error when make converter folder about mnn HOT 6 CLOSED

Could you upgrade protobuf? Probably protobuf version result in this error, you can try protobuf [email protected].

from mnn.

it's not a protobuf problem. we include a copy of protobuf in 3rd_party folder, and which is used in generating headers from protos.

run schema/generate.sh before building converter, and make sure no error occurs.

from mnn.

it's not a protobuf problem. we include a copy of protobuf in 3rd_party folder, and which is used in generating headers from protos.

run schema/generate.sh before building converter, and make sure no error occurs.

I run this script and got these errors

./schema/generate.sh: line 9: pushd: ../3rd_party/flatbuffers: No such file or directory
./schema/generate.sh: line 17: popd: directory stack empty
*** cleaning up ***
*** generating fbs under default ***
find: ‘../default/*.fbs’: No such file or directory
../../3rd_party/flatbuffers/tmp/flatc: error: missing input files

but I can still compile MNN root folder without any problem, this issue only occurs in converter folder

from mnn.

I see ...

please follow install.md's guide -- "cd schema && ./generate.sh && cd ..",

not "./schema/generate.sh", so that path error will be fixed.

consider to make the shell compatible later.

from mnn.

I see ...

please follow install.md's guide -- "cd schema && ./generate.sh && cd ..",

not "./schema/generate.sh", so that path error will be fixed.

consider to make the shell compatible later.

thks, now cmake in root folder makes no misktake ,BUT the converter folder cannot be compiled yet, just as i wrote in this issue, MNN/tools/converter/source/tensorflow/SplitTf.cpp:106:53: error: no match for ‘operator[]’ (operand types are ‘const google::protobuf::RepeatedField<int>’ and ‘int’), it looks definitely like a protobuf problem

from mnn.

solved this problem by changing file tools\converter\source\tensorflow\SplitTf.cpp line 106

splitvParam->axis = value.tensor().int_val()[0];

to

splitvParam->axis = value.tensor().int_val(0);

now everything works.

from mnn.