Skip to content

Remove reduce_range as it is not relevant for HTP #14559

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

Merged
merged 2 commits into from
Sep 26, 2025
Merged

Remove reduce_range as it is not relevant for HTP #14559

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
291121c
Enable quantization for bf16 model (#14558)
navsud Sep 24, 2025
07e6a73
Remove reduce_range as it is not relevant for HTP
navsud Sep 24, 2025
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
2 changes: 1 addition & 1 deletion backends/qualcomm/quantizer/annotators.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -68,7 +68,7 @@ def _is_float_tensor(node: Node):
or not isinstance(node.meta["val"], FakeTensor)
):
return False
return node.meta["val"].dtype == torch.float32
return node.meta["val"].dtype in (torch.bfloat16, torch.float32)


def _mark_nodes_as_annotated(nodes: List[Node]):
Expand Down
9 changes: 0 additions & 9 deletions backends/qualcomm/quantizer/qconfig.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -205,7 +205,6 @@ def get_16a8w_qnn_qat_config(
quant_min=torch.iinfo(torch.uint16).min,
quant_max=torch.iinfo(torch.uint16).max,
qscheme=torch.per_tensor_affine,
reduce_range=True,
observer=act_observer.with_args(**extra_args),
)
act_quantization_spec = QuantizationSpec(
Expand All @@ -220,7 +219,6 @@ def get_16a8w_qnn_qat_config(
quant_min=torch.iinfo(torch.int8).min + 1,
quant_max=torch.iinfo(torch.int8).max,
qscheme=torch.per_tensor_symmetric,
reduce_range=True,
observer=MovingAverageMinMaxObserver,
)
weight_quantization_spec = QuantizationSpec(
Expand Down Expand Up @@ -421,7 +419,6 @@ def get_8a8w_qnn_qat_config(
quant_min=torch.iinfo(torch.int8).min + 1,
quant_max=torch.iinfo(torch.int8).max,
qscheme=torch.per_tensor_symmetric,
reduce_range=True,
observer=MovingAverageMinMaxObserver,
)
weight_quantization_spec = QuantizationSpec(
Expand All @@ -438,7 +435,6 @@ def get_8a8w_qnn_qat_config(
quant_min=torch.iinfo(torch.int32).min,
quant_max=torch.iinfo(torch.int32).max,
qscheme=torch.per_tensor_symmetric,
reduce_range=True,
observer=MovingAverageMinMaxObserver,
)
bias_quantization_spec = QuantizationSpec(
Expand Down Expand Up @@ -467,7 +463,6 @@ def get_16a4w_qnn_qat_config(
quant_min=torch.iinfo(torch.uint16).min,
quant_max=torch.iinfo(torch.uint16).max,
qscheme=torch.per_tensor_affine,
reduce_range=True,
observer=act_observer,
)
act_quantization_spec = QuantizationSpec(
Expand All @@ -484,7 +479,6 @@ def get_16a4w_qnn_qat_config(
quant_max=7,
qscheme=torch.per_tensor_symmetric,
ch_axis=0,
reduce_range=True,
observer=MovingAverageMinMaxObserver,
)
weight_quantization_spec = QuantizationSpec(
Expand All @@ -501,7 +495,6 @@ def get_16a4w_qnn_qat_config(
quant_min=torch.iinfo(torch.int32).min,
quant_max=torch.iinfo(torch.int32).max,
qscheme=torch.per_tensor_symmetric,
reduce_range=True,
observer=MovingAverageMinMaxObserver,
)
bias_quantization_spec = QuantizationSpec(
Expand Down Expand Up @@ -551,7 +544,6 @@ def get_qat_per_channel_quant_config(
act_fake_quant_ctr = FakeQuantize.with_args(
dtype=torch.int32 if act_dtype == torch.uint16 else act_dtype,
qscheme=torch.per_tensor_symmetric,
reduce_range=True,
observer=act_observer,
)
act_quantization_spec = QuantizationSpec(
Expand All @@ -566,7 +558,6 @@ def get_qat_per_channel_quant_config(
quant_min=torch.iinfo(act_dtype).min,
quant_max=torch.iinfo(act_dtype).max,
qscheme=torch.per_tensor_affine,
reduce_range=True,
observer=act_observer,
)
act_quantization_spec = QuantizationSpec(
Expand Down
Loading