release v1.5.4

.github/workflows/release.yml

@@ -109,6 +109,7 @@ jobs:
           go install github.com/wailsapp/wails/v2/cmd/wails@latest
           rm ./backend-python/rwkv_pip/wkv_cuda.pyd
           rm ./backend-python/rwkv_pip/rwkv5.pyd
+          rm ./backend-python/rwkv_pip/rwkv6.pyd
           rm ./backend-python/rwkv_pip/beta/wkv_cuda.pyd
           rm ./backend-python/get-pip.py
           make
@@ -141,6 +142,7 @@ jobs:
           go install github.com/wailsapp/wails/v2/cmd/wails@latest
           rm ./backend-python/rwkv_pip/wkv_cuda.pyd
           rm ./backend-python/rwkv_pip/rwkv5.pyd
+          rm ./backend-python/rwkv_pip/rwkv6.pyd
           rm ./backend-python/rwkv_pip/beta/wkv_cuda.pyd
           rm ./backend-python/get-pip.py
           make

.gitignore

@@ -18,7 +18,7 @@ __pycache__
 /cmd-helper.bat
 /install-py-dep.bat
 /backend-python/wkv_cuda
-/backend-python/rwkv5
+/backend-python/rwkv*
 *.exe
 *.old
 .DS_Store

CURRENT_CHANGE.md

@@ -1,18 +1,12 @@
 ## Changes
 
-### Changes
-
-- add webgpu nf4
-- add sidePanel for Chat page
-- add hf-mirror for cn users
-- use local API when it's working, even if a custom API URL is provided
-- add rwkv API URL Option
-- macos: change default webgpu backend to aarch64-apple-darwin and fix the permission issue
-- strict default_stop
-- fix state_cache of deploy mode
-- fix a finetune bug
-- improve lazy loading ui
-- display lastModelName at the top (WorkHeader)
+- upgrade to rwkv 0.8.22 (rwkv6 support)
+- allow reading attachments even if the model is offline
+- improve launch flow of webgpu mode
+- allow safetensors converter on macOS
+- fix fs watcher of macOS
+- update defaultConfigs
+- update manifest
 - chores
 
 ## Install

backend-golang/app.go

@@ -69,8 +69,8 @@ func (a *App) OnBeforeClose(ctx context.Context) bool {
 func (a *App) watchFs() {
 	watcher, err := fsnotify.NewWatcher()
 	if err == nil {
-		watcher.Add("./lora-models")
-		watcher.Add("./models")
+		watcher.Add(a.exDir + "./lora-models")
+		watcher.Add(a.exDir + "./models")
 		go func() {
 			for {
 				select {

backend-python/rwkv_pip/cuda/rwkv6.cu

@@ -0,0 +1,87 @@
+#include <stdio.h>
+#include <assert.h>
+#include "ATen/ATen.h"
+typedef at::BFloat16 bf16;
+typedef at::Half fp16;
+typedef float fp32;
+
+template <typename F>
+__global__ void kernel_forward(const int B, const int T, const int C, const int H, float *__restrict__ _state,
+                               const F *__restrict__ const _r, const F *__restrict__ const _k, const F *__restrict__ const _v, const float *__restrict__ _w, const F *__restrict__ _u,
+                               F *__restrict__ const _y)
+{
+    const int b = blockIdx.x / H;
+    const int h = blockIdx.x % H;
+    const int i = threadIdx.x;
+    _u += h*_N_;
+    _state += h*_N_*_N_ + i*_N_; // wrong if B > 1 !!!
+
+    __shared__ float r[_N_], k[_N_], u[_N_], w[_N_];
+    
+    float state[_N_];
+    #pragma unroll
+    for (int j = 0; j < _N_; j++)
+        state[j] = _state[j];
+
+    __syncthreads();
+    u[i] = float(_u[i]);
+    __syncthreads();
+
+    for (int t = b*T*C + h*_N_ + i; t < (b+1)*T*C + h*_N_ + i; t += C)
+    {
+        __syncthreads();
+        w[i] = _w[t];
+        r[i] = float(_r[t]);
+        k[i] = float(_k[t]);
+        __syncthreads();
+
+        const float v = float(_v[t]);
+        float y = 0;
+
+        #pragma unroll
+        for (int j = 0; j < _N_; j+=4)
+        {
+            const float4& r_ = (float4&)(r[j]);
+            const float4& k_ = (float4&)(k[j]);
+            const float4& w_ = (float4&)(w[j]);
+            const float4& u_ = (float4&)(u[j]);
+            float4& s = (float4&)(state[j]);
+            float4 x;
+
+            x.x = k_.x * v;
+            x.y = k_.y * v;
+            x.z = k_.z * v;
+            x.w = k_.w * v;
+
+            y += r_.x * (u_.x * x.x + s.x);
+            y += r_.y * (u_.y * x.y + s.y);
+            y += r_.z * (u_.z * x.z + s.z);
+            y += r_.w * (u_.w * x.w + s.w);
+
+            s.x = s.x * w_.x + x.x;
+            s.y = s.y * w_.y + x.y;
+            s.z = s.z * w_.z + x.z;
+            s.w = s.w * w_.w + x.w;
+        }
+        _y[t] = F(y);
+    }
+    #pragma unroll
+    for (int j = 0; j < _N_; j++)
+        _state[j] = state[j];
+}
+
+void cuda_forward_bf16(int B, int T, int C, int H, float *state, bf16 *r, bf16 *k, bf16 *v, float *w, bf16 *u, bf16 *y)
+{
+    assert(H*_N_ == C);
+    kernel_forward<<<dim3(B * H), dim3(_N_)>>>(B, T, C, H, state, r, k, v, w, u, y);
+}
+void cuda_forward_fp16(int B, int T, int C, int H, float *state, fp16 *r, fp16 *k, fp16 *v, float *w, fp16 *u, fp16 *y)
+{
+    assert(H*_N_ == C);
+    kernel_forward<<<dim3(B * H), dim3(_N_)>>>(B, T, C, H, state, r, k, v, w, u, y);
+}
+void cuda_forward_fp32(int B, int T, int C, int H, float *state, fp32 *r, fp32 *k, fp32 *v, float *w, fp32 *u, fp32 *y)
+{
+    assert(H*_N_ == C);
+    kernel_forward<<<dim3(B * H), dim3(_N_)>>>(B, T, C, H, state, r, k, v, w, u, y);
+}

