Skip to content

[Draft] Qualcomm AI Engine Direct - Unexpected graph for mutable buffer after export during Quantization #11309

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

Draft
wants to merge 2 commits into
base: main
Choose a base branch
from
+19 −2
Draft

[Draft] Qualcomm AI Engine Direct - Unexpected graph for mutable buffer after export during Quantization #11309

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ef8594a
draw graph for export and export with patch
shewu-quic Jun 3, 2025
5985058
draw graph after convert pt2e
shewu-quic Jun 3, 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
21 changes: 19 additions & 2 deletions backends/qualcomm/tests/utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,6 +51,7 @@
prepare_pt2e,
prepare_qat_pt2e,
)
from unittest.mock import patch


def generate_context_binary(
Expand Down Expand Up @@ -516,9 +517,20 @@ def get_qdq_module(
block_size_map: Dict[str, Tuple] = None,
submodule_qconfig_list: Optional[List[Tuple[Callable, ModuleQConfig]]] = None,
) -> torch.fx.GraphModule:
from executorch.backends.qualcomm.utils.utils import draw_graph
with patch.object(
torch._utils_internal,
"export_training_ir_rollout_check",
return_value=False,
):
m_with_patch = torch.export.export(
module, inputs, dynamic_shapes=dynamic_shapes, strict=False
).module()
draw_graph("export_with_patch", ".", m_with_patch)
m = torch.export.export(
module, inputs, dynamic_shapes=dynamic_shapes, strict=True
).module()
module, inputs, dynamic_shapes=dynamic_shapes, strict=False
).module()
draw_graph("export", ".", m)

quantizer = make_quantizer(
quant_dtype=quant_dtype,
Expand All @@ -529,6 +541,9 @@ def get_qdq_module(
)
if block_size_map is not None:
quantizer.set_block_size_map(block_size_map)
prepared_with_patch = prepare_pt2e(m_with_patch, quantizer)
prepared_with_patch(*inputs)
quantized_module_with_patch = convert_pt2e(prepared_with_patch)
prepared = prepare_pt2e(m, quantizer)
prepared(*inputs)
quantized_module = convert_pt2e(prepared)
Expand All @@ -543,6 +558,8 @@ def get_qdq_module(
}
if not bypass_check:
self.assertTrue(nodes.intersection(q_and_dq))
draw_graph("convert_pt2e_with_patch", ".", quantized_module_with_patch)
draw_graph("convert_pt2e", ".", quantized_module)
return quantized_module

def get_prepared_qat_module(
Expand Down
Loading