I have configured the environment, and the RPC can be used normally as follows,And tvm sample code android_rpc_test.py can run successfully.
But when I deploy ReID to android device using TVM, I get the following error.
/usr/bin/python3.8 /home/d516/wbj/project/android/tvm-android/reid/ReID_tvm_Android.py
One or more operators have not been tuned. Please tune your model for better performance. Use DEBUG logging level to see more details.
Traceback (most recent call last):
File "/home/d516/wbj/project/android/tvm-android/reid/ReID_tvm_Android.py", line 65, in <module>
module.run()
File "/home/d516/.local/lib/python3.8/site-packages/tvm-0.8.0-py3.8-linux-x86_64.egg/tvm/contrib/graph_executor.py", line 207, in run
self._run()
File "/home/d516/.local/lib/python3.8/site-packages/tvm-0.8.0-py3.8-linux-x86_64.egg/tvm/_ffi/_ctypes/packed_func.py", line 237, in __call__
raise get_last_ffi_error()
tvm._ffi.base.TVMError: Traceback (most recent call last):
4: TVMFuncCall
3: _ZNSt17_Function_handlerIFvN3tvm7runtime7TVMArgsEPNS1_11TVMR
2: tvm::runtime::RPCWrappedFunc::operator()(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
1: tvm::runtime::RPCClientSession::CallFunc(void*, TVMValue const*, int const*, int, std::function<void (tvm::runtime::TVMArgs)> const&)
0: tvm::runtime::RPCEndpoint::CallFunc(void*, TVMValue const*, int const*, int, std::function<void (tvm::runtime::TVMArgs)>)
File "/home/d516/wbj/project/android/tvm-0.8/src/runtime/rpc/rpc_endpoint.cc", line 801
TVMError:
---------------------------------------------------------------
An error occurred during the execution of TVM.
For more information, please see: https://tvm.apache.org/docs/errors.html
---------------------------------------------------------------
Check failed: (code == RPCCode::kReturn) is false: code=kShutdown
free(): invalid pointer
进程已结束，退出代码为 134 (interrupted by signal 6: SIGABRT)
This my code,I have marked the error code location in the code, and the error location is at “module.run()”
import argparse
import os
import sys
import tvm
import tvm.relay as relay
from tvm import rpc
from tvm.contrib import utils, ndk, graph_executor as runtime
from common import *
from models.model import BNNproAtt_IBN
path = sys.path[0]
parser = argparse.ArgumentParser('wbj_reid_tvm')
parser.add_argument('--channel', type=int, default=3)
parser.add_argument('--width', type=int, default=128)
parser.add_argument('--height', type=int, default=384)
parser.add_argument('--batchsize', type=int, default=32)
parser.add_argument('--device', type=str, default='cuda')
parser.add_argument('--torch_weights', type=str, default=path + '/weights/BNNproAtt0710.pt')
args = parser.parse_args()
if __name__ == '__main__':
model = BNNproAtt_IBN()
model.to(args.device)
model.load_state_dict(torch.load(args.torch_weights, map_location=args.device), False)
model.eval()
arch = "arm64"
target = tvm.target.Target("llvm -mtriple=%s-linux-android" % arch)
# ------------get tvm run lib-------------------------------
input_data, input_shape = tvm_pre_data(args.batchsize, args.channel, args.height, args.width, args.device)
scripted_model = torch.jit.trace(model, input_data).eval()
input_name = "input0"
shape_list = [(input_name, input_shape)]
mod, params = relay.frontend.from_pytorch(scripted_model, shape_list)
with tvm.transform.PassContext(opt_level=3):
lib = relay.build(mod, target=target, params=params)
tmp = utils.tempdir()
lib_fname = tmp.relpath("net.so")
local_demo = False
fcompile = ndk.create_shared if not local_demo else None
lib.export_library(lib_fname, fcompile)
# 通过RPC部署安卓设备
# tracker_host = os.environ.get("TVM_TRACKER_HOST", "127.0.0.1")
# tracker_port = int(os.environ.get("TVM_TRACKER_POST", 9190))
tracker_host = "0.0.0.0"
tracker_port = int(9190)
key = "android"
tracker = rpc.connect_tracker(tracker_host, tracker_port)
remote = tracker.request(key, priority=0, session_timeout=60)
dev = remote.cpu(0)
remote.upload(lib_fname)
rlib = remote.load_module("net.so")
module = runtime.GraphModule(rlib["default"](dev))
# 执行模型推理测试
input_test = tvm.nd.array((np.random.uniform(size=input_shape)).astype("float32"))
module.set_input(input_name, input_test)
# !!!!!!!!!!!!!!!!!!!!this is error code location!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!1
module.run()
print("**Evaluate inference time cost...**")
print(module.benchmark(dev, number=1, repeat=3))
So does anyone know how to fix it?
