【error】run tvm.contrib.graph_runtime.creat(xx)，error show:Check failed:!is->fail() Expect number

Questions
#1

run the official tf example(from_tensorflow.py), error occurs when executing to graph_runtime.create(graph,lib,ctx)

Error msg:

WechatIMG7
WechatIMG7.jpeg2016×1512 1.86 MB

Due to the implementation in the internal LAN, it is difficult to put the information out, the picture of the information is not good.I write part of it by hand.

Traceback (most recent call last):
     File "from_tensorflow.py",line 167  
                m = graph_runtime.create(graph,lib,ctx)
     File "tvm/python/tvm/contrib/graph_runtime.py",line 62,in create
               return  GraphModule(fcreate(graph_json_str, libmod, *device_type_id))
    File "tvm/python/tvm/_ffi/_ctypes/function.py",line 207,in __call__
               raise get_last_ffi_error()
tvm._ffi.base.TVMError: Traceback (most recent call last):
    xxxx
tvm/3rdparty/dmlc-core/include/dmlc/json.h:717:Check failed:!is_->fail(): 
                              Error at Line 28.around ^''.Expect number
#2

I was on the verge of breaking down，checked it all night.Then i print the parsing of the json line by line，a strange phenomenon was discovered.

code:

3
3.jpeg4006×833 2.94 MB
run graph_runtime.create(graph,lib,ctx) , print result:
2
2.jpeg4016×387 1.37 MB
print graph show:
5
5.jpeg1286×712 90.3 KB

All the previous prints were correct ，why *is_ >> *out_value error ?

#3

I change the code(tvm/3rdparty/dmlc-core/include/dmlc/json) in bad ways，then the result is good…

image
image.jpg1440×1080 455 KB

It took me this long to use TVM for the first time，but I don’t have to break down anymore…

I guess there’s something wrong with my c++ environment