search

mlc-ai / relax

Apache License 2.0
137 stars 69 forks source link

Did you forget to bind? #274

Open njuhang opened 11 months ago

njuhang commented 11 months ago

Expected behavior

generate a .so model file

Actual behavior

report error: tvm._ffi.base.TVMError: Traceback (most recent call last): [bt] (8) /home/mlc-relax/relax/build/libtvm.so(tvm::ApplyPasses(tvm::IRModule, tvm::transform::Sequential)+0x42) [0x7f5212c4d512] [bt] (7) /home/mlc-relax/relax/build/libtvm.so(tvm::transform::Pass::operator()(tvm::IRModule) const+0x56) [0x7f5212d21dd6] [bt] (6) /home/mlc-relax/relax/build/libtvm.so(tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const+0x347) [0x7f5212d21aa7] [bt] (5) /home/mlc-relax/relax/build/libtvm.so(tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const+0x44a) [0x7f5212d2429a] [bt] (4) /home/mlc-relax/relax/build/libtvm.so(tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const+0x347) [0x7f5212d21aa7] [bt] (3) /home/mlc-relax/relax/build/libtvm.so(tvm::transform::ModulePassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const+0x1f4) [0x7f5212d22b84] [bt] (2) /home/mlc-relax/relax/build/libtvm.so(+0x256a123) [0x7f5213820123] [bt] (1) /home/mlc-relax/relax/build/libtvm.so(tvm::runtime::detail::LogFatal::Entry::Finalize()+0x3d) [0x7f521298bfcd] [bt] (0) /home/mlc-relax/relax/build/libtvm.so(tvm::runtime::Backtrace[abi:cxx11]()+0x2c) [0x7f5214c849dc] Did you forget to bind? Variable A is directly accessed by host memory (it is not contained in a thread environment or in the function arguments. Variable T_relu is directly accessed by host memory (it is not contained in a thread environment or in the function arguments. File "/home/mlc-relax/relax/src/tir/analysis/verify_memory.cc", line 205 RuntimeError: Memory verification failed with the following errors: # from tvm.script import tir as T

Environment

Ubuntu 20.04, mlc-relax up to date, python3.8

Steps to reproduce

cross_compiler = "/home/AndroidSdk/ndk/23.2.8568313/toolchains/llvm/prebuilt/linux-x86_64/bin/aarch64-linux-android28-clang++" target = tvm.target.Target("opencl --device=mali", host="llvm --mtriple=aarch64-linux-gnu")

target = tvm.target.Target("llvm --num-cores=4 --mtriple=aarch64-linux-android --mattr=+neon")

model_out = "./libs/mobilenet.so" relay_mod, relayparam,, _ = get_network("mobilenet", 1) relax_mod = relay_translator.from_relay(relay_mod["main"], target, relay_param) ex = relax.build(relax_mod, target) ex.export_library(model_out, cc=cross_compiler)

in function get_network, it use testing.mobilenet.get_workload to load model.

Triage

*(backend::OpenCL)

raj-khare commented 4 months ago

getting the same err

luismiaresse commented 3 months ago

Getting same error with CUDA 12.1 and ROCM 5.7 in the Web Stable Diffussion notebook demo. Using LLVM target and CPU device I get a different one:

InternalError: Traceback (most recent call last):
  61: tvm::runtime::PackedFuncObj::Extractor<tvm::runtime::PackedFuncSubObj<tvm::runtime::TypedPackedFunc<tvm::runtime::Module (tvm::runtime::Map<tvm::Target, tvm::IRModule, void, void> const&, tvm::Target)>::AssignTypedLambda<tvm::__mk_TVM23::{lambda(tvm::runtime::Map<tvm::Target, tvm::IRModule, void, void> const&, tvm::Target)#1}>(tvm::__mk_TVM23::{lambda(tvm::runtime::Map<tvm::Target, tvm::IRModule, void, void> const&, tvm::Target)#1}, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >)::{lambda(tvm::runtime::TVMArgs const&, tvm::runtime::TVMRetValue*)#1}> >::Call(tvm::runtime::PackedFuncObj const*, tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*)
  60: tvm::TIRToRuntime(tvm::runtime::Map<tvm::Target, tvm::IRModule, void, void> const&, tvm::Target const&)
  59: tvm::SplitMixedModule(tvm::IRModule, tvm::Target const&, tvm::Target const&)
  58: tvm::ApplyPasses(tvm::IRModule, tvm::transform::Sequential)
  57: tvm::transform::Pass::operator()(tvm::IRModule) const
  56: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
  55: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
  54: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
  53: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
  52: _ZN3tvm7runtime13PackedFun
  51: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::AssignTypedLambda<tvm::tir::transform::LowerTVMBuiltin()::{lambda(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)#1}>(tvm::tir::transform::LowerTVMBuiltin()::{lambda(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)#1})::{lambda(tvm::runtime::TVMArgs const&, tvm::runtime::TVMRetValue*)#1}::operator()(tvm::runtime::TVMArgs const&, tvm::runtime::TVMRetValue*) const
  50: tvm::tir::BuiltinLower::VisitBodyAndRealizeAlloca(tvm::tir::Stmt)
  49: tvm::tir::BuiltinLower::GetMaxStack(tvm::tir::Stmt)
  48: tvm::tir::BuiltinLower::VisitStmt(tvm::tir::Stmt const&)
  47: tvm::tir::StmtFunctor<tvm::tir::Stmt (tvm::tir::Stmt const&)>::VisitStmt(tvm::tir::Stmt const&)
  46: _ZZN3tvm3tir11StmtFunctorIFNS0_4StmtERKS2_EE10InitVTableEvENUlRKNS_
  45: tvm::tir::BuiltinLower::VisitStmt(tvm::tir::Stmt const&)
  44: tvm::tir::StmtFunctor<tvm::tir::Stmt (tvm::tir::Stmt const&)>::VisitStmt(tvm::tir::Stmt const&)
  43: _ZZN3tvm3tir11StmtFunctorIFNS0_4StmtERKS2_EE10InitVTableEvENUlRKNS_
  42: tvm::tir::BuiltinLower::VisitStmt(tvm::tir::Stmt const&)
  41: tvm::tir::StmtFunctor<tvm::tir::Stmt (tvm::tir::Stmt const&)>::VisitStmt(tvm::tir::Stmt const&)
  40: _ZZN3tvm3tir11StmtFunctorIFNS0_4StmtERKS2_EE10InitVTableEvENUlRKNS_
  39: tvm::tir::BuiltinLower::VisitStmt(tvm::tir::Stmt const&)
  38: tvm::tir::StmtFunctor<tvm::tir::Stmt (tvm::tir::Stmt const&)>::VisitStmt(tvm::tir::Stmt const&)
  37: _ZZN3tvm3tir11StmtFunctorIFNS0_4StmtERKS2_EE10InitVTableEvENUlRKNS_
  36: tvm::tir::BuiltinLower::VisitStmt(tvm::tir::Stmt const&)
  35: tvm::tir::StmtFunctor<tvm::tir::Stmt (tvm::tir::Stmt const&)>::VisitStmt(tvm::tir::Stmt const&)
  34: _ZZN3tvm3tir11StmtFunctorIFNS0_4StmtERKS2_EE10InitVTableEvENUlRKNS_
  33: tvm::tir::BuiltinLower::VisitStmt(tvm::tir::Stmt const&)
  32: tvm::tir::StmtFunctor<tvm::tir::Stmt (tvm::tir::Stmt const&)>::VisitStmt(tvm::tir::Stmt const&)
  31: _ZZN3tvm3tir11StmtFunctorIFNS0_4StmtERKS2_EE10InitVTableEvENUlRKNS_
