💄 style: patching models info (#2269)

lobehub · Apr 29, 2024 · 03bcb06 · 03bcb06
1 parent 637489b
commit 03bcb06
Show file tree

Hide file tree

Showing 16 changed files with 64 additions and 43 deletions.
diff --git a/src/config/modelProviders/anthropic.ts b/src/config/modelProviders/anthropic.ts
@@ -1,6 +1,6 @@
 import { ModelProviderCard } from '@/types/llm';
 
-// ref https://docs.anthropic.com/claude/docs/models-overview
+// ref https://docs.anthropic.com/claude/docs/models-overview#model-comparison
 const Anthropic: ModelProviderCard = {
   chatModels: [
     {

diff --git a/src/config/modelProviders/azure.ts b/src/config/modelProviders/azure.ts
@@ -1,5 +1,6 @@
 import { ModelProviderCard } from '@/types/llm';
 
+// ref https://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/models
 const Azure: ModelProviderCard = {
   chatModels: [
     {

diff --git a/src/config/modelProviders/bedrock.ts b/src/config/modelProviders/bedrock.ts
@@ -56,14 +56,14 @@ const Bedrock: ModelProviderCard = {
       displayName: 'Llama 2 Chat 13B',
       enabled: true,
       id: 'meta.llama2-13b-chat-v1',
-      tokens: 4000,
+      tokens: 4096,
     },
     {
       description: 'Llama 2 Chat 70B v1，上下文大小为 4k，Llama 2 模型的对话用例优化变体。',
       displayName: 'Llama 2 Chat 70B',
       enabled: true,
       id: 'meta.llama2-70b-chat-v1',
-      tokens: 4000,
+      tokens: 4096,
     },
   ],
   id: 'bedrock',

diff --git a/src/config/modelProviders/google.ts b/src/config/modelProviders/google.ts
@@ -1,35 +1,37 @@
 import { ModelProviderCard } from '@/types/llm';
 
 // ref https://ai.google.dev/models/gemini
+// api https://ai.google.dev/api/rest/v1beta/models/list
 const Google: ModelProviderCard = {
   chatModels: [
     {
       description: 'A legacy text-only model optimized for chat conversations',
       displayName: 'PaLM 2 Chat (Legacy)',
       id: 'chat-bison-001',
       maxOutput: 1024,
+      // tokens: 4096 + 1024, // none tokens test
     },
     {
       description: 'A legacy model that understands text and generates text as an output',
       displayName: 'PaLM 2 (Legacy)',
       id: 'text-bison-001',
       maxOutput: 1024,
-      tokens: 9220,
+      tokens: 8196 + 1024,
     },
     {
       description: 'The best model for scaling across a wide range of tasks',
       displayName: 'Gemini 1.0 Pro',
       enabled: true,
       id: 'gemini-pro',
       maxOutput: 2048,
-      tokens: 32_768,
+      tokens: 30_720 + 2048,
     },
     {
       description: 'The best image understanding model to handle a broad range of applications',
       displayName: 'Gemini 1.0 Pro Vision',
       id: 'gemini-1.0-pro-vision-latest',
       maxOutput: 4096,
-      tokens: 16_384,
+      tokens: 12_288 + 4096,
       vision: true,
     },
     {
@@ -38,7 +40,7 @@ const Google: ModelProviderCard = {
       enabled: true,
       id: 'gemini-pro-vision',
       maxOutput: 4096,
-      tokens: 16_384,
+      tokens: 12_288 + 4096,
       vision: true,
     },
     {
@@ -47,23 +49,23 @@ const Google: ModelProviderCard = {
       displayName: 'Gemini 1.0 Pro 001 (Tuning)',
       id: 'gemini-1.0-pro-001',
       maxOutput: 2048,
-      tokens: 32_768,
+      tokens: 30_720 + 2048,
     },
     {
       description:
         'The best model for scaling across a wide range of tasks. This is the latest model.',
       displayName: 'Gemini 1.0 Pro Latest',
       id: 'gemini-1.0-pro-latest',
       maxOutput: 2048,
-      tokens: 32_768,
+      tokens: 30_720 + 2048,
     },
     {
       description: 'Mid-size multimodal model that supports up to 1 million tokens',
       displayName: 'Gemini 1.5 Pro',
       enabled: true,
       id: 'gemini-1.5-pro-latest',
       maxOutput: 8192,
-      tokens: 1_056_768,
+      tokens: 1_048_576 + 8192,
       vision: true,
     },
     {