backend-python/rwkv_pip/cuda/rwkv6_op.cpp

@@ -0,0 +1,34 @@
+#include <torch/extension.h>
+#include "ATen/ATen.h"
+#include <c10/cuda/CUDAGuard.h>
+typedef at::BFloat16 bf16;
+typedef at::Half fp16;
+typedef float fp32;
+
+void cuda_forward_bf16(int B, int T, int C, int H, float *state, bf16 *r, bf16 *k, bf16 *v, float *w, bf16 *u, bf16 *y);
+void cuda_forward_fp16(int B, int T, int C, int H, float *state, fp16 *r, fp16 *k, fp16 *v, float *w, fp16 *u, fp16 *y);
+void cuda_forward_fp32(int B, int T, int C, int H, float *state, fp32 *r, fp32 *k, fp32 *v, float *w, fp32 *u, fp32 *y);
+
+void forward_bf16(int64_t B, int64_t T, int64_t C, int64_t H, torch::Tensor &state, torch::Tensor &r, torch::Tensor &k, torch::Tensor &v, torch::Tensor &w, torch::Tensor &u, torch::Tensor &y) {
+    const at::cuda::OptionalCUDAGuard device_guard(device_of(state));
+    cuda_forward_bf16(B, T, C, H, state.data_ptr<float>(), r.data_ptr<bf16>(), k.data_ptr<bf16>(), v.data_ptr<bf16>(), w.data_ptr<float>(), u.data_ptr<bf16>(), y.data_ptr<bf16>());
+}
+void forward_fp16(int64_t B, int64_t T, int64_t C, int64_t H, torch::Tensor &state, torch::Tensor &r, torch::Tensor &k, torch::Tensor &v, torch::Tensor &w, torch::Tensor &u, torch::Tensor &y) {
+    const at::cuda::OptionalCUDAGuard device_guard(device_of(state));
+    cuda_forward_fp16(B, T, C, H, state.data_ptr<float>(), r.data_ptr<fp16>(), k.data_ptr<fp16>(), v.data_ptr<fp16>(), w.data_ptr<float>(), u.data_ptr<fp16>(), y.data_ptr<fp16>());
+}
+void forward_fp32(int64_t B, int64_t T, int64_t C, int64_t H, torch::Tensor &state, torch::Tensor &r, torch::Tensor &k, torch::Tensor &v, torch::Tensor &w, torch::Tensor &u, torch::Tensor &y) {
+    const at::cuda::OptionalCUDAGuard device_guard(device_of(state));
+    cuda_forward_fp32(B, T, C, H, state.data_ptr<float>(), r.data_ptr<fp32>(), k.data_ptr<fp32>(), v.data_ptr<fp32>(), w.data_ptr<float>(), u.data_ptr<fp32>(), y.data_ptr<fp32>());
+}
+
+PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) {
+    m.def("forward_bf16", &forward_bf16, "rwkv6 forward_bf16");
+    m.def("forward_fp16", &forward_fp16, "rwkv6 forward_fp16");
+    m.def("forward_fp32", &forward_fp32, "rwkv6 forward_fp32");
+}
+TORCH_LIBRARY(rwkv6, m) {
+    m.def("forward_bf16", forward_bf16);
+    m.def("forward_fp16", forward_fp16);
+    m.def("forward_fp32", forward_fp32);
+}

frontend/package.json

@@ -21,6 +21,7 @@
     "i18next": "^22.4.15",
     "mobx": "^6.9.0",
     "mobx-react-lite": "^3.4.3",
+    "pdfjs-dist": "^4.0.189",
     "react": "^18.2.0",
     "react-beautiful-dnd": "^13.1.1",
     "react-chartjs-2": "^5.2.0",
@@ -50,6 +51,7 @@
     "sass": "^1.62.1",
     "tailwindcss": "^3.3.2",
     "typescript": "^5.0.4",
-    "vite": "^4.3.6"
+    "vite": "^4.3.6",
+    "vite-plugin-top-level-await": "^1.3.1"
   }
 }

frontend/src/_locales/ja/main.json

@@ -271,5 +271,6 @@
   "You can increase the number of stored layers in Configs page to improve performance": "パフォーマンスを向上させるために、保存されるレイヤーの数を設定ページで増やすことができます",
   "Failed to load model, try to increase the virtual memory (Swap of WSL) or use a smaller base model.": "モデルの読み込みに失敗しました、仮想メモリ (WSL Swap) を増やすか小さなベースモデルを使用してみてください。",
   "Save Conversation": "会話を保存",
-  "Use Hugging Face Mirror": "Hugging Faceミラーを使用"
+  "Use Hugging Face Mirror": "Hugging Faceミラーを使用",
+  "File is empty": "ファイルが空です"
 }

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

@@ -271,5 +271,6 @@
   "You can increase the number of stored layers in Configs page to improve performance": "你可以在配置页面增加载入显存层数以提升性能",
   "Failed to load model, try to increase the virtual memory (Swap of WSL) or use a smaller base model.": "模型载入失败，尝试增加虚拟内存(WSL Swap)，或使用一个更小规模的基底模型",
   "Save Conversation": "保存对话",
-  "Use Hugging Face Mirror": "使用Hugging Face镜像源"
+  "Use Hugging Face Mirror": "使用Hugging Face镜像源",
+  "File is empty": "文件为空"
 }

frontend/src/components/RunButton.tsx

@@ -17,6 +17,7 @@ import { ToolTipButton } from './ToolTipButton';
 import { Play16Regular, Stop16Regular } from '@fluentui/react-icons';
 import { useNavigate } from 'react-router';
 import { WindowShow } from '../../wailsjs/runtime';
+import { convertToSt } from '../utils/convert-to-st';
 
 const mainButtonText = {
   [ModelStatus.Offline]: 'Run',
@@ -57,13 +58,36 @@ export const RunButton: FC<{ onClickRun?: MouseEventHandler, iconMode?: boolean
         return;
       }
 
+      const currentModelSource = commonStore.modelSourceList.find(item => item.name === modelName);
+
+      const showDownloadPrompt = (promptInfo: string, downloadName: string) => {
+        toastWithButton(promptInfo, t('Download'), () => {
+          const downloadUrl = currentModelSource?.downloadUrl;
+          if (downloadUrl) {
+            toastWithButton(`${t('Downloading')} ${downloadName}`, t('Check'), () => {
+                navigate({ pathname: '/downloads' });
+              },
+              { autoClose: 3000 });
+            AddToDownloadList(modelPath, getHfDownloadUrl(downloadUrl));
+          } else {
+            toast(t('Can not find download url'), { type: 'error' });
+          }
+        });
+      };
+
       if (webgpu) {
         if (!['.st', '.safetensors'].some(ext => modelPath.endsWith(ext))) {
           const stModelPath = modelPath.replace(/\.pth$/, '.st');
           if (await FileExists(stModelPath)) {
             modelPath = stModelPath;
+          } else if (!await FileExists(modelPath)) {
+            showDownloadPrompt(t('Model file not found'), modelName);
+            commonStore.setStatus({ status: ModelStatus.Offline });
+            return;
           } else {
-            toast(t('Please convert model to safe tensors format first'), { type: 'error' });
+            toastWithButton(t('Please convert model to safe tensors format first'), t('Convert'), () => {
+              convertToSt(navigate, modelConfig);
+            });
             commonStore.setStatus({ status: ModelStatus.Offline });
             return;
           }
@@ -84,23 +108,6 @@ export const RunButton: FC<{ onClickRun?: MouseEventHandler, iconMode?: boolean
           return;
       }
 
-      const currentModelSource = commonStore.modelSourceList.find(item => item.name === modelName);
-
-      const showDownloadPrompt = (promptInfo: string, downloadName: string) => {
-        toastWithButton(promptInfo, t('Download'), () => {
-          const downloadUrl = currentModelSource?.downloadUrl;
-          if (downloadUrl) {
-            toastWithButton(`${t('Downloading')} ${downloadName}`, t('Check'), () => {
-                navigate({ pathname: '/downloads' });
-              },
-              { autoClose: 3000 });
-            AddToDownloadList(modelPath, getHfDownloadUrl(downloadUrl));
-          } else {
-            toast(t('Can not find download url'), { type: 'error' });
-          }
-        });
-      };
-
       if (!await FileExists(modelPath)) {
         showDownloadPrompt(t('Model file not found'), modelName);
         commonStore.setStatus({ status: ModelStatus.Offline });

frontend/src/pages/Chat.tsx

@@ -34,6 +34,7 @@ import { botName, ConversationMessage, MessageType, userName, welcomeUuid } from
 import { Labeled } from '../components/Labeled';
 import { ValuedSlider } from '../components/ValuedSlider';
 import { PresetsButton } from './PresetsManager/PresetsButton';
+import { webOpenOpenFileDialog } from '../utils/web-file-operations';
 
