🐛 [Bug] Unknown output type. Only a single tensor or a TensorList type is supported.

[kiddyjinjin]

Bug Description

I meet some error msg when trying to trans a fastbert model using trtorch:

^[[1;35mDEBUG: ^[[0m[TRTorch Conversion Context] - Evaluating %4 : Tensor?[] = prim::ListConstruct(%rel_diff_idxs.13)
^[[1;35mDEBUG: ^[[0m[TRTorch Conversion Context] - Found the value to be: [<torch.torch.classes._trtorch_eval_ivalue_types.TensorContainer object at 0x876f0f10>]
terminate called after throwing an instance of 'trtorch::Error'
what(): [enforce fail at core/conversion/conversion.cpp:196] Unknown output type. Only a single tensor or a TensorList type is supported.

Expected behavior

support different kinds of graph output. not only tensor and tensorlist.

Environment

Build information about the TRTorch compiler can be found by turning on debug messages

• TRTorch Version (e.g. 0.2.0): latest master version
• PyTorch Version (e.g. 1.0): 1.8.1
• CPU Architecture:
• OS (e.g., Linux): linux centos6u3
• How you installed PyTorch (conda, pip, libtorch, source): source
• Build command you used (if compiling from source):
• Are you using local sources or building from archives: local sources
• Python version: python 3.6.5
• CUDA version: cuda 10.2
• GPU models and configuration: fastbert
• Any other relevant information:

[letian-jiang]

I met the same error when I tried to compile torchvision.models.inception_v3().

[github-actions]

This issue has not seen activity for 90 days, Remove stale label or comment or this will be closed in 10 days