diff --git a/expe_both/kraken.log b/expe_both/kraken.log
new file mode 100644
index 0000000..00893c4
--- /dev/null
+++ b/expe_both/kraken.log
@@ -0,0 +1,61 @@
+┏━━━━┳━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┓
+┃    ┃ Name      ┃ Type                     ┃ Params ┃                 In sizes ┃                Out sizes ┃
+┡━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━┩
+│ 0  │ val_cer   │ CharErrorRate            │      0 │                        ? │                        ? │
+│ 1  │ net       │ MultiParamSequential     │  4.0 M │  [[1, 1, 120, 400], '?'] │   [[1, 121, 1, 50], '?'] │
+│ 2  │ net.C_0   │ ActConv2D                │  1.3 K │  [[1, 1, 120, 400], '?'] │ [[1, 32, 120, 400], '?'] │
+│ 3  │ net.Do_1  │ Dropout                  │      0 │ [[1, 32, 120, 400], '?'] │ [[1, 32, 120, 400], '?'] │
+│ 4  │ net.Mp_2  │ MaxPool                  │      0 │ [[1, 32, 120, 400], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 5  │ net.C_3   │ ActConv2D                │ 40.0 K │  [[1, 32, 60, 200], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 6  │ net.Do_4  │ Dropout                  │      0 │  [[1, 32, 60, 200], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 7  │ net.Mp_5  │ MaxPool                  │      0 │  [[1, 32, 60, 200], '?'] │  [[1, 32, 30, 100], '?'] │
+│ 8  │ net.C_6   │ ActConv2D                │ 55.4 K │  [[1, 32, 30, 100], '?'] │  [[1, 64, 30, 100], '?'] │
+│ 9  │ net.Do_7  │ Dropout                  │      0 │  [[1, 64, 30, 100], '?'] │  [[1, 64, 30, 100], '?'] │
+│ 10 │ net.Mp_8  │ MaxPool                  │      0 │  [[1, 64, 30, 100], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 11 │ net.C_9   │ ActConv2D                │  110 K │   [[1, 64, 15, 50], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 12 │ net.Do_10 │ Dropout                  │      0 │   [[1, 64, 15, 50], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 13 │ net.S_11  │ Reshape                  │      0 │   [[1, 64, 15, 50], '?'] │   [[1, 960, 1, 50], '?'] │
+│ 14 │ net.L_12  │ TransposedSummarizingRNN │  1.9 M │   [[1, 960, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 15 │ net.Do_13 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 16 │ net.L_14  │ TransposedSummarizingRNN │  963 K │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 17 │ net.Do_15 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 18 │ net.L_16  │ TransposedSummarizingRNN │  963 K │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 19 │ net.Do_17 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 20 │ net.O_18  │ LinSoftmax               │ 48.5 K │   [[1, 400, 1, 50], '?'] │   [[1, 121, 1, 50], '?'] │
+└────┴───────────┴──────────────────────────┴────────┴──────────────────────────┴──────────────────────────┘
+Trainable params: 4.0 M                                                                                                                                                                                                                                                                 
+Non-trainable params: 0                                                                                                                                                                                                                                                                 
+Total params: 4.0 M                                                                                                                                                                                                                                                                     
+Total estimated model params size (MB): 16                                                                                                                                                                                                                                              
+stage 0/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:40 val_accuracy: 0.82457  early_stopping: 0/10 0.82457
+stage 1/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:40 val_accuracy: 0.84399  early_stopping: 0/10 0.84399
+stage 2/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:40 val_accuracy: 0.84616  early_stopping: 0/10 0.84616
+stage 3/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:40 val_accuracy: 0.86558  early_stopping: 0/10 0.86558
+stage 4/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:39 val_accuracy: 0.86369  early_stopping: 1/10 0.86558
+stage 5/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:40 val_accuracy: 0.89819  early_stopping: 0/10 0.89819
+stage 6/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:40 val_accuracy: 0.89097  early_stopping: 1/10 0.89819
+stage 7/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:40 val_accuracy: 0.89955  early_stopping: 0/10 0.89955
+stage 8/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:39 val_accuracy: 0.90316  early_stopping: 0/10 0.90316
+stage 9/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:40 val_accuracy: 0.89892  early_stopping: 1/10 0.90316
+stage 10/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:40 val_accuracy: 0.89205  early_stopping: 2/10 0.90316
+stage 11/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:41 val_accuracy: 0.88889  early_stopping: 3/10 0.90316
+stage 12/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:42 val_accuracy: 0.90506  early_stopping: 0/10 0.90506
+stage 13/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:42 val_accuracy: 0.91364  early_stopping: 0/10 0.91364
+stage 14/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:42 val_accuracy: 0.90108  early_stopping: 1/10 0.91364
+stage 15/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:42 val_accuracy: 0.90199  early_stopping: 2/10 0.91364
+stage 16/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:43 val_accuracy: 0.90867  early_stopping: 3/10 0.91364
+stage 17/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:42 val_accuracy: 0.91418  early_stopping: 0/10 0.91418
+stage 18/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:42 val_accuracy: 0.90687  early_stopping: 1/10 0.91418
+stage 19/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:42 val_accuracy: 0.89494  early_stopping: 2/10 0.91418
+stage 20/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:43 val_accuracy: 0.92385  early_stopping: 0/10 0.92385
+stage 21/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:43 val_accuracy: 0.90262  early_stopping: 1/10 0.92385
+stage 22/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:43 val_accuracy: 0.90750  early_stopping: 2/10 0.92385
+stage 23/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:42 val_accuracy: 0.91518  early_stopping: 3/10 0.92385
+stage 24/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:43 val_accuracy: 0.90081  early_stopping: 4/10 0.92385
+stage 25/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:42 val_accuracy: 0.91048  early_stopping: 5/10 0.92385
+stage 26/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:42 val_accuracy: 0.91915  early_stopping: 6/10 0.92385
+stage 27/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:42 val_accuracy: 0.90018  early_stopping: 7/10 0.92385
+stage 28/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:42 val_accuracy: 0.90696  early_stopping: 8/10 0.92385
+stage 29/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:42 val_accuracy: 0.89756  early_stopping: 9/10 0.92385
+stage 30/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 569/569 0:00:00 0:00:42 val_accuracy: 0.90903  early_stopping: 10/10 0.92385
+Moving best model /home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/peraire2_ft_MMCFR_20.mlmodel (0.9238482713699341) to /home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/peraire2_ft_MMCFR_best.mlmodel
diff --git a/expe_both/kraken_B.log b/expe_both/kraken_B.log
new file mode 100644
index 0000000..f36c15c
--- /dev/null
+++ b/expe_both/kraken_B.log
@@ -0,0 +1,68 @@
+┏━━━━┳━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┓
+┃    ┃ Name      ┃ Type                     ┃ Params ┃                 In sizes ┃                Out sizes ┃
+┡━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━┩
+│ 0  │ val_cer   │ CharErrorRate            │      0 │                        ? │                        ? │
+│ 1  │ net       │ MultiParamSequential     │  4.0 M │  [[1, 1, 120, 400], '?'] │   [[1, 102, 1, 50], '?'] │
+│ 2  │ net.C_0   │ ActConv2D                │  1.3 K │  [[1, 1, 120, 400], '?'] │ [[1, 32, 120, 400], '?'] │
+│ 3  │ net.Do_1  │ Dropout                  │      0 │ [[1, 32, 120, 400], '?'] │ [[1, 32, 120, 400], '?'] │
+│ 4  │ net.Mp_2  │ MaxPool                  │      0 │ [[1, 32, 120, 400], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 5  │ net.C_3   │ ActConv2D                │ 40.0 K │  [[1, 32, 60, 200], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 6  │ net.Do_4  │ Dropout                  │      0 │  [[1, 32, 60, 200], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 7  │ net.Mp_5  │ MaxPool                  │      0 │  [[1, 32, 60, 200], '?'] │  [[1, 32, 30, 100], '?'] │
+│ 8  │ net.C_6   │ ActConv2D                │ 55.4 K │  [[1, 32, 30, 100], '?'] │  [[1, 64, 30, 100], '?'] │
+│ 9  │ net.Do_7  │ Dropout                  │      0 │  [[1, 64, 30, 100], '?'] │  [[1, 64, 30, 100], '?'] │
+│ 10 │ net.Mp_8  │ MaxPool                  │      0 │  [[1, 64, 30, 100], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 11 │ net.C_9   │ ActConv2D                │  110 K │   [[1, 64, 15, 50], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 12 │ net.Do_10 │ Dropout                  │      0 │   [[1, 64, 15, 50], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 13 │ net.S_11  │ Reshape                  │      0 │   [[1, 64, 15, 50], '?'] │   [[1, 960, 1, 50], '?'] │
+│ 14 │ net.L_12  │ TransposedSummarizingRNN │  1.9 M │   [[1, 960, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 15 │ net.Do_13 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 16 │ net.L_14  │ TransposedSummarizingRNN │  963 K │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 17 │ net.Do_15 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 18 │ net.L_16  │ TransposedSummarizingRNN │  963 K │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 19 │ net.Do_17 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 20 │ net.O_18  │ LinSoftmax               │ 40.9 K │   [[1, 400, 1, 50], '?'] │   [[1, 102, 1, 50], '?'] │
+└────┴───────────┴──────────────────────────┴────────┴──────────────────────────┴──────────────────────────┘
+Trainable params: 4.0 M                                                                                                                                                                                                                                                                 
+Non-trainable params: 0                                                                                                                                                                                                                                                                 
+Total params: 4.0 M                                                                                                                                                                                                                                                                     
+Total estimated model params size (MB): 16                                                                                                                                                                                                                                              
+stage 0/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.84189  early_stopping: 0/10 0.84189
+stage 1/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:16 val_accuracy: 0.86503  early_stopping: 0/10 0.86503
+stage 2/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.85900  early_stopping: 1/10 0.86503
+stage 3/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.83225  early_stopping: 2/10 0.86503
+stage 4/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.84189  early_stopping: 3/10 0.86503
+stage 5/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.86430  early_stopping: 4/10 0.86503
+stage 6/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.89275  early_stopping: 0/10 0.89275
+stage 7/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.85900  early_stopping: 1/10 0.89275
+stage 8/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.88551  early_stopping: 2/10 0.89275
+stage 9/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.87539  early_stopping: 3/10 0.89275
+stage 10/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.85852  early_stopping: 4/10 0.89275
+stage 11/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.88286  early_stopping: 5/10 0.89275
+stage 12/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.89588  early_stopping: 0/10 0.89588
+stage 13/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.87033  early_stopping: 1/10 0.89588
+stage 14/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.89058  early_stopping: 2/10 0.89588
+stage 15/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.89467  early_stopping: 3/10 0.89588
+stage 16/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.87684  early_stopping: 4/10 0.89588
+stage 17/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.89998  early_stopping: 0/10 0.89998
+stage 18/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.90697  early_stopping: 0/10 0.90697
+stage 19/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.88407  early_stopping: 1/10 0.90697
+stage 20/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.87660  early_stopping: 2/10 0.90697
+stage 21/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.89034  early_stopping: 3/10 0.90697
+stage 22/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.90046  early_stopping: 4/10 0.90697
+stage 23/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.86961  early_stopping: 5/10 0.90697
+stage 24/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.91661  early_stopping: 0/10 0.91661
+stage 25/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.91275  early_stopping: 1/10 0.91661
+stage 26/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.91347  early_stopping: 2/10 0.91661
+stage 27/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.91685  early_stopping: 0/10 0.91685
+stage 28/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:16 val_accuracy: 0.91203  early_stopping: 1/10 0.91685
+stage 29/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.91130  early_stopping: 2/10 0.91685
+stage 30/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.90359  early_stopping: 3/10 0.91685
+stage 31/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.88961  early_stopping: 4/10 0.91685
+stage 32/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.90986  early_stopping: 5/10 0.91685
+stage 33/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.90793  early_stopping: 6/10 0.91685
+stage 34/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:16 val_accuracy: 0.90239  early_stopping: 7/10 0.91685
+stage 35/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.88792  early_stopping: 8/10 0.91685
+stage 36/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.91179  early_stopping: 9/10 0.91685
+stage 37/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 257/257 0:00:00 0:00:17 val_accuracy: 0.90721  early_stopping: 10/10 0.91685
+Moving best model /home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/peraireB_ft_MMCFR_27.mlmodel (0.9168474078178406) to /home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/peraireB_ft_MMCFR_best.mlmodel
diff --git a/expe_both/kraken_D.log b/expe_both/kraken_D.log
new file mode 100644
index 0000000..fc9a8cf
--- /dev/null
+++ b/expe_both/kraken_D.log
@@ -0,0 +1,74 @@
+┏━━━━┳━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┓
+┃    ┃ Name      ┃ Type                     ┃ Params ┃                 In sizes ┃                Out sizes ┃
+┡━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━┩