 let chatSseControllers: {
   [id: string]: AbortController
@@ -515,6 +516,7 @@ const ChatPanel: FC = observer(() => {
     <div className="flex h-full grow pt-4 overflow-hidden">
       <div className="relative flex flex-col w-full grow gap-4 overflow-hidden">
         <Button className="absolute top-1 right-1" size="medium" shape="circular" appearance="subtle"
+          style={{ zIndex: 1 }}
           icon={commonStore.sidePanelCollapsed ? <TextAlignJustify24Regular /> : <TextAlignJustifyRotate9024Regular />}
           onClick={() => commonStore.setSidePanelCollapsed(!commonStore.sidePanelCollapsed)} />
         <div ref={bodyRef} className="grow overflow-y-scroll overflow-x-hidden pr-2">
@@ -559,43 +561,42 @@ const ChatPanel: FC = observer(() => {
                     : <Attach16Regular />}
                   size="small" shape="circular" appearance="secondary"
                   onClick={() => {
-                    if (commonStore.status.status === ModelStatus.Offline && !commonStore.settings.apiUrl && commonStore.platform !== 'web') {
-                      toast(t('Please click the button in the top right corner to start the model'), { type: 'warning' });
-                      return;
-                    }
-
                     if (commonStore.attachmentUploading)
                       return;
 
-                    OpenOpenFileDialog('*.txt;*.pdf').then(async filePath => {
-                      if (!filePath)
-                        return;
+                    const filterPattern = '*.txt;*.pdf';
+                    const setUploading = () => commonStore.setAttachmentUploading(true);
+                    // actually, status of web platform is always Offline
+                    if (commonStore.platform === 'web' || commonStore.status.status === ModelStatus.Offline) {
+                      webOpenOpenFileDialog({ filterPattern, fnStartLoading: setUploading }).then(webReturn => {
+                        if (webReturn.content)
+                          commonStore.setCurrentTempAttachment(
+                            {
+                              name: webReturn.blob.name,
+                              size: webReturn.blob.size,
+                              content: webReturn.content
+                            });
+                        else
+                          toast(t('File is empty'), {
+                            type: 'info',
+                            autoClose: 1000
+                          });
+                      }).catch(e => {
+                        toast(t('Error') + ' - ' + (e.message || e), { type: 'error', autoClose: 2500 });
+                      }).finally(() => {
+                        commonStore.setAttachmentUploading(false);
+                      });
+                    } else {
+                      OpenOpenFileDialog(filterPattern).then(async filePath => {
+                        if (!filePath)
+                          return;
 
-                      commonStore.setAttachmentUploading(true);
+                        setUploading();
 
-                      let blob: Blob;
-                      let attachmentName: string | undefined;
-                      let attachmentContent: string | undefined;
-                      if (commonStore.platform === 'web') {
-                        const webReturn = filePath as any;
-                        blob = webReturn.blob;
-                        attachmentName = blob.name;
-                        attachmentContent = webReturn.content;
-                      } else {
                         // Both are slow. Communication between frontend and backend is slow. Use AssetServer Handler to read the file.
                         // const blob = new Blob([atob(info.content as unknown as string)]); // await fetch(`data:application/octet-stream;base64,${info.content}`).then(r => r.blob());
-                        blob = await fetch(absPathAsset(filePath)).then(r => r.blob());
-                        attachmentName = filePath.split(/[\\/]/).pop();
-                      }
-                      if (attachmentContent) {
-                        commonStore.setCurrentTempAttachment(
-                          {
-                            name: attachmentName!,
-                            size: blob.size,
-                            content: attachmentContent
-                          });
-                        commonStore.setAttachmentUploading(false);
-                      } else {
+                        const blob = await fetch(absPathAsset(filePath)).then(r => r.blob());
+                        const attachmentName = filePath.split(/[\\/]/).pop();
                         const urlPath = `/file-to-text?file_name=${attachmentName}`;
                         const bodyForm = new FormData();
                         bodyForm.append('file_data', blob, attachmentName);
@@ -605,31 +606,38 @@ const ChatPanel: FC = observer(() => {
                         }).then(async r => {
                             if (r.status === 200) {
                               const pages = (await r.json()).pages as any[];
+                              let attachmentContent: string;
                               if (pages.length === 1)
                                 attachmentContent = pages[0].page_content;
                               else
                                 attachmentContent = pages.map((p, i) => `Page ${i + 1}:\n${p.page_content}`).join('\n\n');
-                              commonStore.setCurrentTempAttachment(
-                                {
-                                  name: attachmentName!,
-                                  size: blob.size,
-                                  content: attachmentContent!
+                              if (attachmentContent)
+                                commonStore.setCurrentTempAttachment(
+                                  {
+                                    name: attachmentName!,
+                                    size: blob.size,
+                                    content: attachmentContent!
+                                  });
+                              else
+                                toast(t('File is empty'), {
+                                  type: 'info',
+                                  autoClose: 1000
                                 });
                             } else {
                               toast(r.statusText + '\n' + (await r.text()), {
                                 type: 'error'
                               });
                             }
-                            commonStore.setAttachmentUploading(false);
                           }
                         ).catch(e => {
-                          commonStore.setAttachmentUploading(false);
                           toast(t('Error') + ' - ' + (e.message || e), { type: 'error', autoClose: 2500 });
+                        }).finally(() => {
+                          commonStore.setAttachmentUploading(false);
                         });
-                      }
-                    }).catch(e => {
-                      toast(t('Error') + ' - ' + (e.message || e), { type: 'error', autoClose: 2500 });
-                    });
+                      }).catch(e => {
+                        toast(t('Error') + ' - ' + (e.message || e), { type: 'error', autoClose: 2500 });
+                      });
+                    }
                   }}
                 /> :
                 <div>

frontend/src/pages/Configs.tsx

@@ -26,7 +26,7 @@ import { Page } from '../components/Page';
 import { useNavigate } from 'react-router';
 import { RunButton } from '../components/RunButton';
 import { updateConfig } from '../apis';
-import { ConvertModel, ConvertSafetensors, FileExists, GetPyError } from '../../wailsjs/go/backend_golang/App';
+import { ConvertModel, FileExists, GetPyError } from '../../wailsjs/go/backend_golang/App';
 import { checkDependencies, getStrategy } from '../utils';
 import { useTranslation } from 'react-i18next';
 import { WindowShow } from '../../wailsjs/runtime';
@@ -35,6 +35,7 @@ import strategyZhImg from '../assets/images/strategy_zh.jpg';
 import { ResetConfigsButton } from '../components/ResetConfigsButton';
 import { useMediaQuery } from 'usehooks-ts';
 import { ApiParameters, Device, ModelParameters, Precision } from '../types/configs';
+import { convertToSt } from '../utils/convert-to-st';
 
