 @@ type FileInfo struct {
|
||||
}
|
||||
|
||||
func (a *App) ReadFileInfo(fileName string) (FileInfo, error) {
|
||||
info, err := os.Stat(fileName)
|
||||
info, err := os.Stat(a.exDir + fileName)
|
||||
if err != nil {
|
||||
return FileInfo{}, err
|
||||
}
|
||||
@@ -65,7 +65,7 @@ func (a *App) ReadFileInfo(fileName string) (FileInfo, error) {
|
||||
}
|
||||
|
||||
func (a *App) ListDirFiles(dirPath string) ([]FileInfo, error) {
|
||||
files, err := os.ReadDir(dirPath)
|
||||
files, err := os.ReadDir(a.exDir + dirPath)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
@@ -87,7 +87,7 @@ func (a *App) ListDirFiles(dirPath string) ([]FileInfo, error) {
|
||||
}
|
||||
|
||||
func (a *App) DeleteFile(path string) error {
|
||||
err := os.Remove(path)
|
||||
err := os.Remove(a.exDir + path)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
@@ -95,18 +95,18 @@ func (a *App) DeleteFile(path string) error {
|
||||
}
|
||||
|
||||
func (a *App) CopyFile(src string, dst string) error {
|
||||
sourceFile, err := os.Open(src)
|
||||
sourceFile, err := os.Open(a.exDir + src)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
defer sourceFile.Close()
|
||||
|
||||
err = os.MkdirAll(dst[:strings.LastIndex(dst, "/")], 0755)
|
||||
err = os.MkdirAll(a.exDir+dst[:strings.LastIndex(dst, "/")], 0755)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
destFile, err := os.Create(dst)
|
||||
destFile, err := os.Create(a.exDir + dst)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
@@ -120,7 +120,7 @@ func (a *App) CopyFile(src string, dst string) error {
|
||||
}
|
||||
|
||||
func (a *App) OpenFileFolder(path string) error {
|
||||
absPath, err := filepath.Abs(path)
|
||||
absPath, err := filepath.Abs(a.exDir + path)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
@@ -131,10 +131,21 @@ func (a *App) OpenFileFolder(path string) error {
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
return nil
|
||||
case "darwin":
|
||||
fmt.Println("Running on macOS")
|
||||
cmd := exec.Command("open", "-R", absPath)
|
||||
err := cmd.Run()
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
return nil
|
||||
case "linux":
|
||||
fmt.Println("Running on Linux")
|
||||
cmd := exec.Command("xdg-open", absPath)
|
||||
err := cmd.Run()
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
return nil
|
||||
}
|
||||
return nil
|
||||
return errors.New("unsupported OS")
|
||||
}
|
||||
|
||||
@@ -37,7 +37,7 @@ func (a *App) DepCheck(python string) error {
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
out, err := exec.Command(python, "./backend-python/dep_check.py").CombinedOutput()
|
||||
out, err := exec.Command(python, a.exDir+"./backend-python/dep_check.py").CombinedOutput()
|
||||
if err != nil {
|
||||
return errors.New("DepCheck Error: " + string(out))
|
||||
}
|
||||
@@ -63,7 +63,11 @@ func (a *App) InstallPyDep(python string, cnMirror bool) (string, error) {
|
||||
if err != nil {
|
||||
return "", err
|
||||
}
|
||||
_, err = Cmd(python, "-m", "pip", "install", "torch==1.13.1", "torchvision==0.14.1", "torchaudio==0.13.1", "--index-url", "https://download.pytorch.org/whl/cu117")
|
||||
if runtime.GOOS == "windows" {
|
||||
_, err = Cmd(python, "-m", "pip", "install", "torch==1.13.1", "torchvision==0.14.1", "torchaudio==0.13.1", "--index-url", "https://download.pytorch.org/whl/cu117")
|
||||
} else {
|
||||
_, err = Cmd(python, "-m", "pip", "install", "torch", "torchvision", "torchaudio")
|
||||
}
|
||||
if err != nil {
|
||||
return "", err
|
||||
}
|
||||
|
||||
@@ -15,29 +15,42 @@ import (
|
||||
)
|
||||
|
||||
func Cmd(args ...string) (string, error) {
|
||||
path, err := filepath.Abs(args[0])
|
||||
if err != nil {
|
||||
return "", err
|
||||
}
|
||||
args[0] = path
|
||||
if runtime.GOOS == "windows" {
|
||||
_, err := os.Stat("cmd-helper.bat")
|
||||
if err != nil {
|
||||
if err := os.WriteFile("./cmd-helper.bat", []byte("start %*"), 0644); err != nil {
|
||||
return "", err
|
||||
}
|
||||
switch platform := runtime.GOOS; platform {
|
||||
case "windows":
|
||||
if err := os.WriteFile("./cmd-helper.bat", []byte("start /wait %*"), 0644); err != nil {
|
||||
return "", err
|
||||
}
|
||||
cmdHelper, err := filepath.Abs("./cmd-helper")
|
||||
if err != nil {
|
||||
return "", err
|
||||
}
|
||||
|
||||
for _, arg := range args {
|
||||
if strings.Contains(arg, " ") && strings.Contains(cmdHelper, " ") {
|
||||
return "", errors.New("path contains space") // golang bug https://github.com/golang/go/issues/17149#issuecomment-473976818
|
||||
}
|
||||
}
|
||||
|
||||
cmd := exec.Command(cmdHelper, args...)
|
||||
out, err := cmd.CombinedOutput()
|
||||
if err != nil {
|
||||
return "", err
|
||||
}
|
||||
return string(out), nil
|
||||
} else {
|
||||
case "darwin":
|
||||
ex, err := os.Executable()
|
||||
if err != nil {
|
||||
return "", err
|
||||
}
|
||||
exDir := filepath.Dir(ex) + "/../../../"
|
||||
cmd := exec.Command("osascript", "-e", `tell application "Terminal" to do script "`+"cd "+exDir+" && "+strings.Join(args, " ")+`"`)
|
||||
err = cmd.Start()
|
||||
if err != nil {
|
||||
return "", err
|
||||
}
|
||||
cmd.Wait()
|
||||
return "", nil
|
||||
case "linux":
|
||||
cmd := exec.Command(args[0], args[1:]...)
|
||||
err := cmd.Start()
|
||||
if err != nil {
|
||||
@@ -46,9 +59,19 @@ func Cmd(args ...string) (string, error) {
|
||||
cmd.Wait()
|
||||
return "", nil
|
||||
}
|
||||
return "", errors.New("unsupported OS")
|
||||
}
|
||||
|
||||
func CopyEmbed(efs embed.FS) error {
|
||||
prefix := ""
|
||||
if runtime.GOOS == "darwin" {
|
||||
ex, err := os.Executable()
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
prefix = filepath.Dir(ex) + "/../../../"
|
||||
}
|
||||
|
||||
err := fs.WalkDir(efs, ".", func(path string, d fs.DirEntry, err error) error {
|
||||
if d.IsDir() {
|
||||
return nil
|
||||
@@ -61,6 +84,7 @@ func CopyEmbed(efs embed.FS) error {
|
||||
return err
|
||||
}
|
||||
|
||||
path = prefix + path
|
||||
err = os.MkdirAll(path[:strings.LastIndex(path, "/")], 0755)
|
||||
if err != nil {
|
||||
return err
|
||||
|
||||
@@ -1,8 +1,6 @@
|
||||
import cyac
|
||||
import GPUtil
|
||||
import torch
|
||||
import rwkv
|
||||
import langchain
|
||||
import fastapi
|
||||
import uvicorn
|
||||
import sse_starlette
|
||||
|
||||
@@ -4,17 +4,18 @@ import sys
|
||||
sys.path.append(os.path.dirname(os.path.realpath(__file__)))
|
||||
|
||||
import psutil
|
||||
from fastapi import FastAPI
|
||||
from fastapi import Depends, FastAPI
|
||||
from fastapi.middleware.cors import CORSMiddleware
|
||||
import uvicorn
|
||||
|
||||
from utils.rwkv import *
|
||||
from utils.torch import *
|
||||
from utils.ngrok import *
|
||||
from utils.log import log_middleware
|
||||
from routes import completion, config, state_cache
|
||||
import global_var
|
||||
|
||||
app = FastAPI()
|
||||
app = FastAPI(dependencies=[Depends(log_middleware)])
|
||||
|
||||
app.add_middleware(
|
||||
CORSMiddleware,
|
||||
@@ -42,7 +43,7 @@ def init():
|
||||
|
||||
@app.get("/")
|
||||
def read_root():
|
||||
return {"Hello": "World!", "pid": os.getpid()}
|
||||
return {"Hello": "World!"}
|
||||
|
||||
|
||||
@app.post("/exit")
|
||||
@@ -60,7 +61,7 @@ def debug():
|
||||
strategy="cuda fp16",
|
||||
tokens_path="20B_tokenizer.json",
|
||||
)
|
||||
d = model.tokenizer.decode([])
|
||||
d = model.pipeline.decode([])
|
||||
print(d)
|
||||
|
||||
|
||||
|
||||
BIN
backend-python/requirements_without_cyac.txt
Normal file
@@ -7,6 +7,7 @@ from fastapi import APIRouter, Request, status, HTTPException
|
||||
from sse_starlette.sse import EventSourceResponse
|
||||
from pydantic import BaseModel
|
||||
from utils.rwkv import *
|
||||
from utils.log import quick_log
|
||||
import global_var
|
||||
|
||||
router = APIRouter()
|
||||
@@ -26,6 +27,8 @@ class ChatCompletionBody(ModelConfigBody):
|
||||
|
||||
completion_lock = Lock()
|
||||
|
||||
requests_num = 0
|
||||
|
||||
|
||||
@router.post("/v1/chat/completions")
|
||||
@router.post("/chat/completions")
|
||||
@@ -106,12 +109,32 @@ The following is a coherent verbose detailed conversation between a girl named {
|
||||
completion_text += f"{bot}{interface}"
|
||||
|
||||
async def eval_rwkv():
|
||||
global requests_num
|
||||
requests_num = requests_num + 1
|
||||
quick_log(request, None, "Start Waiting. RequestsNum: " + str(requests_num))
|
||||
while completion_lock.locked():
|
||||
if await request.is_disconnected():
|
||||
requests_num = requests_num - 1
|
||||
print(f"{request.client} Stop Waiting (Lock)")
|
||||
quick_log(
|
||||
request,
|
||||
None,
|
||||
"Stop Waiting (Lock). RequestsNum: " + str(requests_num),
|
||||
)
|
||||
return
|
||||
await asyncio.sleep(0.1)
|
||||
else:
|
||||
completion_lock.acquire()
|
||||
if await request.is_disconnected():
|
||||
completion_lock.release()
|
||||
requests_num = requests_num - 1
|
||||
print(f"{request.client} Stop Waiting (Lock)")
|
||||
quick_log(
|
||||
request,
|
||||
None,
|
||||
"Stop Waiting (Lock). RequestsNum: " + str(requests_num),