diff --git a/src/config/modelProviders/groq.ts b/src/config/modelProviders/groq.ts
@@ -29,7 +29,6 @@ const Groq: ModelProviderCard = {
     },
     {
       displayName: 'LLaMA2-70b-chat',
-      enabled: true,
       id: 'llama2-70b-4096',
       tokens: 4096,
     },

diff --git a/src/config/modelProviders/minimax.ts b/src/config/modelProviders/minimax.ts
@@ -1,7 +1,22 @@
 import { ModelProviderCard } from '@/types/llm';
 
+// ref https://www.minimaxi.com/document/guides/chat-model/pro/api
 const Minimax: ModelProviderCard = {
   chatModels: [
+    {
+      description: '复杂场景，例如应用题计算、科学计算等场景',
+      displayName: 'abab6.5',
+      enabled: true,
+      id: 'abab6.5-chat',
+      tokens: 8192,
+    },
+    {
+      description: '通用场景',
+      displayName: 'abab6.5s',
+      enabled: true,
+      id: 'abab6.5s-chat',
+      tokens: 245_760,
+    },
     {
       description: '更复杂的格式化文本生成',
       displayName: 'abab6',

diff --git a/src/config/modelProviders/mistral.ts b/src/config/modelProviders/mistral.ts
@@ -1,6 +1,6 @@
 import { ModelProviderCard } from '@/types/llm';
 
-// ref https://docs.mistral.ai/platform/pricing/#chat-completions-api
+// ref https://docs.mistral.ai/getting-started/models/
 const Mistral: ModelProviderCard = {
   chatModels: [
     {

diff --git a/src/config/modelProviders/moonshot.ts b/src/config/modelProviders/moonshot.ts
@@ -1,5 +1,6 @@
 import { ModelProviderCard } from '@/types/llm';
 
+// ref https://platform.moonshot.cn/docs/intro#模型列表
 const Moonshot: ModelProviderCard = {
   chatModels: [
     {

