xinference 1.2.2__py3-none-any.whl → 1.3.0.post1__py3-none-any.whl

xinference/model/llm/llm_family.json

@@ -6772,6 +6772,151 @@
     "stop_token_ids": [],
     "stop": []
   },
+  {
+    "version": 1,
+    "context_length": 16384,
+    "model_name": "InternVL2.5",
+    "model_lang": [
+        "en",
+        "zh"
+    ],
+    "model_ability": [
+        "chat",
+        "vision"
+    ],
+    "model_description": "InternVL 2.5 is an open-source multimodal large language model (MLLM) to bridge the capability gap between open-source and proprietary commercial models in multimodal understanding. ",
+    "model_specs": [
+      {
+          "model_format": "pytorch",
+          "model_size_in_billions": 1,
+          "quantizations": [
+            "4-bit",
+            "8-bit",
+            "none"
+          ],
+          "model_id": "OpenGVLab/InternVL2_5-1B"
+        },
+        {
+          "model_format": "awq",
+          "model_size_in_billions": 1,
+          "quantizations": [
+            "Int4"
+          ],
+          "model_id": "OpenGVLab/InternVL2_5-1B-AWQ"
+        },
+        {
+          "model_format": "pytorch",
+          "model_size_in_billions": 2,
+          "quantizations": [
+            "4-bit",
+            "8-bit",
+            "none"
+          ],
+          "model_id": "OpenGVLab/InternVL2_5-2B"
+        },
+        {
+          "model_format": "awq",
+          "model_size_in_billions": 2,
+          "quantizations": [
+            "Int4"
+          ],
+          "model_id": "OpenGVLab/InternVL2_5-2B-AWQ"
+        },
+        {
+          "model_format": "pytorch",
+          "model_size_in_billions": 4,
+          "quantizations": [
+            "4-bit",
+            "8-bit",
+            "none"
+          ],
+          "model_id": "OpenGVLab/InternVL2_5-4B"
+        },
+        {
+          "model_format": "awq",
+          "model_size_in_billions": 4,
+          "quantizations": [
+            "Int4"
+          ],
+          "model_id": "OpenGVLab/InternVL2_5-4B-AWQ"
+        },
+        {
+          "model_format": "pytorch",
+          "model_size_in_billions": 8,
+          "quantizations": [
+            "4-bit",
+            "8-bit",
+            "none"
+          ],
+          "model_id": "OpenGVLab/InternVL2_5-8B"
+        },
+        {
+          "model_format": "awq",
+          "model_size_in_billions": 8,
+          "quantizations": [
+            "Int4"
+          ],
+          "model_id": "OpenGVLab/InternVL2_5-8B-AWQ"
+        },
+        {
+          "model_format": "pytorch",
+          "model_size_in_billions": 26,
+          "quantizations": [
+            "4-bit",
+            "8-bit",
+            "none"
+          ],
+          "model_id": "OpenGVLab/InternVL2_5-26B"
+        },
+        {
+          "model_format": "awq",
+          "model_size_in_billions": 26,
+          "quantizations": [
+            "Int4"
+          ],
+          "model_id": "OpenGVLab/InternVL2_5-26B-AWQ"
+        },
+        {
+          "model_format": "pytorch",
+          "model_size_in_billions": 38,
+          "quantizations": [
+            "4-bit",
+            "8-bit",
+            "none"
+          ],
+          "model_id": "OpenGVLab/InternVL2_5-38B"
+        },
+        {
+          "model_format": "awq",
+          "model_size_in_billions": 38,
+          "quantizations": [
+            "Int4"
+          ],
+          "model_id": "OpenGVLab/InternVL2_5-38B-AWQ"
+        },
+        {
+          "model_format": "pytorch",
+          "model_size_in_billions": 78,
+          "quantizations": [
+            "4-bit",
+            "8-bit",
+            "none"
+          ],
+          "model_id": "OpenGVLab/InternVL2_5-78B"