|
||||
)
|
||||
return
|
||||
set_rwkv_config(model, global_var.get(global_var.Model_Config))
|
||||
set_rwkv_config(model, body)
|
||||
if body.stream:
|
||||
@@ -135,9 +158,21 @@ The following is a coherent verbose detailed conversation between a girl named {
|
||||
}
|
||||
)
|
||||
# torch_gc()
|
||||
requests_num = requests_num - 1
|
||||
completion_lock.release()
|
||||
if await request.is_disconnected():
|
||||
print(f"{request.client} Stop Waiting")
|
||||
quick_log(
|
||||
request,
|
||||
body,
|
||||
response + "\nStop Waiting. RequestsNum: " + str(requests_num),
|
||||
)
|
||||
return
|
||||
quick_log(
|
||||
request,
|
||||
body,
|
||||
response + "\nFinished. RequestsNum: " + str(requests_num),
|
||||
)
|
||||
yield json.dumps(
|
||||
{
|
||||
"response": response,
|
||||
@@ -161,9 +196,21 @@ The following is a coherent verbose detailed conversation between a girl named {
|
||||
if await request.is_disconnected():
|
||||
break
|
||||
# torch_gc()
|
||||
requests_num = requests_num - 1
|
||||
completion_lock.release()
|
||||
if await request.is_disconnected():
|
||||
print(f"{request.client} Stop Waiting")
|
||||
quick_log(
|
||||
request,
|
||||
body,
|
||||
response + "\nStop Waiting. RequestsNum: " + str(requests_num),
|
||||
)
|
||||
return
|
||||
quick_log(
|
||||
request,
|
||||
body,
|
||||
response + "\nFinished. RequestsNum: " + str(requests_num),
|
||||
)
|
||||
yield {
|
||||
"response": response,
|
||||
"model": "rwkv",
|
||||
@@ -182,7 +229,10 @@ The following is a coherent verbose detailed conversation between a girl named {
|
||||
if body.stream:
|
||||
return EventSourceResponse(eval_rwkv())
|
||||
else:
|
||||
return await eval_rwkv().__anext__()
|
||||
try:
|
||||
return await eval_rwkv().__anext__()
|
||||
except StopAsyncIteration:
|
||||
return None
|
||||
|
||||
|
||||
class CompletionBody(ModelConfigBody):
|
||||
@@ -203,12 +253,32 @@ async def completions(body: CompletionBody, request: Request):
|
||||
raise HTTPException(status.HTTP_400_BAD_REQUEST, "prompt not found")
|
||||
|
||||
async def eval_rwkv():
|
||||
global requests_num
|
||||
requests_num = requests_num + 1
|
||||
quick_log(request, None, "Start Waiting. RequestsNum: " + str(requests_num))
|
||||
while completion_lock.locked():
|
||||
if await request.is_disconnected():
|
||||
requests_num = requests_num - 1
|
||||
print(f"{request.client} Stop Waiting (Lock)")
|
||||
quick_log(
|
||||
request,
|
||||
None,
|
||||
"Stop Waiting (Lock). RequestsNum: " + str(requests_num),
|
||||
)
|
||||
return
|
||||
await asyncio.sleep(0.1)
|
||||
else:
|
||||
completion_lock.acquire()
|
||||
if await request.is_disconnected():
|
||||
completion_lock.release()
|
||||
requests_num = requests_num - 1
|
||||
print(f"{request.client} Stop Waiting (Lock)")
|
||||
quick_log(
|
||||
request,
|
||||
None,
|
||||
"Stop Waiting (Lock). RequestsNum: " + str(requests_num),
|
||||
)
|
||||
return
|
||||
set_rwkv_config(model, global_var.get(global_var.Model_Config))
|
||||
set_rwkv_config(model, body)
|
||||
if body.stream:
|
||||
@@ -229,9 +299,21 @@ async def completions(body: CompletionBody, request: Request):
|
||||
}
|
||||
)
|
||||
# torch_gc()
|
||||
requests_num = requests_num - 1
|
||||
completion_lock.release()
|
||||
if await request.is_disconnected():
|
||||
print(f"{request.client} Stop Waiting")
|
||||
quick_log(
|
||||
request,
|
||||
body,
|
||||
response + "\nStop Waiting. RequestsNum: " + str(requests_num),
|
||||
)
|
||||
return
|
||||
quick_log(
|
||||
request,
|
||||
body,
|
||||
response + "\nFinished. RequestsNum: " + str(requests_num),
|
||||
)
|
||||
yield json.dumps(
|
||||
{
|
||||
"response": response,
|
||||
@@ -252,9 +334,21 @@ async def completions(body: CompletionBody, request: Request):
|
||||
if await request.is_disconnected():
|
||||
break
|
||||
# torch_gc()
|
||||
requests_num = requests_num - 1
|
||||
completion_lock.release()
|
||||
if await request.is_disconnected():
|
||||
print(f"{request.client} Stop Waiting")
|
||||
quick_log(
|
||||
request,
|
||||
body,
|
||||
response + "\nStop Waiting. RequestsNum: " + str(requests_num),
|
||||
)
|
||||
return
|
||||
quick_log(
|
||||
request,
|
||||
body,
|
||||
response + "\nFinished. RequestsNum: " + str(requests_num),
|
||||
)
|
||||
yield {
|
||||
"response": response,
|
||||
"model": "rwkv",
|
||||
@@ -270,4 +364,7 @@ async def completions(body: CompletionBody, request: Request):
|
||||
if body.stream:
|
||||
return EventSourceResponse(eval_rwkv())
|
||||
else:
|
||||
return await eval_rwkv().__anext__()
|
||||
try:
|
||||
return await eval_rwkv().__anext__()
|
||||
except StopAsyncIteration:
|
||||
return None
|
||||
|
||||
@@ -2,7 +2,6 @@ import pathlib
|
||||
|
||||
from fastapi import APIRouter, HTTPException, Response, status as Status
|
||||
from pydantic import BaseModel
|
||||
from langchain.llms import RWKV
|
||||
from utils.rwkv import *
|
||||
from utils.torch import *
|
||||
import global_var
|
||||
|
||||
@@ -3,6 +3,7 @@ from fastapi import APIRouter, HTTPException, Response, status
|
||||
from pydantic import BaseModel
|
||||
import gc
|
||||
import copy
|
||||
import torch
|
||||
|
||||
router = APIRouter()
|
||||
|
||||
@@ -41,10 +42,14 @@ def add_state(body: AddStateBody):
|
||||
raise HTTPException(status.HTTP_400_BAD_REQUEST, "trie not loaded")
|
||||
|
||||
id = trie.insert(body.prompt)
|
||||
device = body.state[0].device
|
||||
dtrie[id] = {
|
||||
"tokens": copy.deepcopy(body.tokens),
|
||||
"state": copy.deepcopy(body.state),
|
||||
"state": [tensor.cpu() for tensor in body.state]
|
||||
if device != torch.device("cpu")
|
||||
else copy.deepcopy(body.state),
|
||||
"logits": copy.deepcopy(body.logits),
|
||||
"device": device,
|
||||
}
|
||||
|
||||
return "success"
|
||||
@@ -77,11 +82,15 @@ def longest_prefix_state(body: LongestPrefixStateBody):
|
||||
pass
|
||||
if id != -1:
|
||||
v = dtrie[id]
|
||||
device = v["device"]
|
||||
return {
|
||||
"prompt": trie[id],
|
||||
"tokens": v["tokens"],
|
||||
"state": v["state"],
|
||||
"state": [tensor.to(device) for tensor in v["state"]]
|
||||
if device != torch.device("cpu")
|
||||
else v["state"],
|
||||
"logits": v["logits"],
|
||||
"device": device,
|
||||
}
|
||||
else:
|
||||
return {"prompt": "", "tokens": [], "state": None, "logits": None}
|
||||
|
||||
32
backend-python/utils/log.py
Normal file
@@ -0,0 +1,32 @@
|
||||
import json
|
||||
import logging
|
||||
from typing import Any
|
||||
from fastapi import Request
|
||||
|
||||
|
||||
logger = logging.getLogger()
|
||||
logger.setLevel(logging.INFO)
|
||||
formatter = logging.Formatter("%(asctime)s - %(levelname)s\n%(message)s")
|
||||
fh = logging.handlers.RotatingFileHandler(
|
||||
"api.log", mode="a", maxBytes=3 * 1024 * 1024, backupCount=3
|
||||
)
|
||||
fh.setFormatter(formatter)
|
||||
logger.addHandler(fh)
|
||||
|
||||
|
||||
def quick_log(request: Request, body: Any, response: str):
|
||||
logger.info(
|
||||
f"Client: {request.client}\nUrl: {request.url}\n"
|
||||
+ (
|
||||
f"Body: {json.dumps(body.__dict__, default=vars, ensure_ascii=False)}\n"
|
||||
if body
|
||||
else ""
|
||||
)
|
||||
+ (f"Response:\n{response}\n" if response else "")
|
||||
)
|
||||
|
||||
|
||||
async def log_middleware(request: Request):
|
||||
logger.info(
|
||||
f"Client: {request.client}\nUrl: {request.url}\nBody: {await request.body()}\n"
|
||||
)
|
||||
@@ -204,7 +204,10 @@ def set_rwkv_config(model: RWKV, body: ModelConfigBody):
|
||||
if body.max_tokens is not None:
|
||||
model.max_tokens_per_generation = body.max_tokens
|
||||
if body.temperature is not None:
|
||||
model.temperature = body.temperature
|
||||
if body.temperature < 0.1:
|
||||
model.temperature = 0.1
|
||||
else:
|
||||
model.temperature = body.temperature
|
||||
if body.top_p is not None:
|
||||
model.top_p = body.top_p
|
||||
if body.presence_penalty is not None:
|
||||
|
||||
BIN
build/appicon.png
vendored
|
Before Width: | Height: | Size: 102 KiB After Width: | Height: | Size: 120 KiB |
2
build/darwin/Info.dev.plist
vendored
@@ -8,7 +8,7 @@
|
||||
<key>CFBundleExecutable</key>
|
||||
<string>{{.Name}}</string>
|
||||
<key>CFBundleIdentifier</key>
|
||||
<string>com.wails.{{.Name}}</string>
|
||||
<string>dev.josStorer.RWKV-Runner</string>
|
||||
<key>CFBundleVersion</key>
|
||||
<string>{{.Info.ProductVersion}}</string>
|
||||
<key>CFBundleGetInfoString</key>
|
||||
|
||||
2
build/darwin/Info.plist
vendored
@@ -8,7 +8,7 @@
|
||||
<key>CFBundleExecutable</key>
|
||||
<string>{{.Name}}</string>
|
||||
<key>CFBundleIdentifier</key>
|
||||
<string>com.wails.{{.Name}}</string>
|
||||
<string>dev.josStorer.RWKV-Runner</string>
|
||||
<key>CFBundleVersion</key>
|
||||
<string>{{.Info.ProductVersion}}</string>
|
||||
<key>CFBundleGetInfoString</key>
|
||||
|
||||
9
build/darwin/Readme_Install.txt
vendored
Normal file
@@ -0,0 +1,9 @@
|
||||
Please execute this program in an empty directory. All related dependencies will be placed in this directory.