diff --git a/src/config/modelProviders/ollama.ts b/src/config/modelProviders/ollama.ts
@@ -6,7 +6,7 @@ const Ollama: ModelProviderCard = {
       displayName: 'Llama3 8B',
       enabled: true,
       id: 'llama3',
-      tokens: 8000,
+      tokens: 8000, // https://huggingface.co/blog/zh/llama3#llama-3-的新进展
     },
     {
       displayName: 'Llama3 70B',
@@ -17,48 +17,48 @@ const Ollama: ModelProviderCard = {
       displayName: 'Command R 35B',
       enabled: true,
       id: 'command-r',
-      tokens: 128_000,
+      tokens: 131_072, // https://huggingface.co/CohereForAI/c4ai-command-r-v01/blob/main/config.json
     },
     {
       displayName: 'Command R+ 104B (Q2_K)',
       id: 'command-r-plus:104b-q2_K',
-      tokens: 128_000,
+      tokens: 131_072, // https://huggingface.co/CohereForAI/c4ai-command-r-plus/blob/main/config.json
     },
     {
       displayName: 'Gemma 7B',
       enabled: true,
       id: 'gemma',
-      tokens: 4000,
+      tokens: 8192, // https://huggingface.co/google/gemma-7b-it/discussions/73#65e9678c0cda621164a95bad
     },
     {
       displayName: 'Gemma 2B',
       id: 'gemma:2b',
-      tokens: 4000,
+      tokens: 8192,
     },
     {
       displayName: 'Llama2 Chat 13B',
       id: 'llama2:13b',
-      tokens: 4000,
+      tokens: 4096, // https://llama.meta.com/llama2/
     },
     {
       displayName: 'Llama2 Chat 7B',
       id: 'llama2',
-      tokens: 4000,
+      tokens: 4096,
     },
     {
       displayName: 'Llama2 Chat 70B',
       id: 'llama2:70b',
-      tokens: 4000,
+      tokens: 4096,
     },
     {
       displayName: 'Llama2 CN 13B',
       id: 'llama2-chinese:13b',
-      tokens: 4000,
+      tokens: 4096,
     },
     {
       displayName: 'Llama2 CN 7B',
       id: 'llama2-chinese',
-      tokens: 4000,
+      tokens: 4096,
     },
     {
       displayName: 'WizardLM 2 7B',
@@ -74,45 +74,45 @@ const Ollama: ModelProviderCard = {
     {
       displayName: 'Code Llama 7B',
       id: 'codellama',
-      tokens: 16_000,
+      tokens: 16_384, // https://huggingface.co/codellama/CodeLlama-7b-hf/blob/main/config.json
     },
     {
       displayName: 'Code Llama 34B',
       id: 'codellama:34b',
-      tokens: 16_000,
+      tokens: 16_384,
     },
     {
       displayName: 'Code Llama 70B',
       id: 'codellama:70b',
-      tokens: 16_000,
+      tokens: 16_384,
     },
     {
       displayName: 'Code Llama 7B (Python)',
       id: 'codellama:python',
-      tokens: 16_000,
+      tokens: 16_384,
     },
     {
       displayName: 'Phi3-Instruct 3.8B',
       enabled: true,
       id: 'phi3:instruct',
-      tokens: 128_000,
+      tokens: 131_072, // https://huggingface.co/microsoft/Phi-3-mini-128k-instruct/blob/main/config.json
     },
     {
       displayName: 'Mistral',
       enabled: true,
       id: 'mistral',
-      tokens: 4800,
+      tokens: 32_768, // https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2/blob/main/config.json
     },
     {
       displayName: 'Mixtral 8x7B',
       enabled: true,
       id: 'mixtral',
-      tokens: 32_000,
+      tokens: 32_768,
     },
     {
       displayName: 'Mixtral 8x22B',
       id: 'mixtral:8x22b',
-      tokens: 64_000,
+      tokens: 65_536, // https://huggingface.co/mistralai/Mixtral-8x22B-v0.1/blob/main/config.json
     },
     {
       displayName: 'Qwen Chat 4B',
@@ -138,19 +138,19 @@ const Ollama: ModelProviderCard = {
     {
       displayName: 'LLaVA 7B',
       id: 'llava',
-      tokens: 4000,
+      tokens: 4096, // https://huggingface.co/llava-hf/llava-1.5-7b-hf/blob/main/config.json
       vision: true,
     },
     {
       displayName: 'LLaVA 13B',
       id: 'llava:13b',
-      tokens: 4000,
+      tokens: 4096,
       vision: true,
     },
     {
       displayName: 'LLaVA 34B',
       id: 'llava:34b',
-      tokens: 4000,
+      tokens: 4096,
       vision: true,
     },
   ],

diff --git a/src/config/modelProviders/openai.ts b/src/config/modelProviders/openai.ts
@@ -1,6 +1,6 @@
 import { ModelProviderCard } from '@/types/llm';
 
-// refs to: https://platform.openai.com/docs/models/gpt-4-turbo-and-gpt-4
+// ref https://platform.openai.com/docs/models
 const OpenAI: ModelProviderCard = {
   chatModels: [
     {

diff --git a/src/config/modelProviders/togetherai.ts b/src/config/modelProviders/togetherai.ts
@@ -1,6 +1,6 @@
 import { ModelProviderCard } from '@/types/llm';
 