 const Configs: FC = observer(() => {
   const { t } = useTranslation();
@@ -272,41 +273,7 @@ const Configs: FC = observer(() => {
                       }} /> :
                     <ToolTipButton text={t('Convert To Safe Tensors Format')}
                       desc=""
-                      onClick={async () => {
-                        if (commonStore.platform === 'darwin') {
-                          toast(t('MacOS is not yet supported for performing this operation, please do it manually.') + ' (backend-python/convert_safetensors.py)', { type: 'info' });
-                          return;
-                        } else if (commonStore.platform === 'linux') {
-                          toast(t('Linux is not yet supported for performing this operation, please do it manually.') + ' (backend-python/convert_safetensors.py)', { type: 'info' });
-                          return;
-                        }
-
-                        const ok = await checkDependencies(navigate);
-                        if (!ok)
-                          return;
-
-                        const modelPath = `${commonStore.settings.customModelsPath}/${selectedConfig.modelParameters.modelName}`;
-                        if (await FileExists(modelPath)) {
-                          toast(t('Start Converting'), { autoClose: 1000, type: 'info' });
-                          const newModelPath = modelPath.replace(/\.pth$/, '.st');
-                          ConvertSafetensors(commonStore.settings.customPythonPath, modelPath, newModelPath).then(async () => {
-                            if (!await FileExists(newModelPath)) {
-                              toast(t('Convert Failed') + ' - ' + await GetPyError(), { type: 'error' });
-                            } else {
-                              toast(`${t('Convert Success')} - ${newModelPath}`, { type: 'success' });
-                            }
-                          }).catch(e => {
-                            const errMsg = e.message || e;
-                            if (errMsg.includes('path contains space'))
-                              toast(`${t('Convert Failed')} - ${t('File Path Cannot Contain Space')}`, { type: 'error' });
-                            else
-                              toast(`${t('Convert Failed')} - ${e.message || e}`, { type: 'error' });
-                          });
-                          setTimeout(WindowShow, 1000);
-                        } else {
-                          toast(`${t('Model Not Found')} - ${modelPath}`, { type: 'error' });
-                        }
-                      }} />
+                      onClick={() => convertToSt(navigate, selectedConfig)} />
                 }
                 <Labeled label={t('Strategy')} content={
                   <Dropdown style={{ minWidth: 0 }} className="grow" value={t(selectedConfig.modelParameters.device)!}

frontend/src/pages/defaultConfigs.ts

@@ -111,7 +111,7 @@ export const defaultPresets: CompletionPreset[] = [{
 
 export const defaultModelConfigsMac: ModelConfig[] = [
   {
-    name: 'MAC-0.1B-World',
+    name: 'GPU-2G-1B5-World',
     apiParameters: {
       apiPort: 8000,
       maxResponseToken: 4100,
@@ -121,16 +121,15 @@ export const defaultModelConfigsMac: ModelConfig[] = [
       frequencyPenalty: 1
     },
     modelParameters: {
-      modelName: 'RWKV-4-World-0.1B-v1-20230520-ctx4096.pth',
-      device: 'MPS',
-      precision: 'fp32',
+      modelName: 'RWKV-5-World-1B5-v2-20231025-ctx4096.pth',
+      device: 'WebGPU',
+      precision: 'nf4',
       storedLayers: 41,
-      maxStoredLayers: 41,
-      customStrategy: 'mps fp32'
+      maxStoredLayers: 41
     }
   },
   {
-    name: 'MAC-0.4B-World',
+    name: 'GPU-4G-3B-World',
     apiParameters: {
       apiPort: 8000,
       maxResponseToken: 4100,
@@ -140,12 +139,101 @@ export const defaultModelConfigsMac: ModelConfig[] = [
       frequencyPenalty: 1
     },
     modelParameters: {
-      modelName: 'RWKV-4-World-0.4B-v1-20230529-ctx4096.pth',
-      device: 'MPS',
+      modelName: 'RWKV-5-World-3B-v2-20231118-ctx16k.pth',
+      device: 'WebGPU',
+      precision: 'nf4',
+      storedLayers: 41,
+      maxStoredLayers: 41
+    }
+  },
+  {
+    name: 'GPU-4G-3B-CN',
+    apiParameters: {
+      apiPort: 8000,
+      maxResponseToken: 4100,
+      temperature: 1.0,
+      topP: 0.3,
+      presencePenalty: 0,
+      frequencyPenalty: 1
+    },
+    modelParameters: {
+      modelName: 'RWKV-4-World-CHNtuned-3B-v1-20230625-ctx4096.pth',
+      device: 'WebGPU',
+      precision: 'nf4',
+      storedLayers: 41,
+      maxStoredLayers: 41
+    }
+  },
+  {
+    name: 'GPU-7G-7B-World',
+    apiParameters: {
+      apiPort: 8000,
+      maxResponseToken: 4100,
+      temperature: 1.0,
+      topP: 0.3,
+      presencePenalty: 0,
+      frequencyPenalty: 1
+    },
+    modelParameters: {
+      modelName: 'RWKV-4-World-7B-v1-20230626-ctx4096.pth',
+      device: 'WebGPU',
+      precision: 'nf4',
+      storedLayers: 41,
+      maxStoredLayers: 41
+    }
+  },
+  {
+    name: 'GPU-7G-7B-CN',
+    apiParameters: {
+      apiPort: 8000,
+      maxResponseToken: 4100,
+      temperature: 1.0,
+      topP: 0.3,
+      presencePenalty: 0,
+      frequencyPenalty: 1
+    },
+    modelParameters: {
+      modelName: 'RWKV-4-World-CHNtuned-7B-v1-20230709-ctx4096.pth',
+      device: 'WebGPU',
+      precision: 'nf4',
+      storedLayers: 41,
+      maxStoredLayers: 41
+    }
+  },
+  {
+    name: 'CPU-120M-Music',
+    apiParameters: {
+      apiPort: 8000,
+      maxResponseToken: 4100,
+      temperature: 1.0,
+      topP: 0.8,
+      presencePenalty: 0,
+      frequencyPenalty: 1
+    },
+    modelParameters: {
+      modelName: 'RWKV-4-MIDI-120M-v1-20230714-ctx4096.pth',
+      device: 'CPU',
       precision: 'fp32',
       storedLayers: 41,
-      maxStoredLayers: 41,
-      customStrategy: 'mps fp32'
+      maxStoredLayers: 41
+    }
+  },
+  {
+    name: 'CPU-560M-Music',
+    apiParameters: {
+      apiPort: 8000,
+      maxResponseToken: 4100,
+      temperature: 1.0,
+      topP: 0.8,
+      presencePenalty: 0,
+      frequencyPenalty: 1
+    },
+    modelParameters: {
+      modelName: 'RWKV-4-MIDI-560M-v1-20230717-ctx4096.pth',
+      device: 'CPU',
+      precision: 'fp32',
+      storedLayers: 41,
+      maxStoredLayers: 41
     }
   },
   {
@@ -159,26 +247,7 @@ export const defaultModelConfigsMac: ModelConfig[] = [
       frequencyPenalty: 1
     },
     modelParameters: {
-      modelName: 'RWKV-4-World-1.5B-v1-fixed-20230612-ctx4096.pth',
-      device: 'MPS',
-      precision: 'fp32',
-      storedLayers: 41,
-      maxStoredLayers: 41,
-      customStrategy: 'mps fp32'
-    }
-  },
-  {
-    name: 'MAC-1B5-EN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-Raven-1B5-v12-Eng98%-Other2%-20230520-ctx4096.pth',
+      modelName: 'RWKV-5-World-1B5-v2-20231025-ctx4096.pth',
       device: 'MPS',
       precision: 'fp32',
       storedLayers: 41,
@@ -197,26 +266,7 @@ export const defaultModelConfigsMac: ModelConfig[] = [
       frequencyPenalty: 1
     },
     modelParameters: {
-      modelName: 'RWKV-4-World-3B-v1-20230619-ctx4096.pth',
-      device: 'MPS',
-      precision: 'fp32',
-      storedLayers: 41,
-      maxStoredLayers: 41,
-      customStrategy: 'mps fp32'
-    }
-  },
-  {
-    name: 'MAC-3B-EN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-Raven-3B-v12-Eng98%-Other2%-20230520-ctx4096.pth',
+      modelName: 'RWKV-5-World-3B-v2-20231118-ctx16k.pth',
       device: 'MPS',
       precision: 'fp32',
       storedLayers: 41,
@@ -261,190 +311,29 @@ export const defaultModelConfigsMac: ModelConfig[] = [
       maxStoredLayers: 41,
       customStrategy: 'mps fp32'
     }
-  },
-  {
-    name: 'CPU-120M-Music',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.8,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-MIDI-120M-v1-20230714-ctx4096.pth',
-      device: 'CPU',
-      precision: 'fp32',
-      storedLayers: 41,
-      maxStoredLayers: 41
-    }
-  },
-  {
-    name: 'CPU-560M-Music',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.8,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-MIDI-560M-v1-20230717-ctx4096.pth',
-      device: 'CPU',
-      precision: 'fp32',
-      storedLayers: 41,
-      maxStoredLayers: 41
-    }
-  },
-  {
-    name: 'CPU-6G-1B5-World',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-World-1.5B-v1-fixed-20230612-ctx4096.pth',
-      device: 'CPU',
-      precision: 'fp32',
-      storedLayers: 41,
-      maxStoredLayers: 41
-    }
-  },
-  {
-    name: 'CPU-6G-1B5-EN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-Raven-1B5-v12-Eng98%-Other2%-20230520-ctx4096.pth',
-      device: 'CPU',
-      precision: 'fp32',
-      storedLayers: 41,
-      maxStoredLayers: 41
-    }
-  },
-  {
-    name: 'CPU-12G-3B-World',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-World-3B-v1-20230619-ctx4096.pth',
-      device: 'CPU',
-      precision: 'fp32',
-      storedLayers: 41,
-      maxStoredLayers: 41
-    }
-  },
-  {
-    name: 'CPU-12G-3B-EN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-Raven-3B-v12-Eng98%-Other2%-20230520-ctx4096.pth',
-      device: 'CPU',
-      precision: 'fp32',
-      storedLayers: 41,
-      maxStoredLayers: 41
-    }
-  },
-  {
-    name: 'CPU-12G-3B-CN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-World-CHNtuned-3B-v1-20230625-ctx4096.pth',
-      device: 'CPU',
-      precision: 'fp32',
-      storedLayers: 41,
-      maxStoredLayers: 41
-    }
-  },
-  {
-    name: 'CPU-28G-7B-World',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-World-7B-v1-20230626-ctx4096.pth',
-      device: 'CPU',
-      precision: 'fp32',
-      storedLayers: 41,
-      maxStoredLayers: 41
-    }
-  },
-  {
-    name: 'CPU-28G-7B-EN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-Raven-7B-v12-Eng98%-Other2%-20230521-ctx8192.pth',
-      device: 'CPU',
-      precision: 'fp32',
-      storedLayers: 41,
-      maxStoredLayers: 41
-    }
-  },
-  {
-    name: 'CPU-28G-7B-CN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-World-CHNtuned-7B-v1-20230709-ctx4096.pth',
-      device: 'CPU',
-      precision: 'fp32',
-      storedLayers: 41,
-      maxStoredLayers: 41
-    }
   }
 ];
 
