I was trying to deploy the FSRCNN model（https://github.com/yjn870/FSRCNN-pytorch） on CPU/GPU/VTA using TVM. To be precise:
#load model
onnx_model = onnx.load('fsrcnn.onnx')
#load input image
from PIL import Image
img_url = "https://github.com/dmlc/mxnet.js/blob/main/data/cat.png?raw=true"
img_path = download_testdata(img_url, "cat.png", module="data")
img = Image.open(img_path)
img_rgb = img.convert("RGB")
image_width = (img.width // 3) * 3
image_height = (img.height // 3) * 3
hr = img.resize((image_width, image_height), resample=Image.BICUBIC)
lr = hr.resize((hr.width // 3, hr.height // 3), resample=Image.BICUBIC)
bicubic = lr.resize((lr.width * 3, lr.height * 3), resample=Image.BICUBIC)
lr, _ = preprocess(lr, 'cpu')
hr, _ = preprocess(hr, 'cpu')
_, ycbcr = preprocess(bicubic, 'cpu')
#compile the model with relay
target = "llvm"
input_name = "0"
print (lr.shape)
shape_dict = {input_name:lr.shape}
mod, params = relay.frontend.from_onnx(onnx_model, shape_dict)
I got the errors：
torch.Size([1, 1, 85, 85])
Incompatible broadcast type TensorType([56], float32) and TensorType([1, 56, 85, 85], float32)
Traceback (most recent call last):
File "deploy_fsrcnn_simple.py", line 41, in <module>
mod, params = relay.frontend.from_onnx(onnx_model, shape_dict)
File "/home/tvm/python/tvm/relay/frontend/onnx.py", line 5745, in from_onnx
mod, params = g.from_onnx(graph, opset)
File "/home/tvm/python/tvm/relay/frontend/onnx.py", line 5415, in from_onnx
self._construct_nodes(graph)
File "/home/tvm/python/tvm/relay/frontend/onnx.py", line 5527, in _construct_nodes
op = self._convert_operator(op_name, inputs, attr, self.opset)
File "/home/tvm/python/tvm/relay/frontend/onnx.py", line 5638, in _convert_operator
sym = convert_map[op_name](inputs, attrs, self._params)
File "/home/tvm/python/tvm/relay/frontend/onnx.py", line 599, in _impl_v1
input_shape = infer_shape(data)
File "/home/tvm/python/tvm/relay/frontend/common.py", line 526, in infer_shape
out_type = infer_type(inputs, mod=mod)
File "/home/tvm/python/tvm/relay/frontend/common.py", line 501, in infer_type
new_mod = _transform.InferType()(new_mod)
File "/home/tvm/python/tvm/ir/transform.py", line 161, in __call__
return _ffi_transform_api.RunPass(self, mod)
File "/home/tvm/python/tvm/_ffi/_ctypes/packed_func.py", line 237, in __call__
raise get_last_ffi_error()
tvm.error.DiagnosticError: Traceback (most recent call last):
6: TVMFuncCall
5: 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<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)
4: tvm::transform::Pass::operator()(tvm::IRModule) const
3: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
2: tvm::transform::ModulePassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
1: _ZN3tvm7runtime13PackedFuncObj9ExtractorINS0_16PackedFuncSubObjIZNS0_15TypedPackedFuncIFNS_8IRModuleES5_NS_9transform11PassContextEEE17AssignTypedLambdaIZNS_5relay9transform9InferTypeEvEUlS5_RKS7_E_EEvT_EUlRKNS0_7TVMArgsEPNS0_11TVMRetValueEE_EEE4CallEPKS1_SH_SL_
0: tvm::DiagnosticContext::Render()
File "/home/tvm/src/ir/diagnostic.cc", line 105
DiagnosticError: one or more error diagnostics were emitted, please check diagnostic render for output.
The TensorType([56], float32) and TensorType([1, 56, 85, 85], float32) comes from the output of the first Conv Layer and the PRelu：
anybody have any idea to fix this bug？any help will be appreciated！