clstm

clstm copied to clipboard

Many NaN errors with the new Tensor-based version

2

For many models (especially deep ones with many parameters e.g. bidi2), I keep getting the following error clstm.cc:664: void ocropus::GenericNPLSTM::backward() [with int F = 1; int G = 2; int...

ASDen

@tmbdev Hi ! when I use default clstm models like arabic-beirut-200.clstm model, it's ok and convert successfully: user@user ~/Desktop/mags $ kraken -i images/tt1.jpg image.txt binarize segment ocr -m arabic-beirut- 200.clstm...

Hadi58

Segmentation fault when running clstmocr on pre-trained model

1

I am trying to build clstm on an ubuntu 16.04 machine. My steps of installation are `sudo apt-get install scons libprotobuf-dev libprotobuf9v5 protobuf-compiler libpng-dev libeigen3-dev swig` `scons` `sudo scons install`...

drdrsh

Training two layer networks fails with gradient errors

2

I have added a createBidi2 method in the CLSTMText class function, but training 2 layer networks fails with a gradient assertion error, regardless of the used learning rate and hidden...

mihaelacr-google

I am getting the following error when using uw3-500 to train my data. The date used in the training are sequence of characters (line-level samples). Your prompt help is appreciated...

samansarraf

Arabic 800,000 model cant go below Error Rate 0.5

4

I have been training an Arabic language model from scratch for days now, reaching +800,000 epochs, the error rate wont go below 0.5 and thats very bad. I have used...

ghost

Always missing one character in the output

3

I used pyclstm for trainning my dataset on date but the output always missing the character 9. All the rest character is Ok. Hope for help !

manhcuogntin4

Sort and Sed commands are causing the model not to train (ERROR 1/OUT is empty)

2

Peace Be Upon you, I am training an Arabic model from scratch, reaching about 270,000 epochs in 32 hours, the ERROR is still 1, and the OUT is empty. The...

ghost

Missing last char of the line.

4

If I train "lstm1" instead of "bidi", it is almost always missing the last character in the output sentence. Is it expected behaviour?

ShubhamAgarwal12

usemat acceleration in current version?

17

In the previous version, one can accelerate calculation using "usemat=1" during installation, which use matrix instead of tensor. But in the current version, the matrix alternative is not supported. Is...

crazylyf