+        },
+        {
+          "model_format": "awq",
+          "model_size_in_billions": 78,
+          "quantizations": [
+            "Int4"
+          ],
+          "model_id": "OpenGVLab/InternVL2_5-78B-AWQ"
+        }
+    ],
+    "chat_template": "{% for message in messages %}{% if loop.first and messages[0]['role'] != 'system' %}{{ '<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n' }}{% endif %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}",
+    "stop_token_ids": [],
+    "stop": []
+  },
   {
     "version": 1,
     "context_length": 8192,
@@ -7472,6 +7617,370 @@
       "<｜end▁of▁sentence｜>"
     ]
   },
+  {
+    "version": 1,
+    "context_length": 163840,
+    "model_name": "deepseek-v3",
+    "model_lang": [
+      "en",
+      "zh"
+    ],
+    "model_ability": [
+      "chat"
+    ],
+    "model_description": "DeepSeek-V3, a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token. ",
+    "model_specs": [
+      {
+        "model_format": "pytorch",
+        "model_size_in_billions": 671,
+        "quantizations": [
+          "4-bit",
+          "8-bit",
+          "none"
+        ],
+        "model_id": "deepseek-ai/DeepSeek-V3",
+        "model_revision": "1d044fd82b15f1cedb197a288e50cc96a2c27205"
+      },
+      {
+        "model_format": "awq",
+        "model_size_in_billions": 671,
+        "quantizations": [
+          "Int4"
+        ],
+        "model_id": "cognitivecomputations/DeepSeek-V3-AWQ"
+      },
+      {
+        "model_format": "ggufv2",
+        "model_size_in_billions": 671,
+        "quantizations": [
+          "Q2_K_L",
+          "Q2_K_XS",
+          "Q3_K_M",
+          "Q4_K_M",
+          "Q5_K_M",
+          "Q6_K",
+          "Q8_0"
+        ],
+        "model_id": "unsloth/DeepSeek-V3-GGUF",
+        "model_file_name_template": "DeepSeek-V3-{quantization}/DeepSeek-V3-{quantization}.gguf",
+        "model_file_name_split_template": "DeepSeek-V3-{quantization}/DeepSeek-V3-{quantization}-{part}.gguf",
+        "quantization_parts": {
+          "Q2_K_L": [
+            "00001-of-00005",
+            "00002-of-00005",
+            "00003-of-00005",
+            "00004-of-00005",
+            "00005-of-00005"
+          ],
+          "Q2_K_XS": [
+            "00001-of-00005",
+            "00002-of-00005",
+            "00003-of-00005",
+            "00004-of-00005",
+            "00005-of-00005"
+          ],
+          "Q3_K_M": [
+            "00001-of-00007",
+            "00002-of-00007",
+            "00003-of-00007",
+            "00004-of-00007",
+            "00005-of-00007",
+            "00006-of-00007",
+            "00007-of-00007"
+          ],
+          "Q4_K_M": [
+            "00001-of-00009",
+            "00002-of-00009",
+            "00003-of-00009",
+            "00004-of-00009",
+            "00005-of-00009",
+            "00006-of-00009",
+            "00007-of-00009",
+            "00008-of-00009",
+            "00009-of-00009"
+          ],
+          "Q5_K_M": [
+            "00001-of-00010",
+            "00002-of-00010",
+            "00003-of-00010",
+            "00004-of-00010",
+            "00005-of-00010",
+            "00006-of-00010",
+            "00007-of-00010",
+            "00008-of-00010",
+            "00009-of-00010",
+            "00010-of-00010"
+          ],
+          "Q6_K": [
+            "00001-of-00012",
+            "00002-of-00012",
+            "00003-of-00012",
+            "00004-of-00012",
+            "00005-of-00012",
+            "00006-of-00012",
+            "00007-of-00012",
+            "00008-of-00012",
+            "00009-of-00012",
+            "00010-of-00012",
+            "00011-of-00012",
