2.3.0版本onnx转rknn错误

[LiuFYv]

这是一个并非来自于官方zoo中的模型，在2.3.0版本中转换错误，在1.6.0中可以转换，但是对于1280的输入时1.6.0会报错：E RKNN: [10:01:26.811] REGTASK: The bit width of field value exceeds the limit, target: v2, offset: 0x5014, shift = 0, limit: 0x1fff, value: 0x31ff
尝试降级到1248可以正常转换

I rknn-toolkit2 version: 2.3.0
--> Config model
done
--> Loading model
W load_onnx: If you don't need to crop the model, don't set 'inputs'/'input_size_list'/'outputs'!
I Loading : 100%|██████████████████████████████████████████████| 224/224 [00:00<00:00, 24979.37it/s]
done
--> Building model
I OpFusing 0: 100%|██████████████████████████████████████████████| 100/100 [00:00<00:00, 326.05it/s]
I OpFusing 1 : 99%|█████████████████████████████████████████████▌| 99/100 [00:00<00:00, 244.36it/s]
W build: Show op fuse match nodes:
Rule: squeeze_1_in_nd_transpose
Subgraph:
Op type: Transpose
Op name: /model.24/Transpose
Input:
/model.24/Reshape_output_0 : [1, 3, 9, 160, 160]
Output:
/model.24/Transpose_output_0 : [1, 3, 160, 160, 9]
Attribute:
perm : [0, 1, 3, 4, 2]

W You can add disable_rules=['squeeze_1_in_nd_transpose'] in rknn.config() to temporarily disable the error rule!

E build: Traceback (most recent call last):
File "rknn/api/rknn_log.py", line 344, in rknn.api.rknn_log.error_catch_decorator.error_catch_wrapper
File "rknn/api/rknn_base.py", line 1962, in rknn.api.rknn_base.RKNNBase.build
File "rknn/api/graph_optimizer.py", line 2184, in rknn.api.graph_optimizer.GraphOptimizer.fuse_ops
File "rknn/api/ir_graph.py", line 537, in rknn.api.ir_graph.IRGraph.rebuild
File "rknn/api/ir_graph.py", line 311, in rknn.api.ir_graph.IRGraph._clean_model
File "rknn/api/ir_graph.py", line 125, in rknn.api.ir_graph.IRGraph.infer_shapes
File "rknn/api/ir_graph.py", line 138, in rknn.api.ir_graph.IRGraph.infer_shapes
File "rknn/base/extend_op/extend_op.py", line 192, in rknn.base.extend_op.extend_op.update_model_value_info
KeyError: '/model.6/m/m.0/Gather_output_0-rs'

W build: ===================== WARN(3) =====================
E rknn-toolkit2 version: 2.3.0
Traceback (most recent call last):
File "rknn/api/rknn_log.py", line 344, in rknn.api.rknn_log.error_catch_decorator.error_catch_wrapper
File "rknn/api/rknn_base.py", line 1962, in rknn.api.rknn_base.RKNNBase.build
File "rknn/api/graph_optimizer.py", line 2184, in rknn.api.graph_optimizer.GraphOptimizer.fuse_ops
File "rknn/api/ir_graph.py", line 537, in rknn.api.ir_graph.IRGraph.rebuild
File "rknn/api/ir_graph.py", line 311, in rknn.api.ir_graph.IRGraph._clean_model
File "rknn/api/ir_graph.py", line 125, in rknn.api.ir_graph.IRGraph.infer_shapes
File "rknn/api/ir_graph.py", line 138, in rknn.api.ir_graph.IRGraph.infer_shapes
File "rknn/base/extend_op/extend_op.py", line 192, in rknn.base.extend_op.extend_op.update_model_value_info
KeyError: '/model.6/m/m.0/Gather_output_0-rs'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "/home/abc/moudle/YPLNet-main/yolov5-master/convert.py", line 61, in
ret = rknn.build(do_quantization=do_quant, dataset=DATASET_PATH)
File "/home/abc/anaconda3/envs/rknn2b0/lib/python3.9/site-packages/rknn/api/rknn.py", line 192, in build
return self.rknn_base.build(do_quantization=do_quantization, dataset=dataset, expand_batch_size=rknn_batch_size)
File "rknn/api/rknn_log.py", line 349, in rknn.api.rknn_log.error_catch_decorator.error_catch_wrapper
File "rknn/api/rknn_log.py", line 95, in rknn.api.rknn_log.RKNNLog.e
ValueError: Traceback (most recent call last):
File "rknn/api/rknn_log.py", line 344, in rknn.api.rknn_log.error_catch_decorator.error_catch_wrapper
File "rknn/api/rknn_base.py", line 1962, in rknn.api.rknn_base.RKNNBase.build
File "rknn/api/graph_optimizer.py", line 2184, in rknn.api.graph_optimizer.GraphOptimizer.fuse_ops
File "rknn/api/ir_graph.py", line 537, in rknn.api.ir_graph.IRGraph.rebuild
File "rknn/api/ir_graph.py", line 311, in rknn.api.ir_graph.IRGraph._clean_model
File "rknn/api/ir_graph.py", line 125, in rknn.api.ir_graph.IRGraph.infer_shapes
File "rknn/api/ir_graph.py", line 138, in rknn.api.ir_graph.IRGraph.infer_shapes
File "rknn/base/extend_op/extend_op.py", line 192, in rknn.base.extend_op.extend_op.update_model_value_info
KeyError: '/model.6/m/m.0/Gather_output_0-rs'

W __init__: rknn-toolkit2 version: 1.6.0+81f21f4d
--> Config model
done
--> Loading model
W load_onnx: If you don't need to crop the model, don't set 'inputs'/'input_size_list'/'outputs'!
W load_onnx: It is recommended onnx opset 19, but your onnx model opset is 12!
W load_onnx: Model converted from pytorch, 'opset_version' should be set 19 in torch.onnx.export for successful convert!
Loading : 100%|████████████████████████████████████████████████| 263/263 [00:00<00:00, 23057.67it/s]
done
--> Building model
GraphPreparing : 100%|██████████████████████████████████████████| 371/371 [00:00<00:00, 8076.56it/s]
Quantizating 1/2: 100%|███████████████████████████████████████████| 371/371 [00:07<00:00, 46.82it/s]
Quantizating 2/2: 100%|███████████████████████████████████████████| 371/371 [00:04<00:00, 77.19it/s]
W build: The default input dtype of 'images' is changed from 'float32' to 'int8' in rknn model for performance!
                       Please take care of this change when deploy rknn model with Runtime API!
W build: The default output dtype of '/model.24/Transpose_output_0' is changed from 'float32' to 'int8' in rknn model for performance!
                      Please take care of this change when deploy rknn model with Runtime API!
W build: The default output dtype of '/model.24/Transpose_1_output_0' is changed from 'float32' to 'int8' in rknn model for performance!
                      Please take care of this change when deploy rknn model with Runtime API!
W build: The default output dtype of '/model.24/Transpose_2_output_0' is changed from 'float32' to 'int8' in rknn model for performance!
                      Please take care of this change when deploy rknn model with Runtime API!
E RKNN: [10:01:26.811] REGTASK: The bit width of field value exceeds the limit, target: v2, offset: 0x5014, shift = 0, limit: 0x1fff, value: 0x31ff
done
--> Export rknn model
done