CODEDUMP_NOTES Telegram 626
#人工智能
#开源项目
前两天的Deepseek v3发布,效果有点震惊,看论文所说,只用了2048张H800显卡,2个月不到时间训练完毕,计算预算只有不到600万美元就训练出来了新的模型。

与其他大模型玩家不同的是,deepseek背后是一家叫幻方量化的公司,不是传统的互联网企业,前几个月看到的采访《揭秘DeepSeek:一个更极致的中国技术理想主义故事》,更早的一篇采访《疯狂的幻方:一家隐形AI巨头的大模型之路》。



tgoop.com/codedump_notes/626
Create:
Last Update:

#人工智能
#开源项目
前两天的Deepseek v3发布,效果有点震惊,看论文所说,只用了2048张H800显卡,2个月不到时间训练完毕,计算预算只有不到600万美元就训练出来了新的模型。

与其他大模型玩家不同的是,deepseek背后是一家叫幻方量化的公司,不是传统的互联网企业,前几个月看到的采访《揭秘DeepSeek:一个更极致的中国技术理想主义故事》,更早的一篇采访《疯狂的幻方:一家隐形AI巨头的大模型之路》。

BY codedump的电报频道




Share with your friend now:
tgoop.com/codedump_notes/626

View MORE
Open in Telegram


Telegram News

Date: |

Ng Man-ho, a 27-year-old computer technician, was convicted last month of seven counts of incitement charges after he made use of the 100,000-member Chinese-language channel that he runs and manages to post "seditious messages," which had been shut down since August 2020. Ng, who had pleaded not guilty to all charges, had been detained for more than 20 months. His channel was said to have contained around 120 messages and photos that incited others to vandalise pro-government shops and commit criminal damage targeting police stations. The Channel name and bio must be no more than 255 characters long To upload a logo, click the Menu icon and select “Manage Channel.” In a new window, hit the Camera icon. According to media reports, the privacy watchdog was considering “blacklisting” some online platforms that have repeatedly posted doxxing information, with sources saying most messages were shared on Telegram.
from us


Telegram codedump的电报频道
FROM American