Merge pull request #829 from janhq/update-model-hub

Update model hub

models/capybara-34b/model.json

@@ -2,22 +2,22 @@
     "source_url": "https://huggingface.co/TheBloke/Nous-Capybara-34B-GGUF/resolve/main/nous-capybara-34b.Q5_K_M.gguf",
     "id": "capybara-34b",
     "object": "model",
-    "name": "Capybara 200k 34B",
-    "version": 1.0,
+    "name": "Capybara 200k 34B Q5",
+    "version": "1.0",
     "description": "Nous Capybara 34B, a variant of the Yi-34B model, is the first Nous model with a 200K context length, trained for three epochs on the innovative Capybara dataset.",
     "format": "gguf",
     "settings": {
-      "ctx_len": 4096,
+      "ctx_len": 2048,
       "system_prompt": "",
-      "user_prompt": "USER: ",
-      "ai_prompt": "ASSISTANT: "
+      "user_prompt": "USER:\n",
+      "ai_prompt": "ASSISTANT:\n"
     },
     "parameters": {
-      "max_tokens": 4096
+      "max_tokens": 2048
     },
     "metadata": {
       "author": "NousResearch, The Bloke",
-      "tags": ["General", "Big Context Length"],
+      "tags": ["Recommended", "Medium", "Finetuned"],
       "size": 24320000000
     }
   }

models/deepseek-coder-1.3b/model.json

@@ -2,22 +2,22 @@
     "source_url": "https://huggingface.co/TheBloke/deepseek-coder-1.3b-base-GGUF/resolve/main/deepseek-coder-1.3b-base.Q4_K_M.gguf",
     "id": "deepseek-coder-1.3b",
     "object": "model",
-    "name": "Deepseek Coder 1.3B",
+    "name": "Deepseek Coder 1.3B Q4",
     "version": "1.0",
-    "description": "",
+    "description": "Deepseek Coder trained on 2T tokens (87% code, 13% English/Chinese), excelling in project-level code completion with advanced capabilities across multiple programming languages.",
     "format": "gguf",
     "settings": {
-      "ctx_len": 4096,
+      "ctx_len": 2048,
       "system_prompt": "",
       "user_prompt": "",
       "ai_prompt": ""
     },
     "parameters": {
-      "max_tokens": 4096
+      "max_tokens": 2048
     },
     "metadata": {
-      "author": "deepseek, The Bloke",
-      "tags": ["Code"],
+      "author": "Deepseek, The Bloke",
+      "tags": ["Tiny", "Foundational Model"],
       "size": 870000000
     }
   }

models/deepseek-coder-34b/model.json

@@ -0,0 +1,24 @@
+{
+    "source_url": "https://huggingface.co/TheBloke/deepseek-coder-33B-instruct-GGUF/resolve/main/deepseek-coder-33b-instruct.Q5_K_M.gguf",
+    "id": "deepseek-coder-34b",
+    "object": "model",
+    "name": "Deepseek Coder 33B Q5",
+    "version": "1.0",
+    "description": "Deepseek Coder trained on 2T tokens (87% code, 13% English/Chinese), excelling in project-level code completion with advanced capabilities across multiple programming languages.",
+    "format": "gguf",
+    "settings": {
+      "ctx_len": 2048,
+      "system_prompt": "",
+      "user_prompt": "### Instruction:\n",
+      "ai_prompt": "### Response:\n"
+    },
+    "parameters": {
+      "max_tokens": 2048
+    },
+    "metadata": {
+      "author": "Deepseek, The Bloke",
+      "tags": ["Recommended", "Medium", "Foundational Model"],
+      "size": 26040000000
+    }
+  }
+

