release v1.4.5

add default_stop when stop is null

.github/workflows/release.yml

@@ -57,6 +57,8 @@ jobs:
         with:
           args: install upx
       - run: |
+          Start-BitsTransfer https://github.com/josStorer/LibreHardwareMonitor.Console/releases/download/v0.1.0/LibreHardwareMonitor.Console.zip ./LibreHardwareMonitor.Console.zip
+          Expand-Archive ./LibreHardwareMonitor.Console.zip -DestinationPath ./components/LibreHardwareMonitor.Console
           Start-BitsTransfer https://www.python.org/ftp/python/3.10.11/python-3.10.11-embed-amd64.zip ./python-3.10.11-embed-amd64.zip
           Expand-Archive ./python-3.10.11-embed-amd64.zip -DestinationPath ./py310
           $content=Get-Content "./py310/python310._pth"; $content | ForEach-Object {if ($_.ReadCount -eq 3) {"Lib\\site-packages"} else {$_}} | Set-Content ./py310/python310._pth
@@ -71,6 +73,7 @@ jobs:
           mv ./target/release/ai00_server.exe ../backend-rust/webgpu_server.exe
           cd ..
           go install github.com/wailsapp/wails/v2/cmd/wails@latest
+          (Get-Content -Path ./backend-golang/app.go) -replace "//go:custom_build windows ", "" | Set-Content -Path ./backend-golang/app.go
           make
           Rename-Item -Path "build/bin/RWKV-Runner.exe" -NewName "RWKV-Runner_windows_x64.exe"
 

CURRENT_CHANGE.md

@@ -1,10 +1,14 @@
 ## Changes
 
-- webgpu support (AMD, Intel, Nvidia, Apple)
-- add rwkv-cuda-beta support (faster)
-- add misc API (`/models` and `/dashboard/billing/credit_grants`)
-- allow multiple systems
-- allow completions input to be null
+- frontend: update manifest (a lot of new models)
+- frontend: correct Preset UI description
+- frontend: add HardwareMonitor (Windows Only)
+- lora finetune: fix max_epochs (#170)
+- python-backend: allow message content to be empty
+- python-backend: extra ChatCompletionBody params (`raw`, `presystem`)
+- python-backend: add default_stop when stop is null
+- webgpu: fix webgpu_server file permissions of linux and macos
+- chore
 
 ## Install
 

backend-golang/app.go

@@ -1,6 +1,7 @@
 package backend_golang
 
 import (
+	"bufio"
 	"context"
 	"errors"
 	"net/http"
@@ -8,6 +9,7 @@ import (
 	"os/exec"
 	"path/filepath"
 	"runtime"
+	"syscall"
 
 	"github.com/fsnotify/fsnotify"
 	"github.com/minio/selfupdate"
@@ -41,6 +43,7 @@ func (a *App) OnStartup(ctx context.Context) {
 		a.cmdPrefix = "cd " + a.exDir + " && "
 	}
 
+	os.Chmod("./backend-rust/webgpu_server", 0777)
 	os.Mkdir(a.exDir+"models", os.ModePerm)
 	os.Mkdir(a.exDir+"lora-models", os.ModePerm)
 	os.Mkdir(a.exDir+"finetune/json2binidx_tool/data", os.ModePerm)
@@ -50,7 +53,18 @@ func (a *App) OnStartup(ctx context.Context) {
 	}
 
 	a.downloadLoop()
+	a.watchFs()
+	a.monitorHardware()
+}
 
+func (a *App) OnBeforeClose(ctx context.Context) bool {
+	if monitor != nil {
+		monitor.Process.Kill()
+	}
+	return false
+}
+
+func (a *App) watchFs() {
 	watcher, err := fsnotify.NewWatcher()
 	if err == nil {
 		watcher.Add("./lora-models")
@@ -62,7 +76,7 @@ func (a *App) OnStartup(ctx context.Context) {
 					if !ok {
 						return
 					}
-					wruntime.EventsEmit(ctx, "fsnotify", event.Name)
+					wruntime.EventsEmit(a.ctx, "fsnotify", event.Name)
 				case _, ok := <-watcher.Errors:
 					if !ok {
 						return
@@ -73,6 +87,37 @@ func (a *App) OnStartup(ctx context.Context) {
 	}
 }
 
