Hi, I tested a local autotune with nvidia K80. first, the error messages are like:
GFLOPS: 0.00/0.00
result: MeasureResult(costs=(RuntimeError('Traceback (most recent call last):\n [bt] (5)
libtvm.so(TVMSynchronize+0x4a) [0x7fb40b8a12ea]\n [bt] (4)
libtvm.so(tvm::runtime::RPCDeviceAPI::StreamSync(DLContext, void*)+0x3bb) [0x7fb40b91fe1b]\n [bt] (3)
libtvm.so(tvm::runtime::RPCClientSession::StreamSync(DLContext, void*)+0x9e) [0x7fb40b8fb7ce]\n [bt] (2)
libtvm.so(+0x1569da4) [0x7fb40b8f9da4]\n [bt] (1)
libtvm.so(+0x1569cee) [0x7fb40b8f9cee]\n [bt] (0)
libtvm.so(dmlc::LogMessageFatal::~LogMessageFatal()+0x',),), error_no=4, all_cost=1.2601652145385742, timestamp=1589867856.6892421)
[('tile_f', [-1, 2, 16, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 8]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6679987
when i try to trace the “DLContext” above, I found ctx.device_type==130, i dont understand that code.
And one thing, my gpu machine is network limited, with max 10 free port to use. I’d modified rpc port config in python related things.
ah… help please!