[#263] Fixes Keras Conv2D model error

[lancelafontaine]

[#263]

ValueError: Negative dimension… size caused by subtracting 3 from 1 for 'conv2d_1/convolution' (op: 'Conv2D') with input shapes: [?,1,28,28], [3,3,28,32]

The weights_epoch variable was also undefined.

[codecov[bot]]

Codecov Report

Merging #366 into 263/character_recognition will increase coverage by 11.23%. The diff coverage is n/a.

@@                      Coverage Diff                       @@
##           263/character_recognition     #366       +/-   ##
==============================================================
+ Coverage                      66.51%   77.74%   +11.23%     
==============================================================
  Files                             53       11       -42     
  Lines                           2129      310     -1819     
  Branches                          16       16               
==============================================================
- Hits                            1416      241     -1175     
+ Misses                           711       67      -644     
  Partials                           2        2

Impacted Files	Coverage Δ
src/nlp_service/rasa/rasa_classifier.py
src/nlp_service/services/fact_service.py
src/nlp_service/controllers/nlp_controller.py
src/ml_service/web/ml_controller.py
...raction/post_processing/regex/regex_test_helper.py
handwriting_recognition_ml.py
controllers/ocr_controller.py
...ce/model_training/similar_finder/similar_finder.py
...le_output_regression/additional_indemnity_money.py
src/nlp_service/init_rasa.py
... and 32 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update e03aeeb...8cefe67. Read the comment docs.