Upload tokenizer_config.json with huggingface_hub
Browse files- tokenizer_config.json +6 -13
tokenizer_config.json
CHANGED
|
@@ -1,25 +1,18 @@
|
|
| 1 |
{
|
| 2 |
-
"added_tokens_decoder": {},
|
| 3 |
-
"additional_special_tokens": [
|
| 4 |
-
"<|startoftext|>",
|
| 5 |
-
"<|extra_0|>",
|
| 6 |
-
"<|extra_4|>",
|
| 7 |
-
"<|extra_5|>",
|
| 8 |
-
"<|eos|>"
|
| 9 |
-
],
|
| 10 |
"architectures": [
|
| 11 |
"GPT2LMHeadModel"
|
| 12 |
],
|
|
|
|
|
|
|
| 13 |
"auto_map": {
|
| 14 |
"AutoTokenizer": [
|
| 15 |
"tokenization_hy.HYTokenizer",
|
| 16 |
null
|
| 17 |
-
|
| 18 |
},
|
| 19 |
-
"
|
| 20 |
-
"clean_up_tokenization_spaces": true,
|
| 21 |
-
"model_max_length": 1048576,
|
| 22 |
"model_type": "gpt2",
|
|
|
|
| 23 |
"pad_token": "<|pad|>",
|
| 24 |
-
"
|
| 25 |
}
|
|
|
|
| 1 |
{
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
"architectures": [
|
| 3 |
"GPT2LMHeadModel"
|
| 4 |
],
|
| 5 |
+
"model_max_length": 1048576,
|
| 6 |
+
"tokenizer_class": "HYTokenizer",
|
| 7 |
"auto_map": {
|
| 8 |
"AutoTokenizer": [
|
| 9 |
"tokenization_hy.HYTokenizer",
|
| 10 |
null
|
| 11 |
+
]
|
| 12 |
},
|
| 13 |
+
"eos_token": "<|eos|>",
|
|
|
|
|
|
|
| 14 |
"model_type": "gpt2",
|
| 15 |
+
"additional_special_tokens": ["<|startoftext|>", "<|extra_0|>", "<|extra_4|>", "<|extra_5|>", "<|eos|>"],
|
| 16 |
"pad_token": "<|pad|>",
|
| 17 |
+
"chat_template": "{% set loop_messages = messages %}\n{% if tools %}\n {% set weekday_map = {'Monday': '星期一', 'Tuesday': '星期二', 'Wednesday': '星期三', 'Thursday': '星期四', 'Friday': '星期五', 'Saturday': '星期六', 'Sunday': '星期日'} %}\n {% set weekday_cn = weekday_map[strftime_now('%A')] %}\n {% set datetime_str = strftime_now('%Y-%m-%d %H:%M:%S') %}\n {% set datetime_str = datetime_str + ' ' + weekday_cn %}\n {% for message in loop_messages %}\n {% if 'content' in message %}\n {% set content = message['content'] %}\n {% else %}\n {% set content = '' %}\n {% endif %}\n {% if loop.index0 == 0 %}\n {% set content_tmp = '你是一位函数组合专家。你会得到一个问题和一组可能的函数。根据问题，你需要进行一个或多个函数/工具调用以实现目的。\n如果没有一个函数可以使用，请直接使用自然语言回复用户，以助手：开头。\n如果给定的问题缺少函数所需的参数，请使用自然语言进行提问，向用户询问必要信息，以助手：开头。\n如果调用结果已经足够回答用户问题，请对历史结果进行总结，使用自然语言回复用户，以助手：开头。\n你应该只在工具调用部分返回函数调用。如果你决定调用任何函数，你必须将其格式化为<tool_calls>[{\"name\": \"func_name1\", \"arguments\": {\"argument1\": \"value1\", \"argument2\": \"value2\"}},...]</tool_calls>。你不应该在回复中包含任何其他文本。以下是你可以调用的函数列表，格式为JSON。\n' %}\n {% set content_tmp = content_tmp + '\n' + tools | tojson + '\n' %}\n {% if message['role'] == 'system' %}\n {% set content_tmp = content_tmp + '\n额外要求：\n' + content + '\n\n如果你决定返回函数调用，请将其格式化为<tool_calls>[{\"name\": \"func_name1\", \"arguments\": {\"argument1\": \"value1\", \"argument2\": \"value2\"}},...]</tool_calls>，不得包含其他文本。如果额外要求里有格式要求，请忽略，以此处为准。\n否则，请参考开头说的三种情况，以助手：开头进行回复。