Linly icon indicating copy to clipboard operation
Linly copied to clipboard

训练时 gpu的耗时能透露一下吗

Open zhangxueren9 opened this issue 2 years ago • 8 comments

zhangxueren9 avatar Mar 31 '23 05:03 zhangxueren9

目前版本训练大概用了2天

ydli-ai avatar Mar 31 '23 06:03 ydli-ai

是8张A100吗

zhangxueren9 avatar Mar 31 '23 06:03 zhangxueren9

目前版本训练大概用了2天

大佬 可以留一个wx吗

zhangxueren9 avatar Mar 31 '23 06:03 zhangxueren9

32


发件人: Knihht @.> 发送时间: Friday, March 31, 2023 2:45:35 PM 收件人: ydli-ai/Chinese-ChatLLaMA @.> 抄送: Li Yudong @.>; Comment @.> 主题: Re: [ydli-ai/Chinese-ChatLLaMA] 训练时 gpu的耗时能透露一下吗 (Issue #7)

是8张A100吗

― Reply to this email directly, view it on GitHubhttps://github.com/ydli-ai/Chinese-ChatLLaMA/issues/7#issuecomment-1491396213, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AE3SPV72RSGIW4ZGRICHI2TW6Z4Q7ANCNFSM6AAAAAAWOGBZRE. You are receiving this because you commented.Message ID: @.***>

ydli-ai avatar Mar 31 '23 06:03 ydli-ai

后面会考虑建用户群


发件人: Knihht @.> 发送时间: Friday, March 31, 2023 2:46:11 PM 收件人: ydli-ai/Chinese-ChatLLaMA @.> 抄送: Li Yudong @.>; Comment @.> 主题: Re: [ydli-ai/Chinese-ChatLLaMA] 训练时 gpu的耗时能透露一下吗 (Issue #7)

目前版本训练大概用了2天

大佬 可以留一个wx吗

― Reply to this email directly, view it on GitHubhttps://github.com/ydli-ai/Chinese-ChatLLaMA/issues/7#issuecomment-1491396739, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AE3SPV7AY2UDLY42TW4SZRTW6Z4THANCNFSM6AAAAAAWOGBZRE. You are receiving this because you commented.Message ID: @.***>

ydli-ai avatar Mar 31 '23 06:03 ydli-ai

后面会考虑建用户群 ________________________________ 发件人: Knihht @.> 发送时间: Friday, March 31, 2023 2:46:11 PM 收件人: ydli-ai/Chinese-ChatLLaMA @.> 抄送: Li Yudong @.>; Comment @.> 主题: Re: [ydli-ai/Chinese-ChatLLaMA] 训练时 gpu的耗时能透露一下吗 (Issue #7) 目前版本训练大概用了2天 大佬 可以留一个wx吗 ― Reply to this email directly, view it on GitHub<#7 (comment)>, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AE3SPV7AY2UDLY42TW4SZRTW6Z4THANCNFSM6AAAAAAWOGBZRE. You are receiving this because you commented.Message ID: @.***>

这个有建用户群吗?

baketbek avatar Apr 10 '23 12:04 baketbek

@ydli-ai 老哥, 这个不同的大模型预训练,对gpu需求是线性的吗? 比方说您7B的模型,32xA100用两天, 那么如果用65B的模型(按10倍算),用32xA100需要20天吗?或者用320xA100需要2天吗?

谢谢

ze00ro avatar Apr 21 '23 08:04 ze00ro

@ydli-ai 老哥, 这个不同的大模型预训练,对gpu需求是线性的吗? 比方说您7B的模型,32xA100用两天, 那么如果用65B的模型(按10倍算),用32xA100需要20天吗?或者用320xA100需要2天吗?

谢谢

+1 同问

JingxinLee avatar Jul 08 '23 06:07 JingxinLee