+var monitor *exec.Cmd
+
+func (a *App) monitorHardware() {
+	if runtime.GOOS != "windows" {
+		return
+	}
+
+	monitor = exec.Command("./components/LibreHardwareMonitor.Console/LibreHardwareMonitor.Console.exe")
+	stdout, err := monitor.StdoutPipe()
+	if err != nil {
+		monitor = nil
+		return
+	}
+
+	go func() {
+		reader := bufio.NewReader(stdout)
+		for {
+			line, _, err := reader.ReadLine()
+			if err != nil {
+				wruntime.EventsEmit(a.ctx, "monitorerr", err.Error())
+				break
+			}
+			wruntime.EventsEmit(a.ctx, "monitor", string(line))
+		}
+	}()
+
+	monitor.SysProcAttr = &syscall.SysProcAttr{}
+	//go:custom_build windows monitor.SysProcAttr.HideWindow = true
+	monitor.Start()
+}
+
 func (a *App) UpdateApp(url string) (broken bool, err error) {
 	resp, err := http.Get(url)
 	if err != nil {

backend-golang/rwkv.go

@@ -155,7 +155,6 @@ func (a *App) InstallPyDep(python string, cnMirror bool) (string, error) {
 			"exit"
 		if !cnMirror {
 			installScript = strings.Replace(installScript, " -i https://pypi.tuna.tsinghua.edu.cn/simple", "", -1)
-			installScript = strings.Replace(installScript, "requirements.txt", "requirements_versions.txt", -1)
 		}
 		err = os.WriteFile("./install-py-dep.bat", []byte(installScript), 0644)
 		if err != nil {

backend-python/routes/completion.py

@@ -25,32 +25,44 @@ class Role(Enum):
 
 class Message(BaseModel):
     role: Role
-    content: str = Field(min_length=1)
+    content: str = Field(min_length=0)
+    raw: bool = Field(False, description="Whether to treat content as raw text")
+
+
+default_stop = [
+    "\n\nUser",
+    "\n\nQuestion",
+    "\n\nQ",
+    "\n\nHuman",
+    "\n\nBob",
+]
 
 
 class ChatCompletionBody(ModelConfigBody):
     messages: Union[List[Message], None]
     model: str = "rwkv"
     stream: bool = False
-    stop: Union[str, List[str], None] = [
-        "\n\nUser",
-        "\n\nQuestion",
-        "\n\nQ",
-        "\n\nHuman",
-        "\n\nBob",
-    ]
-    user_name: Union[str, None] = None
-    assistant_name: Union[str, None] = None
+    stop: Union[str, List[str], None] = default_stop
+    user_name: Union[str, None] = Field(None, description="Internal user name")
+    assistant_name: Union[str, None] = Field(
+        None, description="Internal assistant name"
+    )
+    presystem: bool = Field(
+        True, description="Whether to insert default system prompt at the beginning"
+    )
 
     class Config:
         schema_extra = {
             "example": {
-                "messages": [{"role": Role.User.value, "content": "hello"}],
+                "messages": [
+                    {"role": Role.User.value, "content": "hello", "raw": False}
+                ],
                 "model": "rwkv",
                 "stream": False,
                 "stop": None,
                 "user_name": None,
                 "assistant_name": None,
+                "presystem": True,
                 "max_tokens": 1000,
                 "temperature": 1.2,
                 "top_p": 0.5,
@@ -233,10 +245,6 @@ async def chat_completions(body: ChatCompletionBody, request: Request):
     if body.messages is None or body.messages == []:
         raise HTTPException(status.HTTP_400_BAD_REQUEST, "messages not found")
 
-    basic_system: str = ""
-    if body.messages[0].role == Role.System:
-        basic_system = body.messages[0].content
-
     interface = model.interface
     user = model.user if body.user_name is None else body.user_name
     bot = model.bot if body.assistant_name is None else body.assistant_name
@@ -244,46 +252,54 @@ async def chat_completions(body: ChatCompletionBody, request: Request):
     is_raven = model.rwkv_type == RWKVType.Raven
 
