sheirving

非常感谢您的回答！按您的提示运行“python tools/converter.py ....”脚本的tensoflow版本改为1.8.0, 模型训练和PB生成版本为1.12，但是现在运行python tools/converter.py ....”（tf r1.8.0）出现另一个问题：`tensorflow.python.framework.errors_impl.InvalidArgumentError: The node 'MobileNet_CRelu_V2_Text/RNN/BiLstm_1/bw/bw/while/lstm_cell/MatMul' has inputs from different frames. The input 'ConstantFolding/MobileNet_CRelu_V2_Text/RNN/BiLstm_1/bw/lstm_cell/kernel_enter' is in frame 'MobileNet_CRelu_V2_Text/RNN/BiLstm_1/bw/bw/while/while_context'. The input 'MobileNet_CRelu_V2_Text/RNN/BiLstm_1/bw/bw/while/lstm_cell/concat' is in frame ''.`？ 我尝试用tensroflow...

统一为版本1.12.0后遇到新问题： `F tensorflow/compiler/jit/deadness_analysis.cc:639] Check failed: it != predicate_map_.end() _SINK`，似乎google没有解决办法，求教？感谢

1. Tf所有版本：r1.12.0; Bazel: 0.17.2; 其他依赖版本：基本一致 2. 步骤：python tools/converter.py convert --config=./my_model.yml; 3. 日志：`Starting local Bazel server and connecting to it... INFO: Analysed target //mace/python/tools:converter (21 packages loaded). INFO: Found 1 target......

In train.py: +from tensorflow.core.protobuf import saver_pb2 and in function of "def __init__": self.saver = tf.train.Saver(max_to_keep=100,write_version=saver_pb2.SaverDef.V1) you can see the ".ckpt"

训练配置：https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.6/configs/rec/rec_svtrnet_ch.yml 基于这个适配了训练集的输入（64， 512），训练集200万样本，4卡3090训练1周 训练实例图： 没有使用STN，我**先将Local mixer = Conv，这个实验跑完，看看结果吧**。但目前来看以（7， 11）的核， mixer前面6个Local训练难收敛。

好的，谢谢。其他地方没有变化，我这边调整下输入大小再看看效果，再反馈下。

@gdelab while testing,I have a problem about 'pickle'. so ,i change it to open files in byte as you put it,however, another question occured"Ran out of input": File "./faster_rcnn/../lib/datasets/voc_eval.py", line...

Thankyou very much for share! I had train your model by using features, which extracted by my code. The result is lower than using your features data, so could you...

Thankyou very much for your reply! On the IAM database, I got 95.498%. And for Diagram-FA, the accuracy is 98.426%, for Diagram-FC, the accyracy is 97.431%. I trained the three...

> > Thankyou very much for your reply! On the IAM database, I got 95.498%. And for Diagram-FA, the accuracy is 98.426%, for Diagram-FC, the accuracy is 97.431%. I trained...