Skip to content

decrease memory use by remove unnecessary module inference #631

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: master
Choose a base branch
from
Open

decrease memory use by remove unnecessary module inference #631

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
d86a0ea
decrease memory use by remove unnecessary module inference
wwqgtxx Sep 20, 2021
d68b3e0
delete and empty cuda cache before build_cuda_engine to provide more …
wwqgtxx Sep 21, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 13 additions & 9 deletions torch2trt/torch2trt.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -522,18 +522,16 @@ def torch2trt(module,
inputs = tuple(inputs)
if not isinstance(inputs, tuple):
inputs = (inputs,)

# run once to get num outputs
outputs = module(*inputs)
if not isinstance(outputs, tuple) and not isinstance(outputs, list):
outputs = (outputs,)

if input_names is None:
input_names = default_input_names(len(inputs))
if output_names is None:
output_names = default_output_names(len(outputs))


if use_onnx:
# run once to get num outputs
outputs = module(*inputs)
if not isinstance(outputs, tuple) and not isinstance(outputs, list):
outputs = (outputs,)
if output_names is None:
output_names = default_output_names(len(outputs))

f = io.BytesIO()
torch.onnx.export(module, inputs, f, input_names=input_names, output_names=output_names)
Expand All @@ -553,6 +551,8 @@ def torch2trt(module,

if not isinstance(outputs, tuple) and not isinstance(outputs, list):
outputs = (outputs,)
if output_names is None:
output_names = default_output_names(len(outputs))
ctx.mark_outputs(outputs, output_names)

builder.max_workspace_size = max_workspace_size
Expand All @@ -575,7 +575,11 @@ def torch2trt(module,
inputs, int8_calib_dataset, batch_size=int8_calib_batch_size, algorithm=int8_calib_algorithm
)

del inputs
del outputs
torch.cuda.empty_cache()
engine = builder.build_cuda_engine(network)
torch.cuda.empty_cache()

module_trt = TRTModule(engine, input_names, output_names)

Expand Down