+│ 0  │ val_cer   │ CharErrorRate            │      0 │                        ? │                        ? │
+│ 1  │ net       │ MultiParamSequential     │  4.0 M │  [[1, 1, 120, 400], '?'] │   [[1, 117, 1, 50], '?'] │
+│ 2  │ net.C_0   │ ActConv2D                │  1.3 K │  [[1, 1, 120, 400], '?'] │ [[1, 32, 120, 400], '?'] │
+│ 3  │ net.Do_1  │ Dropout                  │      0 │ [[1, 32, 120, 400], '?'] │ [[1, 32, 120, 400], '?'] │
+│ 4  │ net.Mp_2  │ MaxPool                  │      0 │ [[1, 32, 120, 400], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 5  │ net.C_3   │ ActConv2D                │ 40.0 K │  [[1, 32, 60, 200], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 6  │ net.Do_4  │ Dropout                  │      0 │  [[1, 32, 60, 200], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 7  │ net.Mp_5  │ MaxPool                  │      0 │  [[1, 32, 60, 200], '?'] │  [[1, 32, 30, 100], '?'] │
+│ 8  │ net.C_6   │ ActConv2D                │ 55.4 K │  [[1, 32, 30, 100], '?'] │  [[1, 64, 30, 100], '?'] │
+│ 9  │ net.Do_7  │ Dropout                  │      0 │  [[1, 64, 30, 100], '?'] │  [[1, 64, 30, 100], '?'] │
+│ 10 │ net.Mp_8  │ MaxPool                  │      0 │  [[1, 64, 30, 100], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 11 │ net.C_9   │ ActConv2D                │  110 K │   [[1, 64, 15, 50], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 12 │ net.Do_10 │ Dropout                  │      0 │   [[1, 64, 15, 50], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 13 │ net.S_11  │ Reshape                  │      0 │   [[1, 64, 15, 50], '?'] │   [[1, 960, 1, 50], '?'] │
+│ 14 │ net.L_12  │ TransposedSummarizingRNN │  1.9 M │   [[1, 960, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 15 │ net.Do_13 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 16 │ net.L_14  │ TransposedSummarizingRNN │  963 K │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 17 │ net.Do_15 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 18 │ net.L_16  │ TransposedSummarizingRNN │  963 K │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 19 │ net.Do_17 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 20 │ net.O_18  │ LinSoftmax               │ 46.9 K │   [[1, 400, 1, 50], '?'] │   [[1, 117, 1, 50], '?'] │
+└────┴───────────┴──────────────────────────┴────────┴──────────────────────────┴──────────────────────────┘
+Trainable params: 4.0 M                                                                                                                                                                                                                                                                 
+Non-trainable params: 0                                                                                                                                                                                                                                                                 
+Total params: 4.0 M                                                                                                                                                                                                                                                                     
+Total estimated model params size (MB): 16                                                                                                                                                                                                                                              
+stage 0/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.87184  early_stopping: 0/10 0.87184
+stage 1/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:26 val_accuracy: 0.87122  early_stopping: 1/10 0.87184
+stage 2/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.87906  early_stopping: 0/10 0.87906
+stage 3/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.90494  early_stopping: 0/10 0.90494
+stage 4/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.90353  early_stopping: 1/10 0.90494
+stage 5/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.91655  early_stopping: 0/10 0.91655
+stage 6/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.92329  early_stopping: 0/10 0.92329
+stage 7/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.88220  early_stopping: 1/10 0.92329
+stage 8/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:26 val_accuracy: 0.89725  early_stopping: 2/10 0.92329
+stage 9/∞  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:26 val_accuracy: 0.91122  early_stopping: 3/10 0.92329
+stage 10/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:26 val_accuracy: 0.92816  early_stopping: 0/10 0.92816
+stage 11/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:26 val_accuracy: 0.93349  early_stopping: 0/10 0.93349
+stage 12/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.92016  early_stopping: 1/10 0.93349
+stage 13/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.90761  early_stopping: 2/10 0.93349
+stage 14/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:26 val_accuracy: 0.93725  early_stopping: 0/10 0.93725
+stage 15/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:26 val_accuracy: 0.91655  early_stopping: 1/10 0.93725
+stage 16/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.90839  early_stopping: 2/10 0.93725
+stage 17/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:26 val_accuracy: 0.93412  early_stopping: 3/10 0.93725
+stage 18/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.91545  early_stopping: 4/10 0.93725
+stage 19/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.91294  early_stopping: 5/10 0.93725
+stage 20/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.91467  early_stopping: 6/10 0.93725
+stage 21/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.90855  early_stopping: 7/10 0.93725
+stage 22/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.92251  early_stopping: 8/10 0.93725
+stage 23/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:26 val_accuracy: 0.92329  early_stopping: 9/10 0.93725
+stage 24/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:26 val_accuracy: 0.93898  early_stopping: 0/10 0.93898
+stage 25/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:26 val_accuracy: 0.94729  early_stopping: 0/10 0.94729
+stage 26/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:26 val_accuracy: 0.94259  early_stopping: 1/10 0.94729
+stage 27/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:26 val_accuracy: 0.92518  early_stopping: 2/10 0.94729
+stage 28/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.92941  early_stopping: 3/10 0.94729
+stage 29/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:26 val_accuracy: 0.92329  early_stopping: 4/10 0.94729
+stage 30/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.93443  early_stopping: 5/10 0.94729
+stage 31/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.92659  early_stopping: 6/10 0.94729
+stage 32/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.93239  early_stopping: 7/10 0.94729
+stage 33/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.95435  early_stopping: 0/10 0.95435
+stage 34/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:26 val_accuracy: 0.93067  early_stopping: 1/10 0.95435
+stage 35/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.94290  early_stopping: 2/10 0.95435
+stage 36/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:26 val_accuracy: 0.93522  early_stopping: 3/10 0.95435
+stage 37/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.94165  early_stopping: 4/10 0.95435
+stage 38/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:26 val_accuracy: 0.92988  early_stopping: 5/10 0.95435
+stage 39/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.95263  early_stopping: 6/10 0.95435
+stage 40/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.94463  early_stopping: 7/10 0.95435
+stage 41/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.93537  early_stopping: 8/10 0.95435
+stage 42/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.92533  early_stopping: 9/10 0.95435
+stage 43/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 313/313 0:00:00 0:00:25 val_accuracy: 0.90808  early_stopping: 10/10 0.95435
+Moving best model /home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/peraireD_ft_MMCFR_33.mlmodel (0.9543529152870178) to /home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/peraireD_ft_MMCFR_best.mlmodel
diff --git a/expe_both/readme.md b/expe_both/readme.md
new file mode 100644
index 0000000..0fbf7b8
--- /dev/null
+++ b/expe_both/readme.md
@@ -0,0 +1,268 @@
+`Peraire_ft_MMCFR` was trained finetuning the `Manu McFrench` model with the data stored in data/train/ using the eScriptorium web application and Kraken v. 3.0.13.
+
+
+`Peraire_ft_MMCFR` accuracy during training is 93.0%. It reaches an accuracy of 93.3% in the test files.
+
+## KaMI-App evaluation
+
+> (ran on August 12 with kamilib 0.1.1-light)
+
+|FIELD1                                |Default|Ignoring digits|Ignoring case|Ignoring punctuation|Ignoring diacritics|Combining all options|
+|--------------------------------------|-------|---------------|-------------|--------------------|-------------------|---------------------|
+|Levensthein Distance (Char.)          |378    |373            |373          |319                 |336                |266                  |
+|Levensthein Distance (Words)          |256    |251            |252          |211                 |227                |169                  |
+|Hamming Distance                      |Ø      |Ø              |Ø            |Ø                   |Ø                  |Ø                    |
+|Word Error Rate (WER in %)            |26.947 |26.873         |26.526       |22.542              |23.894             |18.51                |
+|Char. Error Rate (CER in %)           |6.723  |6.731          |6.634        |5.909               |6.095              |5.108                |
+|Word Accuracy (Wacc in %)             |73.052 |73.126         |73.473       |77.457              |76.105             |81.489               |
+|Match Error Rate (MER in %)           |6.645  |6.652          |6.557        |5.853               |6.031              |5.065                |
+|Char. Information Lost (CIL in %)     |10.52  |10.45          |10.352       |9.328               |9.863              |8.327                |
+|Char. Information Preserved (CIP in %)|89.479 |89.549         |89.647       |90.671              |90.136             |91.672               |
+|Hits                                  |5310   |5234           |5315         |5131                |5235               |4985                 |
+|Substitutions                         |229    |221            |224          |196                 |222                |177                  |
+|Deletions                             |83     |86             |83           |71                  |55                 |45                   |
+|Insertions                            |66     |66             |66           |52                  |59                 |44                   |
+|Total char. in reference              |5622   |5541           |5622         |5398                |5512               |5207                 |
+|Total char. in prediction             |5605   |5521           |5605         |5379                |5516               |5206                 |
+
+
+<!--
+
+## Reference:
+```
+Je suis reçu par la SUPERIEURE, une réligieuse d'
+une cinquantaine d'année aussi et une française. Elle me
+demande le but de mon voyage, et ne peut arriver à
+croire que je puisse ainsi me faire comprendre grâce
+à l'Esperanto. Elle doute beaucoup, qu'un jour
+l'Esperanto, sera obligatoirement enseigné dans toutes
+les Ecoles des pays civilisés, helas! Elle est réaliste.
+NOTA 1930 et 1966
+TEL EST POURTANT NOTRE BUT: tant au point
+de vue RATIONEL, qu#'IDEOLOGIQUE. Etant une SYNTHESE
+des langues modernes vivantes, l'étude de l'Esperanto éviterait
+des pertes INSOUPÇONNÉES de temps, que l'on pourrait
+dédier ainsi à une vraie CULTURE GENERALE INTER-
+-PEUPLES ET INTERPOPULAIRE. Je suis toujours convain-
+-cu que cette mesure serait un FACTEUR IMPORTANT
+de PAIX INTERNATIONALE. MAIS LA VEUX-T-ON
+CETTE PAIX REELLEMENT ? (fin du nota)
+Pour ma part, me dit-elle, vous soulevez
+un voile devant mes yeux. Et en tout cas vous
+obtiendrez un résultat INESPÉRÉ si vous pouviez
+REELLEMENT CONVAINCRE LE PAPE et son entourage
+(J'en suis convaincu aussi, mais comme je l'ai dit
+haut
+plus x veut-on réellement dans tous les milieux
+ORGANISER LA PAIX ? Si oui qui veut la
+fin, veut les moyens, et l'Esperanto peut-être
+sûrement UN DE CES MOYENS.).
+46
+esperantigo numenata
+MANIFOLD N°18 commencé à Libourne le 3-11-68.
+Fini
+le 24-4-69
+la 21an. 1.75a
+extrait du carnet steno N°16 (suite de la page 50 du manifold N°17)
+TCHAN-DGIA LE LUNDI 7 DECEMBRE 1930. Pour conclure cette soirée....
+maussade, Mr LOU qui me quitte vers les 20h, me dit : - "En
+réalité en Chine, présentement l'on se distrait en cercles fermés
+dans les réceptions, que s'offrent les notables entre-eux ; le
+peuple fait de même, mais rien n'est prévu ni organisé
+présentement, pour le menu peuple ; nous avons trop à faire
+par ailleurs.".
+Hier dimanche, vers les 11 heures, Mr OU-I. le chef
+du KUO-MIN-TANG, m'amena chez le général en chef de la
+province. Je suis reçu, par un homme en uniforme, mais
+sans aucun insigne apparent: Notre premier contact est
+sympathique, il me reçoit sans pompe, sans céremonie, tout
+simplement. La résidence était gardée par tout un
+contingent de soldats, qui bien entendu, nous empêchent
+d'entrer. Apres quelques mots de Mr OU-I, ceux-ci se
+figent en un garde à vous impeccable; et nous introduisent
+dans le bureau d'un capitaine, qui nous saluent
+respectueusement et nous présente des sièges. Par telé,
+il avertit le général de notre visite. C'est un grand
+honneur que l'on me fait, et cependant, j'étais tres calme
+et me disais: - "Il te faut jouer serré". De ces quelques
+minutes dépend peut-être ma vie, ma liberté où tout au
+1
+il devra acheter tous ses bagages et papiers ; à moi il
+ne me demande rien. Plus loin, je croise une troupe de
+dames qui repend dans l'air un flot de parfums suaves et
+qui nous devisagerent d'un air moqueur. (Nous étions passablement
+crottés et mouillés).
+Dans un hôtel, on veut bien nous vendre
+un morceau de pain noir pour 7 kouronnes mais on refuse de
+nous loger pour la nuit. Nous continuons donc notre route.
+La nuit commence à tomber. Dans un pré, nous avisons
+quelques femmes qui gardent les vaches. Nous leur faisons
+comprendre que nous desirerions trouve un endroit sur pour
+pouvoir camper. Le premier geste de méfiance passé, ces braves
+paysannes nous acceptent. et nous font entrer dans la ferme. et
+complètent notre souper avec du lait. à peine sommes-nous
+couchés dans la grange qu'un formidable orage eclate et
+durera toute la nuit. Heureusement que nous avons 2 couvertures
+car le froid arrive très vif.
+MARDI 25 SEPTEMBRE 1928. Décidément nous garderons un souvenir
+profond de cette régions des Carpathes. La pluie continue
+et nous ne pouvons repartir que vers les 9h. Quand nous
+voulons dedommager cette brave femme, elle nous repond : "Le
+bon Dieu me le rendra". Sous la pluie qui diminue
+une 1/2h après P crève sa roue AV, et ça nous fait perdre
+1h pour réparer. Entre temps le soleil s'est levé, et vers
+10h nous arrivons au village de JASINA. Bientôt la
+route est coupée par les eaux. Nous sommes obligés d'accomplir
+de véritables tous d'acrobatie pour traverses ce goulet.
+5
+les boiseries du poste d'équipage sont en bois
+exotiques vernis. Le parquet est en bois huilé et le
+plafond peint en blanc. Les 2/3 du salon sont séparés par
+une petite barriere et bois peint en blanc pour amenager dans
+l'autre tiers, le "coin de Lénine" ; c'est un cabinet avec piano
+des bancs et la radio
+Les 3 rangées de table de la salle à manger sont
+couvertes par des toiles cirées et le service est assuré par une
+femme. A cet effet, tout à côté il y a la buanderie, la
+cuisine (chauffée au mazout) (en 1929).
+J'ai egalement visité, l'atelier de mécanique et de menuiserie
+amenagés à bord.
