Toybrick

标题: Tensorflow mobilenet-ssd 转 Rknn 模型失败 [打印本页]

作者: 摇太阳 时间: 2019-7-11 15:58
标题: Tensorflow mobilenet-ssd 转 Rknn 模型失败
开发板系统：fedora 28
Toolkit版本： 1.0.0
Tensorflow版本：1.10.1

1. 根据tensorflow官方教程生成了pb文件

2. 运行了官方的demo
/home/toybrick/RKNN/rknn-toolbrick/pexample/ssd_mobilenet_V2/step1.py
只是替换了模型的文件

import numpy as np
import cv2
from rknn.api import RKNN

if __name__ == '__main__':

# Create RKNN object
rknn = RKNN(verbose = True, verbose_file =  './Error.log')

# pre-process config
print('--> config model')
rknn.config(channel_mean_value='127.5 127.5 127.5 128', reorder_channel='0 1 2', quantized_dtype='asymmetric_quantized-u8')
print('done')

# Load tflite model
print('--> Loading model')
#模型更改的路径
ret = rknn.load_tensorflow(tf_pb='./tensorflow.pb',
                           inputs=['FeatureExtractor/MobilenetV2/MobilenetV2/input'],
                           outputs=['concat_1', 'concat'],
                           input_size_list=[[300,300,3]],
                           predef_file=None)
if ret != 0:
      print('Load model failed!')
      exit(ret)
print('done')

if ret != 0:
      print('Load model failed!')
      exit(ret)
print('done')

# Build model
print('--> hybrid_quantization_step1')
ret = rknn.hybrid_quantization_step1(dataset='./dataset.txt')
if ret != 0:
      print('hybrid_quantization_step1 failed!')
      exit(ret)
print('done')

# Tips
print('Please modify ssd_mobilenet_v2.quantization.cfg!')
print('==================================================================================================')
print('Modify method:')
print('1. delete @FeatureExtractor/MobilenetV2/expanded_conv/depthwise/depthwise_227:weight and its value')
print('2. delete @FeatureExtractor/MobilenetV2/expanded_conv/depthwise/depthwise_227:bias and its value')
print('3. delete @FeatureExtractor/MobilenetV2/Conv/Relu6_228

ut0 and its value')
print('==================================================================================================')
print('Original quantization profile and modified quantization profile diff shows in file quantization_profile.diff')

rknn.release()

