Update multi_head_attn.py

DerryHub / BEVFormer_tensorrt

BEVFormer inference on TensorRT, including INT8 Quantization and Custom TensorRT Plugins (float/half/half2/int8).

Apache License 2.0

430 stars 71 forks source link

Closed DerryHub closed 1 year ago