在执行tools/optimize/yolov5s-opt.py时报错

[BoXingJinQiu]

在对yolov5转的onnx模型进行优化时报错，请问是不是预训练模型更新了，优化的代码没有更新？

---- Tengine YOLOv5 Optimize Tool ----

Input model : yolov5s.v4.onnx
Output model : yolov5s.v4.opt.onnx
Input tensor : 167
Output tensor : 381,420,459
[Quant Tools Info]: Step 0, load original onnx model from yolov5s.v4.onnx.
256
[Quant Tools Info]: Step 1, Remove the focus and postprocess nodes.
Traceback (most recent call last):
File "yolov5s-opt.py", line 222, in
main()
File "yolov5s-opt.py", line 184, in main
new_nodes = cut_focus_output(old_node, in_tensor, out_tensor)
File "yolov5s-opt.py", line 82, in cut_focus_output
output_pass[node_dict[out_name[i]]] = 2
KeyError: '381'

[xukefang]

+1

cmd:python3 yolov5s-opt.py --input yolov5s.v5.onnx --output yolov5s.v5.opt.onnx --in_tensor 167 --out_tensor 397,458,519 --verbose

---- Tengine YOLOv5 Optimize Tool ----

Input model : yolov5s.v5.onnx
Output model : yolov5s.v5.opt.onnx
Input tensor : 167
Output tensor : 397,458,519
[Quant Tools Info]: Step 0, load original onnx model from yolov5s.v5.onnx.
256
[Quant Tools Info]: Step 1, Remove the focus and postprocess nodes.
[Verbose] node_dict: {'130': 0, '135': 1, '140': 2, '145': 3, '150': 4, '155': 5, '160': 6, '165': 7, '166': 8, '167': 9, '168': 10, '169': 11, '170': 12, '171': 13, '172': 14, '173': 15, '174': 16, '175': 17, '176': 18, '177': 19, '178': 20, '179': 21, '180': 22, '181': 23, '182': 24, '183': 25, '184': 26, '185': 27, '186': 28, '187': 29, '188': 30, '189': 31, '190': 32, '191': 33, '192': 34, '193': 35, '194': 36, '195': 37, '196': 38, '197': 39, '198': 40, '199': 41, '200': 42, '201': 43, '202': 44, '203': 45, '204': 46, '205': 47, '206': 48, '207': 49, '208': 50, '209': 51, '210': 52, '211': 53, '212': 54, '213': 55, '214': 56, '215': 57, '216': 58, '217': 59, '218': 60, '219': 61, '220': 62, '221': 63, '222': 64, '223': 65, '224': 66, '225': 67, '226': 68, '227': 69, '228': 70, '229': 71, '230': 72, '231': 73, '232': 74, '233': 75, '234': 76, '235': 77, '236': 78, '237': 79, '238': 80, '239': 81, '240': 82, '241': 83, '242': 84, '243': 85, '244': 86, '245': 87, '246': 88, '247': 89, '248': 90, '249': 91, '250': 92, '251': 93, '252': 94, '253': 95, '254': 96, '255': 97, '256': 98, '257': 99, '258': 100, '259': 101, '260': 102, '261': 103, '262': 104, '263': 105, '264': 106, '265': 107, '266': 108, '267': 109, '268': 110, '269': 111, '270': 112, '271': 113, '272': 114, '273': 115, '274': 116, '275': 117, '276': 118, '277': 119, '278': 120, '279': 121, '280': 122, '281': 123, '282': 124, '283': 125, '284': 126, '285': 127, '286': 128, '287': 129, '288': 130, '289': 131, '294': 132, '295': 133, '296': 134, '297': 135, '298': 136, '299': 137, '300': 138, '301': 139, '302': 140, '303': 141, '304': 142, '305': 143, '306': 144, '307': 145, '308': 146, '309': 147, '310': 148, '311': 149, '312': 150, '313': 151, '314': 152, '319': 153, '320': 154, '321': 155, '322': 156, '323': 157, '324': 158, '325': 159, '326': 160, '327': 161, '328': 162, '329': 163, '330': 164, '331': 165, '332': 166, '333': 167, '334': 168, '335': 169, '336': 170, '337': 171, '338': 172, '339': 173, '340': 174, '341': 175, '342': 176, '343': 177, '344': 178, '345': 179, '346': 180, '347': 181, '348': 182, '349': 183, '350': 184, '351': 185, '352': 186, '353': 187, '354': 188, '355': 189, '356': 190, '357': 191, '358': 192, '359': 193, '360': 194, '361': 195, '362': 196, '363': 197, '364': 198, '365': 199, '366': 200, '367': 201, '368': 202, '369': 203, '370': 204, '371': 205, '372': 206, '373': 207, '374': 208, '375': 209, '376': 210, '377': 211, '400': 212, '401': 213, '402': 214, '407': 215, '409': 216, '411': 217, '413': 218, '415': 219, '420': 220, '422': 221, '424': 222, '426': 223, '431': 224, '432': 225, '442': 226, '443': 227, '466': 228, '467': 229, '468': 230, '473': 231, '475': 232, '477': 233, '479': 234, '481': 235, '486': 236, '488': 237, '490': 238, '492': 239, '497': 240, '498': 241, '508': 242, '509': 243, '532': 244, '533': 245, '534': 246, '539': 247, '541': 248, '543': 249, '545': 250, '547': 251, '552': 252, '554': 253, '556': 254, '558': 255, '563': 256, '564': 257, '574': 258, 'output': 259}
yolov5s-opt.py:80: DeprecationWarning: np.int is a deprecated alias for the builtin int. To silence this warning, use int by itself. Doing this will not modify any behavior and is safe. When replacing np.int, you may wish to use e.g. np.int64 or np.int32 to specify the precision. If you wish to review your current use, check the release note link for additional information.
Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations
output_pass = np.zeros((len(input_node)), dtype=np.int)
Traceback (most recent call last):
File "yolov5s-opt.py", line 221, in
main()
File "yolov5s-opt.py", line 184, in main
new_nodes = cut_focus_output(old_node, in_tensor, out_tensor)
File "yolov5s-opt.py", line 82, in cut_focus_output
output_pass[node_dict[out_name[i]]] = 2
KeyError: '397'

[qintian-319]

把onnx模型用https://netron.app/ 可视化，看这三个输出的代号，然后替换掉out_tensor的参数值
例如 output,352,364