Skip to content

bugfix: avoid duplicate RTN quantization outside blocks #2134

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

Closed
WeiweiZhang1 wants to merge 1 commit into from
+0 −13
Closed

bugfix: avoid duplicate RTN quantization outside blocks #2134

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
13 changes: 0 additions & 13 deletions auto_round/algorithms/quantization/base.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -198,19 +198,6 @@ def quantize_layer_outside_block(
the valid-token loss mask. ``None`` for RTN fallback.
"""
self._quantize_layer_via_rtn(layer, disable_opt_rtn=disable_opt_rtn)
"""Quantize a single layer outside a transformer block using RTN fallback.
Args:
layer: The layer module to quantize. Must have a
``global_name`` attribute for model re-insertion.
fp_inputs: Optional FP calibration inputs; unused in base RTN.
q_inputs: Optional quantized activations; unused in base RTN.
disable_opt_rtn: ``True`` skips optimized-RTN scale/zp search.
``None`` defers to ``self.config.disable_opt_rtn``.
valid_token_mask: Per-sample masks; unused in base RTN.
input_ids: Original FP calibration inputs, same as ``fp_inputs``
when ``q_inputs`` is ``None``; unused in base RTN.
"""
self._quantize_layer_via_rtn(layer, disable_opt_rtn=disable_opt_rtn)

@torch.no_grad()
def _quantize_layer_via_rtn(self, layer: "torch.nn.Module", disable_opt_rtn: "bool | None" = None) -> None:
Expand Down
Loading