     completion_text: str = ""
-    if basic_system == "":
-        completion_text = (
-            f"""
+    basic_system: Union[str, None] = None
+    if body.presystem:
+        if body.messages[0].role == Role.System:
+            basic_system = body.messages[0].content
+
+        if basic_system is None:
+            completion_text = (
+                f"""
 The following is a coherent verbose detailed conversation between a girl named {bot} and her friend {user}. \
 {bot} is very intelligent, creative and friendly. \
 {bot} is unlikely to disagree with {user}, and {bot} doesn't like to ask {user} questions. \
 {bot} likes to tell {user} a lot about herself and her opinions. \
 {bot} usually gives {user} kind, helpful and informative advices.\n
 """
-            if is_raven
-            else (
-                f"{user}{interface} hi\n\n{bot}{interface} Hi. "
-                + "I am your assistant and I will provide expert full response in full details. Please feel free to ask any question and I will always answer it.\n\n"
-            )
-        )
-    elif basic_system != "":
-        completion_text = (
-            (
-                f"The following is a coherent verbose detailed conversation between a girl named {bot} and her friend {user}. "
                 if is_raven
-                else f"{user}{interface} hi\n\n{bot}{interface} Hi. "
+                else (
+                    f"{user}{interface} hi\n\n{bot}{interface} Hi. "
+                    + "I am your assistant and I will provide expert full response in full details. Please feel free to ask any question and I will always answer it.\n\n"
+                )
+            )
+        else:
+            if not body.messages[0].raw:
+                basic_system = (
+                    basic_system.replace("\r\n", "\n")
+                    .replace("\r", "\n")
+                    .replace("\n\n", "\n")
+                    .replace("\n", " ")
+                    .strip()
+                )
+            completion_text = (
+                (
+                    f"The following is a coherent verbose detailed conversation between a girl named {bot} and her friend {user}. "
+                    if is_raven
+                    else f"{user}{interface} hi\n\n{bot}{interface} Hi. "
+                )
+                + basic_system.replace("You are", f"{bot} is" if is_raven else "I am")
+                .replace("you are", f"{bot} is" if is_raven else "I am")
+                .replace("You're", f"{bot} is" if is_raven else "I'm")
+                .replace("you're", f"{bot} is" if is_raven else "I'm")
+                .replace("You", f"{bot}" if is_raven else "I")
+                .replace("you", f"{bot}" if is_raven else "I")
+                .replace("Your", f"{bot}'s" if is_raven else "My")
+                .replace("your", f"{bot}'s" if is_raven else "my")
+                .replace("你", f"{bot}" if is_raven else "我")
+                + "\n\n"
             )
-            + basic_system.replace("\r\n", "\n")
-            .replace("\r", "\n")
-            .replace("\n\n", "\n")
-            .replace("\n", " ")
-            .strip()
-            .replace("You are", f"{bot} is" if is_raven else "I am")
-            .replace("you are", f"{bot} is" if is_raven else "I am")
-            .replace("You're", f"{bot} is" if is_raven else "I'm")
-            .replace("you're", f"{bot} is" if is_raven else "I'm")
-            .replace("You", f"{bot}" if is_raven else "I")
-            .replace("you", f"{bot}" if is_raven else "I")
-            .replace("Your", f"{bot}'s" if is_raven else "My")
-            .replace("your", f"{bot}'s" if is_raven else "my")
-            .replace("你", f"{bot}" if is_raven else "我")
-            + "\n\n"
-        )
 
-    for message in body.messages[(0 if basic_system == "" else 1) :]:
+    for message in body.messages[(0 if basic_system is None else 1) :]:
         append_message: str = ""
         if message.role == Role.User:
             append_message = f"{user}{interface} " + message.content
@@ -291,20 +307,23 @@ The following is a coherent verbose detailed conversation between a girl named {
             append_message = f"{bot}{interface} " + message.content
         elif message.role == Role.System:
             append_message = message.content
-        completion_text += (
-            append_message.replace("\r\n", "\n")
-            .replace("\r", "\n")
-            .replace("\n\n", "\n")
-            .strip()
-            + "\n\n"
-        )
+        if not message.raw:
+            append_message = (
+                append_message.replace("\r\n", "\n")
+                .replace("\r", "\n")
+                .replace("\n\n", "\n")
+                .strip()
+            )
+        completion_text += append_message + "\n\n"
     completion_text += f"{bot}{interface}"
 
     if type(body.stop) == str:
         body.stop = [body.stop, f"\n\n{user}", f"\n\n{bot}"]
-    else:
+    elif type(body.stop) == list:
         body.stop.append(f"\n\n{user}")
         body.stop.append(f"\n\n{bot}")
+    elif body.stop is None:
+        body.stop = default_stop
 
     if body.stream:
         return EventSourceResponse(

backend-python/routes/state_cache.py

@@ -96,7 +96,7 @@ def add_state(body: AddStateBody):
         quick_log(
             None,
             None,
-            f"New Trie Id: {id}\nTrie Len: {len(trie)}\nTrie Buff Size: {trie.buff_size()}\nDtrie Buff Size Of Id: {_get_a_dtrie_buff_size(dtrie[id])}",
+            f"New Trie Id: {id}\nTrie Len: {len(trie)}\nTrie Buff Size: {trie.buff_size()}\nDtrie Buff Size Of Id: {__get_a_dtrie_buff_size(dtrie[id])}",
         )
         return "success"
     except Exception as e:
@@ -124,7 +124,7 @@ class LongestPrefixStateBody(BaseModel):
     prompt: str
 
