trending/Model-Optimizer
1
0
Fork 0
mirror of https://github.com/NVIDIA/Model-Optimizer.git synced 2026-11-13 02:17:13 +08:00
Code Issues Packages Projects Releases Wiki Activity

Added exception for warning caused while creating int4 tensor (#461)

Browse Source 
Signed-off-by: Hrishith Thadicherla <hthadicherla@nvidia.com>
This commit is contained in:
Hrishith Thadicherla
2025-10-28 11:14:42 +05:30
committed by GitHub
parent 41de55ff3e
commit 14fa1e5759
1 changed files with 8 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modelopt/onnx/quantization/gs_patching.py
+8 -1
View File
@@ -39,7 +39,14 @@ def _make_constant(
converted_dtype = (
dtype if isinstance(values, LazyValues) else onnx.helper.tensor_dtype_to_np_dtype(dtype)
)
if values.dtype != converted_dtype:
# Allow int8/uint8 as intermediate representation for INT4/UINT4
# INT4/UINT4 values are stored as int8/uint8 in numpy arrays and packed during export
is_valid_int4_intermediate = (dtype == onnx.TensorProto.INT4 and values.dtype == np.int8) or (
dtype == onnx.TensorProto.UINT4 and values.dtype == np.uint8
)
if not is_valid_int4_intermediate and values.dtype != converted_dtype:
logger.error(
f"Trying to create tensor with incompatible types: `{values.dtype}`, `{dtype}`"
)