🐛 [Bug] Tutorial Using Torch-TensorRT Directly From PyTorch raises KeyError

[zshn25]

Bug Description

Running Tutorial code raises an error

To Reproduce

Steps to reproduce the behavior:

1. Visit https://pytorch.org/TensorRT/tutorials/use_from_pytorch.html
2. Run the codeblocks

---------------------------------------------------------------------------
KeyError                                  Traceback (most recent call last)
Cell In[16], line 5
      1 import torch
      2 import torch_tensorrt
      4 spec = {
----> 5     "forward": torch_tensorrt.ts.TensorRTCompileSpec(
      6         {
      7             "inputs": [torch_tensorrt.Input([1, 3, 480, 768])],
      8             "enabled_precisions": {torch.float, torch.half},
      9             "refit": False,
     10             "debug": False,
     11             "device": {
     12                 "device_type": torch_tensorrt.DeviceType.GPU,
     13                 "gpu_id": 0,
     14                 "dla_core": 0,
     15                 "allow_gpu_fallback": True,
     16             },
     17             "capability": torch_tensorrt.EngineCapability.default,
     18             "num_avg_timing_iters": 1,
     19         }
     20     )
     21 }
     22 scripted_model = torch.jit.script(model, input_image_pytorch)
     23 trt_model = torch._C._jit_to_backend("tensorrt", scripted_model, spec)

File ~/miniconda3/envs/inference/lib/python3.8/site-packages/torch_tensorrt/ts/_compile_spec.py:437, in TensorRTCompileSpec(inputs, input_signature, device, disable_tf32, sparse_weights, enabled_precisions, refit, debug, capability, num_avg_timing_iters, workspace_size, dla_sram_size, dla_local_dram_size, dla_global_dram_size, truncate_long_and_double, calibrator)
    380 """Utility to create a formated spec dictionary for using the PyTorch TensorRT backend
    381 
    382 Keyword Args:
   (...)
    415     torch.classes.tensorrt.CompileSpec: List of methods and formated spec objects to be provided to ``torch._C._jit_to_tensorrt``
    416 """
    418 compile_spec = {
    419     "inputs": inputs,
    420     # "input_signature": input_signature,
   (...)
    434     "truncate_long_and_double": truncate_long_and_double,
    435 }
--> 437 parsed_spec = _parse_compile_spec(compile_spec)
    439 backend_spec = torch.classes.tensorrt.CompileSpec()
    441 if input_signature is not None:

File ~/miniconda3/envs/inference/lib/python3.8/site-packages/torch_tensorrt/ts/_compile_spec.py:239, in _parse_compile_spec(compile_spec_)
    232 if len(compile_spec["inputs"]) > 0:
    233     if not all(
    234         [
    235             isinstance(i, torch.Tensor) or isinstance(i, Input)
    236             for i in compile_spec["inputs"]
    237         ]
    238     ):
--> 239         raise KeyError(
    240             "Input specs should be either torch_tensorrt.Input or torch.Tensor, found types: {}".format(
    241                 [type(i) for i in compile_spec["inputs"]]
    242             )
    243         )
    245     inputs = [
    246         Input.from_tensor(i) if isinstance(i, torch.Tensor) else i
    247         for i in compile_spec["inputs"]
    248     ]
    249     info.inputs = [i._to_internal() for i in inputs]

KeyError: "Input specs should be either torch_tensorrt.Input or torch.Tensor, found types: [<class 'str'>, <class 'str'>, <class 'str'>, <class 'str'>, <class 'str'>, <class 'str'>, <class 'str'>]"

Expected behavior

No error

Environment

Build information about Torch-TensorRT can be found by turning on debug messages

• Torch-TensorRT Version (e.g. 1.0.0):
• PyTorch Version (e.g. 1.0):
• CPU Architecture:
• OS (e.g., Linux):
• How you installed PyTorch (conda, pip, libtorch, source):
• Build command you used (if compiling from source):
• Are you using local sources or building from archives:
• Python version:
• CUDA version:
• GPU models and configuration:
• Any other relevant information:

Additional context