|
||||
请将本程序放在一个空目录内执行, 所有相关依赖均会放置于此目录.
|
||||
このプログラムを空のディレクトリで実行してください。関連するすべての依存関係は、このディレクトリに配置されます。
|
||||
|
||||
Please execute the following command in the terminal to remove the permission restrictions of this app, and then this program can work properly:
|
||||
请在终端执行以下命令解除本app的权限限制, 然后本程序才可以正常工作:
|
||||
このアプリの権限制限を解除するために、ターミナルで以下のコマンドを実行してください。その後、このプログラムは正常に動作するようになります:
|
||||
|
||||
sudo xattr -r -d com.apple.quarantine ./RWKV-Runner.app
|
||||
16
build/darwin/entitlements.plist
vendored
Normal file
@@ -0,0 +1,16 @@
|
||||
<?xml version="1.0" encoding="UTF-8"?>
|
||||
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
|
||||
<plist version="1.0">
|
||||
<dict>
|
||||
<key>com.apple.security.app-sandbox</key>
|
||||
<false/>
|
||||
<key>com.apple.security.network.client</key>
|
||||
<true/>
|
||||
<key>com.apple.security.network.server</key>
|
||||
<true/>
|
||||
<key>com.apple.security.files.user-selected.read-write</key>
|
||||
<true/>
|
||||
<key>com.apple.security.files.downloads.read-write</key>
|
||||
<true/>
|
||||
</dict>
|
||||
</plist>
|
||||
17
build/darwin/gon-sign.json
vendored
Normal file
@@ -0,0 +1,17 @@
|
||||
{
|
||||
"source": [
|
||||
"./build/bin/RWKV-Runner_darwin.app"
|
||||
],
|
||||
"bundle_id": "dev.josStorer.RWKV-Runner",
|
||||
"apple_id": {
|
||||
"username": "joshua1466587594@outlook.com",
|
||||
"password": ""
|
||||
},
|
||||
"sign": {
|
||||
"application_identity": "D00A983569B4EAA2A008B963254F385F42A493FD",
|
||||
"entitlements_file": "./build/darwin/entitlements.plist"
|
||||
},
|
||||
"zip": {
|
||||
"output_path": "./build/bin/RWKV-Runner_darwin.archive.zip"
|
||||
}
|
||||
}
|
||||
13
build/linux/Readme_Install.txt
vendored
Normal file
@@ -0,0 +1,13 @@
|
||||
Please execute this program in an empty directory. All related dependencies will be placed in this directory.
|
||||
请将本程序放在一个空目录内执行, 所有相关依赖均会放置于此目录.
|
||||
このプログラムを空のディレクトリで実行してください。関連するすべての依存関係は、このディレクトリに配置されます。
|
||||
|
||||
On Linux system, this program cannot invoke the terminal for automatic dependency installation. You must manually execute the following commands for installation so that it can be used normally:
|
||||
在Linux系统下, 本程序无法调用终端自动安装依赖, 你必须手动执行以下命令进行安装, 之后方可正常使用:
|
||||
Linuxシステムでは、このプログラムはターミナルを自動的に呼び出して依存関係をインストールすることができません。以下のコマンドを手動で実行する必要があります。それが完了した後に、正常に使用することができます:
|
||||
|
||||
chmod +x ./RWKV-Runner
|
||||
./RWKV-Runner
|
||||
git clone https://github.com/josStorer/RWKV-Runner --depth=1
|
||||
cd RWKV-Runner/backend-python
|
||||
pip3 install -r requirements.txt # or pip3 install -r requirements_without_cyac.txt
|
||||
3
build/windows/Readme_Install.txt
vendored
Normal file
@@ -0,0 +1,3 @@
|
||||
Please execute this program in an empty directory. All related dependencies will be placed in this directory.
|
||||
请将本程序放在一个空目录内执行, 所有相关依赖均会放置于此目录.
|
||||
このプログラムを空のディレクトリで実行してください。関連するすべての依存関係は、このディレクトリに配置されます。
|
||||
BIN
build/windows/icon.ico
vendored
|
Before Width: | Height: | Size: 147 KiB After Width: | Height: | Size: 167 KiB |
@@ -88,6 +88,7 @@
|
||||
"Downloads": "下载",
|
||||
"Pause": "暂停",
|
||||
"Continue": "继续",
|
||||
"Resume": "继续",
|
||||
"Check": "查看",
|
||||
"Model file not found": "模型文件不存在",
|
||||
"Can not find download url": "找不到下载地址",
|
||||
@@ -114,6 +115,7 @@
|
||||
"Catgirl": "猫娘",
|
||||
"Explain Code": "代码解释",
|
||||
"Werewolf": "狼人杀",
|
||||
"Instruction": "指令",
|
||||
"Blank": "空白",
|
||||
"The following is an epic science fiction masterpiece that is immortalized, with delicate descriptions and grand depictions of interstellar civilization wars.\nChapter 1.\n": "以下是不朽的科幻史诗巨著,描写细腻,刻画了宏大的星际文明战争。\n第一章\n",
|
||||
"The following is a conversation between a cat girl and her owner. The cat girl is a humanized creature that behaves like a cat but is humanoid. At the end of each sentence in the dialogue, she will add \"Meow~\". In the following content, Bob represents the owner and Alice represents the cat girl.\n\nBob: Hello.\n\nAlice: I'm here, meow~.\n\nBob: Can you tell jokes?": "以下是一位猫娘的主人和猫娘的对话内容,猫娘是一种拟人化的生物,其行为似猫但类人,在每一句对话末尾都会加上\"喵~\"。以下内容中,Bob代表主人,Alice代表猫娘。\n\nBob: 你好\n\nAlice: 主人我在哦,喵~\n\nBob: 你会讲笑话吗?",
|
||||
@@ -132,5 +134,12 @@
|
||||
"Are you sure you want to reset all configs? This will obtain the latest preset configs, but will override your custom configs and cannot be undone.": "你确定要重置所有配置吗?这会获取最新的预设配置,但会覆盖你的自定义配置,并且无法撤销",
|
||||
"Advanced": "高级",
|
||||
"Custom Python Path": "自定义Python路径",
|
||||
"Custom Models Path": "自定义模型路径"
|
||||
"Custom Models Path": "自定义模型路径",
|
||||
"Microsoft Visual C++ Redistributable is not installed, would you like to download it?": "微软VC++组件未安装, 是否下载?",
|
||||
"Path Cannot Contain Space": "路径不能包含空格",
|
||||
"Failed to switch model, please try starting the program with administrator privileges.": "切换模型失败, 请尝试以管理员权限启动程序",
|
||||
"Current Strategy": "当前Strategy",
|
||||
"MacOS is not yet supported for performing this operation, please do it manually.": "MacOS尚未支持此操作, 请手动执行",
|
||||
"Linux is not yet supported for performing this operation, please do it manually.": "Linux尚未支持此操作, 请手动执行",
|
||||
"On Linux system, you must manually install python dependencies.": "在Linux系统下, 你必须手动安装python依赖"
|
||||
}
|
||||
|
Before Width: | Height: | Size: 4.4 KiB After Width: | Height: | Size: 4.4 KiB |
BIN
frontend/src/assets/images/strategy.jpg
Normal file
|
After Width: | Height: | Size: 132 KiB |
BIN
frontend/src/assets/images/strategy_zh.jpg
Normal file
|
After Width: | Height: | Size: 115 KiB |
46
frontend/src/components/DialogButton.tsx
Normal file
@@ -0,0 +1,46 @@
|
||||
import { FC, ReactElement } from 'react';
|
||||
import {
|
||||
Button,
|
||||
Dialog,
|
||||
DialogActions,
|
||||
DialogBody,
|
||||
DialogContent,
|
||||
DialogSurface,
|
||||
DialogTitle,
|
||||
DialogTrigger
|
||||
} from '@fluentui/react-components';
|
||||
import { ToolTipButton } from './ToolTipButton';
|
||||
import { useTranslation } from 'react-i18next';
|
||||
|
||||
export const DialogButton: FC<{
|
||||
icon: ReactElement,
|
||||
tooltip: string,
|
||||
title: string,
|
||||
contentText: string,
|
||||
onConfirm: () => void
|
||||
}> = ({ tooltip, icon, title, contentText, onConfirm }) => {
|
||||
const { t } = useTranslation();
|
||||
|
||||
return <Dialog>
|
||||
<DialogTrigger disableButtonEnhancement>
|
||||
<ToolTipButton desc={tooltip} icon={icon} />
|
||||
</DialogTrigger>
|
||||
<DialogSurface>
|
||||
<DialogBody>
|
||||
<DialogTitle>{title}</DialogTitle>
|
||||
<DialogContent>
|
||||
{contentText}
|
||||
</DialogContent>
|
||||
<DialogActions>
|
||||
<DialogTrigger disableButtonEnhancement>
|
||||
<Button appearance="secondary">{t('Cancel')}</Button>
|
||||
</DialogTrigger>
|
||||
<DialogTrigger disableButtonEnhancement>
|
||||
<Button appearance="primary" onClick={onConfirm}>{t('Confirm')}
|
||||
</Button>
|
||||
</DialogTrigger>
|
||||
</DialogActions>
|
||||
</DialogBody>
|
||||
</DialogSurface>
|
||||
</Dialog>;
|
||||
};
|
||||
@@ -5,17 +5,23 @@ import classnames from 'classnames';
|
||||
export const Labeled: FC<{
|
||||
label: string;
|
||||
desc?: string | null,
|
||||
descComponent?: ReactElement,
|
||||
content: ReactElement,
|
||||
flex?: boolean,
|
||||
spaceBetween?: boolean,
|
||||
breakline?: boolean
|
||||
breakline?: boolean,
|
||||
onMouseEnter?: () => void
|
||||
onMouseLeave?: () => void
|
||||
}> = ({
|
||||
label,
|
||||
desc,
|
||||
descComponent,
|
||||
content,
|
||||
flex,
|
||||
spaceBetween,
|
||||
breakline
|
||||
breakline,
|
||||
onMouseEnter,
|
||||
onMouseLeave
|
||||
}) => {
|
||||
return (
|
||||
<div className={classnames(
|
||||
@@ -24,11 +30,11 @@ export const Labeled: FC<{
|
||||
breakline ? 'flex-col' : '',
|
||||
spaceBetween && 'justify-between')
|
||||
}>
|
||||
{desc ?