  30: tvm::tir::BuiltinLower::VisitStmt(tvm::tir::Stmt const&)
  29: tvm::tir::StmtFunctor<tvm::tir::Stmt (tvm::tir::Stmt const&)>::VisitStmt(tvm::tir::Stmt const&)
  28: _ZZN3tvm3tir11StmtFunctorIFNS0_4StmtERKS2_EE10InitVTableEvENUlRKNS_
  27: tvm::tir::BuiltinLower::VisitStmt(tvm::tir::Stmt const&)
  26: tvm::tir::StmtFunctor<tvm::tir::Stmt (tvm::tir::Stmt const&)>::VisitStmt(tvm::tir::Stmt const&)
  25: _ZZN3tvm3tir11StmtFunctorIFNS0_4StmtERKS2_EE10InitVTableEvENUlRKNS_7ru
  24: tvm::runtime::ObjectPtr<tvm::runtime::Object> tvm::runtime::Array<tvm::tir::Stmt, void>::MapHelper<tvm::tir::StmtMutator::Internal::Mutate(tvm::tir::StmtMutator*, tvm::runtime::Array<tvm::tir::Stmt, void> const&)::{lambda(tvm::tir::Stmt const&)#1}, tvm::tir::Stmt>(tvm::runtime::ObjectPtr<tvm::runtime::Object>, tvm::tir::StmtMutator::Internal::Mutate(tvm::tir::StmtMutator*, tvm::runtime::Array<tvm::tir::Stmt, void> const&)::{lambda(tvm::tir::Stmt const&)#1})
  23: tvm::tir::BuiltinLower::VisitStmt(tvm::tir::Stmt const&)
  22: tvm::tir::StmtFunctor<tvm::tir::Stmt (tvm::tir::Stmt const&)>::VisitStmt(tvm::tir::Stmt const&)
  21: _ZZN3tvm3tir11StmtFunctorIFNS0_4StmtERKS2_EE10InitVTableEvENUlRKNS_
  20: tvm::tir::BuiltinLower::VisitStmt_(tvm::tir::AttrStmtNode const*)
  19: tvm::tir::BuiltinLower::VisitStmt(tvm::tir::Stmt const&)
  18: tvm::tir::StmtFunctor<tvm::tir::Stmt (tvm::tir::Stmt const&)>::VisitStmt(tvm::tir::Stmt const&)
  17: _ZZN3tvm3tir11StmtFunctorIFNS0_4StmtERKS2_EE10InitVTableEvENUlRKNS_
  16: tvm::tir::BuiltinLower::VisitStmt_(tvm::tir::AllocateNode const*)
  15: tvm::tir::BuiltinLower::VisitStmt(tvm::tir::Stmt const&)
  14: tvm::tir::StmtFunctor<tvm::tir::Stmt (tvm::tir::Stmt const&)>::VisitStmt(tvm::tir::Stmt const&)
  13: _ZZN3tvm3tir11StmtFunctorIFNS0_4StmtERKS2_EE10InitVTableEvENUlRKNS_
  12: tvm::tir::BuiltinLower::VisitStmt_(tvm::tir::AllocateNode const*)
  11: tvm::tir::BuiltinLower::VisitStmt(tvm::tir::Stmt const&)
  10: tvm::tir::StmtFunctor<tvm::tir::Stmt (tvm::tir::Stmt const&)>::VisitStmt(tvm::tir::Stmt const&)
  9: _ZZN3tvm3tir11StmtFunctorIFNS0_4StmtERKS2_EE10InitVTableEvENUlRKNS_
  8: tvm::tir::BuiltinLower::VisitStmt_(tvm::tir::AllocateNode const*)
  7: tvm::tir::BuiltinLower::VisitStmt(tvm::tir::Stmt const&)
  6: tvm::tir::StmtFunctor<tvm::tir::Stmt (tvm::tir::Stmt const&)>::VisitStmt(tvm::tir::Stmt const&)
  5: _ZZN3tvm3tir11StmtFunctorIFNS0_4StmtERKS2_EE10InitVTableEvENUlRKNS_
  4: tvm::tir::BuiltinLower::VisitStmt_(tvm::tir::AttrStmtNode const*)
  3: tvm::tir::BuiltinLower::VisitStmt(tvm::tir::Stmt const&)
  2: tvm::tir::StmtFunctor<tvm::tir::Stmt (tvm::tir::Stmt const&)>::VisitStmt(tvm::tir::Stmt const&)
  1: _ZZN3tvm3tir11StmtFunctorIFNS0_4StmtERKS2_EE10InitVTableEvENUlRKNS_
  0: tvm::tir::BuiltinLower::VisitStmt_(tvm::tir::AllocateNode const*)
  File "/home/luismi/Documents/relax/src/tir/transforms/lower_tvm_builtin.cc", line 248
InternalError: Check failed: (device_type_) is false: Unknown device type in current IR

Using mlc-ai-nightly-cu121 and mlc-ai-nightly-rocm57 TVM, relax.build fails

EDIT: Same behaviour happens on Google Collab (Tesla T4 GPU with CUDA 12.1, mlc-ai-nightly-cu121, pytorch nightly 2.4 cpu). This is probably a regression, or a failure to bind correctly some variables.