🐛 [Bug] MutableTorchTensorRTModule.save error out in low_vram_mode

Bug Description

python examples/apps/flux_demo.py --dtype fp16 --low_vram_mode --load_or_save save

WARNING:torch_tensorrt.dynamo.runtime._MutableTorchTensorRTModule:Trying to move the original PyTorch model. This will cause CPU offloading failing and increase GPU memory usage.If this is absolute necessary, please call module.pytorch_model.to(...)
The model is still on the original device.
Traceback (most recent call last):
File "C:\Users\local-lanl\git\py310\TensorRT\examples\apps\flux_demo.py", line 319, in
main(args)
File "C:\Users\local-lanl\git\py310\TensorRT\examples\apps\flux_demo.py", line 270, in main
torch_tensorrt.MutableTorchTensorRTModule.save(trt_gm, "mutable_trt_gm.pkl")
File "C:\Users\local-lanl\git\venv_py310\lib\site-packages\torch_tensorrt\dynamo\runtime_MutableTorchTensorRTModule.py", line 709, in save
torch.save(module, path, pickle_protocol=4)
File "C:\Users\local-lanl\git\venv_py310\lib\site-packages\torch\serialization.py", line 967, in save
_save(
File "C:\Users\local-lanl\git\venv_py310\lib\site-packages\torch\serialization.py", line 1213, in _save
pickler.dump(obj)
MemoryError

To Reproduce

Steps to reproduce the behavior:

Expected behavior

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

Start with examples/apps/flux_demo.py around lines 270 and 319, then inspect torch_tensorrt/dynamo/runtime/_MutableTorchTensorRTModule.py around line 709 where save calls torch.save. Reproduce the command with low_vram_mode and compare the module state at save time. Done means saving completes without MemoryError and the saved module remains usable.