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

[Convert] fix dimension mismatch issue in qcdq2qonnx conversion #91

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

[Convert] fix dimension mismatch issue in qcdq2qonnx conversion #91

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
35 changes: 23 additions & 12 deletions src/qonnx/transformation/qcdq_to_qonnx.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -109,6 +109,7 @@ def apply(self, model: ModelWrapper) -> Tuple[ModelWrapper, bool]:
dq_init = model.get_initializer(dq_inp)
dq_scale_v = model.get_initializer(dq_scale)
dq_zeropt_v = model.get_initializer(dq_zeropt)
axis = get_by_name(dq_node.attribute, "axis")
if quant_candidates is None and dq_init is None:
continue
if any([x is None for x in [dq_scale_v, dq_zeropt_v]]):
Expand All @@ -123,13 +124,24 @@ def apply(self, model: ModelWrapper) -> Tuple[ModelWrapper, bool]:
# read quantized weight dtype for standalone deqnt
q_vi = model.get_tensor_valueinfo(dq_inp)
(bitwidth, signed, narrow) = extract_elem_type(q_vi.type.tensor_type.elem_type)
scale_factor, zeropt = dq_scale, dq_zeropt
# fix scale factor for Quant (different shape expectations wrt broadcasting)
if not (axis is None):
axis_i = axis.i
ishape = model.get_tensor_shape(dq_inp)
desired_shp = [1] * len(ishape)
desired_shp[axis_i] = dq_scale_v.shape[0]
dq_scale_v = dq_scale_v.reshape(desired_shp)
dq_zeropt_v = dq_zeropt_v.reshape(desired_shp)
model.set_initializer(scale_factor, dq_scale_v)
model.set_initializer(zeropt, dq_zeropt_v)
# overwrite DQ initializer with scaled version
scaled_qnt_t = (dq_init - dq_zeropt_v) * dq_scale_v
scaled_qnt_t = scaled_qnt_t.astype(np.float32)
model.set_initializer(dq_inp, scaled_qnt_t)
q_inp = dq_inp
final_out = dq_node.output[0]
scale_factor, zeropt = dq_scale, dq_zeropt

nodes_to_remove.append(dq_node)
elif quant_candidates[0].op_type in ["QuantizeLinear", "Clip"]:
clip_range = None
Expand Down Expand Up @@ -167,20 +179,19 @@ def apply(self, model: ModelWrapper) -> Tuple[ModelWrapper, bool]:
value_info = model.get_tensor_valueinfo(quant_node.output[0])
(bitwidth, signed, narrow) = extract_elem_type(value_info.type.tensor_type.elem_type, clip_range)
scale_factor, zeropt = q_scale, q_zeropt
if not (axis is None):
axis_i = axis.i
ishape = model.get_tensor_shape(dq_inp)
desired_shp = [1] * len(ishape)
desired_shp[axis_i] = dq_scale_v.shape[0]
dq_scale_v = dq_scale_v.reshape(desired_shp)
dq_zeropt_v = dq_zeropt_v.reshape(desired_shp)
model.set_initializer(scale_factor, dq_scale_v)
model.set_initializer(zeropt, dq_zeropt_v)
else:
# handle all other cases, skip
continue
axis = get_by_name(dq_node.attribute, "axis")
# fix scale factor for Quant (different shape expectations wrt broadcasting)
if not (axis is None):
axis_i = axis.i
ishape = model.get_tensor_shape(dq_inp)
desired_shp = [1] * len(ishape)
desired_shp[axis_i] = dq_scale_v.shape[0]
dq_scale_v = dq_scale_v.reshape(desired_shp)
dq_zeropt_v = dq_zeropt_v.reshape(desired_shp)
model.set_initializer(scale_factor, dq_scale_v)
model.set_initializer(zeropt, dq_zeropt_v)

# create new Quant node for suitable cases
new_q_node_name = "Quant_" + q_inp
bw_tensor_name = f"{new_q_node_name}_bitwidth"
Expand Down
Loading