|
|
|
|
|
by dtquad
591 days ago
|
|
China can train 70B and 400B LLM models using pre-sanction Nvidia GPUs. The only reason they are loudly announcing "Hey our military is using YOUR open source models" is to trigger fearmongering and overzealous regulation in the West and make us less competitive in AI. |
|
This was used by Tencent for their SOTA open source model (Hunyuan).