[Improvements] Support TorchAllocator as TensorRT Gpu Allocator

[grimoire]

Thanks for your contribution and we appreciate it a lot. The following instructions would make your pull request more healthy and more easily receiving feedbacks. If you do not understand some items, don't worry, just make the pull request and seek help from maintainers.

Motivation

PyTorch memory pool wastes lots of memories. The PR share the pool with TensorRT through custom IGpuAllocator.

a = torch.rand(large_memory_size, device='cuda')

del a

# torch would not release the pool, cudaMalloc might failed with OOM.
print(torch.cuda.memory_summary())

model = TRTWrapper(engine_path)

x = torch.rand(1, 3, 224, 224, device='cuda')
out = model(dict(input=x))

print(torch.cuda.memory_summary())

Modification

• Custom IGpuAllocator wraps torch.cuda.caching_allocator_xxx.
• TRTWrapper with custom allocator.
• Fix bugs of MDCN

Got an error when applying the allocator on the Builder.

BC-breaking (Optional)

Does the modification introduce changes that break the backward-compatibility of the downstream repositories?
If so, please describe how it breaks the compatibility and how the downstream projects should modify their code to keep compatibility with this PR.

Use cases (Optional)

If this PR introduces a new feature, it is better to list some use cases here, and update the documentation.

Checklist

1. Pre-commit or other linting tools are used to fix the potential lint issues.
2. The modification is covered by complete unit tests. If not, please add more unit tests to ensure the correctness.
3. If the modification has a dependency on downstream projects of a newer version, this PR should be tested with all supported versions of downstream projects.
4. The documentation has been modified accordingly, like docstring or example tutorials.

[AllentDan]

LGTM

[CLAassistant]

All committers have signed the CLA.

[lzhangzz]

LGTM