github/ColossalAI
1
0
Fork 0
mirror of https://github.com/hpcaitech/ColossalAI.git synced 2025-09-16 06:30:41 +00:00
Code Issues Packages Projects Releases Wiki Activity

upgrade colossal-chat support tp_group>1, add sp for sft

Browse Source
This commit is contained in:
YeAnbang
2024-05-27 05:55:57 +00:00
parent 73e88a5553
commit 7a7e86987d
33 changed files with 7574 additions and 105 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
applications/ColossalChat/coati/dataset/conversation.py
View File

@@ -17,6 +17,7 @@ class Conversation:
system_message: str
chat_template: str
stop_ids: List[int]
end_of_assistant: str
@classmethod
def from_config(cls, tokenizer: PreTrainedTokenizer, config: Dict):
@@ -24,7 +25,7 @@ class Conversation:
Setup the conversation template from config
"""
tokenizer.chat_template = config["chat_template"]
conv = cls(tokenizer, config["system_message"], config["chat_template"], config["stop_ids"])
conv = cls(tokenizer, config["system_message"], config["chat_template"], config["stop_ids"], config["end_of_assistant"])
conv.clear()
return conv
@@ -109,6 +110,8 @@ def setup_conversation_template(
"""
if any([s not in chat_template_config.keys() for s in Conversation.get_conversation_template_keys()]):
# Try to automatically set up conversation template, if fail, it throws an error that you need to do it manually
if "end_of_assistant" not in chat_template_config:
raise ValueError("Please set the end of assistant token.")
if "system_message" not in chat_template_config:
logger.warning("No system message is provided, will not use system message.")
if "chat_template" not in chat_template_config: