BlackBearBiscuit

Results 6 comments of BlackBearBiscuit

无法导入paddle

解决了吗？老铁

The ChineseReverseDictionary data file is empty

有人可以分享一下中文数据集吗？万分感谢！

[REQUEST] ZeRO stage 3 support for mixture-of-experts (MoE) layer

> Hi @ranggihwang -- thank you for your interest in DeepSpeed and ZeRO-3. > > We do have some rationale for why we only support ZeRO-Stage-2 and MoE together. I...

78k evolved code instructions

+1 to this

data in modelscope for more convenient to download?

+1, please upload it to modelscope!

Applying FA3 in qwen2 model fine-tuning is slower than FA2

> OK, I'll use this commit to test it again. How about the performance? When I pretrain deepseek-v2 in H100-80G, I met the same（FA3 is slower than FA2）