+LUNDI 20 MARS 1929. Hier j'ai visité à environ 10k au
+sud de Novorassisk, au bord de la mer, un magnifique
+vignoble. Il y a un petit village de paysans où vivent
+là
+une cinquantaine de personnes. Avant la Revolution ce
+vignoble appartenait à un propriétaire, et il est devenu proprié-
+-té d'Etat. Il y a là quelque 200 à 300 hectares de
+terres bien exposées au sud. sur le flanc d'une colline
+située au Nord. Elle est travaillée par les anciens ouvriers
+agricoles et par des "artels" = coopé au premier degré, d'ouvriers
+Le "chateau" - comme en France est une superbe
+(toute neuve)
+bâtisse située au milieu du vignoble, et la vinification.
+comme en France - se fait dans des bâtiments qu'il
+entourent
+En ce moment, le "chateau" est occupe par
+uune colonie de "BEZPRIOZORNII" - anciens enfants
+43
+```
+
+## Prediction
+
+```
+Je suis reçu par la SUPERIEVRE, une réligieuse d
+une conquantaine d'anée aussi, et une française. Alle me
+demande le but de mon voyage, et ne peut arriver à
+croué que je puuisse ainsi me faue comprendre grace
+à l'Esperanto. Elle doute beaucoup, qu'un jour
+l'Eperante, sera obligatoirement enseigné dans toutes
+les Ecoles des pays civlisés, helas. Elle est réaliste.
+NOTA 1930 et 1906
+TEL EST POURTANT NOTRE BOT, tant au point
+do vuie RATIONEL, Que IDEOLOGIQUE. Etant une SUNTHESE
+Des lanques modernes vivantes, l'étude de l'Espesanto éviteran
+des pertes INSOUECONNCES de temps, que l'on pourrait
+dedier ainsi à une vade CIITUNE GEMEARLE IMTED-
+PEIPLES ET INTERPOEULAIRE. Je suis toujours convain-
+c, que cette mesure serait un FACTEUR IMIPORTANT
+de PAIK ITITERHATIONALE. MAIS LA VEUK-T-ON
+CETTE PAIK REELLEMENT ? (fun du nota)
+Pour ma part, me dit-elle, vous soulevez
+un voile devant mes yeux. Et en tout cas vous
+obtiendriez un résultat MHESPERE. Si vous pouviez
+REEHEHENT CONVAINCRE LE PAVE et son entourage
+J'en suis convainen aussi, mais, comme je l'ai dit
+hauit
+plus rveut-on reellement dans tous les milieux
+ORGANISER LA DAIK . Si oui qui veut la
+fon, veut les moyens, et l'Esperanto peut être
+surement UN DE CES MIOVENS.).
+46
+esperankigs sumenita
+MANICOLO N° 18, commencé à Lbourse le 3-1168.
+Sini
+le 24-4-68
+la 210-1.750
+extrait du carnet steno N°16- (Suite de la page 50 du manifold N°17)
+(HAN- DENA LE LUNOI 7 DECENBRE 1930. Pour conclure cette soirée...
+maussade. Mr 2Oi qui me quitte vers les 20h, me dit: -lEn
+réalité en Chine, présentement l'on se distrait en cercles fermes
+dans les réceptions, que s'offrent les notables entre eux: le
+peuple fait de même, mais rien n'est prévu m'organisé
+présentement, pour le menu, peuple; nous avans trop à faire
+par ailleurs ?.
+Mier dimanche, vers les, 11 heures, Mr OU -1- le chef
+du KUO-HIN-TANE, m'amena chez le général en chef de la
+province. Je suis reçu, par un homme en uniforme, mais
+sans aucun insigne apparent. Notre premier contait est
+sympathique, il me reçoit sans pompe, sans ceremonié, tout
+simplement. La résidence était gardée par tout un
+contingent de soldats, qui bien entendu, nous empêchent
+d'entrer. Après quelques mots de Mr OU-1-, ceux- ci se
+figent en un garde à vous impeccable., et nous introduisent
+dans le bureau d'un capitaine, qui, nous saluent
+respectueusement, et nous presente des sièges. Par telé,
+il avertit le général de notre visite. C'est un grand
+honneur que l'on me fait, et cependant, j'étois tres calme
+et me disais: -" Il te faut jouer serré". De ces quelques
+minutes, dépend peut être mu rie, ma leiberte où tout au
+1
+il dera extuber tous ses bagages et papiers; à moi il
+ne me temande aien. Plus loin p croise une troupe de
+anses qui apenit beons l'anr, un pot de parfirms sueuves et
+ui nous devisagerent d'un air moqueur. (Nous etions possablement
+crottés et mouillés).
+Dans un hôtel  veut bien nous venitre
+un morceau de pain nour pour I kouronnes mais on refusse de
+nous lager pour la nuit. Nous continuons dine notre route.
+La nuit commence à tomber. Dans un pre, nous avisons
+quilques femmes qui gardent les vaches. Nous deur faisons
+comprendre que nous désirerions trouver un enirois see pour
+pouvour camper.. Le premier geste de mifiane passé, ces traves
+paysannes nous acceptent. et nous font entrer dans la ferme, et
+completent notre souper auc du lait. a peiné sommes nous
+couches (dans la gronge qu'un formidable orage eclate, et
+detrera toute la niit, Neurcusement que nous avons 2 couvertures
+car le froid arrive très vyj..
+MARN 25 SEPTENSRE 1928. Dieullemment, nous gartiron un souvenier
+profanndt de cette regions des Carpathes.. La pluie continne
+et nous ne pouvons réparter que vers les 5h. Qurandt nous.
+poulons dedommager cette brave femme, elle nous repond. Le
+bon Dieu me le rendrà. Sous la pluie qui diminue
+une 12 h après P crève su roue N. et ca nous fait perdre
+1h pour reparer. Entre semps le solul s'est levé, et vers
+1oh nous arrivons au village de JASIMA. Bientot la
+route est coupée par les caux. Nous sommes obliges dte'auomplir
+de veritabtes tours l'acrobatie pour travorsir ce goulet.
+5
+Les bosseries du poste d'équipage sont en bois
+exotiques vernis. Le parquet est en bois huité et le
+plafend peint en blanc.. Les 2/3 du salon sont séparés par
+une patite barrière et bois peint en blane pour amenager dans
+l'autre tions, le "coin de Léniné" ; c'est un cabinet avec piano
+des banes et la nadio
+Les 3 rangées de table de la salle à manger sont
+couvertes par des toiles corces et le service est assuré par une
+femme. A cet effet, tout à coté il y a la buanclerie, la
+cuisine (chauffée au mazant) Cen 1929)
+J'ai egalement vrsuté. l'atelier de mécanique et de menuusert
+aménages à bort.
+LONOI 20 MARS 1929. Mier j'ai visité à environ 10k au
+suit de illivorassisk, au bord de la mer, un magnifique
+vignoble. Il y a un petit village de paysans où ritent
+là
+une cinquantaine de personnes. Avant la Révolution ce
+vignoble appartenait à un propriétaire, et il est divenu proprie-
+-te d'Etat. Il j'a là quelque 200 à 300 hectares de
+terres bien expotes au sud. sur le flane d'une colliné
+située au evord... Elle est travaillée par les anciens ouvriers
+agucoles et par des "'artils - coapé au premier degré, d'ouvriers
+Le "chateau" - comme en France est une superte
+toute neuve)
+batisse située au milicu du vignoble, et la vinification,
+comme en Frence - se fait dans des bâtiments quil
+entourent
+En ce moment, le "chateau" est occupe par
+uune cotonie de SEEPRIEORNI - anciens enfants
+43
+```
\ No newline at end of file
diff --git a/expe_both/test.files b/expe_both/test.files
new file mode 100644
index 0000000..03e60f4
--- /dev/null
+++ b/expe_both/test.files
@@ -0,0 +1,8 @@
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/test/test_B.12.japon3_0048.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/test/test_B.18.chine4_0003.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/test/test_B.3.pologne-urss1_0007.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/test/test_B.5.urss3_0045.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/test/test_D.3.eurasie-5-6_0018.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/test/test_D.4.eurasie-7-8_0077.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/test/test_D.7.eurasie-13-14_0056.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/test/test_D.8.eurasie-15_0062.xml
diff --git a/expe_both/train.files b/expe_both/train.files
new file mode 100644
index 0000000..950765c
--- /dev/null
+++ b/expe_both/train.files
@@ -0,0 +1,67 @@
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.1.intro-eurasie_0005.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.10.chine2_0020.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.11.japon2_0043.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.12.japon3_0014.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.13.japon3b_0048.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.14.japon4_0005.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.15.japon5_0009.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.16.chine2b_0014.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.17.chine3_0003.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.18.chine4_0024.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.19.chine5_0035.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.1b.europe-centrale_0005.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.2.europe-orientale_0007.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.2.europe-orientale_0029.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.20.chine6_0051.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.21.chine7_0022.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.22.indochine2_0025.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.23.indochine3_0018.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.24.cambodge_0037.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.25.siam_0003.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.26.malais_0048.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.27.java1_0006.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.28.java2_0017.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.29.fin_0025.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.3.pologne-urss1_0026.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.30.tdm_0014.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.30.tdm_0023.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.4.urss2_0017.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.5.urss3_0003.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.6.urss4_0050.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.7.urss5_0019.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.8.urss6_0049.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/B.9.chine1_0052.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.1.notes-1-et-2_0010.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.1.notes-1-et-2_0011.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.1.notes-1-et-2_0012.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.1.notes-1-et-2_0013.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.2.eurasie-3-4_0016.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.2.eurasie-3-4_0017.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.2.eurasie-3-4_0018.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.2.eurasie-3-4_0019.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.2.eurasie-3-4_0020.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.3.eurasie-5-6_0017.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.3.eurasie-5-6_0019.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.3.eurasie-5-6_0020.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.4.eurasie-7-8_0078.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.4.eurasie-7-8_0079.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.4.eurasie-7-8_0080.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.5.eurasie-9-10_0058.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.5.eurasie-9-10_0059.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.5.eurasie-9-10_0060.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.5.eurasie-9-10_0061.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.6.eurasie-11-12_0004.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.6.eurasie-11-12_0005.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.6.eurasie-11-12_0006.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.6.eurasie-11-12_0007.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.7.eurasie-13-14_0055.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.7.eurasie-13-14_0057.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.7.eurasie-13-14_0058.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.8.eurasie-15_0061.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.8.eurasie-15_0063.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.8.eurasie-15_0064.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.9.eurasie-16-17_0013.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.9.eurasie-16-17_0014.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.9.eurasie-16-17_0058.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.9.eurasie-16-17_0059.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/models/../data/train/D.9.eurasie-16-17_0100.xml
diff --git a/expe_both/train.sh b/expe_both/train.sh
new file mode 100755
index 0000000..051ab01
--- /dev/null
+++ b/expe_both/train.sh
@@ -0,0 +1,102 @@
+#!/bin/bash
+
+source /home/ROCQ/almanach/achague/anaconda3/etc/profile.d/conda.sh # for conda envs
+
+echo "Stage 1 of 4: Identifying the datasets"
+
+#playground=$(pwd)/playground
+echo "Building playground if necessary"
+#mkdir -p $playground
+
+# test if the playground directory exists
+#if [ ! -d "$playground" ]; then
+#    echo "The playground directory does not exist. Please run step 1 of the workflow first."
+#    exit 1
+#fi
+
+ls -d $PWD/../data/train/*.xml > "$PWD/train.files"
+ls -d $PWD/../data/test/*.xml > "$PWD/test.files"
+
+echo "Stage 1 of 4: Identifying the datasets - DONE"
+
+# -------------------------------------------
+
+echo "Stage 2 of 4: Compiling the datasets into arrow files"
+echo "This may take a while."
+
+n_workers=16 
+batch_size=2 # the greater the number, the more general the models will be
+
+echo "Activating Kraken conda environment"
+conda activate kraken
+
+# if "kraken --version" does not work, stop the script
+if ! kraken --version; then
+    echo "Couldn't find kraken. Please check that the conda environment is activated."
+    exit 1
+fi
+
+echo "Working with $(kraken --version)"
+
+echo "Compiling the datasets into arrow files"
+
+echo "Compiling the train and dev files into arrow files"
+cat $PWD/train.files | xargs ketos compile --workers $n_workers -o "$PWD/train.arrow" --random-split .9 .1 .0 --save-splits -f alto
+
+echo "Compiling the test files into arrow files"
+cat $PWD/test.files | xargs ketos compile --workers $n_workers -o "$PWD/test.arrow" --random-split .0 .0 1.0 --save-splits -f alto
+
+#for file in $PWD/*.files
+#do
+#    base=$(basename $file .files)
+#    #cat $file | xargs ketos compile --workers $n_workers -o "$playground/$base.arrow" --random-split .9 .1 .0 --save-splits -f alto 
+#    if [[ $base != *"test" ]]
+#    then
+#        echo "Compiling the train and dev files into arrow files"
+#        cat $file | xargs ketos compile --workers $n_workers -o "$PWD/$base.arrow" --random-split .9 .1 .0 --save-splits -f alto 
+#    fi
+#    if [[ $base == *"test" ]]
+#    then
+#        echo "Compiling the test files into arrow files"
+#        cat $file | xargs ketos compile --workers $n_workers -o "$PWD/$base.arrow" --random-split .0 .0 1.0 --save-splits -f alto 
+#    fi
+#done
+
+echo "Making a list of dev and train arrow files for each datasets" # cf. issue #397 of kraken
+ls -d $PWD/*.arrow > "arrow.files"
+
+echo "Stage 2 of 4: Compiling the datasets into arrow files - DONE"
+
+# -------------------------------------------
+
+echo "Stage 3 of 4: Training Kraken models"
+echo "This may take a while."