+            "00012-of-00012"
+          ],
+          "Q8_0": [
+            "00001-of-00016",
+            "00002-of-00016",
+            "00003-of-00016",
+            "00004-of-00016",
+            "00005-of-00016",
+            "00006-of-00016",
+            "00007-of-00016",
+            "00008-of-00016",
+            "00009-of-00016",
+            "00010-of-00016",
+            "00011-of-00016",
+            "00012-of-00016",
+            "00013-of-00016",
+            "00014-of-00016",
+            "00015-of-00016",
+            "00016-of-00016"
+          ]
+        }
+      },
+      {
+        "model_format": "mlx",
+        "model_size_in_billions": 671,
+        "quantizations": [
+          "3bit",
+          "4bit"
+        ],
+        "model_id": "mlx-community/DeepSeek-V3-{quantization}"
+      }
+    ],
+    "chat_template": "{% if not add_generation_prompt is defined %}{% set add_generation_prompt = false %}{% endif %}{% set ns = namespace(is_first=false, is_tool=false, is_output_first=true, system_prompt='', is_first_sp=true) %}{%- for message in messages %}{%- if message['role'] == 'system' %}{%- if ns.is_first_sp %}{% set ns.system_prompt = ns.system_prompt + message['content'] %}{% set ns.is_first_sp = false %}{%- else %}{% set ns.system_prompt = ns.system_prompt + '\\n\\n' + message['content'] %}{%- endif %}{%- endif %}{%- endfor %}{{bos_token}}{{ns.system_prompt}}{%- for message in messages %}{%- if message['role'] == 'user' %}{%- set ns.is_tool = false -%}{{'<｜User｜>' + message['content']}}{%- endif %}{%- if message['role'] == 'assistant' and message['content'] is none %}{%- set ns.is_tool = false -%}{%- for tool in message['tool_calls']%}{%- if not ns.is_first %}{{'<｜Assistant｜><｜tool▁calls▁begin｜><｜tool▁call▁begin｜>' + tool['type'] + '<｜tool▁sep｜>' + tool['function']['name'] + '\\n' + '```json' + '\\n' + tool['function']['arguments'] + '\\n' + '```' + '<｜tool▁call▁end｜>'}}{%- set ns.is_first = true -%}{%- else %}{{'\\n' + '<｜tool▁call▁begin｜>' + tool['type'] + '<｜tool▁sep｜>' + tool['function']['name'] + '\\n' + '```json' + '\\n' + tool['function']['arguments'] + '\\n' + '```' + '<｜tool▁call▁end｜>'}}{{'<｜tool▁calls▁end｜><｜end▁of▁sentence｜>'}}{%- endif %}{%- endfor %}{%- endif %}{%- if message['role'] == 'assistant' and message['content'] is not none %}{%- if ns.is_tool %}{{'<｜tool▁outputs▁end｜>' + message['content'] + '<｜end▁of▁sentence｜>'}}{%- set ns.is_tool = false -%}{%- else %}{{'<｜Assistant｜>' + message['content'] + '<｜end▁of▁sentence｜>'}}{%- endif %}{%- endif %}{%- if message['role'] == 'tool' %}{%- set ns.is_tool = true -%}{%- if ns.is_output_first %}{{'<｜tool▁outputs▁begin｜><｜tool▁output▁begin｜>' + message['content'] + '<｜tool▁output▁end｜>'}}{%- set ns.is_output_first = false %}{%- else %}{{'\\n<｜tool▁output▁begin｜>' + message['content'] + '<｜tool▁output▁end｜>'}}{%- endif %}{%- endif %}{%- endfor -%}{% if ns.is_tool %}{{'<｜tool▁outputs▁end｜>'}}{% endif %}{% if add_generation_prompt and not ns.is_tool %}{{'<｜Assistant｜>'}}{% endif %}",
+    "stop_token_ids": [
+      1
+    ],
+    "stop": [
+      "<｜end▁of▁sentence｜>"
+    ]
+  },
+  {
+    "version": 1,
+    "context_length": 163840,
+    "model_name": "deepseek-r1",
+    "model_lang": [
+      "en",
+      "zh"
+    ],
+    "model_ability": [
+      "chat",
+      "reasoning"
+    ],
+    "model_description": "DeepSeek-R1, which incorporates cold-start data before RL. DeepSeek-R1 achieves performance comparable to OpenAI-o1 across math, code, and reasoning tasks.",
+    "model_specs": [
+      {
+        "model_format": "pytorch",
+        "model_size_in_billions": 671,
+        "quantizations": [
+          "4-bit",
+          "8-bit",
