Zhexin Li

icon indicating an email [email protected]

icon company DAMO, Alibaba-inc icon location Hangzhou, China icon location Master of National Laboratory of Pattern Recognition (NLPR), Institute of Automation (IA), Chinese Academy of Sciences (CAS)

Results 4 comments of Zhexin Li

SiLU(Swish) Quantization with QDQ

Hi, I'm new to TensorRT and I can't answer your question. Would you mind telling me how you get the tensorrt engine visualization image? It seems very useful.

confused with an undefined function

Got it! Thanks for your trouble. It helped a lot.

Is there a way to activate int8 MHA_v2 kernel when SeqLen > 512?

Thanks. Is there any possibility that we can surpass the MHA's seq_len limitation? Or does NV have plan to extend mha_v2 to larger dimention so that diffusion-based model can benefit...

[question] Myelin: attention fusion and FlashAttention

> The INT8 MHA fused kernels are already integrated in TRT 8.6. The only caveat is that SeqLen must be 512 or below. > > It does use flash attention...