+
+
+n_workers=16 
+#n_epochs=50 # 4 for test purposes, 50 for real
+batch_size=2 # the greater the number, the more general the models will be
+#seeds="42 123 456 789 1011"
+
+
+echo "Start training the models"
+
+#cat $file | xargs ketos -s $seed train -o $playground/$base/models/$seed-$base -f binary --device cuda:0 -u NFC -s '[1,120,0,1 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 S1(1x0)1,3 Lbx200 Do0.1,2 Lbx200 Do.1,2 Lbx200 Do]' -r 0.0001 -B $batch_size --augment --workers $n_workers --epochs $n_epochs -q dumb --logger tensorboard --log-dir $playground/$base > "$playground/$base/logs/$seed-$base.log"
+#ls "$PWD/train.arrow" | xargs ketos train -o $PWD/peraire2_ft_MMCFR -f binary --device cuda:0 -u NFC -s '[1,120,0,1 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 S1(1x0)1,3 Lbx200 Do0.1,2 Lbx200 Do.1,2 Lbx200 Do]' -r 0.0001 -B $batch_size --augment --workers $n_workers > "$PWD/kraken.log"
+ls "$PWD/train.arrow" | xargs ketos train -o $PWD/peraire2_ft_MMCFR --load HTR-United-Manu_McFrench.mlmodel --resize add  -f binary --device cuda:0 -u NFC -s '[1,120,0,1 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 S1(1x0)1,3 Lbx200 Do0.1,2 Lbx200 Do.1,2 Lbx200 Do]' -r 0.0001 -B $batch_size --augment --workers $n_workers --lag 15 > "$PWD/kraken.log"
+
+conda deactivate
+
+echo "Stage 3 of 4: Training Kraken models - DONE"
+
+# -------------------------------------------
+
+echo "Stage 4 of 4: Testing the models"
+
+for model in $PWD/*.mlmodel
+do
+    basemodel=$(basename $model .mlmodel)
+    echo "Testing $model"
+    #ketos test -m $model -e "$playground/test_files.txt" -f alto --device cuda:0 -u NFC -B $batch_size --workers $n_workers > "$folder/tests/model_$basemodel.test"
+    ls "$PWD/test.arrow" | xargs ketos test -m $model -f binary --device cuda:0 -u NFC -B $batch_size --workers $n_workers > "$PWD/model_$basemodel.test"
+done
diff --git a/kraken_BD_add.log b/kraken_BD_add.log
new file mode 100644
index 0000000..0f0ae0a
--- /dev/null
+++ b/kraken_BD_add.log
@@ -0,0 +1,112 @@
+┏━━━━┳━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┓
+┃    ┃ Name      ┃ Type                     ┃ Params ┃                 In sizes ┃                Out sizes ┃
+┡━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━┩
+│ 0  │ val_cer   │ _CharErrorRate           │      0 │                        ? │                        ? │
+│ 1  │ val_wer   │ _WordErrorRate           │      0 │                        ? │                        ? │
+│ 2  │ net       │ MultiParamSequential     │  4.1 M │  [[1, 1, 120, 400], '?'] │   [[1, 166, 1, 50], '?'] │
+│ 3  │ net.C_0   │ ActConv2D                │  1.3 K │  [[1, 1, 120, 400], '?'] │ [[1, 32, 120, 400], '?'] │
+│ 4  │ net.Do_1  │ Dropout                  │      0 │ [[1, 32, 120, 400], '?'] │ [[1, 32, 120, 400], '?'] │
+│ 5  │ net.Mp_2  │ MaxPool                  │      0 │ [[1, 32, 120, 400], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 6  │ net.C_3   │ ActConv2D                │ 40.0 K │  [[1, 32, 60, 200], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 7  │ net.Do_4  │ Dropout                  │      0 │  [[1, 32, 60, 200], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 8  │ net.Mp_5  │ MaxPool                  │      0 │  [[1, 32, 60, 200], '?'] │  [[1, 32, 30, 100], '?'] │
+│ 9  │ net.C_6   │ ActConv2D                │ 55.4 K │  [[1, 32, 30, 100], '?'] │  [[1, 64, 30, 100], '?'] │
+│ 10 │ net.Do_7  │ Dropout                  │      0 │  [[1, 64, 30, 100], '?'] │  [[1, 64, 30, 100], '?'] │
+│ 11 │ net.Mp_8  │ MaxPool                  │      0 │  [[1, 64, 30, 100], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 12 │ net.C_9   │ ActConv2D                │  110 K │   [[1, 64, 15, 50], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 13 │ net.Do_10 │ Dropout                  │      0 │   [[1, 64, 15, 50], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 14 │ net.S_11  │ Reshape                  │      0 │   [[1, 64, 15, 50], '?'] │   [[1, 960, 1, 50], '?'] │
+│ 15 │ net.L_12  │ TransposedSummarizingRNN │  1.9 M │   [[1, 960, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 16 │ net.Do_13 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 17 │ net.L_14  │ TransposedSummarizingRNN │  963 K │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 18 │ net.Do_15 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 19 │ net.L_16  │ TransposedSummarizingRNN │  963 K │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 20 │ net.Do_17 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 21 │ net.O_18  │ LinSoftmax               │ 66.6 K │   [[1, 400, 1, 50], '?'] │   [[1, 166, 1, 50], '?'] │
+└────┴───────────┴──────────────────────────┴────────┴──────────────────────────┴──────────────────────────┘
+Trainable params: 4.1 M                                                                                                                                                                                                                                                                    
+Non-trainable params: 0                                                                                                                                                                                                                                                                    
+Total params: 4.1 M                                                                                                                                                                                                                                                                        
+Total estimated model params size (MB): 16                                                                                                                                                                                                                                                 
+stage 0/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.37it/s val_accuracy: 0.863 val_word_accuracy: 0.533  early_stopping: 0/10 0.86251
+stage 1/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.22it/s val_accuracy: 0.88 val_word_accuracy: 0.563  early_stopping: 0/10 0.87991
+stage 2/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.29it/s val_accuracy: 0.888 val_word_accuracy: 0.58  early_stopping: 0/10 0.88789
+stage 3/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.22it/s val_accuracy: 0.898 val_word_accuracy: 0.603  early_stopping: 0/10 0.89804
+stage 4/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.36it/s val_accuracy: 0.908 val_word_accuracy: 0.636  early_stopping: 0/10 0.90805
+stage 5/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.22it/s val_accuracy: 0.916 val_word_accuracy: 0.649  early_stopping: 0/10 0.91588
+stage 6/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.28it/s val_accuracy: 0.919 val_word_accuracy: 0.665  early_stopping: 0/10 0.91922
+stage 7/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.23it/s val_accuracy: 0.923 val_word_accuracy: 0.685  early_stopping: 0/10 0.92313
+stage 8/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.15it/s val_accuracy: 0.927 val_word_accuracy: 0.688  early_stopping: 0/10 0.92690
+stage 9/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.34it/s val_accuracy: 0.93 val_word_accuracy: 0.695  early_stopping: 0/10 0.93009
+stage 10/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.39it/s val_accuracy: 0.932 val_word_accuracy: 0.701  early_stopping: 0/10 0.93169
+stage 11/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.23it/s val_accuracy: 0.935 val_word_accuracy: 0.719  early_stopping: 0/10 0.93474
+stage 12/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.14it/s val_accuracy: 0.936 val_word_accuracy: 0.718  early_stopping: 0/10 0.93619
+stage 13/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.15it/s val_accuracy: 0.938 val_word_accuracy: 0.731  early_stopping: 0/10 0.93836
+stage 14/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.20it/s val_accuracy: 0.938 val_word_accuracy: 0.728  early_stopping: 1/10 0.93836
+stage 15/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.12it/s val_accuracy: 0.939 val_word_accuracy: 0.733  early_stopping: 0/10 0.93938
+stage 16/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.18it/s val_accuracy: 0.941 val_word_accuracy: 0.74  early_stopping: 0/10 0.94054
+stage 17/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.34it/s val_accuracy: 0.942 val_word_accuracy: 0.741  early_stopping: 0/10 0.94155
+stage 18/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.23it/s val_accuracy: 0.941 val_word_accuracy: 0.742  early_stopping: 1/10 0.94155
+stage 19/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.43it/s val_accuracy: 0.944 val_word_accuracy: 0.758  early_stopping: 0/10 0.94387
+stage 20/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.28it/s val_accuracy: 0.944 val_word_accuracy: 0.751  early_stopping: 0/10 0.94416
+stage 21/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.20it/s val_accuracy: 0.944 val_word_accuracy: 0.751  early_stopping: 0/10 0.94445
+stage 22/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.24it/s val_accuracy: 0.946 val_word_accuracy: 0.757  early_stopping: 0/10 0.94619
+stage 23/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.14it/s val_accuracy: 0.946 val_word_accuracy: 0.762  early_stopping: 0/10 0.94648
+stage 24/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.35it/s val_accuracy: 0.945 val_word_accuracy: 0.758  early_stopping: 1/10 0.94648
+stage 25/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.24it/s val_accuracy: 0.945 val_word_accuracy: 0.757  early_stopping: 2/10 0.94648
+stage 26/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.30it/s val_accuracy: 0.945 val_word_accuracy: 0.763  early_stopping: 3/10 0.94648
+stage 27/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.24it/s val_accuracy: 0.947 val_word_accuracy: 0.763  early_stopping: 0/10 0.94692
+stage 28/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.31it/s val_accuracy: 0.947 val_word_accuracy: 0.763  early_stopping: 1/10 0.94692
+stage 29/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.27it/s val_accuracy: 0.948 val_word_accuracy: 0.766  early_stopping: 0/10 0.94764
+stage 30/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.36it/s val_accuracy: 0.947 val_word_accuracy: 0.766  early_stopping: 1/10 0.94764
+stage 31/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.20it/s val_accuracy: 0.95 val_word_accuracy: 0.78  early_stopping: 0/10 0.94982
+stage 32/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.27it/s val_accuracy: 0.949 val_word_accuracy: 0.773  early_stopping: 1/10 0.94982
+stage 33/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.14it/s val_accuracy: 0.948 val_word_accuracy: 0.77  early_stopping: 2/10 0.94982
+stage 34/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.26it/s val_accuracy: 0.947 val_word_accuracy: 0.77  early_stopping: 3/10 0.94982
+stage 35/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.25it/s val_accuracy: 0.949 val_word_accuracy: 0.77  early_stopping: 4/10 0.94982
+stage 36/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.20it/s val_accuracy: 0.95 val_word_accuracy: 0.78  early_stopping: 0/10 0.95025
+stage 37/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.19it/s val_accuracy: 0.95 val_word_accuracy: 0.782  early_stopping: 0/10 0.95040
+stage 38/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.31it/s val_accuracy: 0.952 val_word_accuracy: 0.782  early_stopping: 0/10 0.95170
+stage 39/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.34it/s val_accuracy: 0.951 val_word_accuracy: 0.78  early_stopping: 1/10 0.95170
+stage 40/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.29it/s val_accuracy: 0.952 val_word_accuracy: 0.79  early_stopping: 0/10 0.95228
+stage 41/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.37it/s val_accuracy: 0.952 val_word_accuracy: 0.79  early_stopping: 1/10 0.95228
+stage 42/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.29it/s val_accuracy: 0.952 val_word_accuracy: 0.787  early_stopping: 2/10 0.95228
+stage 43/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.36it/s val_accuracy: 0.952 val_word_accuracy: 0.785  early_stopping: 0/10 0.95243
+stage 44/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.09it/s val_accuracy: 0.952 val_word_accuracy: 0.78  early_stopping: 1/10 0.95243
+stage 45/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.22it/s val_accuracy: 0.952 val_word_accuracy: 0.781  early_stopping: 2/10 0.95243
+stage 46/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.26it/s val_accuracy: 0.95 val_word_accuracy: 0.78  early_stopping: 3/10 0.95243
+stage 47/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.31it/s val_accuracy: 0.951 val_word_accuracy: 0.782  early_stopping: 4/10 0.95243
+stage 48/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.16it/s val_accuracy: 0.951 val_word_accuracy: 0.786  early_stopping: 5/10 0.95243
+stage 49/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.11it/s val_accuracy: 0.952 val_word_accuracy: 0.786  early_stopping: 6/10 0.95243
+stage 50/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.04it/s val_accuracy: 0.952 val_word_accuracy: 0.785  early_stopping: 7/10 0.95243
+stage 51/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.17it/s val_accuracy: 0.95 val_word_accuracy: 0.779  early_stopping: 8/10 0.95243
+stage 52/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.35it/s val_accuracy: 0.952 val_word_accuracy: 0.789  early_stopping: 9/10 0.95243
+stage 53/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.22it/s val_accuracy: 0.953 val_word_accuracy: 0.791  early_stopping: 0/10 0.95272
+stage 54/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.19it/s val_accuracy: 0.953 val_word_accuracy: 0.793  early_stopping: 1/10 0.95272
+stage 55/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.12it/s val_accuracy: 0.953 val_word_accuracy: 0.792  early_stopping: 0/10 0.95301
+stage 56/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.21it/s val_accuracy: 0.953 val_word_accuracy: 0.796  early_stopping: 1/10 0.95301
+stage 57/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.13it/s val_accuracy: 0.952 val_word_accuracy: 0.793  early_stopping: 2/10 0.95301
+stage 58/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.26it/s val_accuracy: 0.951 val_word_accuracy: 0.787  early_stopping: 3/10 0.95301
+stage 59/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.21it/s val_accuracy: 0.953 val_word_accuracy: 0.791  early_stopping: 4/10 0.95301
+stage 60/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.29it/s val_accuracy: 0.954 val_word_accuracy: 0.8  early_stopping: 0/10 0.95446
+stage 61/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.24it/s val_accuracy: 0.954 val_word_accuracy: 0.796  early_stopping: 1/10 0.95446
+stage 62/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.30it/s val_accuracy: 0.954 val_word_accuracy: 0.792  early_stopping: 2/10 0.95446
+stage 63/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.37it/s val_accuracy: 0.954 val_word_accuracy: 0.795  early_stopping: 3/10 0.95446
+stage 64/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.12it/s val_accuracy: 0.954 val_word_accuracy: 0.791  early_stopping: 4/10 0.95446
+stage 65/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.26it/s val_accuracy: 0.954 val_word_accuracy: 0.794  early_stopping: 5/10 0.95446
+stage 66/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.28it/s val_accuracy: 0.953 val_word_accuracy: 0.787  early_stopping: 6/10 0.95446
+stage 67/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.32it/s val_accuracy: 0.953 val_word_accuracy: 0.794  early_stopping: 7/10 0.95446