 export const defaultModelConfigs: ModelConfig[] = [
+  {
+    name: 'GPU-2G-1B5-World',
+    apiParameters: {
+      apiPort: 8000,
+      maxResponseToken: 4100,
+      temperature: 1.0,
+      topP: 0.3,
+      presencePenalty: 0,
+      frequencyPenalty: 1
+    },
+    modelParameters: {
+      modelName: 'RWKV-5-World-1B5-v2-20231025-ctx4096.pth',
+      device: 'CUDA',
+      precision: 'int8',
+      storedLayers: 41,
+      maxStoredLayers: 41,
+      useCustomCuda: true
+    }
+  },
   {
     name: 'GPU-2G-3B-World',
     apiParameters: {
@@ -456,7 +345,7 @@ export const defaultModelConfigs: ModelConfig[] = [
       frequencyPenalty: 1
     },
     modelParameters: {
-      modelName: 'RWKV-4-World-3B-v1-20230619-ctx4096.pth',
+      modelName: 'RWKV-5-World-3B-v2-20231118-ctx16k.pth',
       device: 'CUDA',
       precision: 'int8',
       storedLayers: 6,
@@ -464,61 +353,6 @@ export const defaultModelConfigs: ModelConfig[] = [
       useCustomCuda: true
     }
   },
-  {
-    name: 'GPU-2G-0.1B-World',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-World-0.1B-v1-20230520-ctx4096.pth',
-      device: 'CUDA',
-      precision: 'fp32', // using fp16 will disable state cache (->)
-      storedLayers: 41,
-      maxStoredLayers: 41
-    }
-  },
-  {
-    name: 'GPU-2G-1B5-EN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-Raven-1B5-v12-Eng98%-Other2%-20230520-ctx4096.pth',
-      device: 'CUDA',
-      precision: 'int8',
-      storedLayers: 4,
-      maxStoredLayers: 41,
-      useCustomCuda: true
-    }
-  },
-  {
-    name: 'GPU-4G-0.4B-World',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-World-0.4B-v1-20230529-ctx4096.pth',
-      device: 'CUDA',
-      precision: 'fp32',
-      storedLayers: 41,
-      maxStoredLayers: 41
-    }
-  },
   {
     name: 'GPU-4G-1B5-World',
     apiParameters: {
@@ -530,27 +364,9 @@ export const defaultModelConfigs: ModelConfig[] = [
       frequencyPenalty: 1
     },
     modelParameters: {
-      modelName: 'RWKV-4-World-1.5B-v1-fixed-20230612-ctx4096.pth',
+      modelName: 'RWKV-5-World-1B5-v2-20231025-ctx4096.pth',
       device: 'CUDA',
-      precision: 'fp32',
-      storedLayers: 8,
-      maxStoredLayers: 41
-    }
-  },
-  {
-    name: 'GPU-4G-1B5-EN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-Raven-1B5-v12-Eng98%-Other2%-20230520-ctx4096.pth',
-      device: 'CUDA',
-      precision: 'int8',
+      precision: 'fp16',
       storedLayers: 41,
       maxStoredLayers: 41,
       useCustomCuda: true
@@ -567,26 +383,7 @@ export const defaultModelConfigs: ModelConfig[] = [
       frequencyPenalty: 1
     },
     modelParameters: {
-      modelName: 'RWKV-4-World-3B-v1-20230619-ctx4096.pth',
-      device: 'CUDA',
-      precision: 'int8',
-      storedLayers: 24,
-      maxStoredLayers: 41,
-      useCustomCuda: true
-    }
-  },
-  {
-    name: 'GPU-4G-3B-EN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-Raven-3B-v12-Eng98%-Other2%-20230520-ctx4096.pth',
+      modelName: 'RWKV-5-World-3B-v2-20231118-ctx16k.pth',
       device: 'CUDA',
       precision: 'int8',
       storedLayers: 24,
@@ -632,25 +429,6 @@ export const defaultModelConfigs: ModelConfig[] = [
       useCustomCuda: true
     }
   },
-  {
-    name: 'GPU-4G-7B-EN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-Raven-7B-v12-Eng98%-Other2%-20230521-ctx8192.pth',
-      device: 'CUDA',
-      precision: 'int8',
-      storedLayers: 8,
-      maxStoredLayers: 41,
-      useCustomCuda: true
-    }
-  },
   {
     name: 'GPU-4G-7B-CN',
     apiParameters: {
@@ -670,25 +448,6 @@ export const defaultModelConfigs: ModelConfig[] = [
       useCustomCuda: true
     }
   },
-  {
-    name: 'GPU-6G-1B5-EN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-Raven-1B5-v12-Eng98%-Other2%-20230520-ctx4096.pth',
-      device: 'CUDA',
-      precision: 'fp16',
-      storedLayers: 41,
-      maxStoredLayers: 41,
-      useCustomCuda: true
-    }
-  },
   {
     name: 'GPU-6G-3B-World',
     apiParameters: {
@@ -700,26 +459,7 @@ export const defaultModelConfigs: ModelConfig[] = [
       frequencyPenalty: 1
     },
     modelParameters: {
-      modelName: 'RWKV-4-World-3B-v1-20230619-ctx4096.pth',
-      device: 'CUDA',
-      precision: 'int8',
-      storedLayers: 41,
-      maxStoredLayers: 41,
-      useCustomCuda: true
-    }
-  },
-  {
-    name: 'GPU-6G-3B-EN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-Raven-3B-v12-Eng98%-Other2%-20230520-ctx4096.pth',
+      modelName: 'RWKV-5-World-3B-v2-20231118-ctx16k.pth',
       device: 'CUDA',
       precision: 'int8',
       storedLayers: 41,
@@ -765,25 +505,6 @@ export const defaultModelConfigs: ModelConfig[] = [
       useCustomCuda: true
     }
   },
-  {
-    name: 'GPU-6G-7B-EN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-Raven-7B-v12-Eng98%-Other2%-20230521-ctx8192.pth',
-      device: 'CUDA',
-      precision: 'int8',
-      storedLayers: 18,
-      maxStoredLayers: 41,
-      useCustomCuda: true
-    }
-  },
   {
     name: 'GPU-6G-7B-CN',
     apiParameters: {
@@ -803,24 +524,6 @@ export const defaultModelConfigs: ModelConfig[] = [
       useCustomCuda: true
     }
   },
-  {
-    name: 'GPU-8G-1B5-World',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-World-1.5B-v1-fixed-20230612-ctx4096.pth',
-      device: 'CUDA',
-      precision: 'fp32',
-      storedLayers: 41,
-      maxStoredLayers: 41
-    }
-  },
   {
     name: 'GPU-8G-3B-World',
     apiParameters: {
@@ -832,26 +535,7 @@ export const defaultModelConfigs: ModelConfig[] = [
       frequencyPenalty: 1
     },
     modelParameters: {
-      modelName: 'RWKV-4-World-3B-v1-20230619-ctx4096.pth',
-      device: 'CUDA',
-      precision: 'fp16',
-      storedLayers: 41,
-      maxStoredLayers: 41,
-      useCustomCuda: true
-    }
-  },
-  {
-    name: 'GPU-8G-3B-EN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-Raven-3B-v12-Eng98%-Other2%-20230520-ctx4096.pth',
+      modelName: 'RWKV-5-World-3B-v2-20231118-ctx16k.pth',
       device: 'CUDA',
       precision: 'fp16',
       storedLayers: 41,
@@ -897,25 +581,6 @@ export const defaultModelConfigs: ModelConfig[] = [
       useCustomCuda: true
     }
   },
-  {