-// ref https://api.together.xyz/models
+// ref https://docs.together.ai/docs/inference-models
 const TogetherAI: ModelProviderCard = {
   chatModels: [
     {

diff --git a/src/config/modelProviders/zeroone.ts b/src/config/modelProviders/zeroone.ts
@@ -1,29 +1,30 @@
 import { ModelProviderCard } from '@/types/llm';
 
+// ref https://platform.lingyiwanwu.com/
 const ZeroOne: ModelProviderCard = {
   chatModels: [
     {
       description: '支持聊天、问答、对话、写作、翻译等功能。',
       displayName: 'YI 34B Chat',
       enabled: true,
       id: 'yi-34b-chat-0205',
-      tokens: 4000,
+      tokens: 4096, // https://huggingface.co/01-ai/Yi-34B-Chat/blob/main/config.json
     },
     {
       description:
         '支持通用图片问答、图表理解、OCR、视觉推理，能处理高分辨率（1024*1024）的图像，能在复杂视觉任务上提供优秀性能，同时支持多种语言。',
       displayName: 'YI Vision Plus',
       enabled: true,
       id: 'yi-vl-plus',
-      tokens: 4000,
+      tokens: 4096,
       vision: true,
     },
     {
       description: '增强了问答对话交互和深度内容创作能力。文档问答和构建知识库小能手。',
       displayName: 'YI 34B Chat 200k',
       enabled: true,
       id: 'yi-34b-chat-200k',
-      tokens: 200_000,
+      tokens: 200_000, // https://huggingface.co/01-ai/Yi-34B-200K/blob/main/config.json
     },
   ],
   id: 'zeroone',

diff --git a/src/config/modelProviders/zhipu.ts b/src/config/modelProviders/zhipu.ts
@@ -1,8 +1,10 @@
 import { ModelProviderCard } from '@/types/llm';
 
 // TODO: 等待 ZhiPu 修复 API 问题后开启 functionCall
-// refs: https://github.com/lobehub/lobe-chat/discussions/737#discussioncomment-8315815
 // 暂时不透出 GLM 系列的 function_call 功能
+// refs https://github.com/lobehub/lobe-chat/discussions/737#discussioncomment-8315815
+
+// ref https://open.bigmodel.cn/dev/howuse/model
 const ZhiPu: ModelProviderCard = {
   chatModels: [
     {
@@ -19,7 +21,7 @@ const ZhiPu: ModelProviderCard = {
       displayName: 'GLM-4 Vision',
       enabled: true,
       id: 'glm-4v',
-      tokens: 128_000,
+      tokens: 2000,
       vision: true,
     },
     {

diff --git a/src/migrations/FromV3ToV4/fixtures/ollama-output-v4.json b/src/migrations/FromV3ToV4/fixtures/ollama-output-v4.json
@@ -48,7 +48,7 @@
               "displayName": "LLaVA 7B",
               "enabled": true,
               "id": "llava",
-              "tokens": 4000,
+              "tokens": 4096,
               "vision": true
             }
           ]

diff --git a/src/store/user/slices/settings/actions/llm.test.ts b/src/store/user/slices/settings/actions/llm.test.ts
@@ -109,7 +109,7 @@ describe('LLMSettingsSliceAction', () => {
         displayName: 'LLaVA 7B',
         enabled: true,
         id: 'llava',
-        tokens: 4000,
+        tokens: 4096,
         vision: true,
       });
     });

diff --git a/src/types/llm.ts b/src/types/llm.ts
@@ -32,7 +32,7 @@ export interface ChatModelCard {
   legacy?: boolean;
   maxOutput?: number;
   /**
-   * the context window
+   * the context window (or input + output tokens limit)
    */
   tokens?: number;
   /**