models/llama2-chat-70b-q4/model.json

@@ -0,0 +1,24 @@
+{
+    "source_url": "https://huggingface.co/TheBloke/Llama-2-70B-Chat-GGUF/resolve/main/llama-2-70b-chat.Q4_K_M.gguf",
+    "id": "llama2-chat-70b-q4",
+    "object": "model",
+    "name": "Llama 2 Chat 70B Q4",
+    "version": "1.0",
+    "description": "This is a 4-bit quantized version of Meta AI's Llama 2 Chat 70b model.",
+    "format": "gguf",
+    "settings": {
+      "ctx_len": 2048,
+      "system_prompt": "[INST] <<SYS>>\n",
+      "user_prompt": "<</SYS>>\n",
+      "ai_prompt": "[/INST]"
+    },
+    "parameters": {
+      "max_tokens": 2048
+    },
+    "metadata": {
+      "author": "MetaAI, The Bloke",
+      "tags": ["Large", "Foundational Model"],
+      "size": 43920000000
+    }
+  }
+

models/llama2-chat-7b-q4/model.json

@@ -0,0 +1,24 @@
+{
+    "source_url": "https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGUF/resolve/main/llama-2-7b-chat.Q4_K_M.gguf",
+    "id": "llama2-chat-7b-q4",
+    "object": "model",
+    "name": "Llama 2 Chat 7B Q4",
+    "version": "1.0",
+    "description": "This is a 4-bit quantized iteration of Meta AI's Llama 2 Chat 7b model, specifically designed for a comprehensive understanding through training on extensive internet data.",
+    "format": "gguf",
+    "settings": {
+      "ctx_len": 2048,
+      "system_prompt": "[INST] <<SYS>>\n",
+      "user_prompt": "<</SYS>>\n",
+      "ai_prompt": "[/INST]"
+    },
+    "parameters": {
+      "max_tokens": 2048
+    },
+    "metadata": {
+      "author": "MetaAI, The Bloke",
+      "tags": ["Small", "Foundational Model"],
+      "size": 4080000000
+    }
+  }
+

models/llama2-chat-7b-q5/model.json

@@ -0,0 +1,24 @@
+{
+    "source_url": "https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGUF/resolve/main/llama-2-7b-chat.Q5_K_M.gguf",
+    "id": "llama2-chat-7b-q5",
+    "object": "model",
+    "name": "Llama 2 Chat 7B Q5",
+    "version": "1.0",
+    "description": "This is a 5-bit quantized iteration of Meta AI's Llama 2 Chat 7b model, specifically designed for a comprehensive understanding through training on extensive internet data.",
+    "format": "gguf",
+    "settings": {
+      "ctx_len": 2048,
+      "system_prompt": "[INST] <<SYS>>\n",
+      "user_prompt": "<</SYS>>\n",
+      "ai_prompt": "[/INST]"
+    },
+    "parameters": {
+      "max_tokens": 2048
+    },
+    "metadata": {
+      "author": "MetaAI, The Bloke",
+      "tags": ["Small", "Foundational Model"],
+      "size": 4780000000
+    }
+  }
+

models/lzlv-70b/model.json

@@ -2,22 +2,22 @@
     "source_url": "https://huggingface.co/TheBloke/lzlv_70B-GGUF/resolve/main/lzlv_70b_fp16_hf.Q5_K_M.gguf",
     "id": "lzlv-70b",
     "object": "model",
-    "name": "Lzlv 70B",
+    "name": "Lzlv 70B Q4",
     "version": "1.0",
     "description": "lzlv_70B is a sophisticated AI model designed for roleplaying and creative tasks. This merge aims to combine intelligence with creativity, seemingly outperforming its individual components in complex scenarios and creative outputs.",
     "format": "gguf",
     "settings": {
-      "ctx_len": 4096,
+      "ctx_len": 2048,
       "system_prompt": "",
-      "user_prompt": "USER: ",
-      "ai_prompt": "ASSISTANT: "
+      "user_prompt": "USER:\n",
+      "ai_prompt": "ASSISTANT:\n"
     },
     "parameters": {
-      "max_tokens": 4096
+      "max_tokens": 2048
     },
     "metadata": {
-      "author": "lizpreciatior, The Bloke",
-      "tags": ["General Use", "Role-playing"],
+      "author": "Lizpreciatior, The Bloke",
+      "tags": ["Recommended", "Large", "Finetuned"],
       "size": 48750000000
     }
   }

