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

Fix (graph/bias_correction): Fix when layer parameters are offloaded to accelerate #962

Merged
merged 4 commits into from
Jul 8, 2024
Merged

Fix (graph/bias_correction): Fix when layer parameters are offloaded to accelerate #962

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
fb9b257
Fix (graph/bias_correction): Fix when layer parameters are offloaded …
nickfraser May 29, 2024
d264a7c
Fix (bias_correction): Typo fix
nickfraser May 30, 2024
930296e
Fix (bias_correction): Apply accelerate fix to entire `if/elif` block.
nickfraser May 31, 2024
b1d7d5a
fix (bias_corr/accelerate): Added comment
nickfraser Jul 8, 2024
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
7 changes: 7 additions & 0 deletions src/brevitas/graph/calibrate.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -286,11 +286,18 @@ def apply_correction(self, model):
for name, module in model.named_modules():
if name in self.correction_map.keys():
correction = self.correction_map[name] / self.iterations[name]
# When accelerate is enabled, bring tensors onto the device to avoid allocating a meta parameter.
if hasattr(module, 'allocate_params'):
module.allocate_params(module)
if module.bias is not None:
module.bias.data += correction
elif self.skip_if_no_bias is False:
# If accelerate is enabled, bias will be on the same execution device as the weights, but won't be managed properly by accelerate
module.register_parameter(
'bias', nn.Parameter(correction).to(module.weight.device))
# Offload params again
if hasattr(module, 'offload_params'):
module.offload_params(module)

def compute_correct_bias(self, module, inp, name):
inp = self.unpack_input(inp)
Expand Down
Loading