|
||||
<Tooltip content={desc} showDelay={0} hideDelay={0} relationship="description">
|
||||
<Label>{label}</Label>
|
||||
{(desc || descComponent) ?
|
||||
<Tooltip content={descComponent ? descComponent : desc!} showDelay={0} hideDelay={0} relationship="description">
|
||||
<Label onMouseEnter={onMouseEnter} onMouseLeave={onMouseLeave}>{label}</Label>
|
||||
</Tooltip> :
|
||||
<Label>{label}</Label>
|
||||
<Label onMouseEnter={onMouseEnter} onMouseLeave={onMouseLeave}>{label}</Label>
|
||||
}
|
||||
{content}
|
||||
</div>
|
||||
|
||||
18
frontend/src/components/ResetConfigsButton.tsx
Normal file
@@ -0,0 +1,18 @@
|
||||
import React, { FC } from 'react';
|
||||
import { DialogButton } from './DialogButton';
|
||||
import { useTranslation } from 'react-i18next';
|
||||
import { ArrowReset20Regular } from '@fluentui/react-icons';
|
||||
import commonStore from '../stores/commonStore';
|
||||
|
||||
import { defaultModelConfigs, defaultModelConfigsMac } from '../pages/defaultModelConfigs';
|
||||
|
||||
export const ResetConfigsButton: FC<{ afterConfirm?: () => void }> = ({ afterConfirm }) => {
|
||||
const { t } = useTranslation();
|
||||
return <DialogButton icon={<ArrowReset20Regular />} tooltip={t('Reset All Configs')} title={t('Reset All Configs')}
|
||||
contentText={t('Are you sure you want to reset all configs? This will obtain the latest preset configs, but will override your custom configs and cannot be undone.')}
|
||||
onConfirm={() => {
|
||||
commonStore.setModelConfigs(commonStore.platform != 'darwin' ? defaultModelConfigs : defaultModelConfigsMac, false);
|
||||
commonStore.setCurrentConfigIndex(0, true);
|
||||
afterConfirm?.();
|
||||
}} />;
|
||||
};
|
||||
@@ -17,7 +17,7 @@ import { useTranslation } from 'react-i18next';
|
||||
import { ToolTipButton } from './ToolTipButton';
|
||||
import { Play16Regular, Stop16Regular } from '@fluentui/react-icons';
|
||||
import { useNavigate } from 'react-router';
|
||||
import { WindowShow } from '../../wailsjs/runtime/runtime';
|
||||
import { BrowserOpenURL, WindowShow } from '../../wailsjs/runtime/runtime';
|
||||
|
||||
const mainButtonText = {
|
||||
[ModelStatus.Offline]: 'Run',
|
||||
@@ -70,10 +70,22 @@ export const RunButton: FC<{ onClickRun?: MouseEventHandler, iconMode?: boolean
|
||||
AddToDownloadList('python-3.10.11-embed-amd64.zip', 'https://www.python.org/ftp/python/3.10.11/python-3.10.11-embed-amd64.zip');
|
||||
});
|
||||
} else if (depErrorMsg.includes('DepCheck Error')) {
|
||||
toastWithButton(t('Python dependencies are incomplete, would you like to install them?'), t('Install'), () => {
|
||||
InstallPyDep(commonStore.settings.customPythonPath, commonStore.settings.cnMirror);
|
||||
setTimeout(WindowShow, 1000);
|
||||
});
|
||||
if (depErrorMsg.includes('vc_redist')) {
|
||||
toastWithButton(t('Microsoft Visual C++ Redistributable is not installed, would you like to download it?'), t('Download'), () => {
|
||||
BrowserOpenURL('https://aka.ms/vs/16/release/vc_redist.x64.exe');
|
||||
});
|
||||
} else {
|
||||
toast(depErrorMsg, { type: 'info', position: 'bottom-left' });
|
||||
if (commonStore.platform != 'linux')
|
||||
toastWithButton(t('Python dependencies are incomplete, would you like to install them?'), t('Install'), () => {
|
||||
InstallPyDep(commonStore.settings.customPythonPath, commonStore.settings.cnMirror);
|
||||
setTimeout(WindowShow, 1000);
|
||||
});
|
||||
else
|
||||
toastWithButton(t('On Linux system, you must manually install python dependencies.'), t('Check'), () => {
|
||||
BrowserOpenURL('https://github.com/josStorer/RWKV-Runner/blob/master/build/linux/Readme_Install.txt');
|
||||
});
|
||||
}
|
||||
} else {
|
||||
toast(depErrorMsg, { type: 'error' });
|
||||
}
|
||||
@@ -136,23 +148,28 @@ export const RunButton: FC<{ onClickRun?: MouseEventHandler, iconMode?: boolean
|
||||
});
|
||||
|
||||
let customCudaFile = '';
|
||||
if (modelConfig.modelParameters.device != 'CPU' && modelConfig.modelParameters.useCustomCuda) {
|
||||
customCudaFile = getSupportedCustomCudaFile();
|
||||
if (customCudaFile && commonStore.platform === 'windows') {
|
||||
FileExists('./py310/Lib/site-packages/rwkv/model.py').then((exist) => {
|
||||
// defensive measure. As Python has already been launched, will only take effect the next time it runs.
|
||||
if (!exist) CopyFile('./backend-python/wkv_cuda_utils/wkv_cuda_model.py', './py310/Lib/site-packages/rwkv/model.py');
|
||||
});
|
||||
await CopyFile(customCudaFile, './py310/Lib/site-packages/rwkv/wkv_cuda.pyd').catch(() => {
|
||||
FileExists('./py310/Lib/site-packages/rwkv/wkv_cuda.pyd').then((exist) => {
|
||||
if (!exist) {
|
||||
customCudaFile = '';
|
||||
toast(t('Failed to copy custom cuda file'), { type: 'error' });
|
||||
}
|
||||
if ((modelConfig.modelParameters.device === 'CUDA' || modelConfig.modelParameters.device === 'Custom')
|
||||
&& modelConfig.modelParameters.useCustomCuda) {
|
||||
if (commonStore.platform === 'windows') {
|
||||
customCudaFile = getSupportedCustomCudaFile();
|
||||
if (customCudaFile) {
|
||||
FileExists('./py310/Lib/site-packages/rwkv/model.py').then((exist) => {
|
||||
// defensive measure. As Python has already been launched, will only take effect the next time it runs.
