Add tests for prediction code

tests/conftest.py

@@ -48,7 +48,7 @@ def training_config():
     return {
         "dataset_params": {
             "datasets": {
-                "training": "./tests/data/training_page",
+                "training": "./tests/data/training/training_dataset",
             },
             "train": {
                 "name": "training-train",
@@ -165,3 +165,8 @@ def training_config():
             },
         },
     }
+
+
+@pytest.fixture
+def prediction_data_path():
+    return FIXTURES / "prediction"

tests/data/evaluation/predict_training-test_0.txt

-nb_chars: 49
-cer: 1.1429
-nb_words: 9
-wer: 1.0
-nb_words_no_punct: 9
-wer_no_punct: 1.0
-sample_time: 0.8385
-time: 1.6771
-nb_samples: 2

tests/data/evaluation/predict_training-train_0.txt

-nb_chars: 43
-cer: 1.2791
-nb_words: 9
-wer: 1.0
-nb_words_no_punct: 9
-wer_no_punct: 1.0
-sample_time: 0.8334
-time: 1.6668
-nb_samples: 2

tests/data/evaluation/predict_training-val_0.txt

-nb_chars: 41
-cer: 1.2683
-nb_words: 9
-wer: 1.0
-nb_words_no_punct: 9
-wer_no_punct: 1.0
-sample_time: 0.8314
-time: 1.6628
-nb_samples: 2

tests/data/prediction/parameters.yml

+---
+version: 0.0.1
+parameters:
+  mean: [166.8418783515498, 166.8418783515498, 166.8418783515498]
+  std: [34.084189571536385, 34.084189571536385, 34.084189571536385]
+  max_char_prediction: 200
+  encoder:
+    input_channels: 3
+    dropout: 0.5
+  decoder:
+    enc_dim: 256
+    l_max: 15000
+    dec_pred_dropout: 0.1
+    attention_win: 100
+    use_1d_pe: True
+    use_lstm: False
+    vocab_size: 96
+    h_max: 500
+    w_max: 1000
+    dec_num_layers: 8
+    dec_dim_feedforward: 256
+    dec_num_heads: 4
+    dec_att_dropout: 0.1
+    dec_res_dropout: 0.1