Make loading of previously trained model work

tmbdev / clstm

A small C++ implementation of LSTM networks, focused on OCR.

Apache License 2.0

821 stars 224 forks source link

Make loading of previously trained model work #53

Closed amitdo closed 8 years ago

amitdo commented 9 years ago

Fixes #52

amitdo commented 9 years ago

@tmbdev, Please fix the broken CircleCI setup.

This PR compiles fine on my machine, which is Ubuntu 14.04 64bit with Eigen 3.3 alpha.

stexandev commented 8 years ago

Thank you for your quick response and solid fix. Recompiled and tested the fix with three different models and worked flawless. Testing included loading previously trained models, training until next save, loading again. Everything seems fine.

amitdo commented 8 years ago

@stexandev Glad to help!

isaomatsunami commented 8 years ago

Current code still fails with "FATAL missing parameter" this parameter seems to be lrate.

Is this parameter stored in saved .clstm data?

amitdo commented 8 years ago

Tom didn't merge this PR with the master yet. That's why the "current code still fails"...

amitdo commented 8 years ago

Tom, FYI @isaomatsunami confirned elsewhere:

patched code works perfectly.