Skip to content
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

Jump to bottom

DLM gradients #161

Merged
merged 23 commits into from
Jun 7, 2022
Merged

DLM gradients #161

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
23 commits
Select commit Hold shift + click to select a range
f9b5121
DLM gradients hacky example
rtqichen Apr 7, 2022
03060dc
Merge remote-tracking branch 'origin/main' into rtqichen.dlm_gradients
rtqichen Apr 12, 2022
de08895
implement DLM using autograd.Function
rtqichen Apr 13, 2022
b7ce470
make soln a bit more accurate
rtqichen Apr 13, 2022
8741e37
minor; removed unnecessary code
rtqichen Apr 13, 2022
5f8b7e1
lower case for dlm_epsilon
rtqichen Apr 22, 2022
a807f83
backward test for DLM
rtqichen Apr 22, 2022
915fdca
fix imports
rtqichen Apr 22, 2022
5b37ab4
rename and make linter happy
rtqichen Apr 22, 2022
9ddd124
filter for tensors that require grad
rtqichen Apr 22, 2022
eb5a365
Construct the bwd objective only once
rtqichen Apr 22, 2022
3cddda4
minor
rtqichen Apr 22, 2022
c8d57f1
remove print statements
rtqichen Apr 24, 2022
f587cd9
Fix DLM when using gpu; cost function shape; and handle case when no …
rtqichen Apr 24, 2022
141525f
Fix memory leak by removing dict input_data from input arguments
rtqichen Apr 25, 2022
29355ab
preserve ordering
rtqichen Apr 25, 2022
c12dde6
Expand batch dim if possible
rtqichen Apr 25, 2022
6555108
undo
rtqichen Apr 25, 2022
f199719
use lower case
rtqichen Apr 26, 2022
b512d56
Merge remote-tracking branch 'origin/main' into rtqichen.dlm_gradients
rtqichen Apr 26, 2022
d3dff3c
Merge remote-tracking branch 'origin/main' into rtqichen.dlm_gradients
rtqichen May 31, 2022
c0679d5
reduce a bit of python overhead
rtqichen May 31, 2022
18feb22
explicit one step
rtqichen May 31, 2022
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
Prev Previous commit
Next Next commit
lower case for dlm_epsilon
  • Loading branch information
@rtqichen
rtqichen committed Apr 22, 2022
commit 5f8b7e1038cc858c9ab70ab2109a57ccbad28a76
2 changes: 1 addition & 1 deletion examples/backward_modes.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -221,7 +221,7 @@ def fit_x(data_x_np):
"track_best_solution": True,
"verbose": False,
"backward_mode": th.BackwardMode.DLM,
"DLM_epsilon": 1e-2,
"dlm_epsilon": 1e-2,
},
)
start = time.time()
Expand Down
4 changes: 2 additions & 2 deletions theseus/theseus_layer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,13 +40,13 @@ def forward(
)
optimizer_kwargs = optimizer_kwargs or {}
backward_mode = optimizer_kwargs.get("backward_mode", None)
DLM_epsilon = optimizer_kwargs.get("DLM_epsilon", 1e-2)
dlm_epsilon = optimizer_kwargs.get("dlm_epsilon", 1e-2)
if backward_mode == BackwardMode.DLM:
# TODO: instantiate self.bwd_objective here.
names = set(self.objective.aux_vars.keys()).intersection(input_data.keys())
rtqichen marked this conversation as resolved.
Show resolved Hide resolved
tensors = [input_data[n] for n in names]
*vars, info = TheseusLayerDLMForward.apply(
self.objective, self.optimizer, optimizer_kwargs, input_data, DLM_epsilon, *tensors
self.objective, self.optimizer, optimizer_kwargs, input_data, dlm_epsilon, *tensors
)
else:
vars, info = _forward(self.objective, self.optimizer, optimizer_kwargs, input_data)
Expand Down