|
||||
if (!exist) CopyFile('./backend-python/wkv_cuda_utils/wkv_cuda_model.py', './py310/Lib/site-packages/rwkv/model.py');
|
||||
});
|
||||
});
|
||||
} else
|
||||
toast(t('Supported custom cuda file not found'), { type: 'warning' });
|
||||
await CopyFile(customCudaFile, './py310/Lib/site-packages/rwkv/wkv_cuda.pyd').catch(() => {
|
||||
FileExists('./py310/Lib/site-packages/rwkv/wkv_cuda.pyd').then((exist) => {
|
||||
if (!exist) {
|
||||
customCudaFile = '';
|
||||
toast(t('Failed to copy custom cuda file'), { type: 'error' });
|
||||
}
|
||||
});
|
||||
});
|
||||
} else
|
||||
toast(t('Supported custom cuda file not found'), { type: 'warning' });
|
||||
} else {
|
||||
customCudaFile = 'any';
|
||||
}
|
||||
}
|
||||
|
||||
switchModel({
|
||||
@@ -173,7 +190,10 @@ export const RunButton: FC<{ onClickRun?: MouseEventHandler, iconMode?: boolean
|
||||
}
|
||||
}).catch(() => {
|
||||
commonStore.setStatus({ status: ModelStatus.Offline });
|
||||
toast(t('Failed to switch model'), { type: 'error' });
|
||||
if (commonStore.platform === 'windows')
|
||||
toast(t('Failed to switch model, please try starting the program with administrator privileges.'), { type: 'error' });
|
||||
else
|
||||
toast(t('Failed to switch model'), { type: 'error' });
|
||||
});
|
||||
}
|
||||
}).catch(() => {
|
||||
|
||||
@@ -6,7 +6,7 @@ import App from './App';
|
||||
import { HashRouter } from 'react-router-dom';
|
||||
import { startup } from './startup';
|
||||
import './_locales/i18n-react';
|
||||
import { WindowSetAlwaysOnTop } from '../wailsjs/runtime';
|
||||
import { WindowShow } from '../wailsjs/runtime';
|
||||
|
||||
startup().then(() => {
|
||||
const container = document.getElementById('root');
|
||||
@@ -20,6 +20,5 @@ startup().then(() => {
|
||||
);
|
||||
|
||||
// force display the window
|
||||
WindowSetAlwaysOnTop(true);
|
||||
WindowSetAlwaysOnTop(false);
|
||||
WindowShow();
|
||||
});
|
||||
|
||||
@@ -7,7 +7,7 @@ import { v4 as uuid } from 'uuid';
|
||||
import classnames from 'classnames';
|
||||
import { fetchEventSource } from '@microsoft/fetch-event-source';
|
||||
import { ConversationPair, getConversationPairs, Record } from '../utils/get-conversation-pairs';
|
||||
import logo from '../../../build/appicon.jpg';
|
||||
import logo from '../assets/images/logo.jpg';
|
||||
import MarkdownRender from '../components/MarkdownRender';
|
||||
import { ToolTipButton } from '../components/ToolTipButton';
|
||||
import { ArrowCircleUp28Regular, Delete28Regular, RecordStop28Regular } from '@fluentui/react-icons';
|
||||
|
||||
@@ -101,6 +101,19 @@ export const defaultPresets: CompletionPreset[] = [{
|
||||
injectStart: '\\n\\nAlice: ',
|
||||
injectEnd: '\\n\\nBob: '
|
||||
}
|
||||
}, {
|
||||
name: 'Instruction',
|
||||
prompt: 'Below is an instruction that describes a task. Write a response that appropriately completes the request.\n\n# Instruction:\nExplain the following metaphor: Life is like cats.\n\n# Response:\n',
|
||||
params: {
|
||||
maxResponseToken: 500,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4,
|
||||
stop: '',
|
||||
injectStart: '',
|
||||
injectEnd: ''
|
||||
}
|
||||
}, {
|
||||
name: 'Blank',
|
||||
prompt: '',
|
||||
|
||||
@@ -1,26 +1,5 @@
|
||||
import {
|
||||
Button,
|
||||
Dialog,
|
||||
DialogActions,
|
||||
DialogBody,
|
||||
DialogContent,
|
||||
DialogSurface,
|
||||
DialogTitle,
|
||||
DialogTrigger,
|
||||
Dropdown,
|
||||
Input,
|
||||
Label,
|
||||
Option,
|
||||
Select,
|
||||
Switch
|
||||
} from '@fluentui/react-components';
|
||||
import {
|
||||
AddCircle20Regular,
|
||||
ArrowReset20Regular,
|
||||
DataUsageSettings20Regular,
|
||||
Delete20Regular,
|
||||
Save20Regular
|
||||
} from '@fluentui/react-icons';
|
||||
import { Dropdown, Input, Label, Option, Select, Switch, Text } from '@fluentui/react-components';
|
||||
import { AddCircle20Regular, DataUsageSettings20Regular, Delete20Regular, Save20Regular } from '@fluentui/react-icons';
|
||||
import React, { FC } from 'react';
|
||||
import { Section } from '../components/Section';
|
||||
import { Labeled } from '../components/Labeled';
|
||||
@@ -38,6 +17,9 @@ import { ConvertModel, FileExists } from '../../wailsjs/go/backend_golang/App';
|
||||
import { getStrategy, refreshLocalModels } from '../utils';
|
||||
import { useTranslation } from 'react-i18next';
|
||||
import { WindowShow } from '../../wailsjs/runtime/runtime';
|
||||
import strategyImg from '../assets/images/strategy.jpg';
|
||||
import strategyZhImg from '../assets/images/strategy_zh.jpg';
|
||||
import { ResetConfigsButton } from '../components/ResetConfigsButton';
|
||||
|
||||
export type ApiParameters = {
|
||||
apiPort: number
|
||||
@@ -48,7 +30,7 @@ export type ApiParameters = {
|
||||
frequencyPenalty: number;
|
||||
}
|
||||
|
||||
export type Device = 'CPU' | 'CUDA' | 'Custom';
|
||||
export type Device = 'CPU' | 'CUDA' | 'MPS' | 'Custom';
|
||||
export type Precision = 'fp16' | 'int8' | 'fp32';
|
||||
|
||||
export type ModelParameters = {
|
||||
@@ -70,568 +52,11 @@ export type ModelConfig = {
|
||||
modelParameters: ModelParameters
|
||||
}
|
||||
|
||||
export const defaultModelConfigs: ModelConfig[] = [
|
||||
{
|
||||
name: 'GPU-2G-1B5-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-1B5-v12-Eng98%-Other2%-20230520-ctx4096.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 4,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: true,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-4G-1B5-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-1B5-v12-Eng98%-Other2%-20230520-ctx4096.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-4G-3B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-3B-v12-Eng98%-Other2%-20230520-ctx4096.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 24,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: true,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-4G-3B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-3B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230527-ctx4096.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 24,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: true,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-4G-7B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng98%-Other2%-20230521-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 8,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: true,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-4G-7B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230530-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 8,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: true,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-6G-1B5-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-1B5-v12-Eng98%-Other2%-20230520-ctx4096.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'fp16',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-6G-3B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-3B-v12-Eng98%-Other2%-20230520-ctx4096.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-6G-3B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-3B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230527-ctx4096.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-6G-7B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng98%-Other2%-20230521-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 18,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: true,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-6G-7B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230530-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 18,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: true,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-8G-3B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-3B-v12-Eng98%-Other2%-20230520-ctx4096.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'fp16',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-8G-3B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-3B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230527-ctx4096.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'fp16',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-8G-7B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng98%-Other2%-20230521-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 27,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-8G-7B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230530-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 27,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-10G-7B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng98%-Other2%-20230521-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-10G-7B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230530-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-12G-14B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-14B-v12-Eng98%-Other2%-20230523-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 24,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-16G-7B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng98%-Other2%-20230521-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'fp16',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-16G-7B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230530-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'fp16',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-16G-14B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-14B-v12-Eng98%-Other2%-20230523-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 37,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-18G-14B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-14B-v12-Eng98%-Other2%-20230523-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-32G-14B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-14B-v12-Eng98%-Other2%-20230523-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'fp16',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'CPU-6G-1B5-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-1B5-v12-Eng98%-Other2%-20230520-ctx4096.pth',
|
||||
device: 'CPU',
|
||||
precision: 'fp32',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'CPU-12G-3B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-3B-v12-Eng98%-Other2%-20230520-ctx4096.pth',
|
||||
device: 'CPU',
|
||||
precision: 'fp32',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'CPU-12G-3B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-3B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230527-ctx4096.pth',
|
||||
device: 'CPU',
|
||||
precision: 'fp32',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'CPU-28G-7B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng98%-Other2%-20230521-ctx8192.pth',
|
||||
device: 'CPU',
|
||||