+stage 68/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.25it/s val_accuracy: 0.953 val_word_accuracy: 0.793  early_stopping: 8/10 0.95446
+stage 69/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.13it/s val_accuracy: 0.954 val_word_accuracy: 0.799  early_stopping: 9/10 0.95446
+stage 70/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.33it/s val_accuracy: 0.955 val_word_accuracy: 0.803  early_stopping: 0/10 0.95504
+stage 71/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.27it/s val_accuracy: 0.955 val_word_accuracy: 0.799  early_stopping: 1/10 0.95504
+stage 72/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.38it/s val_accuracy: 0.953 val_word_accuracy: 0.791  early_stopping: 2/10 0.95504
+stage 73/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.19it/s val_accuracy: 0.953 val_word_accuracy: 0.791  early_stopping: 3/10 0.95504
+stage 74/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.14it/s val_accuracy: 0.954 val_word_accuracy: 0.799  early_stopping: 4/10 0.95504
+stage 75/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.23it/s val_accuracy: 0.955 val_word_accuracy: 0.797  early_stopping: 5/10 0.95504
+stage 76/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.22it/s val_accuracy: 0.954 val_word_accuracy: 0.792  early_stopping: 6/10 0.95504
+stage 77/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.25it/s val_accuracy: 0.953 val_word_accuracy: 0.797  early_stopping: 7/10 0.95504
+stage 78/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.29it/s val_accuracy: 0.954 val_word_accuracy: 0.799  early_stopping: 8/10 0.95504
+stage 79/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.25it/s val_accuracy: 0.954 val_word_accuracy: 0.798  early_stopping: 9/10 0.95504
+stage 80/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:21 • 0:00:00 13.29it/s val_accuracy: 0.954 val_word_accuracy: 0.798  early_stopping: 10/10 0.95504
+Moving best model /home/ROCQ/almanach/achague/peraire/peraire-ground-truth/finetuned/peraire_BD_add__70.mlmodel (0.9550398588180542) to /home/ROCQ/almanach/achague/peraire/peraire-ground-truth/finetuned/peraire_BD_add__best.mlmodel
diff --git a/kraken_BD_scratch.log b/kraken_BD_scratch.log
new file mode 100644
index 0000000..2baa756
--- /dev/null
+++ b/kraken_BD_scratch.log
@@ -0,0 +1,46 @@
+┏━━━━┳━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┓
+┃    ┃ Name      ┃ Type                     ┃ Params ┃                 In sizes ┃                Out sizes ┃
+┡━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━┩
+│ 0  │ val_cer   │ _CharErrorRate           │      0 │                        ? │                        ? │
+│ 1  │ val_wer   │ _WordErrorRate           │      0 │                        ? │                        ? │
+│ 2  │ net       │ MultiParamSequential     │  4.0 M │  [[1, 1, 120, 400], '?'] │   [[1, 117, 1, 50], '?'] │
+│ 3  │ net.C_0   │ ActConv2D                │  1.3 K │  [[1, 1, 120, 400], '?'] │ [[1, 32, 120, 400], '?'] │
+│ 4  │ net.Do_1  │ Dropout                  │      0 │ [[1, 32, 120, 400], '?'] │ [[1, 32, 120, 400], '?'] │
+│ 5  │ net.Mp_2  │ MaxPool                  │      0 │ [[1, 32, 120, 400], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 6  │ net.C_3   │ ActConv2D                │ 40.0 K │  [[1, 32, 60, 200], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 7  │ net.Do_4  │ Dropout                  │      0 │  [[1, 32, 60, 200], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 8  │ net.Mp_5  │ MaxPool                  │      0 │  [[1, 32, 60, 200], '?'] │  [[1, 32, 30, 100], '?'] │
+│ 9  │ net.C_6   │ ActConv2D                │ 55.4 K │  [[1, 32, 30, 100], '?'] │  [[1, 64, 30, 100], '?'] │
+│ 10 │ net.Do_7  │ Dropout                  │      0 │  [[1, 64, 30, 100], '?'] │  [[1, 64, 30, 100], '?'] │
+│ 11 │ net.Mp_8  │ MaxPool                  │      0 │  [[1, 64, 30, 100], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 12 │ net.C_9   │ ActConv2D                │  110 K │   [[1, 64, 15, 50], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 13 │ net.Do_10 │ Dropout                  │      0 │   [[1, 64, 15, 50], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 14 │ net.S_11  │ Reshape                  │      0 │   [[1, 64, 15, 50], '?'] │   [[1, 960, 1, 50], '?'] │
+│ 15 │ net.L_12  │ TransposedSummarizingRNN │  1.9 M │   [[1, 960, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 16 │ net.Do_13 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 17 │ net.L_14  │ TransposedSummarizingRNN │  963 K │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 18 │ net.Do_15 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 19 │ net.L_16  │ TransposedSummarizingRNN │  963 K │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 20 │ net.Do_17 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 21 │ net.O_18  │ LinSoftmax               │ 46.9 K │   [[1, 400, 1, 50], '?'] │   [[1, 117, 1, 50], '?'] │
+└────┴───────────┴──────────────────────────┴────────┴──────────────────────────┴──────────────────────────┘
+Trainable params: 4.0 M                                                                                                                                                                                                                                                                    
+Non-trainable params: 0                                                                                                                                                                                                                                                                    
+Total params: 4.0 M                                                                                                                                                                                                                                                                        
+Total estimated model params size (MB): 16                                                                                                                                                                                                                                                 
+stage 0/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:23 • 0:00:00 12.92it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 0/10 0.00000
+stage 1/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.88it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 1/10 0.00000
+stage 2/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:23 • 0:00:00 12.42it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 2/10 0.00000
+stage 3/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:23 • 0:00:00 12.55it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 3/10 0.00000
+stage 4/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:24 • 0:00:00 12.50it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 4/10 0.00000
+stage 5/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:23 • 0:00:00 12.55it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 5/10 0.00000
+stage 6/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:23 • 0:00:00 12.51it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 6/10 0.00000
+stage 7/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:23 • 0:00:00 12.60it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 7/10 0.00000
+stage 8/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.71it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 8/10 0.00000
+stage 9/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:23 • 0:00:00 12.65it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 9/10 0.00000
+stage 10/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:23 • 0:00:00 12.62it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 10/10 0.00000
+stage 11/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:23 • 0:00:00 12.55it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 11/10 0.00000
+stage 12/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:23 • 0:00:00 12.67it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 12/10 0.00000
+stage 13/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:23 • 0:00:00 12.61it/s val_accuracy: 0.113 val_word_accuracy: 0.0  early_stopping: 0/10 0.11282
+stage 14/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:23 • 0:00:00 12.73it/s val_accuracy: 0.133 val_word_accuracy: 0.0  early_stopping: 0/10 0.13282
+Moving best model /home/ROCQ/almanach/achague/peraire/peraire-ground-truth/scratch/peraire_BD_scratch_14.mlmodel (0.1328219175338745) to /home/ROCQ/almanach/achague/peraire/peraire-ground-truth/scratch/peraire_BD_scratch_best.mlmodel
diff --git a/kraken_B_add.log b/kraken_B_add.log
new file mode 100644
index 0000000..377f0e0
--- /dev/null
+++ b/kraken_B_add.log
@@ -0,0 +1,100 @@
+┏━━━━┳━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┓
+┃    ┃ Name      ┃ Type                     ┃ Params ┃                 In sizes ┃                Out sizes ┃
+┡━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━┩
+│ 0  │ val_cer   │ _CharErrorRate           │      0 │                        ? │                        ? │
+│ 1  │ val_wer   │ _WordErrorRate           │      0 │                        ? │                        ? │
+│ 2  │ net       │ MultiParamSequential     │  4.1 M │  [[1, 1, 120, 400], '?'] │   [[1, 155, 1, 50], '?'] │
+│ 3  │ net.C_0   │ ActConv2D                │  1.3 K │  [[1, 1, 120, 400], '?'] │ [[1, 32, 120, 400], '?'] │
+│ 4  │ net.Do_1  │ Dropout                  │      0 │ [[1, 32, 120, 400], '?'] │ [[1, 32, 120, 400], '?'] │
+│ 5  │ net.Mp_2  │ MaxPool                  │      0 │ [[1, 32, 120, 400], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 6  │ net.C_3   │ ActConv2D                │ 40.0 K │  [[1, 32, 60, 200], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 7  │ net.Do_4  │ Dropout                  │      0 │  [[1, 32, 60, 200], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 8  │ net.Mp_5  │ MaxPool                  │      0 │  [[1, 32, 60, 200], '?'] │  [[1, 32, 30, 100], '?'] │
+│ 9  │ net.C_6   │ ActConv2D                │ 55.4 K │  [[1, 32, 30, 100], '?'] │  [[1, 64, 30, 100], '?'] │
+│ 10 │ net.Do_7  │ Dropout                  │      0 │  [[1, 64, 30, 100], '?'] │  [[1, 64, 30, 100], '?'] │
+│ 11 │ net.Mp_8  │ MaxPool                  │      0 │  [[1, 64, 30, 100], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 12 │ net.C_9   │ ActConv2D                │  110 K │   [[1, 64, 15, 50], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 13 │ net.Do_10 │ Dropout                  │      0 │   [[1, 64, 15, 50], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 14 │ net.S_11  │ Reshape                  │      0 │   [[1, 64, 15, 50], '?'] │   [[1, 960, 1, 50], '?'] │
+│ 15 │ net.L_12  │ TransposedSummarizingRNN │  1.9 M │   [[1, 960, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 16 │ net.Do_13 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 17 │ net.L_14  │ TransposedSummarizingRNN │  963 K │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 18 │ net.Do_15 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 19 │ net.L_16  │ TransposedSummarizingRNN │  963 K │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 20 │ net.Do_17 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 21 │ net.O_18  │ LinSoftmax               │ 62.2 K │   [[1, 400, 1, 50], '?'] │   [[1, 155, 1, 50], '?'] │
+└────┴───────────┴──────────────────────────┴────────┴──────────────────────────┴──────────────────────────┘
+Trainable params: 4.1 M                                                                                                                                                                                                                                                                    
+Non-trainable params: 0                                                                                                                                                                                                                                                                    
+Total params: 4.1 M                                                                                                                                                                                                                                                                        
+Total estimated model params size (MB): 16                                                                                                                                                                                                                                                 
+stage 0/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.19it/s val_accuracy: 0.857 val_word_accuracy: 0.534  early_stopping: 0/10 0.85652
+stage 1/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.01it/s val_accuracy: 0.876 val_word_accuracy: 0.573  early_stopping: 0/10 0.87585
+stage 2/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 16.93it/s val_accuracy: 0.884 val_word_accuracy: 0.592  early_stopping: 0/10 0.88376
+stage 3/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.20it/s val_accuracy: 0.891 val_word_accuracy: 0.604  early_stopping: 0/10 0.89079
+stage 4/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.03it/s val_accuracy: 0.897 val_word_accuracy: 0.625  early_stopping: 0/10 0.89651
+stage 5/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.13it/s val_accuracy: 0.898 val_word_accuracy: 0.633  early_stopping: 0/10 0.89848
+stage 6/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.11it/s val_accuracy: 0.903 val_word_accuracy: 0.648  early_stopping: 0/10 0.90288
+stage 7/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.11it/s val_accuracy: 0.906 val_word_accuracy: 0.657  early_stopping: 0/10 0.90552
+stage 8/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.17it/s val_accuracy: 0.906 val_word_accuracy: 0.656  early_stopping: 0/10 0.90617
+stage 9/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.30it/s val_accuracy: 0.909 val_word_accuracy: 0.662  early_stopping: 0/10 0.90881
+stage 10/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.08it/s val_accuracy: 0.914 val_word_accuracy: 0.666  early_stopping: 0/10 0.91365
+stage 11/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.07it/s val_accuracy: 0.914 val_word_accuracy: 0.677  early_stopping: 0/10 0.91408
+stage 12/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.12it/s val_accuracy: 0.915 val_word_accuracy: 0.676  early_stopping: 0/10 0.91496
+stage 13/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.02it/s val_accuracy: 0.916 val_word_accuracy: 0.685  early_stopping: 0/10 0.91562
+stage 14/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.03it/s val_accuracy: 0.917 val_word_accuracy: 0.691  early_stopping: 0/10 0.91716
+stage 15/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.05it/s val_accuracy: 0.917 val_word_accuracy: 0.681  early_stopping: 1/10 0.91716
+stage 16/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.12it/s val_accuracy: 0.918 val_word_accuracy: 0.686  early_stopping: 0/10 0.91804
+stage 17/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.19it/s val_accuracy: 0.92 val_word_accuracy: 0.687  early_stopping: 0/10 0.92046
+stage 18/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.18it/s val_accuracy: 0.921 val_word_accuracy: 0.694  early_stopping: 0/10 0.92112
+stage 19/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.10it/s val_accuracy: 0.922 val_word_accuracy: 0.7  early_stopping: 0/10 0.92221
+stage 20/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.09it/s val_accuracy: 0.922 val_word_accuracy: 0.699  early_stopping: 1/10 0.92221