 
-def _get_a_dtrie_buff_size(dtrie_v):
+def __get_a_dtrie_buff_size(dtrie_v):
     # print(sys.getsizeof(dtrie_v["tokens"][0]))  # str
     # print(sys.getsizeof(dtrie_v["tokens"][0]) * len(dtrie_v["tokens"]))
     # print(dtrie_v["state"][0][0].element_size())

build/darwin/Readme_Install.txt

@@ -1,6 +1,6 @@
-For Mac and Linux users, please manually install Python 3.10 (usually the latest systems come with it built-in). You can specify the Python interpreter to use in Settings.
-对于Mac和Linux用户，请手动安装 Python3.10 (通常最新的系统已经内置了). 你可以在设置中指定使用的Python解释器.
-MacおよびLinuxのユーザーの方は、Python3.10を手動でインストールしてください（通常、最新のシステムには既に組み込まれています）。 設定メニューで使用するPythonインタプリタを指定することができます。
+For Mac and Linux users, please manually install Python 3.10 (usually the latest systems come with it built-in). You can specify the Python interpreter to use in Settings. (which python3)
+对于Mac和Linux用户，请手动安装 Python3.10 (通常最新的系统已经内置了). 你可以在设置中指定使用的Python解释器. (which python3)
+MacおよびLinuxのユーザーの方は、Python3.10を手動でインストールしてください（通常、最新のシステムには既に組み込まれています）。 設定メニューで使用するPythonインタプリタを指定することができます。 (which python3)
 
 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.

finetune/install-wsl-dep-and-train.sh

@@ -1,3 +1,5 @@
+echo $@
+
 if [[ ${cnMirror} == 1 ]]; then
   export PIP_INDEX_URL="https://pypi.tuna.tsinghua.edu.cn/simple"
   if grep -q "mirrors.aliyun.com" /etc/apt/sources.list; then

finetune/lora/train.py

@@ -184,7 +184,7 @@ if __name__ == "__main__":
     args.num_sanity_val_steps = 0
     args.check_val_every_n_epoch = int(1e20)
     args.log_every_n_steps = int(1e20)
-    args.max_epochs = -1  # continue forever
+    args.max_epochs = args.epoch_count  # continue forever
     args.betas = (args.beta1, args.beta2)
     args.real_bsz = int(args.num_nodes) * int(args.devices) * args.micro_bsz
     os.environ["RWKV_T_MAX"] = str(args.ctx_len)
@@ -373,7 +373,7 @@ if __name__ == "__main__":
                     for param in module.parameters():
                         param.requires_grad = True
                 elif enable_time_finetune and any(
-                    n.startswith("time") for n, _ in module.named_parameters()
+                        n.startswith("time") for n, _ in module.named_parameters()
                 ):
                     for pname, param in module.named_parameters():
                         if pname.startswith("time"):
@@ -381,7 +381,7 @@ if __name__ == "__main__":
                             param.requires_grad = True
 
     if (
-        len(args.load_model) == 0 or args.my_pile_stage == 1
+            len(args.load_model) == 0 or args.my_pile_stage == 1
     ):  # shall we build the initial weights?
         init_weight_name = f"{args.proj_dir}/rwkv-init.pth"
         generate_init_weight(model, init_weight_name)  # save initial weights
@@ -423,8 +423,8 @@ if __name__ == "__main__":
     )
 
     if (
-        args.lr_init > 1e-4
-        or trainer.world_size * args.micro_bsz * trainer.accumulate_grad_batches < 8
+            args.lr_init > 1e-4
+            or trainer.world_size * args.micro_bsz * trainer.accumulate_grad_batches < 8
     ):
         if "I_KNOW_WHAT_IM_DOING" in os.environ:
             if trainer.global_rank == 0:
@@ -459,10 +459,10 @@ if __name__ == "__main__":
 
     if "deepspeed" in args.strategy:
         trainer.strategy.config["zero_optimization"]["allgather_bucket_size"] = (
-            args.ds_bucket_mb * 1000 * 1000
+                args.ds_bucket_mb * 1000 * 1000
         )
         trainer.strategy.config["zero_optimization"]["reduce_bucket_size"] = (
-            args.ds_bucket_mb * 1000 * 1000
+                args.ds_bucket_mb * 1000 * 1000
         )
 
