Problem with compile yolo tflite model

markii · June 7, 2022, 8:04am

Hi Sirs,

tvm version: 0.9.dev0 (commit: b555bf5481d3eb261427850cea286c162aa3d2e3) I use command to compile my yolo model (tflite) with CMSIS-NN, ad want to porting to STM32H7.

below is the command i used. python3 -m tvm.driver.tvmc compile --target=cmsis-nn,c --target-cmsis-nn-mcpu=cortex-m7 --target-c-mcpu=cortex-m7 --executor=aot --executor-aot-interface-api=c --executor-aot-unpacked-api=1 --runtime=crt --pass-config tir.disable_vectorize=1 -f mlf yolo.tflite

but i got the error message: ################################################################################# /home/markii/Project/tvm/python/tvm/driver/build_module.py:263: UserWarning: target_host parameter is going to be deprecated. Please pass in tvm.target.Target(target, host=target_host) instead. warnings.warn( the function is provided too many arguments expected 1, found 2 Traceback (most recent call last): File “/home/markii/anaconda3/envs/tvm/lib/python3.8/runpy.py”, line 194, in _run_module_as_main return _run_code(code, main_globals, None, File “/home/markii/anaconda3/envs/tvm/lib/python3.8/runpy.py”, line 87, in _run_code exec(code, run_globals) File “/home/markii/Project/tvm/python/tvm/driver/tvmc/main.py”, line 24, in tvmc.main.main() File “/home/markii/Project/tvm/python/tvm/driver/tvmc/main.py”, line 115, in main sys.exit(_main(sys.argv[1:])) File “/home/markii/Project/tvm/python/tvm/driver/tvmc/main.py”, line 103, in _main return args.func(args) File “/home/markii/Project/tvm/python/tvm/driver/tvmc/compiler.py”, line 173, in drive_compile compile_model( File “/home/markii/Project/tvm/python/tvm/driver/tvmc/compiler.py”, line 291, in compile_model mod = partition_function(mod, params, mod_name=mod_name, codegen_from_cli[“opts”]) File “/home/markii/Project/tvm/python/tvm/relay/op/contrib/cmsisnn.py”, line 73, in partition_for_cmsisnn return seq(mod) File “/home/markii/Project/tvm/python/tvm/ir/transform.py”, line 161, in call return _ffi_transform_api.RunPass(self, mod) File “/home/markii/Project/tvm/python/tvm/_ffi/_ctypes/packed_func.py”, line 242, in call raise get_last_ffi_error() tvm.error.DiagnosticError: Traceback (most recent call last): 18: TVMFuncCall 17: tvm::runtime::PackedFuncObj::Extractor<tvm::runtime::PackedFuncSubObj<tvm::runtime::TypedPackedFunc<tvm::IRModule (tvm::transform::Pass, tvm::IRModule)>::AssignTypedLambda<tvm::transform::{lambda(tvm::transform::Pass, tvm::IRModule)#7}>(tvm::transform::{lambda(tvm::transform::Pass, tvm::IRModule)#7}, std::__cxx11::basic_string<char, std::char_traits, std::allocator >)::{lambda(tvm::runtime::TVMArgs const&, tvm::runtime::TVMRetValue)#1}> >::Call(tvm::runtime::PackedFuncObj const, tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) 16: tvm::transform::Pass::operator()(tvm::IRModule) const 15: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const 14: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const 13: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const 12: tvm::transform::ModulePassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const 11: tvm::runtime::PackedFuncObj::Extractor<tvm::runtime::PackedFuncSubObj<tvm::runtime::TypedPackedFunc<tvm::IRModule (tvm::IRModule, tvm::transform::PassContext)>::AssignTypedLambda<tvm::relay::transform::InferType()::{lambda(tvm::IRModule, tvm::transform::PassContext const&)#1}>(tvm::relay::transform::InferType()::{lambda(tvm::IRModule, tvm::transform::PassContext const&)#1})::{lambda(tvm::runtime::TVMArgs const&, tvm::runtime::TVMRetValue*)#1}> >::Call(tvm::runtime::PackedFuncObj const*, tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) 10: tvm::relay::TypeInferencer::Infer(tvm::GlobalVar, tvm::relay::Function) 9: tvm::relay::TypeInferencer::GetType(tvm::RelayExpr const&) 8: void tvm::relay::ExpandDataflow<tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#1}, tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#2}, tvm::relay::ExpandDataflow<tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#1}, tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#2}>(tvm::RelayExpr, tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#1}, tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#2})::{lambda(tvm::RelayExpr const&)#1}>(tvm::RelayExpr, tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#1}, tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#2}, tvm::relay::ExpandDataflow<tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#1}, tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#2}>(tvm::RelayExpr, tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#1}, tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#2})::{lambda(tvm::RelayExpr const&)#1}) 7: ZZN3tvm5relay11ExprFunctorIFNS_4TypeERKNS_9RelayExprEEE10InitVTableEvENUlR 6: tvm::relay::TypeInferencer::VisitExpr(tvm::relay::FunctionNode const*) 5: tvm::relay::TypeInferencer::GetType(tvm::RelayExpr const&) 4: void tvm::relay::ExpandDataflow<tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#1}, tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#2}, tvm::relay::ExpandDataflow<tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#1}, tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#2}>(tvm::RelayExpr, tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#1}, tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#2})::{lambda(tvm::RelayExpr const&)#1}>(tvm::RelayExpr, tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#1}, tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#2}, tvm::relay::ExpandDataflow<tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#1}, tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#2}>(tvm::RelayExpr, tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#1}, tvm::relay::TypeInferencer::VisitExpr(tvm::RelayExpr const&)::{lambda(tvm::RelayExpr const&)#2})::{lambda(tvm::RelayExpr const&)#1}) 3: ZZN3tvm5relay11ExprFunctorIFNS_4TypeERKNS_9RelayExprEEE10InitVTableEvENUlR 2: tvm::relay::TypeInferencer::VisitExpr(tvm::relay::CallNode const*) 1: tvm::relay::TypeInferencer::GeneralCall(tvm::relay::CallNode const*, tvm::runtime::Array<tvm::Type, void>) 0: tvm::DiagnosticContext::Render() File “/home/markii/Project/tvm/src/ir/diagnostic.cc”, line 105 DiagnosticError: one or more error diagnostics were emitted, please check diagnostic render for output. ################################################################################