+stage 21/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 16.92it/s val_accuracy: 0.924 val_word_accuracy: 0.708  early_stopping: 0/10 0.92419
+stage 22/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.18it/s val_accuracy: 0.923 val_word_accuracy: 0.699  early_stopping: 1/10 0.92419
+stage 23/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.10it/s val_accuracy: 0.926 val_word_accuracy: 0.713  early_stopping: 0/10 0.92551
+stage 24/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.30it/s val_accuracy: 0.926 val_word_accuracy: 0.717  early_stopping: 0/10 0.92595
+stage 25/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.27it/s val_accuracy: 0.929 val_word_accuracy: 0.731  early_stopping: 0/10 0.92947
+stage 26/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.17it/s val_accuracy: 0.926 val_word_accuracy: 0.717  early_stopping: 1/10 0.92947
+stage 27/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.23it/s val_accuracy: 0.927 val_word_accuracy: 0.719  early_stopping: 2/10 0.92947
+stage 28/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.17it/s val_accuracy: 0.929 val_word_accuracy: 0.733  early_stopping: 3/10 0.92947
+stage 29/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.16it/s val_accuracy: 0.928 val_word_accuracy: 0.723  early_stopping: 4/10 0.92947
+stage 30/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.08it/s val_accuracy: 0.931 val_word_accuracy: 0.736  early_stopping: 0/10 0.93122
+stage 31/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.11it/s val_accuracy: 0.927 val_word_accuracy: 0.732  early_stopping: 1/10 0.93122
+stage 32/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.03it/s val_accuracy: 0.931 val_word_accuracy: 0.734  early_stopping: 2/10 0.93122
+stage 33/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.20it/s val_accuracy: 0.929 val_word_accuracy: 0.732  early_stopping: 3/10 0.93122
+stage 34/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.07it/s val_accuracy: 0.932 val_word_accuracy: 0.733  early_stopping: 0/10 0.93210
+stage 35/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.14it/s val_accuracy: 0.931 val_word_accuracy: 0.74  early_stopping: 1/10 0.93210
+stage 36/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.19it/s val_accuracy: 0.932 val_word_accuracy: 0.736  early_stopping: 2/10 0.93210
+stage 37/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.17it/s val_accuracy: 0.933 val_word_accuracy: 0.732  early_stopping: 0/10 0.93276
+stage 38/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.15it/s val_accuracy: 0.933 val_word_accuracy: 0.738  early_stopping: 0/10 0.93320
+stage 39/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.18it/s val_accuracy: 0.931 val_word_accuracy: 0.731  early_stopping: 1/10 0.93320
+stage 40/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.08it/s val_accuracy: 0.934 val_word_accuracy: 0.741  early_stopping: 0/10 0.93430
+stage 41/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.20it/s val_accuracy: 0.934 val_word_accuracy: 0.743  early_stopping: 1/10 0.93430
+stage 42/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.08it/s val_accuracy: 0.932 val_word_accuracy: 0.74  early_stopping: 2/10 0.93430
+stage 43/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.15it/s val_accuracy: 0.933 val_word_accuracy: 0.742  early_stopping: 3/10 0.93430
+stage 44/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.10it/s val_accuracy: 0.934 val_word_accuracy: 0.75  early_stopping: 4/10 0.93430
+stage 45/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.01it/s val_accuracy: 0.933 val_word_accuracy: 0.743  early_stopping: 5/10 0.93430
+stage 46/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.17it/s val_accuracy: 0.935 val_word_accuracy: 0.746  early_stopping: 0/10 0.93540
+stage 47/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.13it/s val_accuracy: 0.936 val_word_accuracy: 0.748  early_stopping: 0/10 0.93584
+stage 48/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.32it/s val_accuracy: 0.935 val_word_accuracy: 0.738  early_stopping: 1/10 0.93584
+stage 49/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.18it/s val_accuracy: 0.931 val_word_accuracy: 0.726  early_stopping: 2/10 0.93584
+stage 50/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.15it/s val_accuracy: 0.935 val_word_accuracy: 0.745  early_stopping: 3/10 0.93584
+stage 51/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.05it/s val_accuracy: 0.935 val_word_accuracy: 0.748  early_stopping: 4/10 0.93584
+stage 52/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.15it/s val_accuracy: 0.934 val_word_accuracy: 0.74  early_stopping: 5/10 0.93584
+stage 53/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.11it/s val_accuracy: 0.937 val_word_accuracy: 0.738  early_stopping: 0/10 0.93716
+stage 54/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.19it/s val_accuracy: 0.935 val_word_accuracy: 0.743  early_stopping: 1/10 0.93716
+stage 55/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 16.99it/s val_accuracy: 0.936 val_word_accuracy: 0.742  early_stopping: 2/10 0.93716
+stage 56/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.22it/s val_accuracy: 0.936 val_word_accuracy: 0.74  early_stopping: 3/10 0.93716
+stage 57/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.17it/s val_accuracy: 0.936 val_word_accuracy: 0.741  early_stopping: 4/10 0.93716
+stage 58/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.26it/s val_accuracy: 0.937 val_word_accuracy: 0.743  early_stopping: 0/10 0.93738
+stage 59/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.26it/s val_accuracy: 0.936 val_word_accuracy: 0.741  early_stopping: 1/10 0.93738
+stage 60/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.32it/s val_accuracy: 0.935 val_word_accuracy: 0.738  early_stopping: 2/10 0.93738
+stage 61/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.02it/s val_accuracy: 0.935 val_word_accuracy: 0.747  early_stopping: 3/10 0.93738
+stage 62/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.06it/s val_accuracy: 0.936 val_word_accuracy: 0.743  early_stopping: 4/10 0.93738
+stage 63/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.21it/s val_accuracy: 0.935 val_word_accuracy: 0.734  early_stopping: 5/10 0.93738
+stage 64/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.21it/s val_accuracy: 0.935 val_word_accuracy: 0.741  early_stopping: 6/10 0.93738
+stage 65/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.06it/s val_accuracy: 0.931 val_word_accuracy: 0.732  early_stopping: 7/10 0.93738
+stage 66/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.31it/s val_accuracy: 0.935 val_word_accuracy: 0.75  early_stopping: 8/10 0.93738
+stage 67/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.22it/s val_accuracy: 0.934 val_word_accuracy: 0.737  early_stopping: 9/10 0.93738
+stage 68/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.27it/s val_accuracy: 0.934 val_word_accuracy: 0.731  early_stopping: 10/10 0.93738
+Moving best model /home/ROCQ/almanach/achague/peraire/peraire-ground-truth/finetuned/peraire_B_add_58.mlmodel (0.9373763799667358) to /home/ROCQ/almanach/achague/peraire/peraire-ground-truth/finetuned/peraire_B_add_best.mlmodel
diff --git a/kraken_B_scratch.log b/kraken_B_scratch.log
new file mode 100644
index 0000000..caaeb7c
--- /dev/null
+++ b/kraken_B_scratch.log
@@ -0,0 +1,46 @@
+┏━━━━┳━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┓
+┃    ┃ Name      ┃ Type                     ┃ Params ┃                 In sizes ┃                Out sizes ┃
+┡━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━┩
+│ 0  │ val_cer   │ _CharErrorRate           │      0 │                        ? │                        ? │
+│ 1  │ val_wer   │ _WordErrorRate           │      0 │                        ? │                        ? │
+│ 2  │ net       │ MultiParamSequential     │  4.0 M │  [[1, 1, 120, 400], '?'] │   [[1, 102, 1, 50], '?'] │
+│ 3  │ net.C_0   │ ActConv2D                │  1.3 K │  [[1, 1, 120, 400], '?'] │ [[1, 32, 120, 400], '?'] │
+│ 4  │ net.Do_1  │ Dropout                  │      0 │ [[1, 32, 120, 400], '?'] │ [[1, 32, 120, 400], '?'] │
+│ 5  │ net.Mp_2  │ MaxPool                  │      0 │ [[1, 32, 120, 400], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 6  │ net.C_3   │ ActConv2D                │ 40.0 K │  [[1, 32, 60, 200], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 7  │ net.Do_4  │ Dropout                  │      0 │  [[1, 32, 60, 200], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 8  │ net.Mp_5  │ MaxPool                  │      0 │  [[1, 32, 60, 200], '?'] │  [[1, 32, 30, 100], '?'] │
+│ 9  │ net.C_6   │ ActConv2D                │ 55.4 K │  [[1, 32, 30, 100], '?'] │  [[1, 64, 30, 100], '?'] │
+│ 10 │ net.Do_7  │ Dropout                  │      0 │  [[1, 64, 30, 100], '?'] │  [[1, 64, 30, 100], '?'] │
+│ 11 │ net.Mp_8  │ MaxPool                  │      0 │  [[1, 64, 30, 100], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 12 │ net.C_9   │ ActConv2D                │  110 K │   [[1, 64, 15, 50], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 13 │ net.Do_10 │ Dropout                  │      0 │   [[1, 64, 15, 50], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 14 │ net.S_11  │ Reshape                  │      0 │   [[1, 64, 15, 50], '?'] │   [[1, 960, 1, 50], '?'] │
+│ 15 │ net.L_12  │ TransposedSummarizingRNN │  1.9 M │   [[1, 960, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 16 │ net.Do_13 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 17 │ net.L_14  │ TransposedSummarizingRNN │  963 K │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 18 │ net.Do_15 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 19 │ net.L_16  │ TransposedSummarizingRNN │  963 K │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 20 │ net.Do_17 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 21 │ net.O_18  │ LinSoftmax               │ 40.9 K │   [[1, 400, 1, 50], '?'] │   [[1, 102, 1, 50], '?'] │
+└────┴───────────┴──────────────────────────┴────────┴──────────────────────────┴──────────────────────────┘
+Trainable params: 4.0 M                                                                                                                                                                                                                                                                    
+Non-trainable params: 0                                                                                                                                                                                                                                                                    
+Total params: 4.0 M                                                                                                                                                                                                                                                                        
+Total estimated model params size (MB): 16                                                                                                                                                                                                                                                 
+stage 0/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.04it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 0/10 0.00000
+stage 1/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 16.79it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 1/10 0.00000
+stage 2/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.05it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 2/10 0.00000
+stage 3/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 16.94it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 3/10 0.00000
+stage 4/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.14it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 4/10 0.00000
+stage 5/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.09it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 5/10 0.00000
+stage 6/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.00it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 6/10 0.00000
+stage 7/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:13 • 0:00:00 17.25it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 7/10 0.00000
+stage 8/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 16.89it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 8/10 0.00000
+stage 9/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.19it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 9/10 0.00000
+stage 10/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 16.96it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 10/10 0.00000
+stage 11/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.05it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 11/10 0.00000
+stage 12/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 16.94it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 12/10 0.00000
+stage 13/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 17.12it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 13/10 0.00000
+stage 14/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 231/231 0:00:14 • 0:00:00 16.93it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 14/10 0.00000
+Moving best model /home/ROCQ/almanach/achague/peraire/peraire-ground-truth/scratch/peraire_B_scratch_0.mlmodel (0.044117629528045654) to /home/ROCQ/almanach/achague/peraire/peraire-ground-truth/scratch/peraire_B_scratch_best.mlmodel
diff --git a/kraken_D_add.log b/kraken_D_add.log
new file mode 100644
index 0000000..cdb6f8a
--- /dev/null
+++ b/kraken_D_add.log
@@ -0,0 +1,105 @@
+┏━━━━┳━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┓
+┃    ┃ Name      ┃ Type                     ┃ Params ┃                 In sizes ┃                Out sizes ┃
+┡━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━┩
+│ 0  │ val_cer   │ _CharErrorRate           │      0 │                        ? │                        ? │
+│ 1  │ val_wer   │ _WordErrorRate           │      0 │                        ? │                        ? │
+│ 2  │ net       │ MultiParamSequential     │  4.1 M │  [[1, 1, 120, 400], '?'] │   [[1, 166, 1, 50], '?'] │
+│ 3  │ net.C_0   │ ActConv2D                │  1.3 K │  [[1, 1, 120, 400], '?'] │ [[1, 32, 120, 400], '?'] │
+│ 4  │ net.Do_1  │ Dropout                  │      0 │ [[1, 32, 120, 400], '?'] │ [[1, 32, 120, 400], '?'] │
+│ 5  │ net.Mp_2  │ MaxPool                  │      0 │ [[1, 32, 120, 400], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 6  │ net.C_3   │ ActConv2D                │ 40.0 K │  [[1, 32, 60, 200], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 7  │ net.Do_4  │ Dropout                  │      0 │  [[1, 32, 60, 200], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 8  │ net.Mp_5  │ MaxPool                  │      0 │  [[1, 32, 60, 200], '?'] │  [[1, 32, 30, 100], '?'] │
+│ 9  │ net.C_6   │ ActConv2D                │ 55.4 K │  [[1, 32, 30, 100], '?'] │  [[1, 64, 30, 100], '?'] │
+│ 10 │ net.Do_7  │ Dropout                  │      0 │  [[1, 64, 30, 100], '?'] │  [[1, 64, 30, 100], '?'] │
+│ 11 │ net.Mp_8  │ MaxPool                  │      0 │  [[1, 64, 30, 100], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 12 │ net.C_9   │ ActConv2D                │  110 K │   [[1, 64, 15, 50], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 13 │ net.Do_10 │ Dropout                  │      0 │   [[1, 64, 15, 50], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 14 │ net.S_11  │ Reshape                  │      0 │   [[1, 64, 15, 50], '?'] │   [[1, 960, 1, 50], '?'] │