     # must set shuffle=False, persistent_workers=False (because worker is in another thread)

frontend/src/_locales/ja/main.json

@@ -178,7 +178,7 @@
   "Failed to import. Please copy a preset to the clipboard.": "インポートに失敗しました。プリセットをクリップボードにコピーしてください。",
   "Clipboard is empty.": "クリップボードが空です。",
   "Successfully copied to clipboard.": "クリップボードにコピーしました。",
-  "Edit Messages": "メッセージの編集",
+  "Edit Character Settings": "キャラクター設定を編集",
   "Go Back": "戻る",
   "Description": "説明",
   "Avatar Url": "アバターURL",
@@ -244,5 +244,9 @@
   "Failed to load local sound font, please check if the files exist - assets/sound-font": "ローカルサウンドフォントの読み込みに失敗しました、ファイルが存在するか確認してください - assets/sound-font",
   "Please convert model to safe tensors format first": "モデルを安全なテンソル形式に変換してください",
   "Convert To Safe Tensors Format": "安全なテンソル形式に変換",
-  "Please change Strategy to WebGPU to use safetensors format": "StrategyをWebGPUに変更して、安全なテンソル形式を使用してください"
+  "Please change Strategy to WebGPU to use safetensors format": "StrategyをWebGPUに変更して、安全なテンソル形式を使用してください",
+  "Preview Only": "プレビューのみ",
+  "RAM": "RAM",
+  "VRAM": "VRAM",
+  "GPU Usage": "GPU使用率"
 }

frontend/src/_locales/zh-hans/main.json

@@ -178,7 +178,7 @@
   "Failed to import. Please copy a preset to the clipboard.": "导入失败。请复制一个预设到剪贴板",
   "Clipboard is empty.": "剪贴板没有内容",
   "Successfully copied to clipboard.": "成功复制到剪贴板",
-  "Edit Messages": "编辑对话",
+  "Edit Character Settings": "编辑人设",
   "Go Back": "返回",
   "Description": "描述",
   "Avatar Url": "头像图片地址",
@@ -244,5 +244,9 @@
   "Failed to load local sound font, please check if the files exist - assets/sound-font": "加载本地音色资源失败，请检查文件是否存在 - assets/sound-font",
   "Please convert model to safe tensors format first": "请先将模型转换为Safetensors格式",
   "Convert To Safe Tensors Format": "转换为Safetensors格式",
-  "Please change Strategy to WebGPU to use safetensors format": "请将Strategy改为WebGPU以使用safetensors格式"
+  "Please change Strategy to WebGPU to use safetensors format": "请将Strategy改为WebGPU以使用safetensors格式",
+  "Preview Only": "仅预览",
+  "RAM": "内存",
+  "VRAM": "显存",
+  "GPU Usage": "GPU占用"
 }

frontend/src/pages/Configs.tsx

@@ -130,7 +130,7 @@ export const Configs: FC = observer(() => {
             setSelectedIndex(0);
             setSelectedConfig(commonStore.modelConfigs[0]);
           }} />
-          <ToolTipButton desc={mq ? '' : t('Save Config')} icon={<Save20Regular />} text={mq ? t('Save Config') : ''}
+          <ToolTipButton desc={mq ? '' : t('Save Config')} icon={<Save20Regular />} text={mq ? t('Save Config') : null}
             onClick={onClickSave} />
         </div>
         <div className="flex items-center gap-4">

frontend/src/pages/PresetsManager/PresetsButton.tsx

@@ -250,14 +250,14 @@ export const ChatPresetEditor: FC<{
                 }} />
                 <Button onClick={() => {
                   setEditingMessages(!editingMessages);
-                }}>{!editingMessages ? t('Edit Messages') : t('Go Back')}</Button>
+                }}>{!editingMessages ? t('Edit Character Settings') : t('Go Back')}</Button>
               </div>
             } />
           {
             editingMessages ?
               <MessagesEditor /> :
               <div className="flex flex-col gap-1 p-2 overflow-x-hidden overflow-y-auto">
-                <Labeled flex breakline label={t('Description')}
+                <Labeled flex breakline label={`${t('Description')} (${t("Preview Only")})`}
                   content={
                     <Input value={editingPreset.desc} onChange={(e, data) => {
                       setEditingPreset({