I do some test:

if i mark “cmsis-nn.qnn_mul” in cmsis-nn pattern_table(), the compilation would be ok.
cmsis-nn_mul752×182 44.9 KB
change yolo model to mobilenet (https://storage.googleapis.com/download.tensorflow.org/models/mobilenet_v1_2018_08_02/mobilenet_v1_1.0_224_quant.tgz), the compilation also works.

could you help this issue?

Thanks!

ashutosh-arm · June 7, 2022, 8:34am

Thanks @markii for reporting the issue. Looks like type inference has gone wrong here. Could you please point me to the failing Yolo model so that I can reproduce the issue locally?

markii · June 7, 2022, 9:50am

Hi Sir,

please refer to the below link~ https://sendy.link/TV8NETGI

Thanks~

ashutosh-arm · June 9, 2022, 11:17am

@manupa-arm @grant-arm @Mousius

markii · June 10, 2022, 1:26am

Hi Sirs,

It seems the square function issue. the square op generated by tvm looks like this

the two inputs are the same, it says it needs only one arguement, but found 2 the arguments:

Thanks~

ashutosh-arm · June 14, 2022, 12:19pm

For the first issue that I encountered while trying to reproduce the issue is fixed here: https://github.com/apache/tvm/pull/11701

The next issue I see now after the above fix is this: “The function is provided too many arguments”. I am working on this. Hope once I have resolved this one, I will be able to see the issue you have mentioned at the top.

markii · June 15, 2022, 1:29am

Hi Ashutosh,

yes, I forgot this issue and I also fixed it. I’m also working on this, hope can find the solution soon. Thanks for your help~~

Thanks~~

ashutosh-arm · June 15, 2022, 3:08pm

Thanks @markii for looking into this.

Another fix for the issue we are seeing with type inference is here: https://github.com/apache/tvm/pull/11732

markii · June 16, 2022, 3:23am

Hi Ashutosh,

this commit works!
the arguements error has fixed~
but the inference result is not correct, I’ll working on this~
thanks for your help to solve this issue!

Thanks~