precision: 'fp32',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'CPU-28G-7B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230530-ctx8192.pth',
|
||||
device: 'CPU',
|
||||
precision: 'fp32',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false
|
||||
}
|
||||
}
|
||||
];
|
||||
|
||||
export const Configs: FC = observer(() => {
|
||||
const { t } = useTranslation();
|
||||
const [selectedIndex, setSelectedIndex] = React.useState(commonStore.currentModelConfigIndex);
|
||||
const [selectedConfig, setSelectedConfig] = React.useState(commonStore.modelConfigs[selectedIndex]);
|
||||
const [displayStrategyImg, setDisplayStrategyImg] = React.useState(false);
|
||||
const navigate = useNavigate();
|
||||
const port = selectedConfig.apiParameters.apiPort;
|
||||
|
||||
@@ -700,31 +125,10 @@ export const Configs: FC = observer(() => {
|
||||
commonStore.deleteModelConfig(selectedIndex);
|
||||
updateSelectedIndex(Math.min(selectedIndex, commonStore.modelConfigs.length - 1));
|
||||
}} />
|
||||
<Dialog>
|
||||
<DialogTrigger disableButtonEnhancement>
|
||||
<ToolTipButton desc={t('Reset All Configs')} icon={<ArrowReset20Regular />} />
|
||||
</DialogTrigger>
|
||||
<DialogSurface>
|
||||
<DialogBody>
|
||||
<DialogTitle>{t('Reset All Configs')}</DialogTitle>
|
||||
<DialogContent>
|
||||
{t('Are you sure you want to reset all configs? This will obtain the latest preset configs, but will override your custom configs and cannot be undone.')}
|
||||
</DialogContent>
|
||||
<DialogActions>
|
||||
<DialogTrigger disableButtonEnhancement>
|
||||
<Button appearance="secondary">{t('Cancel')}</Button>
|
||||
</DialogTrigger>
|
||||
<DialogTrigger disableButtonEnhancement>
|
||||
<Button appearance="primary" onClick={() => {
|
||||
commonStore.setModelConfigs(defaultModelConfigs, false); // updateSelectedIndex() will save configs
|
||||
updateSelectedIndex(0);
|
||||
}}>{t('Confirm')}
|
||||
</Button>
|
||||
</DialogTrigger>
|
||||
</DialogActions>
|
||||
</DialogBody>
|
||||
</DialogSurface>
|
||||
</Dialog>
|
||||
<ResetConfigsButton afterConfirm={() => {
|
||||
setSelectedIndex(0);
|
||||
setSelectedConfig(commonStore.modelConfigs[0]);
|
||||
}} />
|
||||
<ToolTipButton desc={t('Save Config')} icon={<Save20Regular />} onClick={onClickSave} />
|
||||
</div>
|
||||
<div className="flex items-center gap-4">
|
||||
@@ -835,6 +239,14 @@ export const Configs: FC = observer(() => {
|
||||
</div>
|
||||
} />
|
||||
<ToolTipButton text={t('Convert')} desc={t('Convert model with these configs')} onClick={async () => {
|
||||
if (commonStore.platform == 'darwin') {
|
||||
toast(t('MacOS is not yet supported for performing this operation, please do it manually.'), { type: 'info' });
|
||||
return;
|
||||
} else if (commonStore.platform == 'linux') {
|
||||
toast(t('Linux is not yet supported for performing this operation, please do it manually.'), { type: 'info' });
|
||||
return;
|
||||
}
|
||||
|
||||
const modelPath = `${commonStore.settings.customModelsPath}/${selectedConfig.modelParameters.modelName}`;
|
||||
if (await FileExists(modelPath)) {
|
||||
const strategy = getStrategy(selectedConfig);
|
||||
@@ -844,7 +256,11 @@ export const Configs: FC = observer(() => {
|
||||
toast(`${t('Convert Success')} - ${newModelPath}`, { type: 'success' });
|
||||
refreshLocalModels({ models: commonStore.modelSourceList }, false);
|
||||
}).catch(e => {
|
||||
toast(`${t('Convert Failed')} - ${e.message || e}`, { type: 'error' });
|
||||
const errMsg = e.message || e;
|
||||
if (errMsg.includes('path contains space'))
|
||||
toast(`${t('Convert Failed')} - ${t('Path Cannot Contain Space')}`, { type: 'error' });
|
||||
else
|
||||
toast(`${t('Convert Failed')} - ${e.message || e}`, { type: 'error' });
|
||||
});
|
||||
setTimeout(WindowShow, 1000);
|
||||
} else {
|
||||
@@ -862,6 +278,7 @@ export const Configs: FC = observer(() => {
|
||||
}
|
||||
}}>
|
||||
<Option value="CPU">CPU</Option>
|
||||
{commonStore.platform === 'darwin' && <Option value="MPS">MPS</Option>}
|
||||
<Option value="CUDA">CUDA</Option>
|
||||
<Option value="Custom">{t('Custom')!}</Option>
|
||||
</Dropdown>
|
||||
@@ -886,9 +303,14 @@ export const Configs: FC = observer(() => {
|
||||
</Dropdown>
|
||||
} />
|
||||
}
|
||||
{selectedConfig.modelParameters.device == 'CUDA' && <div />}
|
||||
{
|
||||
selectedConfig.modelParameters.device == 'CUDA' && <Labeled label={t('Stored Layers')}
|
||||
selectedConfig.modelParameters.device == 'CUDA' &&
|
||||
<Labeled label={t('Current Strategy')}
|
||||
content={<Text> {getStrategy(selectedConfig)} </Text>} />
|
||||
}
|
||||
{
|
||||
selectedConfig.modelParameters.device == 'CUDA' &&
|
||||
<Labeled label={t('Stored Layers')}
|
||||
desc={t('Number of the neural network layers loaded into VRAM, the more you load, the faster the speed, but it consumes more VRAM.')}
|
||||
content={
|
||||
<ValuedSlider value={selectedConfig.modelParameters.storedLayers} min={0}
|
||||
@@ -913,11 +335,19 @@ export const Configs: FC = observer(() => {
|
||||
}} />
|
||||
} />
|
||||
}
|
||||
{
|
||||
displayStrategyImg &&
|
||||
<img style={{ width: '80vh', height: 'auto', zIndex: 100 }} className="fixed left-0 top-0"
|
||||
src={commonStore.settings.language === 'zh' ? strategyZhImg : strategyImg} />
|
||||
}
|
||||
{
|
||||
selectedConfig.modelParameters.device == 'Custom' &&
|
||||
<Labeled label="Strategy" desc="https://github.com/BlinkDL/ChatRWKV/blob/main/ChatRWKV-strategy.png"
|
||||
<Labeled label="Strategy"
|
||||
onMouseEnter={() => setDisplayStrategyImg(true)}
|
||||
onMouseLeave={() => setDisplayStrategyImg(false)}
|
||||
content={
|
||||
<Input className="grow" placeholder="cuda:0 fp16 *20 -> cuda:1 fp16"
|
||||
<Input className="grow"
|
||||
placeholder={commonStore.platform != 'darwin' ? 'cuda:0 fp16 *20 -> cuda:1 fp16' : 'mps fp32'}
|
||||
value={selectedConfig.modelParameters.customStrategy}
|
||||
onChange={(e, data) => {
|
||||
setSelectedConfigModelParams({
|
||||
@@ -928,7 +358,7 @@ export const Configs: FC = observer(() => {
|
||||
}
|
||||
{selectedConfig.modelParameters.device == 'Custom' && <div />}
|
||||
{
|
||||
selectedConfig.modelParameters.device != 'CPU' &&
|
||||
selectedConfig.modelParameters.device != 'CPU' && selectedConfig.modelParameters.device != 'MPS' &&
|
||||
<Labeled label={t('Use Custom CUDA kernel to Accelerate')}
|
||||
desc={t('Enabling this option can greatly improve inference speed, but there may be compatibility issues. If it fails to start, please turn off this option.')}
|
||||
content={
|
||||
|
||||
@@ -53,7 +53,7 @@ export const Downloads: FC = observer(() => {
|
||||
<div className="flex items-center gap-2">
|
||||
<ProgressBar className="grow" value={status.progress} max={100} />
|
||||
{!status.done &&
|
||||
<ToolTipButton desc={status.downloading ? t('Pause') : t('Continue')}
|
||||
<ToolTipButton desc={status.downloading ? t('Pause') : t('Resume')}
|
||||
icon={status.downloading ? <Pause20Regular /> : <Play20Regular />}
|
||||
onClick={() => {
|
||||
if (status.downloading)
|
||||
@@ -62,7 +62,7 @@ export const Downloads: FC = observer(() => {
|
||||
ContinueDownload(status.url);
|
||||
}} />}
|
||||
<ToolTipButton desc={t('Open Folder')} icon={<Folder20Regular />} onClick={() => {
|
||||
OpenFileFolder(status.path);
|
||||
OpenFileFolder(`${commonStore.settings.customModelsPath}/${status.name}`);
|
||||
}} />
|
||||
</div>
|
||||
</Field>
|
||||
|
||||
@@ -17,6 +17,7 @@ import { ConfigSelector } from '../components/ConfigSelector';
|
||||
import MarkdownRender from '../components/MarkdownRender';
|
||||
import commonStore from '../stores/commonStore';
|
||||
import { Completion } from './Completion';
|
||||
import { ResetConfigsButton } from '../components/ResetConfigsButton';
|
||||
|
||||
export type IntroductionContent = { [lang: string]: string }
|
||||
|
||||
@@ -65,7 +66,8 @@ export const Home: FC = observer(() => {
|
||||
|
||||
return (
|
||||
<div className="flex flex-col justify-between h-full">
|
||||
<img className="rounded-xl select-none hidden sm:block" src={banner} />
|
||||
<img className="rounded-xl select-none hidden sm:block"
|
||||
style={{ maxHeight: '40%', margin: '0 auto' }} src={banner} />
|
||||
<div className="flex flex-col gap-2">
|
||||
<Text size={600} weight="medium">{t('Introduction')}</Text>
|
||||
<div className="h-40 overflow-y-auto overflow-x-hidden p-1">
|
||||
@@ -85,6 +87,7 @@ export const Home: FC = observer(() => {
|
||||
<div className="flex flex-col gap-2">
|
||||
<div className="flex flex-row-reverse sm:fixed bottom-2 right-2">
|
||||
<div className="flex gap-3">
|
||||
<ResetConfigsButton />
|
||||
<ConfigSelector />
|
||||
<RunButton />
|
||||
</div>
|
||||
|
||||
@@ -95,7 +95,7 @@ export const Settings: FC = observer(() => {
|
||||
} />
|
||||
}
|
||||
{
|
||||
commonStore.settings.language === 'zh' &&
|
||||
commonStore.settings.language === 'zh' && commonStore.platform != 'linux' &&
|
||||
<Labeled label={t('Use Tsinghua Pip Mirrors')} flex spaceBetween content={
|
||||
<Switch checked={commonStore.settings.cnMirror}
|
||||
onChange={(e, data) => {
|
||||
@@ -118,16 +118,18 @@ export const Settings: FC = observer(() => {
|
||||
<AccordionHeader ref={advancedHeaderRef} size="large">{t('Advanced')}</AccordionHeader>
|
||||
<AccordionPanel>
|
||||
<div className="flex flex-col gap-2 overflow-hidden">
|
||||
<Labeled label={t('Custom Models Path')}
|
||||
content={
|
||||
<Input className="grow" placeholder="./models" value={commonStore.settings.customModelsPath}
|
||||
onChange={(e, data) => {
|
||||
commonStore.setSettings({
|
||||
customModelsPath: data.value
|
||||
});
|
||||
}} />
|
||||
} />
|
||||
<Labeled label={t('Custom Python Path')}
|
||||
{commonStore.platform !== 'darwin' &&
|
||||
<Labeled label={t('Custom Models Path')}
|
||||
content={
|
||||
<Input className="grow" placeholder="./models" value={commonStore.settings.customModelsPath}
|
||||
onChange={(e, data) => {
|
||||
commonStore.setSettings({
|
||||
customModelsPath: data.value
|
||||
});
|
||||
}} />
|
||||
} />
|
||||
}
|
||||
<Labeled label={t('Custom Python Path')} // if set, will not use precompiled cuda kernel
|
||||
content={
|
||||
<Input className="grow" placeholder="./py310/python" value={commonStore.settings.customPythonPath}
|
||||
onChange={(e, data) => {
|
||||
|
||||
720
frontend/src/pages/defaultModelConfigs.ts
Normal file
@@ -0,0 +1,720 @@
|
||||
import { ModelConfig } from './Configs';
|
||||
|
||||
export const defaultModelConfigsMac: ModelConfig[] = [
|
||||
{
|
||||
name: 'MAC-1B5-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-1B5-v12-Eng98%-Other2%-20230520-ctx4096.pth',
|
||||
device: 'MPS',
|
||||
precision: 'fp32',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: false,
|
||||
customStrategy: 'mps fp32'