frontend/src/pages/Train.tsx

@@ -219,7 +219,7 @@ const Terminal: FC = observer(() => {
       WslStart().then(() => {
         addWslMessage('WSL> ' + input);
         setInput('');
-        WslCommand(input).catch(showError);
+        WslCommand(input).then(WindowShow).catch(showError);
       }).catch(showError);
     }
   };

frontend/src/startup.ts

@@ -2,11 +2,12 @@ import commonStore, { Platform } from './stores/commonStore';
 import { GetPlatform, ListDirFiles, ReadJson } from '../wailsjs/go/backend_golang/App';
 import { Cache, checkUpdate, downloadProgramFiles, LocalConfig, refreshLocalModels, refreshModels } from './utils';
 import { getStatus } from './apis';
-import { EventsOn } from '../wailsjs/runtime';
+import { EventsOn, WindowSetTitle } from '../wailsjs/runtime';
 import manifest from '../../manifest.json';
 import { defaultModelConfigs, defaultModelConfigsMac } from './pages/defaultConfigs';
 import { Preset } from './pages/PresetsManager/PresetsButton';
 import { wslHandler } from './pages/Train';
+import { t } from 'i18next';
 
 export async function startup() {
   downloadProgramFiles();
@@ -23,6 +24,8 @@ export async function startup() {
 
   initPresets();
 
+  initHardwareMonitor();
+
   await GetPlatform().then(p => commonStore.setPlatform(p as Platform));
   await initConfig();
 
@@ -117,3 +120,20 @@ async function initLocalModelsNotify() {
       refreshLocalModels({ models: commonStore.modelSourceList }, false); //TODO fix bug that only add models
   });
 }
+
+type monitorData = {
+  usedMemory: number;
+  totalMemory: number;
+  gpuUsage: number;
+  gpuPower: number;
+  usedVram: number;
+  totalVram: number;
+}
+
+async function initHardwareMonitor() {
+  EventsOn('monitor', (data: string) => {
+    const results: monitorData = JSON.parse(data);
+    if (results)
+      WindowSetTitle(`RWKV-Runner (${t('RAM')}: ${results.usedMemory.toFixed(1)}/${results.totalMemory.toFixed(1)} GB, ${t('VRAM')}: ${(results.usedVram / 1024).toFixed(1)}/${(results.totalVram / 1024).toFixed(1)} GB, ${t('GPU Usage')}: ${results.gpuUsage}%)`);
+  });
+}

main.go

@@ -61,6 +61,9 @@ var midi embed.FS
 //go:embed assets/sound-font
 var midiAssets embed.FS
 
+//go:embed components
+var components embed.FS
+
 func main() {
 	if buildInfo, ok := debug.ReadBuildInfo(); !ok || strings.Contains(buildInfo.String(), "-ldflags") {
 		backend.CopyEmbed(cyac)
@@ -70,6 +73,7 @@ func main() {
 		backend.CopyEmbed(finetune)
 		backend.CopyEmbed(midi)
 		backend.CopyEmbed(midiAssets)
+		backend.CopyEmbed(components)
 	}
 
 	// Create an instance of the app structure
@@ -102,7 +106,8 @@ func main() {
 			Assets:  assets,
 			Handler: NewFileLoader(),
 		},
-		OnStartup: app.OnStartup,
+		OnStartup:     app.OnStartup,
+		OnBeforeClose: app.OnBeforeClose,
 		Bind: []any{
 			app,
 		},

