search

DerryHub / BEVFormer_tensorrt

BEVFormer inference on TensorRT, including INT8 Quantization and Custom TensorRT Plugins (float/half/half2/int8).
Apache License 2.0
410 stars 67 forks source link

Update multi_head_attn.py #67

Closed DerryHub closed 1 year ago