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'MAC-3B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-3B-v12-Eng98%-Other2%-20230520-ctx4096.pth',
|
||||
device: 'MPS',
|
||||
precision: 'fp32',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: false,
|
||||
customStrategy: 'mps fp32'
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'MAC-3B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-3B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230527-ctx4096.pth',
|
||||
device: 'MPS',
|
||||
precision: 'fp32',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: false,
|
||||
customStrategy: 'mps fp32'
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'CPU-6G-1B5-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-1B5-v12-Eng98%-Other2%-20230520-ctx4096.pth',
|
||||
device: 'CPU',
|
||||
precision: 'fp32',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'CPU-12G-3B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-3B-v12-Eng98%-Other2%-20230520-ctx4096.pth',
|
||||
device: 'CPU',
|
||||
precision: 'fp32',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'CPU-12G-3B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-3B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230527-ctx4096.pth',
|
||||
device: 'CPU',
|
||||
precision: 'fp32',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'CPU-28G-7B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng98%-Other2%-20230521-ctx8192.pth',
|
||||
device: 'CPU',
|
||||
precision: 'fp32',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'CPU-28G-7B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230530-ctx8192.pth',
|
||||
device: 'CPU',
|
||||
precision: 'fp32',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false
|
||||
}
|
||||
}
|
||||
];
|
||||
|
||||
export const defaultModelConfigs: ModelConfig[] = [
|
||||
{
|
||||
name: 'GPU-2G-1B5-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-1B5-v12-Eng98%-Other2%-20230520-ctx4096.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 4,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: true,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-4G-1B5-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-1B5-v12-Eng98%-Other2%-20230520-ctx4096.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-4G-3B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-3B-v12-Eng98%-Other2%-20230520-ctx4096.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 24,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: true,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-4G-3B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-3B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230527-ctx4096.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 24,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: true,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-4G-7B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng98%-Other2%-20230521-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 8,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: true,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-4G-7B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230530-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 8,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: true,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-6G-1B5-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-1B5-v12-Eng98%-Other2%-20230520-ctx4096.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'fp16',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-6G-3B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-3B-v12-Eng98%-Other2%-20230520-ctx4096.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-6G-3B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-3B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230527-ctx4096.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-6G-7B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng98%-Other2%-20230521-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 18,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: true,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-6G-7B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230530-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 18,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: true,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-8G-3B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-3B-v12-Eng98%-Other2%-20230520-ctx4096.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'fp16',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-8G-3B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-3B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230527-ctx4096.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'fp16',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-8G-7B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng98%-Other2%-20230521-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 27,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-8G-7B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230530-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 27,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-10G-7B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng98%-Other2%-20230521-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-10G-7B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230530-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-12G-14B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-14B-v12-Eng98%-Other2%-20230523-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 24,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-16G-7B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng98%-Other2%-20230521-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'fp16',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-16G-7B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230530-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'fp16',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-16G-14B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-14B-v12-Eng98%-Other2%-20230523-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 37,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-18G-14B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-14B-v12-Eng98%-Other2%-20230523-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'int8',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'GPU-32G-14B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-14B-v12-Eng98%-Other2%-20230523-ctx8192.pth',
|
||||
device: 'CUDA',
|
||||
precision: 'fp16',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false,
|
||||
useCustomCuda: true
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'CPU-6G-1B5-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-1B5-v12-Eng98%-Other2%-20230520-ctx4096.pth',
|
||||
device: 'CPU',
|
||||
precision: 'fp32',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'CPU-12G-3B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-3B-v12-Eng98%-Other2%-20230520-ctx4096.pth',
|
||||
device: 'CPU',
|
||||
precision: 'fp32',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'CPU-12G-3B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-3B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230527-ctx4096.pth',
|
||||
device: 'CPU',
|
||||
precision: 'fp32',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'CPU-28G-7B-EN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng98%-Other2%-20230521-ctx8192.pth',
|
||||
device: 'CPU',
|
||||
precision: 'fp32',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false
|
||||
}
|
||||
},
|
||||
{
|
||||
name: 'CPU-28G-7B-CN',
|
||||
apiParameters: {
|
||||
apiPort: 8000,
|
||||
maxResponseToken: 4100,
|
||||
temperature: 1.2,
|
||||
topP: 0.5,
|
||||
presencePenalty: 0.4,
|
||||
frequencyPenalty: 0.4
|
||||
},
|
||||
modelParameters: {
|
||||
modelName: 'RWKV-4-Raven-7B-v12-Eng49%-Chn49%-Jpn1%-Other1%-20230530-ctx8192.pth',
|
||||
device: 'CPU',
|
||||
precision: 'fp32',
|
||||
storedLayers: 41,
|
||||
maxStoredLayers: 41,
|
||||
enableHighPrecisionForLastLayer: false
|
||||
}
|
||||
}
|
||||
];
|
||||
@@ -3,8 +3,8 @@ import { GetPlatform, ReadJson } from '../wailsjs/go/backend_golang/App';
|
||||
import { Cache, checkUpdate, downloadProgramFiles, LocalConfig, refreshModels } from './utils';
|
||||
import { getStatus } from './apis';
|
||||
import { EventsOn } from '../wailsjs/runtime';
|
||||
import { defaultModelConfigs } from './pages/Configs';
|
||||
import manifest from '../../manifest.json';
|
||||
import { defaultModelConfigs, defaultModelConfigsMac } from './pages/defaultModelConfigs';
|
||||
|
||||
export async function startup() {
|
||||
downloadProgramFiles();
|
||||
@@ -54,7 +54,7 @@ async function initConfig() {
|
||||
configData.currentModelConfigIndex >= 0 && configData.currentModelConfigIndex < configData.modelConfigs.length)
|
||||
commonStore.setCurrentConfigIndex(configData.currentModelConfigIndex, false);
|
||||
}).catch(() => {
|
||||
commonStore.setModelConfigs(defaultModelConfigs, true);
|
||||
commonStore.setModelConfigs(commonStore.platform != 'darwin' ? defaultModelConfigs : defaultModelConfigsMac, true);
|
||||
});
|
||||
}
|
||||
|
||||
|
||||
@@ -2,7 +2,7 @@ import { makeAutoObservable } from 'mobx';
|
||||
import { getUserLanguage, isSystemLightMode, saveConfigs } from '../utils';
|
||||
import { WindowSetDarkTheme, WindowSetLightTheme } from '../../wailsjs/runtime';
|
||||
import manifest from '../../../manifest.json';
|
||||
import { defaultModelConfigs, ModelConfig } from '../pages/Configs';
|
||||
import { ModelConfig } from '../pages/Configs';
|
||||
import { Conversations } from '../pages/Chat';
|
||||
import { ModelSourceItem } from '../pages/Models';
|
||||
import { DownloadStatus } from '../pages/Downloads';
|
||||
@@ -11,6 +11,8 @@ import { IntroductionContent } from '../pages/Home';
|
||||
import { AboutContent } from '../pages/About';
|
||||
import i18n from 'i18next';
|
||||
import { CompletionPreset } from '../pages/Completion';
|
||||
import { defaultModelConfigs, defaultModelConfigsMac } from '../pages/defaultModelConfigs';
|
||||
import commonStore from './commonStore';
|
||||
|
||||
export enum ModelStatus {
|
||||
Offline,
|
||||
@@ -98,7 +100,8 @@ class CommonStore {
|
||||
|
||||
createModelConfig = (config: ModelConfig = defaultModelConfigs[0], saveConfig: boolean = true) => {
|
||||
if (config.name === defaultModelConfigs[0].name) {
|
||||
config = JSON.parse(JSON.stringify(config)); // deep copy
|
||||
// deep copy
|
||||
config = JSON.parse(JSON.stringify(commonStore.platform != 'darwin' ? defaultModelConfigs[0] : defaultModelConfigsMac[0]));
|
||||
config.name = new Date().toLocaleString();
|
||||
}
|
||||
this.modelConfigs.push(config);
