ModuleNotFoundError: No module named 'onnx.mapping'

Error: The console stream is logged into /root/sg_logs/console.log
/content/myenv/lib/python3.10/site-packages/super_gradients/common/environment/cfg_utils.py:6: UserWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html. The pkg_resources package is slated for removal as early as 2025-11-30. Refrain from using this package or pin to Setuptools<81.
import pkg_resources
[2025-10-29 20:26:40] INFO - crash_tips_setup.py - Crash tips is enabled. You can set your environment variable to CRASH_HANDLER=FALSE to disable it
[2025-10-29 20:26:47] INFO - checkpoint_utils.py - Successfully loaded model weights from /content/myenv/checkpoints/my_first_yolonas_run/RUN_20251029_185759_477436/ckpt_best.pth EMA checkpoint.
[2025-10-29 20:27:10] INFO - unused_removal.py - Removed 986 unused nodes
[2025-10-29 20:27:10] INFO - unused_removal.py - No unused functions to remove
[2025-10-29 20:27:10] INFO - _constant_folding.py - Skipping constant folding for node 'node__to_copy' because it is graph input to preserve graph signature
[2025-10-29 20:27:11] INFO - _constant_folding.py - Skipping constant folding for node Node(name='node_Unsqueeze_835', domain='', op_type='Unsqueeze', inputs=(Value(name='expand_1', type=Tensor(FLOAT), shape=Shape([128, 160]), producer='node_Constant_1518', index=0, const_value={Tensor(...)}), SymbolicTensor(name='val_837', type=Tensor(INT64), shape=Shape([1]), producer='node_Constant_834', index=0, const_value={Tensor<INT64,[1]>(array([-1]), name='val_837')})), attributes={}, overload='', outputs=(SymbolicTensor(name='val_838', type=Tensor(FLOAT), shape=Shape([128, 160, 1]), producer='node_Unsqueeze_835', index=0),), version=20, doc_string=None) due to large input sizes: [20480, 1]
[2025-10-29 20:27:11] INFO - _constant_folding.py - Skipping constant folding for node Node(name='node_Unsqueeze_837', domain='', op_type='Unsqueeze', inputs=(Value(name='expand', type=Tensor(FLOAT), shape=Shape([128, 160]), producer='node_Constant_1514', index=0, const_value={Tensor(...)}), SymbolicTensor(name='val_839', type=Tensor(INT64), shape=Shape([1]), producer='node_Constant_836', index=0, const_value={Tensor<INT64,[1]>(array([-1]), name='val_839')})), attributes={}, overload='', outputs=(SymbolicTensor(name='val_840', type=Tensor(FLOAT), shape=Shape([128, 160, 1]), producer='node_Unsqueeze_837', index=0),), version=20, doc_string=None) due to large input sizes: [20480, 1]
[2025-10-29 20:27:11] INFO - _constant_folding.py - Skipping constant folding for node Node(name='node_view_14', domain='', op_type='Reshape', inputs=(Value(name='stack_1', type=Tensor(FLOAT), shape=Shape([64, 80, 2]), producer='node_Constant_1544', index=0, const_value={Tensor(...)}), Value(name='val_880', type=Tensor(INT64), shape=Shape([2]), producer='node_Constant_1545', index=0, const_value={Tensor<INT64,[2]>(array([-1, 2]), name='val_880')})), attributes={'allowzero': Attr('allowzero', INT, 1)}, overload='', outputs=(SymbolicTensor(name='view_14', type=Tensor(FLOAT), shape=Shape([5120, 2]), producer='node_view_14', index=0),), version=20, doc_string=None) due to large input sizes: [10240, 2]
[2025-10-29 20:27:11] INFO - _constant_folding.py - Skipping constant folding for node Node(name='node_cat_17', domain='', op_type='Concat', inputs=(Value(name='full', type=Tensor(FLOAT), shape=Shape([20480, 1]), producer='node_Constant_1523', index=0, const_value={Tensor(...)}), Value(name='full_1', type=Tensor(FLOAT), shape=Shape([5120, 1]), producer='node_Constant_1549', index=0, const_value={Tensor(...)}), Value(name='full_2', type=Tensor(FLOAT), shape=Shape([1280, 1]), producer='node_Constant_1576', index=0, const_value={Tensor(...)})), attributes={'axis': Attr('axis', INT, 0)}, overload='', outputs=(SymbolicTensor(name='cat_17', type=Tensor(FLOAT), shape=Shape([26880, 1]), producer='node_cat_17', index=0),), version=20, doc_string=None) due to large input sizes: [20480, 5120, 1280]
[2025-10-29 20:27:13] INFO - unused_removal.py - Removed 617 unused nodes
Applied 50 of general pattern rewrite rules.
[2025-10-29 20:27:13] INFO - unused_removal.py - No unused functions to remove
[2025-10-29 20:27:13] INFO - _constant_folding.py - Skipping constant folding for node 'node__to_copy' because it is graph input to preserve graph signature
[2025-10-29 20:27:13] INFO - _constant_folding.py - Skipping constant folding for node Node(name='node_Unsqueeze_835', domain='', op_type='Unsqueeze', inputs=(Value(name='expand_1', type=Tensor(FLOAT), shape=Shape([128, 160]), producer='node_Constant_1518', index=0, const_value={Tensor(...)}), SymbolicTensor(name='val_837', type=Tensor(INT64), shape=Shape([1]), producer='node_Constant_834', index=0, const_value={Tensor<INT64,[1]>(array([-1]), name='val_837')})), attributes={}, overload='', outputs=(SymbolicTensor(name='val_838', type=Tensor(FLOAT), shape=Shape([128, 160, 1]), producer='node_Unsqueeze_835', index=0),), version=20, doc_string=None) due to large input sizes: [20480, 1]