models/mistral-ins-7b-q4/model.json

@@ -0,0 +1,24 @@
+{
+    "source_url": "https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.1-GGUF/resolve/main/mistral-7b-instruct-v0.1.Q4_K_M.gguf",
+    "id": "mistral-ins-7b-q4",
+    "object": "model",
+    "name": "Mistral Instruct 7B Q4",
+    "version": "1.0",
+    "description": "This is a 4-bit quantized iteration of MistralAI's Mistral Instruct 7b model, specifically designed for a comprehensive understanding through training on extensive internet data.",
+    "format": "gguf",
+    "settings": {
+      "ctx_len": 2048,
+      "system_prompt": "",
+      "user_prompt": "<s>[INST]",
+      "ai_prompt": "[/INST]"
+    },
+    "parameters": {
+      "max_tokens": 2048
+    },
+    "metadata": {
+      "author": "MistralAI, The Bloke",
+      "tags": ["Small", "Foundational Model"],
+      "size": 4370000000
+    }
+  }
+

models/mistral-ins-7b-q5/model.json

@@ -0,0 +1,24 @@
+{
+    "source_url": "https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.1-GGUF/resolve/main/mistral-7b-instruct-v0.1.Q5_K_M.gguf",
+    "id": "mistral-ins-7b-q5",
+    "object": "model",
+    "name": "Mistral Instruct 7B Q5",
+    "version": "1.0",
+    "description": "This is a 5-bit quantized iteration of MistralAI's Mistral Instruct 7b model, specifically designed for a comprehensive understanding through training on extensive internet data.",
+    "format": "gguf",
+    "settings": {
+      "ctx_len": 2048,
+      "system_prompt": "",
+      "user_prompt": "<s>[INST]",
+      "ai_prompt": "[/INST]"
+    },
+    "parameters": {
+      "max_tokens": 2048
+    },
+    "metadata": {
+      "author": "MistralAI, The Bloke",
+      "tags": ["Small", "Foundational Model"],
+      "size": 5130000000
+    }
+  }
+

models/neural-chat-7b/model.json

@@ -2,22 +2,22 @@
     "source_url": "https://huggingface.co/TheBloke/neural-chat-7B-v3-1-GGUF/resolve/main/neural-chat-7b-v3-1.Q4_K_M.gguf",
     "id": "neural-chat-7b",
     "object": "model",
-    "name": "Neural Chat 7B",
+    "name": "Neural Chat 7B Q4",
     "version": "1.0",
     "description": "The Neural Chat 7B model, developed on the foundation of mistralai/Mistral-7B-v0.1, has been fine-tuned using the Open-Orca/SlimOrca dataset and aligned with the Direct Preference Optimization (DPO) algorithm. It has demonstrated substantial improvements in various AI tasks and performance well on the open_llm_leaderboard.",
     "format": "gguf",
     "settings": {
-      "ctx_len": 4096,
-      "system_prompt": "### System: ",
-      "user_prompt": "### User: ",
-      "ai_prompt": "### Assistant: "
+      "ctx_len": 2048,
+      "system_prompt": "### System:\n",
+      "user_prompt": "### User:\n",
+      "ai_prompt": "### Assistant:\n"
     },
     "parameters": {
-      "max_tokens": 4096
+      "max_tokens": 2048
     },
     "metadata": {
       "author": "Intel, The Bloke",
-      "tags": ["General Use", "Role-playing", "Big Context Length"],
+      "tags": ["Recommended", "Small", "Finetuned"],
       "size": 4370000000
     }
   }