github
/
minimind
mirror of https://github.com/jingyaogong/minimind.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

修正了训练tokenizer中的chat_template中的逻辑,以及修正了tokenizer_config.json相应部分

pull/86/head
matianhao 2024-11-08 01:05:51 +08:00
parent 1240829c89
commit 8669d9c10f
2 changed files with 13 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
model/minimind_tokenizer/tokenizer_config.json
View File

@ -40,5 +40,5 @@
"tokenizer_class": "PreTrainedTokenizerFast", "tokenizer_class": "PreTrainedTokenizerFast",
"unk_token": "<unk>", "unk_token": "<unk>",
"use_default_system_prompt": false, "use_default_system_prompt": false,
"chat_template": "{% if messages[0]['role'] == 'system' %}{% set system_message = messages[0]['content'] %}{% endif %}{% if system_message is defined %}{{ system_message }}{% endif %}{% for message in messages %}{% set content = message['content'] %}{% if message['role'] == 'user' %}{{ '<s>user\\n' + content + '</s>\\n<s>assistant\\n' }}{% elif message['role'] == 'assistant' %}{{ content + '</s>' + '\\n' }}{% endif %}{% endfor %}" "chat_template": "{% for message in messages %}{% if loop.first and messages[0]['role'] != 'system' %}{{ '<s>system\n你是一个乐于助人的助手</s>\n' }}{% endif %}{{'<s>' + message['role'] + '\n' + message['content'] + '</s>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<s>assistant\n' }}{% endif %}"
} }

15
train_tokenizer.py
View File

@ -104,7 +104,7 @@ def train_tokenizer():
"tokenizer_class": "PreTrainedTokenizerFast", "tokenizer_class": "PreTrainedTokenizerFast",
"unk_token": "<unk>", "unk_token": "<unk>",
"use_default_system_prompt": False, "use_default_system_prompt": False,
"chat_template": "{% if messages[0]['role'] == 'system' %}{% set system_message = messages[0]['content'] %}{% endif %}{% if system_message is defined %}{{ system_message }}{% endif %}{% for message in messages %}{% set content = message['content'] %}{% if message['role'] == 'user' %}{{ '<s>user\\n' + content + '</s>\\n<s>assistant\\n' }}{% elif message['role'] == 'assistant' %}{{ content + '</s>' + '\\n' }}{% endif %}{% endfor %}" "chat_template": "{% for message in messages %}{% if loop.first and messages[0]['role'] != 'system' %}{{ '<s>system\n你是一个乐于助人的助手</s>\n' }}{% endif %}{{'<s>' + message['role'] + '\n' + message['content'] + '</s>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<s>assistant\n' }}{% endif %}"
} }
# 保存配置文件 # 保存配置文件
@ -122,14 +122,23 @@ def eval_tokenizer():
messages = [ messages = [
{"role": "system", "content": "你是一个优秀的聊天机器人，总是给我正确的回应！"}, {"role": "system", "content": "你是一个优秀的聊天机器人，总是给我正确的回应！"},
{"role": "user", "content": '你来自哪里？'}, {"role": "user", "content": '你来自哪里？'}
{"role": "assistant", "content": '我来自地球'}
] ]
new_prompt = tokenizer.apply_chat_template( new_prompt = tokenizer.apply_chat_template(
messages, messages,
tokenize=False tokenize=False
) )
add_generation_prompt_messages = [
{"role": "user", "content": '你来自哪里？'}
]
add_generation_prompt_new_prompt = tokenizer.apply_chat_template(
add_generation_prompt_messages,
tokenize=False,
add_generation_prompt=True
)
print(new_prompt) print(new_prompt)
print("----------")
print(add_generation_prompt_new_prompt)
# 获取实际词汇表长度（包括特殊符号） # 获取实际词汇表长度（包括特殊符号）
actual_vocab_size = len(tokenizer) actual_vocab_size = len(tokenizer)