I am running tune_relay_cuda.py in tutorials/autotvm. The device is a TX2. The host is running on Ubuntu 18.04 with 1080Ti. Both are built with llvm-6.0, cuda-10.
The error code is as follows:
DEBUG:autotvm:No: 71 GFLOPS: 0.00/0.00 result: MeasureResult(costs=(RuntimeError('Traceback (most recent call last):\n [bt] (8) /home/nvidia/tvm/build/libtvm.so(tvm::runtime::RPCSession::ServerLoop()+0
x104) [0x7f8c5b8f04]\n [bt] (7) /home/nvidia/tvm/build/libtvm.so(tvm::runtime::RPCSession::HandleUntilReturnEvent(tvm::runtime::TVMRetValue*, bool, tvm::runtime::PackedFunc const*)+0x1b8) [0x7f8c5b8bd8]\n [bt]
(6) /home/nvidia/tvm/build/libtvm.so(tvm::runtime::RPCSession::EventHandler::HandleNextEvent(tvm::runtime::TVMRetValue*, bool, tvm::runtime::PackedFunc const*)+0x554) [0x7f8c5be72c]\n [bt] (5) /home/nvidia/tvm
/build/libtvm.so(tvm::runtime::RPCSession::EventHandler::HandleRecvPackedSeqArg()+0x148) [0x7f8c5bdc78]\n [bt] (4) /home/nvidia/tvm/build/libtvm.so(tvm::runtime::RPCSession::EventHandler::SwitchToState(tvm::run
time::RPCSession::EventHandler::State)+0x320) [0x7f8c5bcb98]\n [bt] (3) /home/nvidia/tvm/build/libtvm.so(tvm::runtime::RPCSession::EventHandler::HandlePackedCall()+0x688) [0x7f8c5b7078]\n [bt] (2) /home/nvidia
/tvm/build/libtvm.so(void tvm::runtime::RPCSession::EventHandler::C',),), error_no=4, all_cost=6.999524354934692, timestamp=1569314585.3646219) [('tile_f', [2, 32, 8, 1]), ('tile_y', [1, 1, 1, 7]), ('tile_x', [1
, 1, 1, 7]), ('tile_rc', [256, 1]), ('tile_ry', [1, 1]), ('tile_rx', [1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],direct,None,35012
The same tuning setting and code on normal laptop works fine. Are there any suggestions? Thanks a lot!