+│ 15 │ net.L_12  │ TransposedSummarizingRNN │  1.9 M │   [[1, 960, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 16 │ net.Do_13 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 17 │ net.L_14  │ TransposedSummarizingRNN │  963 K │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 18 │ net.Do_15 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 19 │ net.L_16  │ TransposedSummarizingRNN │  963 K │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 20 │ net.Do_17 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 21 │ net.O_18  │ LinSoftmax               │ 66.6 K │   [[1, 400, 1, 50], '?'] │   [[1, 166, 1, 50], '?'] │
+└────┴───────────┴──────────────────────────┴────────┴──────────────────────────┴──────────────────────────┘
+Trainable params: 4.1 M                                                                                                                                                                                                                                                                    
+Non-trainable params: 0                                                                                                                                                                                                                                                                    
+Total params: 4.1 M                                                                                                                                                                                                                                                                        
+Total estimated model params size (MB): 16                                                                                                                                                                                                                                                 
+stage 0/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:23 • 0:00:00 12.87it/s val_accuracy: 0.861 val_word_accuracy: 0.523  early_stopping: 0/10 0.86131
+stage 1/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.82it/s val_accuracy: 0.886 val_word_accuracy: 0.567  early_stopping: 0/10 0.88584
+stage 2/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.90it/s val_accuracy: 0.897 val_word_accuracy: 0.592  early_stopping: 0/10 0.89739
+stage 3/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.05it/s val_accuracy: 0.906 val_word_accuracy: 0.613  early_stopping: 0/10 0.90585
+stage 4/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.01it/s val_accuracy: 0.913 val_word_accuracy: 0.632  early_stopping: 0/10 0.91332
+stage 5/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.74it/s val_accuracy: 0.922 val_word_accuracy: 0.66  early_stopping: 0/10 0.92178
+stage 6/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.80it/s val_accuracy: 0.926 val_word_accuracy: 0.672  early_stopping: 0/10 0.92586
+stage 7/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.01it/s val_accuracy: 0.931 val_word_accuracy: 0.687  early_stopping: 0/10 0.93136
+stage 8/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.97it/s val_accuracy: 0.933 val_word_accuracy: 0.692  early_stopping: 0/10 0.93305
+stage 9/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.85it/s val_accuracy: 0.936 val_word_accuracy: 0.704  early_stopping: 0/10 0.93559
+stage 10/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.96it/s val_accuracy: 0.939 val_word_accuracy: 0.715  early_stopping: 0/10 0.93911
+stage 11/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.97it/s val_accuracy: 0.942 val_word_accuracy: 0.721  early_stopping: 0/10 0.94151
+stage 12/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.85it/s val_accuracy: 0.942 val_word_accuracy: 0.722  early_stopping: 0/10 0.94207
+stage 13/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.75it/s val_accuracy: 0.945 val_word_accuracy: 0.738  early_stopping: 0/10 0.94489
+stage 14/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.94it/s val_accuracy: 0.944 val_word_accuracy: 0.737  early_stopping: 1/10 0.94489
+stage 15/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.98it/s val_accuracy: 0.946 val_word_accuracy: 0.745  early_stopping: 0/10 0.94602
+stage 16/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.83it/s val_accuracy: 0.946 val_word_accuracy: 0.748  early_stopping: 1/10 0.94602
+stage 17/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.88it/s val_accuracy: 0.948 val_word_accuracy: 0.751  early_stopping: 0/10 0.94799
+stage 18/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.92it/s val_accuracy: 0.948 val_word_accuracy: 0.748  early_stopping: 1/10 0.94799
+stage 19/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.01it/s val_accuracy: 0.95 val_word_accuracy: 0.761  early_stopping: 0/10 0.95011
+stage 20/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.92it/s val_accuracy: 0.948 val_word_accuracy: 0.752  early_stopping: 1/10 0.95011
+stage 21/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.95it/s val_accuracy: 0.951 val_word_accuracy: 0.765  early_stopping: 0/10 0.95067
+stage 22/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.16it/s val_accuracy: 0.953 val_word_accuracy: 0.775  early_stopping: 0/10 0.95307
+stage 23/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.89it/s val_accuracy: 0.953 val_word_accuracy: 0.773  early_stopping: 1/10 0.95307
+stage 24/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.01it/s val_accuracy: 0.952 val_word_accuracy: 0.773  early_stopping: 2/10 0.95307
+stage 25/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.87it/s val_accuracy: 0.953 val_word_accuracy: 0.774  early_stopping: 3/10 0.95307
+stage 26/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.96it/s val_accuracy: 0.952 val_word_accuracy: 0.772  early_stopping: 4/10 0.95307
+stage 27/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.89it/s val_accuracy: 0.954 val_word_accuracy: 0.781  early_stopping: 0/10 0.95405
+stage 28/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.90it/s val_accuracy: 0.954 val_word_accuracy: 0.785  early_stopping: 0/10 0.95447
+stage 29/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.95it/s val_accuracy: 0.957 val_word_accuracy: 0.791  early_stopping: 0/10 0.95673
+stage 30/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.88it/s val_accuracy: 0.954 val_word_accuracy: 0.783  early_stopping: 1/10 0.95673
+stage 31/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.88it/s val_accuracy: 0.956 val_word_accuracy: 0.787  early_stopping: 2/10 0.95673
+stage 32/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.97it/s val_accuracy: 0.956 val_word_accuracy: 0.794  early_stopping: 3/10 0.95673
+stage 33/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.77it/s val_accuracy: 0.955 val_word_accuracy: 0.783  early_stopping: 4/10 0.95673
+stage 34/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.07it/s val_accuracy: 0.955 val_word_accuracy: 0.778  early_stopping: 5/10 0.95673
+stage 35/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.12it/s val_accuracy: 0.954 val_word_accuracy: 0.783  early_stopping: 6/10 0.95673
+stage 36/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.08it/s val_accuracy: 0.956 val_word_accuracy: 0.791  early_stopping: 7/10 0.95673
+stage 37/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.95it/s val_accuracy: 0.955 val_word_accuracy: 0.788  early_stopping: 8/10 0.95673
+stage 38/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.00it/s val_accuracy: 0.957 val_word_accuracy: 0.793  early_stopping: 0/10 0.95687
+stage 39/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.01it/s val_accuracy: 0.955 val_word_accuracy: 0.792  early_stopping: 1/10 0.95687
+stage 40/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.97it/s val_accuracy: 0.956 val_word_accuracy: 0.79  early_stopping: 2/10 0.95687
+stage 41/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.83it/s val_accuracy: 0.956 val_word_accuracy: 0.788  early_stopping: 3/10 0.95687
+stage 42/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:23 • 0:00:00 12.83it/s val_accuracy: 0.957 val_word_accuracy: 0.798  early_stopping: 4/10 0.95687
+stage 43/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.96it/s val_accuracy: 0.956 val_word_accuracy: 0.788  early_stopping: 5/10 0.95687
+stage 44/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:23 • 0:00:00 12.50it/s val_accuracy: 0.957 val_word_accuracy: 0.788  early_stopping: 6/10 0.95687
+stage 45/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:23 • 0:00:00 12.55it/s val_accuracy: 0.959 val_word_accuracy: 0.798  early_stopping: 0/10 0.95870
+stage 46/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.11it/s val_accuracy: 0.957 val_word_accuracy: 0.793  early_stopping: 1/10 0.95870
+stage 47/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.97it/s val_accuracy: 0.959 val_word_accuracy: 0.8  early_stopping: 0/10 0.95884
+stage 48/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:23 • 0:00:00 12.86it/s val_accuracy: 0.957 val_word_accuracy: 0.798  early_stopping: 1/10 0.95884
+stage 49/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.95it/s val_accuracy: 0.958 val_word_accuracy: 0.792  early_stopping: 2/10 0.95884
+stage 50/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.92it/s val_accuracy: 0.958 val_word_accuracy: 0.798  early_stopping: 3/10 0.95884
+stage 51/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.04it/s val_accuracy: 0.959 val_word_accuracy: 0.8  early_stopping: 0/10 0.95927
+stage 52/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.89it/s val_accuracy: 0.958 val_word_accuracy: 0.801  early_stopping: 1/10 0.95927
+stage 53/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.14it/s val_accuracy: 0.958 val_word_accuracy: 0.798  early_stopping: 2/10 0.95927
+stage 54/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.93it/s val_accuracy: 0.959 val_word_accuracy: 0.802  early_stopping: 3/10 0.95927
+stage 55/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.97it/s val_accuracy: 0.959 val_word_accuracy: 0.802  early_stopping: 4/10 0.95927
+stage 56/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.93it/s val_accuracy: 0.959 val_word_accuracy: 0.803  early_stopping: 5/10 0.95927
+stage 57/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.95it/s val_accuracy: 0.959 val_word_accuracy: 0.808  early_stopping: 6/10 0.95927
+stage 58/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.02it/s val_accuracy: 0.959 val_word_accuracy: 0.798  early_stopping: 7/10 0.95927
+stage 59/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.10it/s val_accuracy: 0.96 val_word_accuracy: 0.805  early_stopping: 0/10 0.95997
+stage 60/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.93it/s val_accuracy: 0.959 val_word_accuracy: 0.804  early_stopping: 1/10 0.95997
+stage 61/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.01it/s val_accuracy: 0.959 val_word_accuracy: 0.81  early_stopping: 2/10 0.95997
+stage 62/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.99it/s val_accuracy: 0.96 val_word_accuracy: 0.815  early_stopping: 3/10 0.95997
+stage 63/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.91it/s val_accuracy: 0.962 val_word_accuracy: 0.813  early_stopping: 0/10 0.96166
+stage 64/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.92it/s val_accuracy: 0.96 val_word_accuracy: 0.803  early_stopping: 1/10 0.96166
+stage 65/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.88it/s val_accuracy: 0.959 val_word_accuracy: 0.801  early_stopping: 2/10 0.96166
+stage 66/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.90it/s val_accuracy: 0.96 val_word_accuracy: 0.813  early_stopping: 3/10 0.96166
+stage 67/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.79it/s val_accuracy: 0.96 val_word_accuracy: 0.812  early_stopping: 4/10 0.96166
+stage 68/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.96it/s val_accuracy: 0.96 val_word_accuracy: 0.81  early_stopping: 5/10 0.96166
+stage 69/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.95it/s val_accuracy: 0.961 val_word_accuracy: 0.815  early_stopping: 6/10 0.96166
+stage 70/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.96it/s val_accuracy: 0.961 val_word_accuracy: 0.81  early_stopping: 7/10 0.96166
+stage 71/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.95it/s val_accuracy: 0.962 val_word_accuracy: 0.814  early_stopping: 8/10 0.96166
+stage 72/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.91it/s val_accuracy: 0.961 val_word_accuracy: 0.815  early_stopping: 9/10 0.96166
+stage 73/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.05it/s val_accuracy: 0.96 val_word_accuracy: 0.81  early_stopping: 10/10 0.96166
+Moving best model /home/ROCQ/almanach/achague/peraire/peraire-ground-truth/finetuned/peraireD_D_add_63.mlmodel (0.9616631269454956) to /home/ROCQ/almanach/achague/peraire/peraire-ground-truth/finetuned/peraireD_D_add_best.mlmodel
diff --git a/kraken_D_scratch.log b/kraken_D_scratch.log
new file mode 100644
index 0000000..3b333e8
--- /dev/null
+++ b/kraken_D_scratch.log
@@ -0,0 +1,159 @@
+┏━━━━┳━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┓
+┃    ┃ Name      ┃ Type                     ┃ Params ┃                 In sizes ┃                Out sizes ┃
+┡━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━┩
+│ 0  │ val_cer   │ _CharErrorRate           │      0 │                        ? │                        ? │
+│ 1  │ val_wer   │ _WordErrorRate           │      0 │                        ? │                        ? │
+│ 2  │ net       │ MultiParamSequential     │  4.0 M │  [[1, 1, 120, 400], '?'] │   [[1, 117, 1, 50], '?'] │
+│ 3  │ net.C_0   │ ActConv2D                │  1.3 K │  [[1, 1, 120, 400], '?'] │ [[1, 32, 120, 400], '?'] │
+│ 4  │ net.Do_1  │ Dropout                  │      0 │ [[1, 32, 120, 400], '?'] │ [[1, 32, 120, 400], '?'] │
+│ 5  │ net.Mp_2  │ MaxPool                  │      0 │ [[1, 32, 120, 400], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 6  │ net.C_3   │ ActConv2D                │ 40.0 K │  [[1, 32, 60, 200], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 7  │ net.Do_4  │ Dropout                  │      0 │  [[1, 32, 60, 200], '?'] │  [[1, 32, 60, 200], '?'] │
+│ 8  │ net.Mp_5  │ MaxPool                  │      0 │  [[1, 32, 60, 200], '?'] │  [[1, 32, 30, 100], '?'] │
+│ 9  │ net.C_6   │ ActConv2D                │ 55.4 K │  [[1, 32, 30, 100], '?'] │  [[1, 64, 30, 100], '?'] │
+│ 10 │ net.Do_7  │ Dropout                  │      0 │  [[1, 64, 30, 100], '?'] │  [[1, 64, 30, 100], '?'] │
+│ 11 │ net.Mp_8  │ MaxPool                  │      0 │  [[1, 64, 30, 100], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 12 │ net.C_9   │ ActConv2D                │  110 K │   [[1, 64, 15, 50], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 13 │ net.Do_10 │ Dropout                  │      0 │   [[1, 64, 15, 50], '?'] │   [[1, 64, 15, 50], '?'] │
+│ 14 │ net.S_11  │ Reshape                  │      0 │   [[1, 64, 15, 50], '?'] │   [[1, 960, 1, 50], '?'] │
+│ 15 │ net.L_12  │ TransposedSummarizingRNN │  1.9 M │   [[1, 960, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 16 │ net.Do_13 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 17 │ net.L_14  │ TransposedSummarizingRNN │  963 K │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 18 │ net.Do_15 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 19 │ net.L_16  │ TransposedSummarizingRNN │  963 K │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 20 │ net.Do_17 │ Dropout                  │      0 │   [[1, 400, 1, 50], '?'] │   [[1, 400, 1, 50], '?'] │
+│ 21 │ net.O_18  │ LinSoftmax               │ 46.9 K │   [[1, 400, 1, 50], '?'] │   [[1, 117, 1, 50], '?'] │
+└────┴───────────┴──────────────────────────┴────────┴──────────────────────────┴──────────────────────────┘