-    name: 'GPU-8G-7B-EN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-Raven-7B-v12-Eng98%-Other2%-20230521-ctx8192.pth',
-      device: 'CUDA',
-      precision: 'int8',
-      storedLayers: 27,
-      maxStoredLayers: 41,
-      useCustomCuda: true
-    }
-  },
   {
     name: 'GPU-8G-7B-CN',
     apiParameters: {
@@ -954,25 +619,6 @@ export const defaultModelConfigs: ModelConfig[] = [
       useCustomCuda: true
     }
   },
-  {
-    name: 'GPU-10G-7B-EN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-Raven-7B-v12-Eng98%-Other2%-20230521-ctx8192.pth',
-      device: 'CUDA',
-      precision: 'int8',
-      storedLayers: 41,
-      maxStoredLayers: 41,
-      useCustomCuda: true
-    }
-  },
   {
     name: 'GPU-10G-7B-CN',
     apiParameters: {
@@ -992,25 +638,6 @@ export const defaultModelConfigs: ModelConfig[] = [
       useCustomCuda: true
     }
   },
-  {
-    name: 'GPU-12G-14B-EN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-Raven-14B-v12-Eng98%-Other2%-20230523-ctx8192.pth',
-      device: 'CUDA',
-      precision: 'int8',
-      storedLayers: 24,
-      maxStoredLayers: 41,
-      useCustomCuda: true
-    }
-  },
   {
     name: 'GPU-16G-7B-World',
     apiParameters: {
@@ -1030,25 +657,6 @@ export const defaultModelConfigs: ModelConfig[] = [
       useCustomCuda: true
     }
   },
-  {
-    name: 'GPU-16G-7B-EN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-Raven-7B-v12-Eng98%-Other2%-20230521-ctx8192.pth',
-      device: 'CUDA',
-      precision: 'fp16',
-      storedLayers: 41,
-      maxStoredLayers: 41,
-      useCustomCuda: true
-    }
-  },
   {
     name: 'GPU-16G-7B-CN',
     apiParameters: {
@@ -1068,63 +676,6 @@ export const defaultModelConfigs: ModelConfig[] = [
       useCustomCuda: true
     }
   },
-  {
-    name: 'GPU-16G-14B-EN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-Raven-14B-v12-Eng98%-Other2%-20230523-ctx8192.pth',
-      device: 'CUDA',
-      precision: 'int8',
-      storedLayers: 37,
-      maxStoredLayers: 41,
-      useCustomCuda: true
-    }
-  },
-  {
-    name: 'GPU-18G-14B-EN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-Raven-14B-v12-Eng98%-Other2%-20230523-ctx8192.pth',
-      device: 'CUDA',
-      precision: 'int8',
-      storedLayers: 41,
-      maxStoredLayers: 41,
-      useCustomCuda: true
-    }
-  },
-  {
-    name: 'GPU-32G-14B-EN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-Raven-14B-v12-Eng98%-Other2%-20230523-ctx8192.pth',
-      device: 'CUDA',
-      precision: 'fp16',
-      storedLayers: 41,
-      maxStoredLayers: 41,
-      useCustomCuda: true
-    }
-  },
   {
     name: 'CPU-120M-Music',
     apiParameters: {
@@ -1162,7 +713,7 @@ export const defaultModelConfigs: ModelConfig[] = [
     }
   },
   {
-    name: 'CPU-6G-1B5-World',
+    name: 'AnyGPU-2G-1B5-World',
     apiParameters: {
       apiPort: 8000,
       maxResponseToken: 4100,
@@ -1172,15 +723,15 @@ export const defaultModelConfigs: ModelConfig[] = [
       frequencyPenalty: 1
     },
     modelParameters: {
-      modelName: 'RWKV-4-World-1.5B-v1-fixed-20230612-ctx4096.pth',
-      device: 'CPU',
-      precision: 'fp32',
+      modelName: 'RWKV-5-World-1B5-v2-20231025-ctx4096.pth',
+      device: 'WebGPU',
+      precision: 'nf4',
       storedLayers: 41,
       maxStoredLayers: 41
     }
   },
   {
-    name: 'CPU-6G-1B5-EN',
+    name: 'AnyGPU-4G-3B-World',
     apiParameters: {
       apiPort: 8000,
       maxResponseToken: 4100,
@@ -1190,51 +741,15 @@ export const defaultModelConfigs: ModelConfig[] = [
       frequencyPenalty: 1
     },
     modelParameters: {
-      modelName: 'RWKV-4-Raven-1B5-v12-Eng98%-Other2%-20230520-ctx4096.pth',
-      device: 'CPU',
-      precision: 'fp32',
+      modelName: 'RWKV-5-World-3B-v2-20231118-ctx16k.pth',
+      device: 'WebGPU',
+      precision: 'nf4',
       storedLayers: 41,
       maxStoredLayers: 41
     }
   },
   {
-    name: 'CPU-12G-3B-World',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-World-3B-v1-20230619-ctx4096.pth',
-      device: 'CPU',
-      precision: 'fp32',
-      storedLayers: 41,
-      maxStoredLayers: 41
-    }
-  },
-  {
-    name: 'CPU-12G-3B-EN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-Raven-3B-v12-Eng98%-Other2%-20230520-ctx4096.pth',
-      device: 'CPU',
-      precision: 'fp32',
-      storedLayers: 41,
-      maxStoredLayers: 41
-    }
-  },
-  {
-    name: 'CPU-12G-3B-CN',
+    name: 'AnyGPU-4G-3B-CN',
     apiParameters: {
       apiPort: 8000,
       maxResponseToken: 4100,
@@ -1245,14 +760,14 @@ export const defaultModelConfigs: ModelConfig[] = [
     },
     modelParameters: {
       modelName: 'RWKV-4-World-CHNtuned-3B-v1-20230625-ctx4096.pth',
-      device: 'CPU',
-      precision: 'fp32',
+      device: 'WebGPU',
+      precision: 'nf4',
       storedLayers: 41,
       maxStoredLayers: 41
     }
   },
   {
-    name: 'CPU-28G-7B-World',
+    name: 'AnyGPU-7G-7B-World',
     apiParameters: {
       apiPort: 8000,
       maxResponseToken: 4100,
@@ -1263,32 +778,14 @@ export const defaultModelConfigs: ModelConfig[] = [
     },
     modelParameters: {
       modelName: 'RWKV-4-World-7B-v1-20230626-ctx4096.pth',
-      device: 'CPU',
-      precision: 'fp32',
+      device: 'WebGPU',
+      precision: 'nf4',
       storedLayers: 41,
       maxStoredLayers: 41
     }
   },
   {
-    name: 'CPU-28G-7B-EN',
-    apiParameters: {
-      apiPort: 8000,
-      maxResponseToken: 4100,
-      temperature: 1.0,
-      topP: 0.3,
-      presencePenalty: 0,
-      frequencyPenalty: 1
-    },
-    modelParameters: {
-      modelName: 'RWKV-4-Raven-7B-v12-Eng98%-Other2%-20230521-ctx8192.pth',
-      device: 'CPU',
-      precision: 'fp32',
-      storedLayers: 41,
-      maxStoredLayers: 41
-    }
-  },
-  {
-    name: 'CPU-28G-7B-CN',
+    name: 'AnyGPU-7G-7B-CN',
     apiParameters: {
       apiPort: 8000,
       maxResponseToken: 4100,
@@ -1299,8 +796,8 @@ export const defaultModelConfigs: ModelConfig[] = [
     },
     modelParameters: {
       modelName: 'RWKV-4-World-CHNtuned-7B-v1-20230709-ctx4096.pth',
-      device: 'CPU',
-      precision: 'fp32',
+      device: 'WebGPU',
+      precision: 'nf4',
       storedLayers: 41,
       maxStoredLayers: 41
     }