[2025-10-29 20:27:13] INFO - _constant_folding.py - Skipping constant folding for node Node(name='node_Unsqueeze_837', domain='', op_type='Unsqueeze', inputs=(Value(name='expand', type=Tensor(FLOAT), shape=Shape([128, 160]), producer='node_Constant_1514', index=0, const_value={Tensor(...)}), SymbolicTensor(name='val_839', type=Tensor(INT64), shape=Shape([1]), producer='node_Constant_836', index=0, const_value={Tensor<INT64,[1]>(array([-1]), name='val_839')})), attributes={}, overload='', outputs=(SymbolicTensor(name='val_840', type=Tensor(FLOAT), shape=Shape([128, 160, 1]), producer='node_Unsqueeze_837', index=0),), version=20, doc_string=None) due to large input sizes: [20480, 1]
[2025-10-29 20:27:13] INFO - _constant_folding.py - Skipping constant folding for node Node(name='node_view_14', domain='', op_type='Reshape', inputs=(Value(name='stack_1', type=Tensor(FLOAT), shape=Shape([64, 80, 2]), producer='node_Constant_1544', index=0, const_value={Tensor(...)}), Value(name='val_880', type=Tensor(INT64), shape=Shape([2]), producer='node_Constant_1545', index=0, const_value={Tensor<INT64,[2]>(array([-1, 2]), name='val_880')})), attributes={'allowzero': Attr('allowzero', INT, 1)}, overload='', outputs=(SymbolicTensor(name='view_14', type=Tensor(FLOAT), shape=Shape([5120, 2]), producer='node_view_14', index=0),), version=20, doc_string=None) due to large input sizes: [10240, 2]
[2025-10-29 20:27:13] INFO - _constant_folding.py - Skipping constant folding for node Node(name='node_cat_17', domain='', op_type='Concat', inputs=(Value(name='full', type=Tensor(FLOAT), shape=Shape([20480, 1]), producer='node_Constant_1523', index=0, const_value={Tensor(...)}), Value(name='full_1', type=Tensor(FLOAT), shape=Shape([5120, 1]), producer='node_Constant_1549', index=0, const_value={Tensor(...)}), Value(name='full_2', type=Tensor(FLOAT), shape=Shape([1280, 1]), producer='node_Constant_1576', index=0, const_value={Tensor(...)})), attributes={'axis': Attr('axis', INT, 0)}, overload='', outputs=(SymbolicTensor(name='cat_17', type=Tensor(FLOAT), shape=Shape([26880, 1]), producer='node_cat_17', index=0),), version=20, doc_string=None) due to large input sizes: [20480, 5120, 1280]
[2025-10-29 20:27:13] INFO - unused_removal.py - Removed 5 unused nodes
[2025-10-29 20:27:13] INFO - unused_removal.py - No unused functions to remove
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_48' with existing initializer 'val_23'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_57' with existing initializer 'val_23'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_103' with existing initializer 'val_70'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_174' with existing initializer 'val_125'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_221' with existing initializer 'val_196'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_239' with existing initializer 'val_183'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_251' with existing initializer 'val_230'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_260' with existing initializer 'val_112'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_269' with existing initializer 'val_112'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_278' with existing initializer 'val_112'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_287' with existing initializer 'val_112'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_296' with existing initializer 'val_112'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_305' with existing initializer 'val_70'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_346' with existing initializer 'val_70'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_355' with existing initializer 'val_112'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_364' with existing initializer 'val_23'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_373' with existing initializer 'val_23'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_382' with existing initializer 'val_23'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_391' with existing initializer 'val_23'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_400' with existing initializer 'val_23'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_409' with existing initializer 'val_70'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_450' with existing initializer 'val_70'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_459' with existing initializer 'val_23'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_468' with existing initializer 'val_23'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_477' with existing initializer 'val_70'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_486' with existing initializer 'val_70'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_495' with existing initializer 'val_70'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_504' with existing initializer 'val_70'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_513' with existing initializer 'val_70'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_522' with existing initializer 'val_70'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_531' with existing initializer 'val_70'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_540' with existing initializer 'val_70'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_549' with existing initializer 