bindings[0] exceed min ~ max range at index 3

[jihad-akl]

I converted a torch model with multiple input but it is not working

input_shape = (1, 3 , 32, 96)
input_tensor = torch.ones(input_shape).cuda()
print("creating engine")
engine = torch2trt(self.net, [input_tensor], min_shape= [(1, 3, 32, 32)],max_shape= [(1, 3, 960, 960)],fp16_mode= True)
print("engine done")
torch.save(engine.state_dict(), 'model.pth')

the model is created but in inference if i give a shape other than (1, 3 , 32, 96) i get an error:

[05/05/2023-07:45:37] [TRT] [E] 3: [executionContext.cpp::setBindingDimensions::966] Error Code 3: API Usage Error (Parameter check failed at: runtime/api/executionContext.cpp::setBindingDimensions::966, condition: profileMinDims.d[i] <= dimensions.d[i]. Supplied binding dimension [1,3,32,64] for bindings[0] exceed min ~ max range at index 3, maximum dimension in profile is 96, minimum dimension in profile is 96, but supplied dimension is 64.

[jihad-akl]

Update: putting s to min_shape and max_shape it result to an error:

creating engine
/opt/conda/lib/python3.8/site-packages/torch/nn/functional.py:3728: UserWarning: nn.functional.upsample is deprecated. Use nn.functional.interpolate instead.
warnings.warn("nn.functional.upsample is deprecated. Use nn.functional.interpolate instead.")
[05/05/2023-11:29:34] [TRT] [E] 3: [network.cpp::addPoolingNd::1025] Error Code 3: API Usage Error (Parameter check failed at: optimizer/api/network.cpp::addPoolingNd::1025, condition: allDimsGtEq(windowSize, 1) && volume(windowSize) < MAX_KERNEL_DIMS_PRODUCT(nbSpatialDims)
)
Traceback (most recent call last):
File "test_to_ocr_torch_tensorrt.py", line 10, in
model = TextSystem(args=ModelArgs(det_model_path=det_model_path, use_gpu=True,
File "/test/PytorchOCR/text_system.py", line 16, in init
self.text_detector = TextDetector(args, **kwargs)
File "/test/PytorchOCR/text_detector.py", line 270, in init
engine = torch2trt(self.net, [input_tensor], min_shapes= [(1, 3, 32, 32)], max_shapes= [(1, 3, 960, 960)],fp8_mode= True)
File "/opt/conda/lib/python3.8/site-packages/torch2trt-0.4.0-py3.8.egg/torch2trt/torch2trt.py", line 778, in torch2trt
outputs = module(*inputs)
File "/opt/conda/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1204, in _call_impl
result = forward_call(*input, **kwargs)
File "/test/PytorchOCR/modeling/architectures/base_model.py", line 88, in forward
x = self.neck(x)
File "/opt/conda/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1204, in _call_impl
result = forward_call(*input, **kwargs)
File "/test/PytorchOCR/modeling/necks/db_fpn.py", line 214, in forward
in5 = self.ins_conv3
File "/opt/conda/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1204, in _call_impl
result = forward_call(*input, **kwargs)
File "/test/PytorchOCR/modeling/necks/db_fpn.py", line 184, in forward
out = x + self.se_block(x)
File "/opt/conda/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1204, in _call_impl
result = forward_call(*input, **kwargs)
File "/test/PytorchOCR/modeling/backbones/det_mobilenet_v3.py", line 71, in forward
outputs = self.avg_pool(inputs)
File "/opt/conda/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1204, in _call_impl
result = forward_call(*input, **kwargs)
File "/opt/conda/lib/python3.8/site-packages/torch2trt-0.4.0-py3.8.egg/torch2trt/torch2trt.py", line 310, in wrapper
converter"converter"
File "/opt/conda/lib/python3.8/site-packages/torch2trt-0.4.0-py3.8.egg/torch2trt/converters/AdaptiveAvgPool2d.py", line 22, in convert_AdaptiveAvgPool2d
layer.stride = stride
AttributeError: 'NoneType' object has no attribute 'stride'

[vjixy]

I am facing the same issue