search

triton-inference-server / fil_backend

FIL backend for the Triton Inference Server
Apache License 2.0
72 stars 36 forks source link

Synchronize stream during TritonTensor sync #85

Closed wphicks closed 3 years ago

wphicks commented 3 years ago

Call cudaStreamSynchronize when syncing TritonTensor with output buffers in order to ensure correct output in cuda shared memory mode

Resolve #80