manifest.json

@@ -1,5 +1,5 @@
 {
-  "version": "1.4.3",
+  "version": "1.4.4",
   "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）。"
@@ -301,6 +301,58 @@
       "url": "https://huggingface.co/BlinkDL/rwkv-4-world/blob/main/RWKV-4-World-7B-v1-20230626-ctx4096.pth",
       "downloadUrl": "https://huggingface.co/BlinkDL/rwkv-4-world/resolve/main/RWKV-4-World-7B-v1-20230626-ctx4096.pth"
     },
+    {
+      "name": "RWKV-claude-4-World-7B-20230805-ctx65k.pth",
+      "desc": {
+        "en": "Global Languages 7B v1 Ctx65k Claude Like",
+        "zh": "全球语言 7B v1 65k上下文 Claude功能",
+        "ja": "グローバル言語 7B v1 65kコンテキスト Claude機能"
+      },
+      "size": 15035391533,
+      "SHA256": "8cd25f8a1ab58965993cc47b3b2f99585836eed008a2e44526c258189ea751a6",
+      "lastUpdated": "2023-08-05T08:52:20",
+      "url": "https://huggingface.co/xiaol/RWKV-claude-4-World-7B-65k/blob/main/RWKV-claude-4-World-7B-20230805-ctx65k.pth",
+      "downloadUrl": "https://huggingface.co/xiaol/RWKV-claude-4-World-7B-65k/resolve/main/RWKV-claude-4-World-7B-20230805-ctx65k.pth"
+    },
+    {
+      "name": "RWKV-toolformer-translation-japanese-chinese-english-7B-World-20230815-ctx128k.pth",
+      "desc": {
+        "en": "Global Languages 7B v1 Ctx128k Toolformer",
+        "zh": "全球语言 7B v1 128k上下文 Toolformer",
+        "ja": "グローバル言語 7B v1 128kコンテキスト Toolformer"
+      },
+      "size": 15035391533,
+      "SHA256": "648a3b21055bdab77021ce278da80fbada8dcaae0b3d41d1eca9aa194c1fd25f",
+      "lastUpdated": "2023-08-15T07:18:23",
+      "url": "https://huggingface.co/xiaol/RWKV-toolformer-translation-japanese-chinese-english-7B-World-128k/blob/main/RWKV-toolformer-translation-japanese-chinese-english-7B-World-20230815-ctx128k.pth",
+      "downloadUrl": "https://huggingface.co/xiaol/RWKV-toolformer-translation-japanese-chinese-english-7B-World-128k/resolve/main/RWKV-toolformer-translation-japanese-chinese-english-7B-World-20230815-ctx128k.pth"
+    },
+    {
+      "name": "RWKV-code-4-World-7B-20230820-ctx32k.pth",
+      "desc": {
+        "en": "Global Languages 7B v1 Ctx32k Code Ability",
+        "zh": "全球语言 7B v1 32k上下文 代码能力",
+        "ja": "グローバル言語 7B v1 32kコンテキスト コード能力"
+      },
+      "size": 15035391533,
+      "SHA256": "19666620437ae3a5fb06e16a52729d67e449fca155fab3d5861ffe9ecf247404",
+      "lastUpdated": "2023-08-20T05:00:17",
+      "url": "https://huggingface.co/xiaol/RWKV-Code-7B-world-32k/blob/main/RWKV-code-4-World-7B-20230820-ctx32k.pth",
+      "downloadUrl": "https://huggingface.co/xiaol/RWKV-Code-7B-world-32k/resolve/main/RWKV-code-4-World-7B-20230820-ctx32k.pth"
+    },
+    {
+      "name": "wizard-rwkv-4-world-ctx32k.pth",
+      "desc": {
+        "en": "Global Languages 7B v1 Ctx32k Wikipedia",
+        "zh": "全球语言 7B v1 32k上下文 维基百科",
+        "ja": "グローバル言語 7B v1 32kコンテキスト ウィキペディア"
+      },
+      "size": 15035391538,
+      "SHA256": "c5d991f315a1676d4bed93dd91f803b1376096e7a4af5bf72b339d055f53bac7",
+      "lastUpdated": "2023-07-29T03:21:47",
+      "url": "https://huggingface.co/xiaol/wizard-rwkv-world-7B-ctx32k/blob/main/wizard-rwkv-4-world-ctx32k.pth",
+      "downloadUrl": "https://huggingface.co/xiaol/wizard-rwkv-world-7B-ctx32k/resolve/main/wizard-rwkv-4-world-ctx32k.pth"
+    },
     {
       "name": "RWKV-4-World-CHNtuned-7B-v1-20230709-ctx4096.pth",
       "desc": {
@@ -327,6 +379,45 @@
       "url": "https://huggingface.co/xiaol/readflow-rwkv-4-world-ctx32k/blob/main/Readflow-RWKV-4-World-CHNtuned-7B-v1-20230709-ctx32k.pth",