3. 报错的日志：

D import clients finished
E Catch exception when loading tensorflow model: ./tensorflow.pb!
T Traceback (most recent call last):
T File "rknn/api/rknn_base.py", line 136, in rknn.api.rknn_base.RKNNBase.load_tensorflow
T File "rknn/base/RKNNlib/converter/convert_tf.py", line 102, in rknn.base.RKNNlib.converter.convert_tf.convert_tf.__init__
T File "rknn/base/RKNNlib/converter/tensorflowloader.py", line 44, in rknn.base.RKNNlib.converter.tensorflowloader.TF_Graph_Preprocess.__init__
T File "rknn/base/RKNNlib/converter/tensorflowloader.py", line 46, in rknn.base.RKNNlib.converter.tensorflowloader.TF_Graph_Preprocess.__init__
T File "/usr/local/lib/python3.6/site-packages/google/protobuf/message.py", line 187, in ParseFromString
T    return self.MergeFromString(serialized)
T File "/usr/local/lib/python3.6/site-packages/google/protobuf/internal/python_message.py", line 1124, in MergeFromString
T    if self._InternalParse(serialized, 0, length) != length:
T File "/usr/local/lib/python3.6/site-packages/google/protobuf/internal/python_message.py", line 1189, in InternalParse
T    pos = field_decoder(buffer, new_pos, end, self, field_dict)
T File "/usr/local/lib/python3.6/site-packages/google/protobuf/internal/decoder.py", line 723, in DecodeField
T    if value._InternalParse(buffer, pos, new_pos) != new_pos:
T File "/usr/local/lib/python3.6/site-packages/google/protobuf/internal/python_message.py", line 1189, in InternalParse
T    pos = field_decoder(buffer, new_pos, end, self, field_dict)
T File "/usr/local/lib/python3.6/site-packages/google/protobuf/internal/decoder.py", line 702, in DecodeRepeatedField
T    if value.add()._InternalParse(buffer, pos, new_pos) != new_pos:
T File "/usr/local/lib/python3.6/site-packages/google/protobuf/internal/python_message.py", line 1189, in InternalParse
T    pos = field_decoder(buffer, new_pos, end, self, field_dict)
T File "/usr/local/lib/python3.6/site-packages/google/protobuf/internal/decoder.py", line 847, in DecodeMap
T    if submsg._InternalParse(buffer, pos, new_pos) != new_pos:
T File "/usr/local/lib/python3.6/site-packages/google/protobuf/internal/python_message.py", line 1176, in InternalParse
T    buffer, new_pos, wire_type)  # pylint: disable=protected-access
T File "/usr/local/lib/python3.6/site-packages/google/protobuf/internal/decoder.py", line 948, in _DecodeUnknownField
T    (data, pos) = _DecodeUnknownFieldSet(buffer, pos)
T File "/usr/local/lib/python3.6/site-packages/google/protobuf/internal/decoder.py", line 927, in _DecodeUnknownFieldSet
T    (data, pos) = _DecodeUnknownField(buffer, pos, wire_type)
T File "/usr/local/lib/python3.6/site-packages/google/protobuf/internal/decoder.py", line 952, in _DecodeUnknownField
T    raise _DecodeError('Wrong wire type in tag.')
T google.protobuf.message.DecodeError: Wrong wire type in tag.

作者: jefferyzhang 时间: 2019-7-12 16:40
首先，你这个错都还在google/protobuf/里，都还没走到rknn_toolkit，这种错你要自己先debug下。
用这个pb你当前系统的tf是否可以运行，这里报的错很怀疑这个pb就不能在这个tf版本上运行

作者: 摇太阳 时间: 2019-7-15 09:25

jefferyzhang 发表于 2019-7-12 16:40
首先，你这个错都还在google/protobuf/里，都还没走到rknn_toolkit，这种错你要自己先debug下。
用这个pb你 ...

喔喔喔，明白。
请教下，你们那个 /RKNN/rknn-toolkit/example/mobilenet-ssd/ssd_mobilenet_v1_coco_2017_11_17.pb训练的跟Tensorflow版本有关系嘛？你们用的是哪个版本啊？
能不能给一个官方训练的教程呀？
谢谢啊

作者: troy 时间: 2019-7-16 08:31

摇太阳发表于 2019-7-15 09:25
喔喔喔，明白。
请教下，你们那个 /RKNN/rknn-toolkit/example/mobilenet-ssd/ssd_mobilenet_v1_coco_201 ...

官方没有重新进行训练，都是使用现有的预训练模型。

作者: 摇太阳 时间: 2019-7-16 15:14

troy 发表于 2019-7-16 08:31
官方没有重新进行训练，都是使用现有的预训练模型。

请问下您有没有用过自己的模型训练过呀？
如果有成功训练出来的能给个教程嘛？
我这里用自己的模型按官方的训练出来的转不欸
谢谢啦

作者: troy 时间: 2019-7-16 16:10

摇太阳发表于 2019-7-16 15:14
请问下您有没有用过自己的模型训练过呀？
如果有成功训练出来的能给个教程嘛？
我这里用自己的模型按官 ...

上面同事有回复，转换失败有可能是pb模型的TensorFlow版本太高导致的

作者: 小宁哥 时间: 2019-11-26 19:19
楼主问题解决了吗，我也遇到了，tf版本和你一样，toolkit1.2

欢迎光临 Toybrick (https://t.rock-chips.com/)