failed to build model to library for Android #133
Description
Dear
I try to build model for android according to the guide.
when run ""
python3 build.py --model vicuna-v1-7b --quantization q4f16_0 --target android --max-seq-len 768
I get below error:
(mlc) wj:~/work/mlc-llm$ python3 build.py --model vicuna-v1-7b --quantization q4f16_0 --target android --max-seq-len 768
Loading checkpoint shards: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 2/2 [27:22<00:00, 821.24s/it]
Total param size: 3.5307693481445312 GB
Start storing to cache dist/vicuna-v1-7b-q4f16_0/params
[0519/0519] saving param_518
All finished, 132 total shards committed, record saved to dist/vicuna-v1-7b-q4f16_0/params/ndarray-cache.json
Save a cached module to dist/vicuna-v1-7b-q4f16_0/mod_cache_before_build_android.pkl.
19 static functions: [I.GlobalVar("take_decode"), I.GlobalVar("reshape"), I.GlobalVar("rms_norm1"), I.GlobalVar("fused_decode1_fused_matmul3_multiply"), I.GlobalVar("squeeze"), I.GlobalVar("fused_decode3_fused_matmul5_cast2"), I.GlobalVar("fused_decode_matmul1"), I.GlobalVar("softmax1"), I.GlobalVar("fused_reshape2_squeeze"), I.GlobalVar("fused_decode1_fused_matmul3_silu"), I.GlobalVar("fused_decode_fused_matmul1_add"), I.GlobalVar("fused_transpose3_reshape4"), I.GlobalVar("transpose1"), I.GlobalVar("rotary_embedding1"), I.GlobalVar("divide1"), I.GlobalVar("slice1"), I.GlobalVar("reshape1"), I.GlobalVar("reshape2"), I.GlobalVar("fused_decode2_fused_matmul4_add")]
26 dynamic functions: [I.GlobalVar("fused_decode4_NT_matmul1"), I.GlobalVar("matmul2"), I.GlobalVar("transpose2"), I.GlobalVar("take_decode1"), I.GlobalVar("reshape7"), I.GlobalVar("full"), I.GlobalVar("fused_softmax2_cast4"), I.GlobalVar("transpose7"), I.GlobalVar("reshape6"), I.GlobalVar("slice"), I.GlobalVar("fused_NT_matmul2_divide2_maximum1_minimum1_cast3"), I.GlobalVar("extend_te"), I.GlobalVar("rms_norm"), I.GlobalVar("rotary_embedding"), I.GlobalVar("squeeze1"), I.GlobalVar("reshape8"), I.GlobalVar("fused_decode6_fused_NT_matmul4_add1"), I.GlobalVar("fused_decode4_fused_NT_matmul1_add1"), I.GlobalVar("matmul10"), I.GlobalVar("fused_decode5_fused_NT_matmul3_silu1"), I.GlobalVar("fused_min_max_triu_te_broadcast_to"), I.GlobalVar("reshape3"), I.GlobalVar("fused_decode5_fused_NT_matmul3_multiply1"), I.GlobalVar("reshape5"), I.GlobalVar("fused_NT_matmul_divide_maximum_minimum_cast"), I.GlobalVar("fused_softmax_cast1")]
Dump static shape TIR to dist/vicuna-v1-7b-q4f16_0/debug/mod_tir_static.py
Dump dynamic shape TIR to dist/vicuna-v1-7b-q4f16_0/debug/mod_tir_dynamic.py
- Dispatch to pre-scheduled op: matmul2
- Dispatch to pre-scheduled op: fused_softmax2_cast4
- Dispatch to pre-scheduled op: fused_NT_matmul2_divide2_maximum1_minimum1_cast3
- Dispatch to pre-scheduled op: rms_norm
- Dispatch to pre-scheduled op: matmul10
- Dispatch to pre-scheduled op: fused_min_max_triu_te_broadcast_to
- Dispatch to pre-scheduled op: fused_softmax_cast1
- Dispatch to pre-scheduled op: fused_NT_matmul_divide_maximum_minimum_cast
╭─────────────────────────────── Traceback (most recent call last) ────────────────────────────────╮
│ /mnt/e/ubuntu_code/mlc-llm/build.py:221 in <module> │
│ │
│ 218 │ │ ) │
│ 219 │ │ mod = pickle.load(open(cache_path, "rb")) │
│ 220 │ dump_split_tir(mod) │
│ ❱ 221 │ build(mod, ARGS) │
│ 222 │
│ │
│ /mnt/e/ubuntu_code/mlc-llm/build.py:158 in build │
│ │
│ 155 │ │
│ 156 │ debug_dump_script(mod_deploy, "mod_build_stage.py", args) │
│ 157 │ │
│ ❱ 158 │ ex = relax.build(mod_deploy, args.target, system_lib=args.system_lib) │
│ 159 │ │
│ 160 │ output_filename = ( │
│ 161 │ │ f"{args.model}-{args.quantization.name}-{target_kind}.{args.lib_format}" │
│ │
│ /home/wj/work/relax/python/tvm/relax/vm_build.py:337 in build │
│ │
│ 334 │ builder = relax.ExecBuilder() │
│ 335 │ leftover_mod = _vmcodegen(builder, new_mod, exec_mode=exec_mode) │
│ 336 │ tir_mod = _filter_tir(leftover_mod) │
