Hi, I have just started using TVM and having issues with running the script for auto-tuning a network for x86. I tried the script from the tutorial here and get the error enclosed below. I was able to successfully build from source using python3.5. Any help is appreciated!
Enclosed
Traceback (most recent call last):
File “tune_relay_x86.py”, line 30, in
from tvm import relay
File “/python/tvm/relay/init.py”, line 24, in
from . import expr_functor
File “/python/tvm/relay/expr_functor.py”, line 24, in
from .op import Op
File “/python/tvm/relay/op/init.py”, line 20, in
from .op import get, register, register_schedule, register_compute, register_gradient,
File “/python/tvm/relay/op/op.py”, line 19, in
import topi
File “/topi/python/topi/init.py”, line 27, in
from . import cuda
File “/topi/python/topi/cuda/init.py”, line 5, in
from . import conv2d, depthwise_conv2d, conv2d_transpose_nchw, deformable_conv2d,
File “/topi/python/topi/cuda/conv2d.py”, line 28, in
from .conv2d_int8 import conv2d_NCHWc_int8, schedule_conv2d_NCHWc_int8
File “/topi/python/topi/cuda/conv2d_int8.py”, line 152, in
_dp4a = dp4a(‘shared’, ‘shared’, ‘local’)
File “/topi/python/topi/cuda/tensor_intrin.py”, line 71, in dp4a
with tvm.build_config(data_alignment=4, offset_factor=1) as cfg:
File “/python/tvm/build_module.py”, line 260, in build_config
config = make.node(“BuildConfig”, **node_args)
File “/python/tvm/make.py”, line 115, in node
return _Node(*args)
File “/python/tvm/_ffi/_ctypes/function.py”, line 209, in call
raise get_last_ffi_error()
tvm._ffi.base.TVMError: Traceback (most recent call last):
[bt] (3) /build/libtvm.so(TVMFuncCall+0x61) [0x7fe275dbb461]
[bt] (2) /build/libtvm.so(+0x9be536) [0x7fe27574b536]
[bt] (1) /build/libtvm.so(+0x9be08d) [0x7fe27574b08d]
[bt] (0) /build/libtvm.so(+0x83f432) [0x7fe2755cc432]
File “/src/lang/reflection.cc”, line 543
TVMError: BuildConfig does not contain field disable_vectorize