+          "none"
+        ],
+        "model_id": "deepseek-ai/DeepSeek-R1",
+        "model_revision": "8a58a132790c9935686eb97f042afa8013451c9f"
+      },
+      {
+        "model_format": "awq",
+        "model_size_in_billions": 671,
+        "quantizations": [
+          "Int4"
+        ],
+        "model_id": "cognitivecomputations/DeepSeek-R1-AWQ"
+      },
+      {
+        "model_format": "ggufv2",
+        "model_size_in_billions": 671,
+        "quantizations": [
+          "UD-IQ1_S",
+          "UD-IQ1_M",
+          "UD-IQ2_XXS",
+          "UD-Q2_K_XL",
+          "Q2_K",
+          "Q2_K_L",
+          "Q2_K_XS",
+          "Q3_K_M",
+          "Q4_K_M",
+          "Q5_K_M",
+          "Q6_K",
+          "Q8_0",
+          "BF16"
+        ],
+        "model_id": "unsloth/DeepSeek-R1-GGUF",
+        "model_file_name_template": "DeepSeek-R1-{quantization}/DeepSeek-R1-{quantization}.gguf",
+        "model_file_name_split_template": "DeepSeek-R1-{quantization}/DeepSeek-R1-{quantization}-{part}.gguf",
+        "quantization_parts": {
+          "UD-IQ1_S": [
+            "00001-of-00003",
+            "00002-of-00003",
+            "00003-of-00003"
+          ],
+          "UD-IQ1_M": [
+            "00001-of-00004",
+            "00002-of-00004",
+            "00003-of-00004",
+            "00004-of-00004"
+          ],
+          "UD-IQ2_XXS": [
+            "00001-of-00004",
+            "00002-of-00004",
+            "00003-of-00004",
+            "00004-of-00004"
+          ],
+          "UD-Q2_K_XL": [
+            "00001-of-00005",
+            "00002-of-00005",
+            "00003-of-00005",
+            "00004-of-00005",
+            "00005-of-00005"
+          ],
+          "Q2_K": [
+            "00001-of-00005",
+            "00002-of-00005",
+            "00003-of-00005",
+            "00004-of-00005",
+            "00005-of-00005"
+          ],
+          "Q2_K_L": [
+            "00001-of-00005",
+            "00002-of-00005",
+            "00003-of-00005",
+            "00004-of-00005",
+            "00005-of-00005"
+          ],
+          "Q2_K_XS": [
+            "00001-of-00005",
+            "00002-of-00005",
+            "00003-of-00005",
+            "00004-of-00005",
+            "00005-of-00005"
+          ],
+          "Q3_K_M": [
+            "00001-of-00007",
+            "00002-of-00007",
+            "00003-of-00007",
+            "00004-of-00007",
+            "00005-of-00007",
+            "00006-of-00007",
+            "00007-of-00007"
+          ],
+          "Q4_K_M": [
+            "00001-of-00009",
+            "00002-of-00009",
+            "00003-of-00009",
+            "00004-of-00009",
+            "00005-of-00009",
+            "00006-of-00009",
+            "00007-of-00009",
+            "00008-of-00009",
+            "00009-of-00009"
+          ],
+          "Q5_K_M": [
+            "00001-of-00010",
+            "00002-of-00010",
+            "00003-of-00010",
+            "00004-of-00010",
+            "00005-of-00010",
+            "00006-of-00010",
+            "00007-of-00010",
+            "00008-of-00010",
+            "00009-of-00010",
+            "00010-of-00010"
+          ],
+          "Q6_K": [
+            "00001-of-00012",
+            "00002-of-00012",
+            "00003-of-00012",
+            "00004-of-00012",
+            "00005-of-00012",
+            "00006-of-00012",
+            "00007-of-00012",
+            "00008-of-00012",
+            "00009-of-00012",
+            "00010-of-00012",
+            "00011-of-00012",
+            "00012-of-00012"
+          ],
+          "Q8_0": [
+            "00001-of-00015",
+            "00002-of-00015",
+            "00003-of-00015",