       "downloadUrl": "https://huggingface.co/xiaol/readflow-rwkv-4-world-ctx32k/resolve/main/Readflow-RWKV-4-World-CHNtuned-7B-v1-20230709-ctx32k.pth"
     },
+    {
+      "name": "novel-RWKV-4-World-CHNtuned-7B-v1-20230709-ctx32k.pth",
+      "desc": {
+        "en": "Global Languages 7B v1 Enhanced Chinese Ctx32k Novel Outline Ability",
+        "zh": "全球语言 7B v1 中文增强 32k上下文 小说大纲扩写",
+        "ja": "グローバル言語 7B v1 中国語強化 32kコンテキスト 小説のあらすじを書く"
+      },
+      "size": 15035391538,
+      "SHA256": "0fe2415ce61af52a8c38c071b475c01b4c9f8a4f2b4aaed6181f0334f3faf7f4",
+      "lastUpdated": "2023-07-28T13:30:59",
+      "url": "https://huggingface.co/xiaol/ruotangwx-rwkv-7b-novel-32k/blob/main/novel-RWKV-4-World-CHNtuned-7B-v1-20230709-ctx32k.pth",
+      "downloadUrl": "https://huggingface.co/xiaol/ruotangwx-rwkv-7b-novel-32k/resolve/main/novel-RWKV-4-World-CHNtuned-7B-v1-20230709-ctx32k.pth"
+    },
+    {
+      "name": "chatgal-RWKV-4-World-CHNtuned-7B-v1-20230709-ctx32k-1000.pth",
+      "desc": {
+        "en": "Global Languages 7B v1 Enhanced Chinese Ctx32k GalGame 1000",
+        "zh": "全球语言 7B v1 中文增强 32k上下文 GalGame 1000",
+        "ja": "グローバル言語 7B v1 中国語強化 32kコンテキスト GalGame 1000"
+      },
+      "size": 15035391543,
+      "SHA256": "aaed29cfd1bddee47c48f564aa800eb001f62fd03290d772647d5678e40d66e8",
+      "lastUpdated": "2023-07-21T08:59:18",
+      "url": "https://huggingface.co/xiaol/chatgal-rwkv-7b-world-32k/blob/main/chatgal-RWKV-4-World-CHNtuned-7B-v1-20230709-ctx32k-1000.pth",
+      "downloadUrl": "https://huggingface.co/xiaol/chatgal-rwkv-7b-world-32k/resolve/main/chatgal-RWKV-4-World-CHNtuned-7B-v1-20230709-ctx32k-1000.pth"
+    },
+    {
+      "name": "chatgal-RWKV-4-World-CHNtuned-7B-v1-20230709-ctx32k-500.pth",
+      "desc": {
+        "en": "Global Languages 7B v1 Enhanced Chinese Ctx32k GalGame 500",
+        "zh": "全球语言 7B v1 中文增强 32k上下文 GalGame 500",
+        "ja": "グローバル言語 7B v1 中国語強化 32kコンテキスト GalGame 500"
+      },
+      "size": 15035391538,
+      "SHA256": "b5d347d5dedb4f398ec31489ab87b75b1dee772ae7d0a34c26635cf5d95c8794",
+      "lastUpdated": "2023-07-21T07:31:05",
+      "url": "https://huggingface.co/xiaol/chatgal-rwkv-7b-world-32k/blob/main/chatgal-RWKV-4-World-CHNtuned-7B-v1-20230709-ctx32k-500.pth",
+      "downloadUrl": "https://huggingface.co/xiaol/chatgal-rwkv-7b-world-32k/resolve/main/chatgal-RWKV-4-World-CHNtuned-7B-v1-20230709-ctx32k-500.pth"
+    },
     {
       "name": "RWKV-4-World-JPNtuned-7B-v1-20230718-ctx4096.pth",
       "desc": {
@@ -340,6 +431,19 @@
       "url": "https://huggingface.co/BlinkDL/rwkv-4-world/blob/main/RWKV-4-World-JPNtuned-7B-v1-20230718-ctx4096.pth",
       "downloadUrl": "https://huggingface.co/BlinkDL/rwkv-4-world/resolve/main/RWKV-4-World-JPNtuned-7B-v1-20230718-ctx4096.pth"
     },
+    {
+      "name": "RWKV-novel-4-World-7B-20230810-ctx128k.pth",
+      "desc": {
+        "en": "Global Languages Writer 7B v1 Ctx128k",
+        "zh": "全球语言写作 7B v1 128k上下文",
+        "ja": "グローバル言語ライター 7B v1 128kコンテキスト"
+      },
+      "size": 15035391533,
+      "SHA256": "5e429c49e4cab2f29a93f87a80635422c8710d70e5b1d962c078e47d957389c8",
+      "lastUpdated": "2023-08-10T06:30:32",
+      "url": "https://huggingface.co/xiaol/rwkv-7B-world-novel-128k/blob/main/RWKV-novel-4-World-7B-20230810-ctx128k.pth",
+      "downloadUrl": "https://huggingface.co/xiaol/rwkv-7B-world-novel-128k/resolve/main/RWKV-novel-4-World-7B-20230810-ctx128k.pth"
+    },
     {
       "name": "RWKV-4-Novel-7B-v1-ChnEng-ChnPro-20230410-ctx4096.pth",
       "desc": {