frontend/src/utils/convert-to-st.ts

@@ -0,0 +1,41 @@
+import { toast } from 'react-toastify';
+import commonStore from '../stores/commonStore';
+import { t } from 'i18next';
+import { checkDependencies } from './index';
+import { ConvertSafetensors, FileExists, GetPyError } from '../../wailsjs/go/backend_golang/App';
+import { WindowShow } from '../../wailsjs/runtime';
+import { NavigateFunction } from 'react-router';
+import { ModelConfig } from '../types/configs';
+
+export const convertToSt = async (navigate: NavigateFunction, selectedConfig: ModelConfig) => {
+  if (commonStore.platform === 'linux') {
+    toast(t('Linux is not yet supported for performing this operation, please do it manually.') + ' (backend-python/convert_safetensors.py)', { type: 'info' });
+    return;
+  }
+
+  const ok = await checkDependencies(navigate);
+  if (!ok)
+    return;
+
+  const modelPath = `${commonStore.settings.customModelsPath}/${selectedConfig.modelParameters.modelName}`;
+  if (await FileExists(modelPath)) {
+    toast(t('Start Converting'), { autoClose: 1000, type: 'info' });
+    const newModelPath = modelPath.replace(/\.pth$/, '.st');
+    ConvertSafetensors(commonStore.settings.customPythonPath, modelPath, newModelPath).then(async () => {
+      if (!await FileExists(newModelPath)) {
+        toast(t('Convert Failed') + ' - ' + await GetPyError(), { type: 'error' });
+      } else {
+        toast(`${t('Convert Success')} - ${newModelPath}`, { type: 'success' });
+      }
+    }).catch(e => {
+      const errMsg = e.message || e;
+      if (errMsg.includes('path contains space'))
+        toast(`${t('Convert Failed')} - ${t('File Path Cannot Contain Space')}`, { type: 'error' });
+      else
+        toast(`${t('Convert Failed')} - ${e.message || e}`, { type: 'error' });
+    });
+    setTimeout(WindowShow, 1000);
+  } else {
+    toast(`${t('Model Not Found')} - ${modelPath}`, { type: 'error' });
+  }
+};

frontend/src/utils/index.tsx

@@ -117,7 +117,7 @@ function initLastUnfinishedModelDownloads() {
         {
           name: item.name,
           path: `${commonStore.settings.customModelsPath}/${item.name}`,
-          url: item.downloadUrl!,
+          url: getHfDownloadUrl(item.downloadUrl!),
           transferred: item.localSize!,
           size: item.size,
           speed: 0,

frontend/src/utils/web-file-operations.ts

@@ -0,0 +1,72 @@
+import { getDocument, GlobalWorkerOptions, PDFDocumentProxy } from 'pdfjs-dist';
+import { TextItem } from 'pdfjs-dist/types/src/display/api';
+
+export function webOpenOpenFileDialog({ filterPattern, fnStartLoading }: { filterPattern: string, fnStartLoading: Function | null }): Promise<{ blob: Blob, content?: string }> {
+  return new Promise((resolve, reject) => {
+    const input = document.createElement('input');
+    input.type = 'file';
+    input.accept = filterPattern
+    .replaceAll('*.txt', 'text/plain')
+    .replaceAll('*.', 'application/')
+    .replaceAll(';', ',');
+
+    input.onchange = async e => {
+      // @ts-ignore
+      const file: Blob = e.target?.files[0];
+      if (fnStartLoading && typeof fnStartLoading === 'function')
+        fnStartLoading();
+      if (!GlobalWorkerOptions.workerSrc)
+        // @ts-ignore
+        GlobalWorkerOptions.workerSrc = await import('pdfjs-dist/build/pdf.worker.min.mjs');
+      if (file.type === 'text/plain') {
+        const reader = new FileReader();
+        reader.readAsText(file, 'UTF-8');
+
+        reader.onload = event => {
+          const content = event.target?.result as string;
+          resolve({
+            blob: file,
+            content: content
+          });
+        };
+        reader.onerror = reject;
+      } else if (file.type === 'application/pdf') {
+        const readPDFPage = async (doc: PDFDocumentProxy, pageNo: number) => {
+          const page = await doc.getPage(pageNo);
+          const tokenizedText = await page.getTextContent();
+          return tokenizedText.items.map(token => (token as TextItem).str).join('');
+        };
+        let reader = new FileReader();
+        reader.readAsArrayBuffer(file);
+
+        reader.onload = async (event) => {
+          try {
+            const doc = await getDocument(event.target?.result!).promise;
+            const pageTextPromises = [];
+            for (let pageNo = 1; pageNo <= doc.numPages; pageNo++) {
+              pageTextPromises.push(readPDFPage(doc, pageNo));
+            }
+            const pageTexts = await Promise.all(pageTextPromises);
+            let content;
+            if (pageTexts.length === 1)
+              content = pageTexts[0];
+            else
+              content = pageTexts.map((p, i) => `Page ${i + 1}:\n${p}`).join('\n\n');
+            resolve({
+              blob: file,
+              content: content
+            });
+          } catch (err) {
+            reject(err);
+          }
+        };
+        reader.onerror = reject;
+      } else {
+        resolve({
+          blob: file
+        });
+      }
+    };
+    input.click();
+  });
+}