+            "00004-of-00015",
+            "00005-of-00015",
+            "00006-of-00015",
+            "00007-of-00015",
+            "00008-of-00015",
+            "00009-of-00015",
+            "00010-of-00015",
+            "00011-of-00015",
+            "00012-of-00015",
+            "00013-of-00015",
+            "00014-of-00015",
+            "00015-of-00015"
+          ],
+          "BF16": [
+            "00001-of-00030",
+            "00002-of-00030",
+            "00003-of-00030",
+            "00004-of-00030",
+            "00005-of-00030",
+            "00006-of-00030",
+            "00007-of-00030",
+            "00008-of-00030",
+            "00009-of-00030",
+            "00010-of-00030",
+            "00011-of-00030",
+            "00012-of-00030",
+            "00013-of-00030",
+            "00014-of-00030",
+            "00015-of-00030",
+            "00016-of-00030",
+            "00017-of-00030",
+            "00018-of-00030",
+            "00019-of-00030",
+            "00020-of-00030",
+            "00021-of-00030",
+            "00022-of-00030",
+            "00023-of-00030",
+            "00024-of-00030",
+            "00025-of-00030",
+            "00026-of-00030",
+            "00027-of-00030",
+            "00028-of-00030",
+            "00029-of-00030",
+            "00030-of-00030"
+          ]
+        }
+      },
+      {
+        "model_format": "mlx",
+        "model_size_in_billions": 671,
+        "quantizations": [
+          "2bit",
+          "3bit",
+          "4bit"
+        ],
+        "model_id": "mlx-community/DeepSeek-R1-{quantization}"
+      }
+    ],
+    "chat_template": "{% if not add_generation_prompt is defined %}{% set add_generation_prompt = false %}{% endif %}{% set ns = namespace(is_first=false, is_tool=false, is_output_first=true, system_prompt='') %}{%- for message in messages %}{%- if message['role'] == 'system' %}{% set ns.system_prompt = message['content'] %}{%- endif %}{%- endfor %}{{bos_token}}{{ns.system_prompt}}{%- for message in messages %}{%- if message['role'] == 'user' %}{%- set ns.is_tool = false -%}{{'<｜User｜>' + message['content']}}{%- endif %}{%- if message['role'] == 'assistant' and message['content'] is none %}{%- set ns.is_tool = false -%}{%- for tool in message['tool_calls']%}{%- if not ns.is_first %}{{'<｜Assistant｜><｜tool▁calls▁begin｜><｜tool▁call▁begin｜>' + tool['type'] + '<｜tool▁sep｜>' + tool['function']['name'] + '\\n' + '```json' + '\\n' + tool['function']['arguments'] + '\\n' + '```' + '<｜tool▁call▁end｜>'}}{%- set ns.is_first = true -%}{%- else %}{{'\\n' + '<｜tool▁call▁begin｜>' + tool['type'] + '<｜tool▁sep｜>' + tool['function']['name'] + '\\n' + '```json' + '\\n' + tool['function']['arguments'] + '\\n' + '```' + '<｜tool▁call▁end｜>'}}{{'<｜tool▁calls▁end｜><｜end▁of▁sentence｜>'}}{%- endif %}{%- endfor %}{%- endif %}{%- if message['role'] == 'assistant' and message['content'] is not none %}{%- if ns.is_tool %}{{'<｜tool▁outputs▁end｜>' + message['content'] + '<｜end▁of▁sentence｜>'}}{%- set ns.is_tool = false -%}{%- else %}{% set content = message['content'] %}{% if '</think>' in content %}{% set content = content.split('</think>')[-1] %}{% endif %}{{'<｜Assistant｜>' + content + '<｜end▁of▁sentence｜>'}}{%- endif %}{%- endif %}{%- if message['role'] == 'tool' %}{%- set ns.is_tool = true -%}{%- if ns.is_output_first %}{{'<｜tool▁outputs▁begin｜><｜tool▁output▁begin｜>' + message['content'] + '<｜tool▁output▁end｜>'}}{%- set ns.is_output_first = false %}{%- else %}{{'\\n<｜tool▁output▁begin｜>' + message['content'] + '<｜tool▁output▁end｜>'}}{%- endif %}{%- endif %}{%- endfor -%}{% if ns.is_tool %}{{'<｜tool▁outputs▁end｜>'}}{% endif %}{% if add_generation_prompt and not ns.is_tool %}{{'<｜Assistant｜><think>\\n'}}{% endif %}",