'val_70'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_558' with existing initializer 'val_70'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_567' with existing initializer 'val_112'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_576' with existing initializer 'val_112'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_585' with existing initializer 'val_125'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_594' with existing initializer 'val_125'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_603' with existing initializer 'val_125'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_612' with existing initializer 'val_125'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_621' with existing initializer 'val_125'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_630' with existing initializer 'val_125'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_639' with existing initializer 'val_125'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_648' with existing initializer 'val_125'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_657' with existing initializer 'val_125'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_666' with existing initializer 'val_125'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_675' with existing initializer 'val_183'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_684' with existing initializer 'val_70'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_693' with existing initializer 'val_70'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_702' with existing initializer 'val_70'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_728' with existing initializer 'val_125'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_737' with existing initializer 'val_125'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_746' with existing initializer 'val_125'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_761' with existing initializer 'val_717'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_771' with existing initializer 'val_196'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_780' with existing initializer 'val_196'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_789' with existing initializer 'val_196'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_804' with existing initializer 'val_717'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_837' with existing initializer 'val_4'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_839' with existing initializer 'val_4'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_880' with existing initializer 'val_844'
[2025-10-29 20:27:13] INFO - initializer_deduplication.py - Replaced initializer 'val_1062' with existing initializer 'val_1052'
Traceback (most recent call last):
File "/content/conversion.py", line 22, in
export_result = model.export("cell_grouping.onnx")
File "/content/myenv/lib/python3.10/site-packages/super_gradients/module_interfaces/exportable_detector.py", line 501, in export
nms_attach_method(
File "/content/myenv/lib/python3.10/site-packages/super_gradients/conversion/onnx/nms.py", line 312, in attach_onnx_nms
graph = gs.import_onnx(onnx.load(onnx_model_path))
File "/content/myenv/lib/python3.10/site-packages/onnx_graphsurgeon/importers/onnx_importer.py", line 337, in import_onnx
return OnnxImporter.import_graph(
File "/content/myenv/lib/python3.10/site-packages/onnx_graphsurgeon/importers/onnx_importer.py", line 283, in import_graph
get_tensor(initializer)
File "/content/myenv/lib/python3.10/site-packages/onnx_graphsurgeon/importers/onnx_importer.py", line 277, in get_tensor
subgraph_tensor_map[onnx_tensor.name] = OnnxImporter.import_tensor(onnx_tensor)
File "/content/myenv/lib/python3.10/site-packages/onnx_graphsurgeon/importers/onnx_importer.py", line 132, in import_tensor
return Constant(name=onnx_tensor.name, values=LazyValues(onnx_tensor), data_location=data_location)
File "/content/myenv/lib/python3.10/site-packages/onnx_graphsurgeon/ir/tensor.py", line 208, in init
self.dtype = get_onnx_tensor_dtype(self.tensor)
File "/content/myenv/lib/python3.10/site-packages/onnx_graphsurgeon/importers/onnx_importer.py", line 99, in get_onnx_tensor_dtype
dtype = get_numpy_type(onnx_type)
File "/content/myenv/lib/python3.10/site-packages/onnx_graphsurgeon/importers/onnx_importer.py", line 86, in get_numpy_type
if onnx_type != onnx.TensorProto.BFLOAT16 and onnx_type in onnx.mapping.TENSOR_TYPE_TO_NP_TYPE:
AttributeError: module 'onnx' has no attribute 'mapping'. Did you mean: '_mapping'?

Code: from super_gradients.common.object_names import Models
from super_gradients.training import models

model = models.get('yolo_nas_l',
num_classes=len(dataset_params['classes']),
checkpoint_path="/content/drive/MyDrive/RUN_20251029_185759_477436/ckpt_best.pth")

Convert model to onnx
export_result = model.export("cell_grouping.onnx")

Start with the reported console.log and the import context in super_gradients/common/environment/cfg_utils.py, then trace where the onnx.mapping import is attempted. Reproduce the failure in the reported Python 3.10 environment and identify the dependency or compatibility condition involved. Done means the reported workflow no longer raises ModuleNotFoundError.