|
||||
|
||||
@@ -16,6 +16,7 @@ import { Button } from '@fluentui/react-components';
|
||||
import { Language, Languages, SettingsType } from '../pages/Settings';
|
||||
import { ModelSourceItem } from '../pages/Models';
|
||||
import { ModelConfig, ModelParameters } from '../pages/Configs';
|
||||
import { BrowserOpenURL } from '../../wailsjs/runtime';
|
||||
|
||||
export type Cache = {
|
||||
models: ModelSourceItem[]
|
||||
@@ -140,6 +141,10 @@ export const getStrategy = (modelConfig: ModelConfig | undefined = undefined) =>
|
||||
if (params.enableHighPrecisionForLastLayer)
|
||||
strategy += ' -> cpu fp32 *1';
|
||||
break;
|
||||
case 'MPS':
|
||||
strategy += 'mps ';
|
||||
strategy += params.precision === 'fp16' ? 'fp16' : params.precision === 'int8' ? 'fp16i8' : 'fp32';
|
||||
break;
|
||||
case 'Custom':
|
||||
strategy = params.customStrategy || '';
|
||||
break;
|
||||
@@ -245,20 +250,24 @@ export async function checkUpdate(notifyEvenLatest: boolean = false) {
|
||||
`https://gitee.com/josc146/RWKV-Runner/releases/download/${versionTag}/${asset.name}`;
|
||||
toastWithButton(t('New Version Available') + ': ' + versionTag, t('Update'), () => {
|
||||
DeleteFile('cache.json');
|
||||
toast(t('Downloading update, please wait. If it is not completed, please manually download the program from GitHub and replace the original program.'), {
|
||||
type: 'info',
|
||||
position: 'bottom-left',
|
||||
autoClose: 10000
|
||||
});
|
||||
setTimeout(() => {
|
||||
UpdateApp(updateUrl).catch((e) => {
|
||||
toast(t('Update Error') + ' - ' + e.message || e, {
|
||||
type: 'error',
|
||||
position: 'bottom-left',
|
||||
autoClose: false
|
||||
});
|
||||
if (commonStore.platform != 'darwin') {
|
||||
toast(t('Downloading update, please wait. If it is not completed, please manually download the program from GitHub and replace the original program.'), {
|
||||
type: 'info',
|
||||
position: 'bottom-left',
|
||||
autoClose: 20000
|
||||
});
|
||||
}, 500);
|
||||
setTimeout(() => {
|
||||
UpdateApp(updateUrl).catch((e) => {
|
||||
toast(t('Update Error') + ' - ' + e.message || e, {
|
||||
type: 'error',
|
||||
position: 'bottom-left',
|
||||
autoClose: false
|
||||
});
|
||||
});
|
||||
}, 500);
|
||||
} else {
|
||||
BrowserOpenURL(updateUrl);
|
||||
}
|
||||
}, {
|
||||
autoClose: false,
|
||||
position: 'bottom-left'
|
||||
@@ -309,9 +318,9 @@ export function toastWithButton(text: string, buttonText: string, onClickButton:
|
||||
}
|
||||
|
||||
export function getSupportedCustomCudaFile() {
|
||||
if ([' 10', ' 16', ' 20', ' 30', 'P40', 'P104', 'P106'].some(v => commonStore.status.device_name.includes(v)))
|
||||
if ([' 10', ' 16', ' 20', ' 30', 'MX', 'Tesla P', 'Quadro P', 'NVIDIA P', 'TITAN X', 'TITAN RTX', 'RTX A'].some(v => commonStore.status.device_name.includes(v)))
|
||||
return './backend-python/wkv_cuda_utils/wkv_cuda10_30.pyd';
|
||||
else if ([' 40'].some(v => commonStore.status.device_name.includes(v)))
|
||||
else if ([' 40', 'RTX TITAN Ada'].some(v => commonStore.status.device_name.includes(v)))
|
||||
return './backend-python/wkv_cuda_utils/wkv_cuda40.pyd';
|
||||
else
|
||||
return '';
|
||||
|
||||
0
frontend/wailsjs/go/backend_golang/App.d.ts
generated
vendored
Normal file → Executable file
0
frontend/wailsjs/go/backend_golang/App.js
generated
Normal file → Executable file
0
frontend/wailsjs/go/models.ts
generated
Normal file → Executable file
@@ -1,5 +1,5 @@
|
||||
{
|
||||
"version": "1.1.4",
|
||||
"version": "1.1.9",
|
||||
"introduction": {
|
||||
"en": "RWKV is an open-source, commercially usable large language model with high flexibility and great potential for development.\n### About This Tool\nThis tool aims to lower the barrier of entry for using large language models, making it accessible to everyone. It provides fully automated dependency and model management. You simply need to click and run, following the instructions, to deploy a local large language model. The tool itself is very compact and only requires a single executable file for one-click deployment.\nAdditionally, this tool offers an interface that is fully compatible with the OpenAI API. This means you can use any ChatGPT client as a client for RWKV, enabling capability expansion beyond just chat functionality.\n### Preset Configuration Rules at the Bottom\nThis tool comes with a series of preset configurations to reduce complexity. The naming rules for each configuration represent the following in order: device - required VRAM/memory - model size - model language.\nFor example, \"GPU-8G-3B-EN\" indicates that this configuration is for a graphics card with 8GB of VRAM, a model size of 3 billion parameters, and it uses an English language model.\nLarger model sizes have higher performance and VRAM requirements. Among configurations with the same model size, those with higher VRAM usage will have faster runtime.\nFor example, if you have 12GB of VRAM but running the \"GPU-12G-7B-EN\" configuration is slow, you can downgrade to \"GPU-8G-3B-EN\" for a significant speed improvement.\n### About RWKV\nRWKV is an RNN with Transformer-level LLM performance, which can also be directly trained like a GPT transformer (parallelizable). And it's 100% attention-free. You only need the hidden state at position t to compute the state at position t+1. You can use the \"GPT\" mode to quickly compute the hidden state for the \"RNN\" mode.<br/>So it's combining the best of RNN and transformer - great performance, fast inference, saves VRAM, fast training, \"infinite\" ctx_len, and free sentence embedding (using the final hidden state).",
|
||||
"zh": "RWKV是一个开源且允许商用的大语言模型,灵活性很高且极具发展潜力。\n### 关于本工具\n本工具旨在降低大语言模型的使用门槛,做到人人可用,本工具提供了全自动化的依赖和模型管理,你只需要直接点击运行,跟随引导,即可完成本地大语言模型的部署,工具本身体积极小,只需要一个exe即可完成一键部署。\n此外,本工具提供了与OpenAI API完全兼容的接口,这意味着你可以把任意ChatGPT客户端用作RWKV的客户端,实现能力拓展,而不局限于聊天。\n### 底部的预设配置规则\n本工具内置了一系列预设配置,以降低使用难度,每个配置名的规则,依次代表着:设备-所需显存/内存-模型规模-模型语言。\n例如,GPU-8G-3B-CN,表示该配置用于显卡,需要8G显存,模型规模为30亿参数,使用的是中文模型。\n模型规模越大,性能要求越高,显存要求也越高,而同样模型规模的配置中,显存占用越高的,运行速度越快。\n例如当你有12G显存,但运行GPU-12G-7B-CN配置速度比较慢,可降级成GPU-8G-3B-CN,将会大幅提速。\n### 关于RWKV\nRWKV是具有Transformer级别LLM性能的RNN,也可以像GPT Transformer一样直接进行训练(可并行化)。而且它是100% attention-free的。你只需在位置t处获得隐藏状态即可计算位置t + 1处的状态。你可以使用“GPT”模式快速计算用于“RNN”模式的隐藏状态。\n因此,它将RNN和Transformer的优点结合起来 - 高性能、快速推理、节省显存、快速训练、“无限”上下文长度以及免费的语句嵌入(使用最终隐藏状态)。"
|
||||
@@ -238,6 +238,18 @@
|
||||
"url": "https://huggingface.co/BlinkDL/rwkv-4-novel/blob/main/RWKV-4-Novel-7B-v1-ChnEng-20230426-ctx8192.pth",
|
||||
"downloadUrl": "https://huggingface.co/BlinkDL/rwkv-4-novel/resolve/main/RWKV-4-Novel-7B-v1-ChnEng-20230426-ctx8192.pth"
|
||||
},
|
||||
{
|
||||
"name": "RWKV-4-World-7B-v1-OnlyForTest_40%_trained-20230601-ctx4096.pth",
|
||||
"desc": {
|
||||
"en": "100+ Languages 7B v1 Test",
|
||||
"zh": "100+ 语言 7B v1 测试"
|
||||
},
|
||||
"size": 15035393581,
|
||||
"SHA256": "63c060c472e45b6c3af2baaaee448ffd95f9b46e3cc6e1ef70ce7ecb1d01bcfa",
|
||||
"lastUpdated": "2023-06-02T00:09:39",
|
||||
"url": "https://huggingface.co/BlinkDL/rwkv-4-world/blob/main/RWKV-4-World-7B-v1-OnlyForTest_40%25_trained-20230601-ctx4096.pth",
|
||||
"downloadUrl": "https://huggingface.co/BlinkDL/rwkv-4-world/resolve/main/RWKV-4-World-7B-v1-OnlyForTest_40%25_trained-20230601-ctx4096.pth"
|
||||
},
|
||||
{
|
||||
"name": "RWKV-4-World-7B-v1-OnlyForTest_30%_trained-20230529-ctx4096.pth",
|
||||
"desc": {
|
||||
@@ -248,7 +260,8 @@
|
||||
"SHA256": "05f91562b2ae8b025226e40b3fb536d6f8eb3c142ac899c0808ee1c9dc189ec4",
|
||||
"lastUpdated": "2023-05-29T13:25:53",
|
||||
"url": "https://huggingface.co/BlinkDL/rwkv-4-world/blob/main/RWKV-4-World-7B-v1-OnlyForTest_30%25_trained-20230529-ctx4096.pth",
|
||||
"downloadUrl": "https://huggingface.co/BlinkDL/rwkv-4-world/resolve/main/RWKV-4-World-7B-v1-OnlyForTest_30%25_trained-20230529-ctx4096.pth"
|
||||
"downloadUrl": "https://huggingface.co/BlinkDL/rwkv-4-world/resolve/main/RWKV-4-World-7B-v1-OnlyForTest_30%25_trained-20230529-ctx4096.pth",
|
||||
"hide": true
|
||||
},
|
||||
{
|
||||
"name": "RWKV-4-World-0.1B-v1-20230520-ctx4096.pth",
|
||||
@@ -272,7 +285,8 @@
|
||||
"SHA256": "ac36770931776c5aa179690918c9a3b0b5f4ebe3301ea3574a7e182209778788",
|
||||
"lastUpdated": "2023-05-29T13:25:53",
|
||||
"url": "https://huggingface.co/BlinkDL/rwkv-4-world/blob/main/RWKV-4-World-1.5B-v1-OnlyForTest_57%25_trained-20230529-ctx4096.pth",
|
||||
"downloadUrl": "https://huggingface.co/BlinkDL/rwkv-4-world/resolve/main/RWKV-4-World-1.5B-v1-OnlyForTest_57%25_trained-20230529-ctx4096.pth"
|
||||
"downloadUrl": "https://huggingface.co/BlinkDL/rwkv-4-world/resolve/main/RWKV-4-World-1.5B-v1-OnlyForTest_57%25_trained-20230529-ctx4096.pth",
|
||||
"hide": true
|
||||
},
|
||||
{
|
||||
"name": "RWKV-4-World-3B-v1-OnlyForTest_35%_trained-20230529-ctx4096.pth",
|
||||
@@ -284,7 +298,32 @@
|
||||
"SHA256": "e4ee6e91a80d56de43bc79841f3a8be3b7b215d7d9788f79c467b9b1f7f03cb8",
|
||||
"lastUpdated": "2023-05-29T13:25:53",
|
||||
"url": "https://huggingface.co/BlinkDL/rwkv-4-world/blob/main/RWKV-4-World-3B-v1-OnlyForTest_35%25_trained-20230529-ctx4096.pth",
|
||||
"downloadUrl": "https://huggingface.co/BlinkDL/rwkv-4-world/resolve/main/RWKV-4-World-3B-v1-OnlyForTest_35%25_trained-20230529-ctx4096.pth"
|
||||
"downloadUrl": "https://huggingface.co/BlinkDL/rwkv-4-world/resolve/main/RWKV-4-World-3B-v1-OnlyForTest_35%25_trained-20230529-ctx4096.pth",
|
||||
"hide": true
|
||||
},
|
||||
{
|
||||
"name": "RWKV-4-World-1.5B-v1-OnlyForTest_81%_trained-20230603-ctx4096.pth",
|
||||
"desc": {
|
||||
"en": "100+ Languages 1.5B v1 Test",
|
||||
"zh": "100+ 语言 1.5B v1 测试"
|
||||
},
|
||||
"size": 3155281581,
|
||||
"SHA256": "044fb10daa71f4c012493ac8ef455c8c3301095b5f009dae58f0f6382a53e23c",
|
||||
"lastUpdated": "2023-06-03T13:57:20",
|
||||
"url": "https://huggingface.co/BlinkDL/rwkv-4-world/blob/main/RWKV-4-World-1.5B-v1-OnlyForTest_81%25_trained-20230603-ctx4096.pth",
|
||||
"downloadUrl": "https://huggingface.co/BlinkDL/rwkv-4-world/resolve/main/RWKV-4-World-1.5B-v1-OnlyForTest_81%25_trained-20230603-ctx4096.pth"
|
||||
},
|
||||
{
|
||||
"name": "RWKV-4-World-3B-v1-OnlyForTest_52%_trained-20230603-ctx4096.pth",
|
||||
"desc": {
|
||||
"en": "100+ Languages 3B v1 Test",
|
||||
"zh": "100+ 语言 3B v1 测试"
|
||||
},
|
||||
"size": 6125597613,
|
||||
"SHA256": "aad3671078a0c686368add4f4b695a76c2ba1ddd505a64c0949bb003beeee9a3",
|
||||
"lastUpdated": "2023-06-03T13:57:20",
|
||||
"url": "https://huggingface.co/BlinkDL/rwkv-4-world/blob/main/RWKV-4-World-3B-v1-OnlyForTest_52%25_trained-20230603-ctx4096.pth",
|
||||
"downloadUrl": "https://huggingface.co/BlinkDL/rwkv-4-world/resolve/main/RWKV-4-World-3B-v1-OnlyForTest_52%25_trained-20230603-ctx4096.pth"
|
||||
},
|
||||
{
|
||||
"name": "RWKV-4-World-0.4B-v1-20230529-ctx4096.pth",
|
||||
|
||||