I am a beginner and I was trying to use model-compiler to get the LLVM IR code.
here is my commond
./model-compiler --backend=CPU --model=/home/…/…/build_Debug/examples/bundles/resnet50/resnet50 -emit-bundle=build
The following are output
WARNING: Logging before InitGoogleLogging() is written to STDERR
F1016 00:45:17.236379 7420 Error.cpp:123] exitOnError(Error) got an unexpected ErrorValue:
Error message: No node under name gpu_0/data
Error return stack:
/home/yinix/EE554/glow/lib/Importer/ProtobufLoader.cpp:175
/home/yinix/EE554/glow/lib/Importer/Caffe2ModelLoader.cpp:445
/home/yinix/EE554/glow/lib/Importer/Caffe2ModelLoader.cpp:2405
/home/yinix/EE554/glow/lib/Importer/Caffe2ModelLoader.cpp:2964
/home/yinix/EE554/glow/lib/Importer/Caffe2ModelLoader.cpp:2981
*** Check failure stack trace: ***
./model-compiler[0xc33f3f]
./model-compiler[0xc321c2]
./model-compiler[0xc34345]
/lib/x86_64-linux-gnu/libpthread.so.0(+0x14420)[0x7fc67fc9f420]
/lib/x86_64-linux-gnu/libc.so.6(gsignal+0xcb)[0x7fc67f78e00b]
/lib/x86_64-linux-gnu/libc.so.6(abort+0x12b)[0x7fc67f76d859]
/lib/x86_64-linux-gnu/libglog.so.0(+0x9e4c)[0x7fc680721e4c]
/lib/x86_64-linux-gnu/libglog.so.0(+0xd1c3)[0x7fc6807251c3]
/lib/x86_64-linux-gnu/libglog.so.0(_ZN6google10LogMessage9SendToLogEv+0x26b)[0x7fc68072a25b]
/lib/x86_64-linux-gnu/libglog.so.0(_ZN6google10LogMessage5FlushEv+0xbf)[0x7fc680724ebf]
/lib/x86_64-linux-gnu/libglog.so.0(_ZN6google15LogMessageFatalD2Ev+0xf)[0x7fc6807256ef]
./model-compiler[0x2bc64ba]
./model-compiler[0x7df01c]
./model-compiler[0x4b131e]
./model-compiler[0x4f0470]
/lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xf3)[0x7fc67f76f083]
./model-compiler[0x4b0bfe]
zsh: abort ./model-compiler --backend=CPU -emit-bundle=build
Then I read for several posts about what is “model-input”. It is said to be where the input will be stored and will be shown in the protobuf file of the model. Then I checked for the predict_net.pb file in /home/…/…/build_Debug/examples/bundles/resnet50/resnet50. It shows that my first operator is indeed gpu_0/data.
Here are the first few lines of predict_net.pb shown in vim
^Hresnet50^Rd
gpu_0/data
^Mgpu_0/conv1_w^R^Kgpu_0/conv1^Z^@"^DConv*
^Fkernel^X^G*^G
^Cpad^X^C*^M
^Eorder"^DNCHW*
^Fstride^X^B2^@:^@^R¹^A
^Kgpu_0/conv1
^Tgpu_0/res_conv1_bn_s
^Tgpu_0/res_conv1_bn_b
^Ugpu_0/res_conv1_bn_rm
Where can I find more information about my problem? What configurations should I check to solve this issue?