frontend/src/webWails.js

@@ -1,3 +1,5 @@
+import { webOpenOpenFileDialog } from './utils/web-file-operations'
+
 function defineRuntime(name, func) {
   window.runtime[name] = func
 }
@@ -107,37 +109,7 @@ if (!window.go) {
   defineApp('ListDirFiles', async () => {
     return []
   })
-  defineApp('OpenOpenFileDialog', async (filterPattern) => {
-    return new Promise((resolve, reject) => {
-      const input = document.createElement('input')
-      input.type = 'file'
-      input.accept = filterPattern
-        .replaceAll('*.txt', 'text/plain')
-        .replaceAll('*.', 'application/')
-        .replaceAll(';', ',')
-
-      input.onchange = e => {
-        const file = e.target?.files[0]
-        if (file.type === 'text/plain') {
-          const reader = new FileReader()
-          reader.readAsText(file, 'UTF-8')
-
-          reader.onload = readerEvent => {
-            const content = readerEvent.target?.result
-            resolve({
-              blob: file,
-              content: content
-            })
-          }
-        } else {
-          resolve({
-            blob: file
-          })
-        }
-      }
-      input.click()
-    })
-  })
+  defineApp('OpenOpenFileDialog', webOpenOpenFileDialog)
   defineApp('OpenSaveFileDialog', async (filterPattern, defaultFileName, savedContent) => {
     const saver = await import('file-saver')
     saver.saveAs(new Blob([savedContent], { type: 'text/plain;charset=utf-8' }), defaultFileName)

frontend/vite.config.ts

@@ -3,6 +3,7 @@ import { dependencies } from './package.json';
 import { defineConfig } from 'vite';
 import react from '@vitejs/plugin-react';
 import { visualizer } from 'rollup-plugin-visualizer';
+import topLevelAwait from 'vite-plugin-top-level-await';
 
 // dependencies that exist anywhere
 const vendor = [
@@ -35,12 +36,18 @@ function renderChunks(deps: Record<string, string>) {
 
 // https://vitejs.dev/config/
 export default defineConfig({
-  plugins: [react(),
+  plugins: [
+    react(),
     visualizer({
       template: 'treemap',
       gzipSize: true,
       brotliSize: true
-    })],
+    }),
+    topLevelAwait({
+      promiseExportName: '__tla',
+      promiseImportName: i => `__tla_${i}`
+    })
+  ],
   build: {
     chunkSizeWarningLimit: 3000,
     rollupOptions: {

manifest.json

@@ -1,5 +1,5 @@
 {
-  "version": "1.5.2",
+  "version": "1.5.3",
   "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）。"
@@ -15,6 +15,32 @@
     }
   ],
   "models": [
+    {
+      "name": "RWKV-5-World-0.1B-v1-20230803-ctx4096.pth",
+      "desc": {
+        "en": "RWKV-5 Global Languages 0.1B v1",
+        "zh": "RWKV-5 全球语言 0.1B v1",
+        "ja": "RWKV-5 グローバル言語 0.1B v1"
+      },
+      "size": 385598386,
+      "SHA256": "c844a3ee05bcb9065848cb05b10c48a3f381f5ac1953aad89e156ecdf31d7703",
+      "lastUpdated": "2023-08-03T15:18:46",
+      "url": "https://huggingface.co/BlinkDL/rwkv-5-world/blob/main/RWKV-5-World-0.1B-v1-20230803-ctx4096.pth?download=true",
+      "downloadUrl": "https://huggingface.co/BlinkDL/rwkv-5-world/resolve/main/RWKV-5-World-0.1B-v1-20230803-ctx4096.pth?download=true"
+    },
+    {
+      "name": "RWKV-5-World-0.4B-v2-20231113-ctx4096.pth",
+      "desc": {
+        "en": "RWKV-5 Global Languages 0.4B v2",
+        "zh": "RWKV-5 全球语言 0.4B v2",
+        "ja": "RWKV-5 グローバル言語 0.4B v2"
+      },
+      "size": 923523954,
+      "SHA256": "5a288c54c7f30b0e2d4af23991133fad2af2d5e59ec7ad850ffe78054a5e4f92",
+      "lastUpdated": "2023-11-14T01:23:49",
+      "url": "https://huggingface.co/BlinkDL/rwkv-5-world/blob/main/RWKV-5-World-0.4B-v2-20231113-ctx4096.pth?download=true",
+      "downloadUrl": "https://huggingface.co/BlinkDL/rwkv-5-world/resolve/main/RWKV-5-World-0.4B-v2-20231113-ctx4096.pth?download=true"
+    },
     {
       "name": "RWKV-5-World-1B5-v2-20231025-ctx4096.pth",
       "desc": {
@@ -28,6 +54,32 @@
       "url": "https://huggingface.co/BlinkDL/rwkv-5-world/blob/main/RWKV-5-World-1B5-v2-20231025-ctx4096.pth",
       "downloadUrl": "https://huggingface.co/BlinkDL/rwkv-5-world/resolve/main/RWKV-5-World-1B5-v2-20231025-ctx4096.pth"
     },
+    {
+      "name": "RWKV-5-World-3B-v2-20231113-ctx4096.pth",
+      "desc": {
+        "en": "RWKV-5 Global Languages 3B v2",
+        "zh": "RWKV-5 全球语言 3B v2",
+        "ja": "RWKV-5 グローバル言語 3B v2"
+      },
+      "size": 6126106674,
+      "SHA256": "a4bd430343c6fd138b85bbc68bb20262d3a2f053ea57dc4b41078269af68ff9c",
+      "lastUpdated": "2023-11-14T01:23:49",
+      "url": "https://huggingface.co/BlinkDL/rwkv-5-world/blob/main/RWKV-5-World-3B-v2-20231113-ctx4096.pth?download=true",
+      "downloadUrl": "https://huggingface.co/BlinkDL/rwkv-5-world/resolve/main/RWKV-5-World-3B-v2-20231113-ctx4096.pth?download=true"
+    },
+    {
+      "name": "RWKV-5-World-3B-v2-20231118-ctx16k.pth",
+      "desc": {
+        "en": "RWKV-5 Global Languages 3B v2 Ctx16k",
+        "zh": "RWKV-5 全球语言 3B v2 16k上下文",
+        "ja": "RWKV-5 グローバル言語 3B v2 16kコンテキスト"
+      },
+      "size": 6126106467,
+      "SHA256": "efa5178d1c824b94ef17c6c9a456674e5581a8be832becbda9aba4dc533f88c2",
+      "lastUpdated": "2023-11-19T04:21:04",
+      "url": "https://huggingface.co/BlinkDL/rwkv-5-world/blob/main/RWKV-5-World-3B-v2-20231118-ctx16k.pth?download=true",
+      "downloadUrl": "https://huggingface.co/BlinkDL/rwkv-5-world/resolve/main/RWKV-5-World-3B-v2-20231118-ctx16k.pth?download=true"
+    },
     {
       "name": "RWKV-4-World-CHNtuned-0.1B-v1-20230617-ctx4096.pth",
       "desc": {