\n\n如果额外要求里有时间信息，就以额外要求里的时间为准，否则，参考当前时间：' + datetime_str %}\n {% set content = '<|startoftext|>' + content_tmp + '<|extra_4|>' %}\n {% elif message['role'] == 'user' %}\n {% set content_tmp = content_tmp + '\n如果你决定返回函数调用，请将其格式化为<tool_calls>[{\"name\": \"func_name1\", \"arguments\": {\"argument1\": \"value1\", \"argument2\": \"value2\"}},...]</tool_calls>，不得包含其他文本。\n否则，请参考开头说的三种情况，以助手：开头进行回复。\n\n当前时间：' + datetime_str %}\n {% set content_tmp = '<|startoftext|>' + content_tmp + '<|extra_4|>'%}\n {% set content = content_tmp + '用户：' + content + '<|extra_0|>' %}\n {% endif %}\n {% else %}\n {% if message['role'] == 'user' %}\n {% set content = '用户：' + content + '<|extra_0|>' %}\n {% elif message['role'] == 'assistant' %}\n {% if 'tool_calls' in message %}\n {% set tool_calls = message['tool_calls'] %}\n {% set ns = namespace(tool_calls=\"[\") %}\n {% for tool_call in tool_calls %}\n {% set function = tool_call['function'] %}\n {% set name = function['name'] %}\n {% set ns.tool_calls = ns.tool_calls + '{\"name\": \"' + name + '\", '%}\n {% set arguments = function['arguments'] %}\n {% if arguments is not string %}\n {% set arguments = arguments | tojson %}\n {% endif %}\n {% set ns.tool_calls = ns.tool_calls + '\"arguments\": ' + arguments + '}' %}\n {% if not loop.last %}\n {% set ns.tool_calls = ns.tool_calls + ', '%}\n {% endif %}\n {% endfor %}\n {% set ns.tool_calls = ns.tool_calls + ']' %}\n {% set content = content + '<tool_calls>' + ns.tool_calls + '</tool_calls>' %}\n {% else %}\n {% set content = '助手：' + content %}\n {% endif %}\n {% set content = content + '<|eos|>' %}\n {% elif message['role'] == 'tool' %}\n {% if content is not string %}\n {set content = content | tojson }\n {% endif %}\n {% set content = '<tool_response>' + content + '</tool_response>' %}\n {% set content = content + '<|extra_0|>' %}\n {% endif %}\n {% endif %}\n {{- content -}}\n {% endfor %}\n{% else %}\n {% set context = {'has_head': true} %}\n {% for message in loop_messages %}\n {% if 'content' in message %}\n {% set content = message['content'] %}\n {% else %}\n {% set content = '' %}\n {% endif %}\n {% if loop.index0 == 0 %}\n {% if content == '' %}\n {% set _ = context.update({'has_head': false}) %}\n {% elif message['role'] == 'system' %}\n {% set content = '<|startoftext|>' + content + '<|extra_4|>' %}\n {% endif %}\n {% endif %}\n {% if message['role'] == 'user' %}\n {% if loop.index0 == 1 and not context.has_head %}\n {% set content = '<|startoftext|>' + content %}\n {% endif %}\n {% if loop.index0 == 1 and context.has_head %}\n {% set content = content + '<|extra_0|>' %}\n {% else %}\n {% set content = '<|startoftext|>' + content + '<|extra_0|>' %}\n {% endif %}\n {% elif message['role'] == 'assistant' %}\n {% set content = content + '<|eos|>' %}\n {% elif message['role'] == 'tool' %}\n {% set content = content + '<|extra_0|>' %}\n {% endif %}\n {{- content -}}\n {% endfor %}\n{% endif %}\n{%- if enable_thinking is defined and enable_thinking is false %}\n {{- '<think>\\n\\n</think>\\n' }}\n{%- endif %}"
|
| 18 |
}
|