+    "stop_token_ids": [
+      1
+    ],
+    "stop": [
+      "<｜end▁of▁sentence｜>"
+    ],
+    "reasoning_start_tag": "<think>",
+    "reasoning_end_tag": "</think>"
+  },
   {
     "version": 1,
     "context_length": 131072,
@@ -8810,7 +9319,8 @@
       "zh"
     ],
     "model_ability": [
-      "chat"
+      "chat",
+      "reasoning"
     ],
     "model_description": "deepseek-r1-distill-qwen is distilled from DeepSeek-R1 based on Qwen",
     "model_specs": [
@@ -9014,13 +9524,15 @@
         "model_id": "mlx-community/DeepSeek-R1-Distill-Qwen-32B-{quantization}"
       }
     ],
-    "chat_template": "{% if not add_generation_prompt is defined %}{% set add_generation_prompt = false %}{% endif %}{% set ns = namespace(is_first=false, is_tool=false, is_output_first=true, system_prompt='') %}{%- for message in messages %}{%- if message['role'] == 'system' %}{% set ns.system_prompt = message['content'] %}{%- endif %}{%- endfor %}{{bos_token}}{{ns.system_prompt}}{%- for message in messages %}{%- if message['role'] == 'user' %}{%- set ns.is_tool = false -%}{{'<｜User｜>' + message['content']}}{%- endif %}{%- if message['role'] == 'assistant' and message['content'] is none %}{%- set ns.is_tool = false -%}{%- for tool in message['tool_calls']%}{%- if not ns.is_first %}{{'<｜Assistant｜><｜tool▁calls▁begin｜><｜tool▁call▁begin｜>' + tool['type'] + '<｜tool▁sep｜>' + tool['function']['name'] + '\\n' + '```json' + '\\n' + tool['function']['arguments'] + '\\n' + '```' + '<｜tool▁call▁end｜>'}}{%- set ns.is_first = true -%}{%- else %}{{'\\n' + '<｜tool▁call▁begin｜>' + tool['type'] + '<｜tool▁sep｜>' + tool['function']['name'] + '\\n' + '```json' + '\\n' + tool['function']['arguments'] + '\\n' + '```' + '<｜tool▁call▁end｜>'}}{{'<｜tool▁calls▁end｜><｜end▁of▁sentence｜>'}}{%- endif %}{%- endfor %}{%- endif %}{%- if message['role'] == 'assistant' and message['content'] is not none %}{%- if ns.is_tool %}{{'<｜tool▁outputs▁end｜>' + message['content'] + '<｜end▁of▁sentence｜>'}}{%- set ns.is_tool = false -%}{%- else %}{% set content = message['content'] %}{% if '</think>' in content %}{% set content = content.split('</think>')[-1] %}{% endif %}{{'<｜Assistant｜>' + content + '<｜end▁of▁sentence｜>'}}{%- endif %}{%- endif %}{%- if message['role'] == 'tool' %}{%- set ns.is_tool = true -%}{%- if ns.is_output_first %}{{'<｜tool▁outputs▁begin｜><｜tool▁output▁begin｜>' + message['content'] + '<｜tool▁output▁end｜>'}}{%- set ns.is_output_first = false %}{%- else %}{{'\\n<｜tool▁output▁begin｜>' + message['content'] + '<｜tool▁output▁end｜>'}}{%- endif %}{%- endif %}{%- endfor -%}{% if ns.is_tool %}{{'<｜tool▁outputs▁end｜>'}}{% endif %}{% if add_generation_prompt and not ns.is_tool %}{{'<｜Assistant｜>'}}{% endif %}",
+    "chat_template": "{% if not add_generation_prompt is defined %}{% set add_generation_prompt = false %}{% endif %}{% set ns = namespace(is_first=false, is_tool=false, is_output_first=true, system_prompt='', is_first_sp=true) %}{%- for message in messages %}{%- if message['role'] == 'system' %}{%- if ns.is_first_sp %}{% set ns.system_prompt = ns.system_prompt + message['content'] %}{% set ns.is_first_sp = false %}{%- else %}{% set ns.system_prompt = ns.system_prompt + '\\n\\n' + message['content'] %}{%- endif %}{%- endif %}{%- endfor %}{{ bos_token }}{{ ns.system_prompt }}{%- for message in messages %}{%- if message['role'] == 'user' %}{%- set ns.is_tool = false -%}{{'<｜User｜>' + message['content']}}{%- endif %}{%- if message['role'] == 'assistant' and 'tool_calls' in message %}{%- set ns.is_tool = false -%}{%- for tool in message['tool_calls'] %}{%- if not ns.is_first %}{%- if message['content'] is none %}{{'<｜Assistant｜><｜tool▁calls▁begin｜><｜tool▁call▁begin｜>' + tool['type'] + '<｜tool▁sep｜>' + tool['function']['name'] + '\\n' + '```json' + '\\n' + tool['function']['arguments'] + '\\n' + '```' + '<｜tool▁call▁end｜>'}}{%- else %}{{'<｜Assistant｜>' + message['content'] + '<｜tool▁calls▁begin｜><｜tool▁call▁begin｜>' + tool['type'] + '<｜tool▁sep｜>' + tool['function']['name'] + '\\n' + '```json' + '\\n' + tool['function']['arguments'] + '\\n' + '```' + '<｜tool▁call▁end｜>'}}{%- endif %}{%- set ns.is_first = true -%}{%- else %}{{'\\n' + '<｜tool▁call▁begin｜>' + tool['type'] + '<｜tool▁sep｜>' + tool['function']['name'] + '\\n' + '```json' + '\\n' + tool['function']['arguments'] + '\\n' + '```' + '<｜tool▁call▁end｜>'}}{%- endif %}{%- endfor %}{{'<｜tool▁calls▁end｜><｜end▁of▁sentence｜>'}}{%- endif %}{%- if message['role'] == 'assistant' and 'tool_calls' not in message %}{%- if ns.is_tool %}{{'<｜tool▁outputs▁end｜>' + message['content'] + '<｜end▁of▁sentence｜>'}}{%- set ns.is_tool = false -%}{%- else %}{% set content = message['content'] %}{% if '</think>' in content %}{% set content = content.split('</think>')[-1] %}{% endif %}{{'<｜Assistant｜>' + content + '<｜end▁of▁sentence｜>'}}{%- endif %}{%- endif %}{%- if message['role'] == 'tool' %}{%- set ns.is_tool = true -%}{%- if ns.is_output_first %}{{'<｜tool▁outputs▁begin｜><｜tool▁output▁begin｜>' + message['content'] + '<｜tool▁output▁end｜>'}}{%- set ns.is_output_first = false %}{%- else %}{{'<｜tool▁output▁begin｜>' + message['content'] + '<｜tool▁output▁end｜>'}}{%- endif %}{%- endif %}{%- endfor -%}{% if ns.is_tool %}{{'<｜tool▁outputs▁end｜>'}}{% endif %}{% if add_generation_prompt and not ns.is_tool %}{{'<｜Assistant｜><think>\\n'}}{% endif %}",
     "stop_token_ids": [
       151643
     ],
     "stop": [
       "<｜end▁of▁sentence｜>"
-    ]
+    ],
+    "reasoning_start_tag": "<think>",
+    "reasoning_end_tag": "</think>"
   },
   {
     "version": 1,
@@ -9031,7 +9543,8 @@
       "zh"
     ],
     "model_ability": [
-      "chat"
+      "chat",
+      "reasoning"
     ],
     "model_description": "deepseek-r1-distill-llama is distilled from DeepSeek-R1 based on Llama",
     "model_specs": [
@@ -9159,13 +9672,15 @@
         "model_id": "mlx-community/DeepSeek-R1-Distill-Llama-70B-{quantization}"
       }
     ],
-    "chat_template": "{% if not add_generation_prompt is defined %}{% set add_generation_prompt = false %}{% endif %}{% set ns = namespace(is_first=false, is_tool=false, is_output_first=true, system_prompt='') %}{%- for message in messages %}{%- if message['role'] == 'system' %}{% set ns.system_prompt = message['content'] %}{%- endif %}{%- endfor %}{{bos_token}}{{ns.system_prompt}}{%- for message in messages %}{%- if message['role'] == 'user' %}{%- set ns.is_tool = false -%}{{'<｜User｜>' + message['content']}}{%- endif %}{%- if message['role'] == 'assistant' and message['content'] is none %}{%- set ns.is_tool = false -%}{%- for tool in message['tool_calls']%}{%- if not ns.is_first %}{{'<｜Assistant｜><｜tool▁calls▁begin｜><｜tool▁call▁begin｜>' + tool['type'] + '<｜tool▁sep｜>' + tool['function']['name'] + '\\n' + '```json' + '\\n' + tool['function']['arguments'] + '\\n' + '```' + '<｜tool▁call▁end｜>'}}{%- set ns.is_first = true -%}{%- else %}{{'\\n' + '<｜tool▁call▁begin｜>' + tool['type'] + '<｜tool▁sep｜>' + tool['function']['name'] + '\\n' + '```json' + '\\n' + tool['function']['arguments'] + '\\n' + '```' + '<｜tool▁call▁end｜>'}}{{'<｜tool▁calls▁end｜><｜end▁of▁sentence｜>'}}{%- endif %}{%- endfor %}{%- endif %}{%- if message['role'] == 'assistant' and message['content'] is not none %}{%- if ns.is_tool %}{{'<｜tool▁outputs▁end｜>' + message['content'] + '<｜end▁of▁sentence｜>'}}{%- set ns.is_tool = false -%}{%- else %}{% set content = message['content'] %}{% if '</think>' in content %}{% set content = content.split('</think>')[-1] %}{% endif %}{{'<｜Assistant｜>' + content + '<｜end▁of▁sentence｜>'}}{%- endif %}{%- endif %}{%- if message['role'] == 'tool' %}{%- set ns.is_tool = true -%}{%- if ns.is_output_first %}{{'<｜tool▁outputs▁begin｜><｜tool▁output▁begin｜>' + message['content'] + '<｜tool▁output▁end｜>'}}{%- set ns.is_output_first = false %}{%- else %}{{'\\n<｜tool▁output▁begin｜>' + message['content'] + '<｜tool▁output▁end｜>'}}{%- endif %}{%- endif %}{%- endfor -%}{% if ns.is_tool %}{{'<｜tool▁outputs▁end｜>'}}{% endif %}{% if add_generation_prompt and not ns.is_tool %}{{'<｜Assistant｜>'}}{% endif %}",
+    "chat_template": "{% if not add_generation_prompt is defined %}{% set add_generation_prompt = false %}{% endif %}{% set ns = namespace(is_first=false, is_tool=false, is_output_first=true, system_prompt='') %}{%- for message in messages %}{%- if message['role'] == 'system' %}{% set ns.system_prompt = message['content'] %}{%- endif %}{%- endfor %}{{bos_token}}{{ns.system_prompt}}{%- for message in messages %}{%- if message['role'] == 'user' %}{%- set ns.is_tool = false -%}{{'<｜User｜>' + message['content']}}{%- endif %}{%- if message['role'] == 'assistant' and message['content'] is none %}{%- set ns.is_tool = false -%}{%- for tool in message['tool_calls']%}{%- if not ns.is_first %}{{'<｜Assistant｜><｜tool▁calls▁begin｜><｜tool▁call▁begin｜>' + tool['type'] + '<｜tool▁sep｜>' + tool['function']['name'] + '\\n' + '```json' + '\\n' + tool['function']['arguments'] + '\\n' + '```' + '<｜tool▁call▁end｜>'}}{%- set ns.is_first = true -%}{%- else %}{{'\\n' + '<｜tool▁call▁begin｜>' + tool['type'] + '<｜tool▁sep｜>' + tool['function']['name'] + '\\n' + '```json' + '\\n' + tool['function']['arguments'] + '\\n' + '```' + '<｜tool▁call▁end｜>'}}{{'<｜tool▁calls▁end｜><｜end▁of▁sentence｜>'}}{%- endif %}{%- endfor %}{%- endif %}{%- if message['role'] == 'assistant' and message['content'] is not none %}{%- if ns.is_tool %}{{'<｜tool▁outputs▁end｜>' + message['content'] + '<｜end▁of▁sentence｜>'}}{%- set ns.is_tool = false -%}{%- else %}{% set content = message['content'] %}{% if '</think>' in content %}{% set content = content.split('</think>')[-1] %}{% endif %}{{'<｜Assistant｜>' + content + '<｜end▁of▁sentence｜>'}}{%- endif %}{%- endif %}{%- if message['role'] == 'tool' %}{%- set ns.is_tool = true -%}{%- if ns.is_output_first %}{{'<｜tool▁outputs▁begin｜><｜tool▁output▁begin｜>' + message['content'] + '<｜tool▁output▁end｜>'}}{%- set ns.is_output_first = false %}{%- else %}{{'\\n<｜tool▁output▁begin｜>' + message['content'] + '<｜tool▁output▁end｜>'}}{%- endif %}{%- endif %}{%- endfor -%}{% if ns.is_tool %}{{'<｜tool▁outputs▁end｜>'}}{% endif %}{% if add_generation_prompt and not ns.is_tool %}{{'<｜Assistant｜><think>\\n'}}{% endif %}",
     "stop_token_ids": [
       151643
     ],
     "stop": [
       "<｜end▁of▁sentence｜>"
-    ]
+    ],
+    "reasoning_start_tag": "<think>",
+    "reasoning_end_tag": "</think>"
   },
   {
     "version": 1,

xinference/model/llm/llm_family.py

@@ -134,7 +134,7 @@ class LLMFamilyV1(BaseModel):
     model_name: str
     model_lang: List[str]
     model_ability: List[
-        Literal["embed", "generate", "chat", "tools", "vision", "audio"]
+        Literal["embed", "generate", "chat", "tools", "vision", "audio", "reasoning"]
     ]
     model_description: Optional[str]
     # reason for not required str here: legacy registration
@@ -143,6 +143,8 @@ class LLMFamilyV1(BaseModel):
     chat_template: Optional[str]
     stop_token_ids: Optional[List[int]]
     stop: Optional[List[str]]
+    reasoning_start_tag: Optional[str]
+    reasoning_end_tag: Optional[str]
 
 
 class CustomLLMFamilyV1(LLMFamilyV1):