│ ❱ 337 │ return _vmlink(builder, target, tir_mod, ext_libs, params, system_lib=system_lib) │
│ 338 │
│ 339 │
│ 340 def _filter_tir(mod: tvm.IRModule) -> tvm.IRModule: │
│ │
│ /home/wj/work/relax/python/tvm/relax/vm_build.py:242 in _vmlink │
│ │
│ 239 │ │ ext_libs = [] │
│ 240 │ lib = None │
│ 241 │ if tir_mod is not None: │
│ ❱ 242 │ │ lib = tvm.build( │
│ 243 │ │ │ tir_mod, target=target, runtime=_autodetect_system_lib_req(target, system_li │
│ 244 │ │ ) │
│ 245 │ return Executable(_ffi_api.VMLink(builder, target, lib, ext_libs, params)) # type: │
│ │
│ /home/wj/work/relax/python/tvm/driver/build_module.py:281 in build │
│ │
│ 278 │ │
│ 279 │ annotated_mods, target_host = Target.canon_target_map_and_host(annotated_mods, targe │
│ 280 │ │
│ ❱ 281 │ rt_mod_host = _driver_ffi.tir_to_runtime(annotated_mods, target_host) │
│ 282 │ │
│ 283 │ annotated_mods, target_host = Target.canon_target_map_and_host(annotated_mods, targe │
│ 284 │
│ │
│ /home/wj/work/relax/python/tvm/_ffi/_ctypes/packed_func.py:238 in __call__ │
│ │
│ 235 │ │ │ ) │
│ 236 │ │ │ != 0 │
│ 237 │ │ ): │
│ ❱ 238 │ │ │ raise get_last_ffi_error() │
│ 239 │ │ _ = temp_args │
│ 240 │ │ _ = args │
│ 241 │ │ return RETURN_SWITCH[ret_tcode.value](ret_val) │
╰──────────────────────────────────────────────────────────────────────────────────────────────────╯
TVMError: Traceback (most recent call last):
11: TVMFuncCall
10: 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::{lambda(tvm::runtime::Map<tvm::Target, tvm::IRModule, void, void> const&, tvm::Target)#6}>(tvm::{lambda(tvm::runtime::Map<tvm::Target, tvm::IRModule, void, void>
const&, tvm::Target)#6}, 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*, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, tvm::runtime::TVMRetValue)
9: tvm::TIRToRuntime(tvm::runtime::Map<tvm::Target, tvm::IRModule, void, void> const&, tvm::Target const&)
8: tvm::codegen::Build(tvm::IRModule, tvm::Target)
7: tvm::runtime::PackedFuncObj::Extractor<tvm::runtime::PackedFuncSubObj<tvm::runtime::TypedPackedFunc<tvm::runtime::Module (tvm::IRModule, tvm::Target)>::AssignTypedLambda<tvm::runtime::Module
(*)(tvm::IRModule, tvm::Target)>(tvm::runtime::Module (*)(tvm::IRModule, tvm::Target), 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*)
6: tvm::codegen::BuildOpenCL(tvm::IRModule, tvm::Target)
5: tvm::runtime::OpenCLModuleCreate(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >,
std::unordered_map<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, tvm::runtime::FunctionInfo, std::hash<std::__cxx11::basic_string<char, std::char_traits<char>,
std::allocator<char> > >, std::equal_to<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > >, std::allocator<std::pair<std::__cxx11::basic_string<char, std::char_traits<char>,
std::allocator<char> > const, tvm::runtime::FunctionInfo> > >, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >)
4: tvm::runtime::OpenCLModuleNode::Init()
3: tvm::runtime::cl::OpenCLWorkspace::Init()
2: tvm::runtime::cl::OpenCLWorkspace::Init(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::__cxx11::basic_string<char, std::char_traits<char>,
std::allocator<char> > const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)
1: tvm::runtime::cl::GetPlatformIDs()
0: clGetPlatformIDs
File "/mnt/e/ubuntu_code/relax/src/runtime/opencl/opencl_wrapper/opencl_wrapper.cc", line 106
TVMError:
---------------------------------------------------------------
An error occurred during the execution of TVM.
For more information, please see: https://tvm.apache.org/docs/errors.html
---------------------------------------------------------------
Check failed: (m_libHandler != nullptr) is false: Error! Cannot open libOpenCL!
free(): invalid pointer
Aborted (core dumped)
I use wsl ubuntu environnment, and ubuntu does not have OpenCL.
But this model is build for Android running on a Android arm phone, do not need ubuntu host OpenCL runtime.
Could you please help check this ?
Thank you very much