+Trainable params: 4.0 M                                                                                                                                                                                                                                                                    
+Non-trainable params: 0                                                                                                                                                                                                                                                                    
+Total params: 4.0 M                                                                                                                                                                                                                                                                        
+Total estimated model params size (MB): 16                                                                                                                                                                                                                                                 
+stage 0/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.94it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 0/10 0.00000
+stage 1/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.85it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 1/10 0.00000
+stage 2/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.03it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 2/10 0.00000
+stage 3/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.92it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 3/10 0.00000
+stage 4/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.06it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 4/10 0.00000
+stage 5/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.93it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 5/10 0.00000
+stage 6/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.96it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 6/10 0.00000
+stage 7/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.04it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 7/10 0.00000
+stage 8/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.91it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 8/10 0.00000
+stage 9/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.94it/s val_accuracy: 0.0 val_word_accuracy: 0.0  early_stopping: 9/10 0.00000
+stage 10/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.07it/s val_accuracy: 0.008 val_word_accuracy: 0.0  early_stopping: 0/10 0.00812
+stage 11/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.92it/s val_accuracy: 0.12 val_word_accuracy: 0.0  early_stopping: 0/10 0.11981
+stage 12/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.99it/s val_accuracy: 0.129 val_word_accuracy: 0.0  early_stopping: 0/10 0.12868
+stage 13/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.93it/s val_accuracy: 0.136 val_word_accuracy: 0.0  early_stopping: 0/10 0.13635
+stage 14/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.92it/s val_accuracy: 0.144 val_word_accuracy: 0.0  early_stopping: 0/10 0.14432
+stage 15/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.00it/s val_accuracy: 0.151 val_word_accuracy: 0.0  early_stopping: 0/10 0.15063
+stage 16/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.84it/s val_accuracy: 0.174 val_word_accuracy: 0.0  early_stopping: 0/10 0.17408
+stage 17/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.88it/s val_accuracy: 0.186 val_word_accuracy: 0.0  early_stopping: 0/10 0.18596
+stage 18/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.95it/s val_accuracy: 0.237 val_word_accuracy: 0.031  early_stopping: 0/10 0.23677
+stage 19/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.94it/s val_accuracy: 0.248 val_word_accuracy: 0.046  early_stopping: 0/10 0.24835
+stage 20/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.95it/s val_accuracy: 0.277 val_word_accuracy: 0.064  early_stopping: 0/10 0.27676
+stage 21/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.85it/s val_accuracy: 0.278 val_word_accuracy: 0.053  early_stopping: 0/10 0.27841
+stage 22/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.94it/s val_accuracy: 0.325 val_word_accuracy: 0.094  early_stopping: 0/10 0.32517
+stage 23/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.00it/s val_accuracy: 0.352 val_word_accuracy: 0.112  early_stopping: 0/10 0.35192
+stage 24/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.92it/s val_accuracy: 0.35 val_word_accuracy: 0.119  early_stopping: 1/10 0.35192
+stage 25/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.89it/s val_accuracy: 0.392 val_word_accuracy: 0.148  early_stopping: 0/10 0.39221
+stage 26/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.94it/s val_accuracy: 0.417 val_word_accuracy: 0.158  early_stopping: 0/10 0.41747
+stage 27/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.99it/s val_accuracy: 0.448 val_word_accuracy: 0.17  early_stopping: 0/10 0.44814
+stage 28/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.84it/s val_accuracy: 0.485 val_word_accuracy: 0.189  early_stopping: 0/10 0.48527
+stage 29/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.93it/s val_accuracy: 0.522 val_word_accuracy: 0.21  early_stopping: 0/10 0.52195
+stage 30/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.95it/s val_accuracy: 0.549 val_word_accuracy: 0.222  early_stopping: 0/10 0.54931
+stage 31/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.97it/s val_accuracy: 0.617 val_word_accuracy: 0.248  early_stopping: 0/10 0.61666
+stage 32/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.01it/s val_accuracy: 0.665 val_word_accuracy: 0.277  early_stopping: 0/10 0.66491
+stage 33/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.78it/s val_accuracy: 0.712 val_word_accuracy: 0.309  early_stopping: 0/10 0.71182
+stage 34/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.87it/s val_accuracy: 0.718 val_word_accuracy: 0.308  early_stopping: 0/10 0.71768
+stage 35/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.93it/s val_accuracy: 0.751 val_word_accuracy: 0.344  early_stopping: 0/10 0.75120
+stage 36/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.87it/s val_accuracy: 0.765 val_word_accuracy: 0.358  early_stopping: 0/10 0.76503
+stage 37/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.07it/s val_accuracy: 0.784 val_word_accuracy: 0.362  early_stopping: 0/10 0.78443
+stage 38/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.92it/s val_accuracy: 0.794 val_word_accuracy: 0.383  early_stopping: 0/10 0.79420
+stage 39/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.87it/s val_accuracy: 0.805 val_word_accuracy: 0.389  early_stopping: 0/10 0.80472
+stage 40/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.92it/s val_accuracy: 0.813 val_word_accuracy: 0.418  early_stopping: 0/10 0.81254
+stage 41/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.77it/s val_accuracy: 0.821 val_word_accuracy: 0.426  early_stopping: 0/10 0.82051
+stage 42/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.85it/s val_accuracy: 0.824 val_word_accuracy: 0.43  early_stopping: 0/10 0.82351
+stage 43/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.91it/s val_accuracy: 0.834 val_word_accuracy: 0.451  early_stopping: 0/10 0.83449
+stage 44/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.01it/s val_accuracy: 0.836 val_word_accuracy: 0.464  early_stopping: 0/10 0.83614
+stage 45/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.95it/s val_accuracy: 0.843 val_word_accuracy: 0.477  early_stopping: 0/10 0.84290
+stage 46/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.96it/s val_accuracy: 0.845 val_word_accuracy: 0.467  early_stopping: 0/10 0.84546
+stage 47/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.00it/s val_accuracy: 0.851 val_word_accuracy: 0.481  early_stopping: 0/10 0.85117
+stage 48/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.00it/s val_accuracy: 0.854 val_word_accuracy: 0.502  early_stopping: 0/10 0.85448
+stage 49/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.03it/s val_accuracy: 0.858 val_word_accuracy: 0.495  early_stopping: 0/10 0.85794
+stage 50/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.04it/s val_accuracy: 0.857 val_word_accuracy: 0.502  early_stopping: 1/10 0.85794
+stage 51/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.96it/s val_accuracy: 0.872 val_word_accuracy: 0.526  early_stopping: 0/10 0.87162
+stage 52/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.98it/s val_accuracy: 0.869 val_word_accuracy: 0.513  early_stopping: 1/10 0.87162
+stage 53/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.98it/s val_accuracy: 0.871 val_word_accuracy: 0.525  early_stopping: 2/10 0.87162
+stage 54/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.88it/s val_accuracy: 0.874 val_word_accuracy: 0.532  early_stopping: 0/10 0.87402
+stage 55/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.01it/s val_accuracy: 0.876 val_word_accuracy: 0.538  early_stopping: 0/10 0.87628
+stage 56/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.81it/s val_accuracy: 0.881 val_word_accuracy: 0.549  early_stopping: 0/10 0.88079
+stage 57/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.98it/s val_accuracy: 0.879 val_word_accuracy: 0.536  early_stopping: 1/10 0.88079
+stage 58/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.91it/s val_accuracy: 0.879 val_word_accuracy: 0.548  early_stopping: 2/10 0.88079
+stage 59/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.06it/s val_accuracy: 0.885 val_word_accuracy: 0.564  early_stopping: 0/10 0.88485
+stage 60/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.94it/s val_accuracy: 0.885 val_word_accuracy: 0.567  early_stopping: 0/10 0.88500
+stage 61/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.99it/s val_accuracy: 0.886 val_word_accuracy: 0.559  early_stopping: 0/10 0.88575
+stage 62/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.05it/s val_accuracy: 0.888 val_word_accuracy: 0.569  early_stopping: 0/10 0.88755
+stage 63/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.00it/s val_accuracy: 0.89 val_word_accuracy: 0.571  early_stopping: 0/10 0.89011
+stage 64/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.09it/s val_accuracy: 0.888 val_word_accuracy: 0.571  early_stopping: 1/10 0.89011
+stage 65/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.94it/s val_accuracy: 0.892 val_word_accuracy: 0.58  early_stopping: 0/10 0.89206
+stage 66/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.96it/s val_accuracy: 0.892 val_word_accuracy: 0.581  early_stopping: 0/10 0.89221
+stage 67/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.02it/s val_accuracy: 0.894 val_word_accuracy: 0.594  early_stopping: 0/10 0.89387
+stage 68/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.00it/s val_accuracy: 0.895 val_word_accuracy: 0.585  early_stopping: 0/10 0.89507
+stage 69/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.95it/s val_accuracy: 0.897 val_word_accuracy: 0.592  early_stopping: 0/10 0.89657
+stage 70/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.02it/s val_accuracy: 0.897 val_word_accuracy: 0.598  early_stopping: 0/10 0.89747
+stage 71/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.95it/s val_accuracy: 0.896 val_word_accuracy: 0.595  early_stopping: 1/10 0.89747
+stage 72/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.08it/s val_accuracy: 0.898 val_word_accuracy: 0.588  early_stopping: 0/10 0.89823
+stage 73/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.79it/s val_accuracy: 0.895 val_word_accuracy: 0.603  early_stopping: 1/10 0.89823
+stage 74/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.02it/s val_accuracy: 0.898 val_word_accuracy: 0.6  early_stopping: 2/10 0.89823
+stage 75/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.02it/s val_accuracy: 0.897 val_word_accuracy: 0.6  early_stopping: 3/10 0.89823
+stage 76/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.03it/s val_accuracy: 0.902 val_word_accuracy: 0.608  early_stopping: 0/10 0.90183
+stage 77/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.04it/s val_accuracy: 0.895 val_word_accuracy: 0.592  early_stopping: 1/10 0.90183
+stage 78/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.95it/s val_accuracy: 0.903 val_word_accuracy: 0.601  early_stopping: 0/10 0.90274
+stage 79/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.75it/s val_accuracy: 0.904 val_word_accuracy: 0.611  early_stopping: 0/10 0.90424
+stage 80/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.01it/s val_accuracy: 0.902 val_word_accuracy: 0.607  early_stopping: 1/10 0.90424
+stage 81/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.02it/s val_accuracy: 0.902 val_word_accuracy: 0.607  early_stopping: 2/10 0.90424
+stage 82/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.95it/s val_accuracy: 0.903 val_word_accuracy: 0.61  early_stopping: 3/10 0.90424
+stage 83/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.98it/s val_accuracy: 0.905 val_word_accuracy: 0.616  early_stopping: 0/10 0.90529
+stage 84/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.99it/s val_accuracy: 0.902 val_word_accuracy: 0.605  early_stopping: 1/10 0.90529
+stage 85/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.93it/s val_accuracy: 0.904 val_word_accuracy: 0.613  early_stopping: 2/10 0.90529
+stage 86/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.95it/s val_accuracy: 0.905 val_word_accuracy: 0.623  early_stopping: 3/10 0.90529
+stage 87/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.93it/s val_accuracy: 0.902 val_word_accuracy: 0.606  early_stopping: 4/10 0.90529
+stage 88/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.01it/s val_accuracy: 0.903 val_word_accuracy: 0.62  early_stopping: 5/10 0.90529
+stage 89/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.05it/s val_accuracy: 0.907 val_word_accuracy: 0.617  early_stopping: 0/10 0.90710
+stage 90/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.03it/s val_accuracy: 0.907 val_word_accuracy: 0.623  early_stopping: 1/10 0.90710
+stage 91/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.95it/s val_accuracy: 0.905 val_word_accuracy: 0.628  early_stopping: 2/10 0.90710
+stage 92/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.08it/s val_accuracy: 0.909 val_word_accuracy: 0.62  early_stopping: 0/10 0.90875
+stage 93/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.93it/s val_accuracy: 0.908 val_word_accuracy: 0.632  early_stopping: 1/10 0.90875
+stage 94/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.93it/s val_accuracy: 0.91 val_word_accuracy: 0.629  early_stopping: 0/10 0.90965
+stage 95/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.08it/s val_accuracy: 0.91 val_word_accuracy: 0.637  early_stopping: 0/10 0.90980
+stage 96/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:23 • 0:00:00 12.67it/s val_accuracy: 0.91 val_word_accuracy: 0.625  early_stopping: 0/10 0.91040
+stage 97/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.95it/s val_accuracy: 0.908 val_word_accuracy: 0.626  early_stopping: 1/10 0.91040
+stage 98/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.01it/s val_accuracy: 0.91 val_word_accuracy: 0.632  early_stopping: 2/10 0.91040
+stage 99/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.00it/s val_accuracy: 0.91 val_word_accuracy: 0.632  early_stopping: 3/10 0.91040
+stage 100/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.04it/s val_accuracy: 0.912 val_word_accuracy: 0.639  early_stopping: 0/10 0.91161
+stage 101/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.97it/s val_accuracy: 0.912 val_word_accuracy: 0.643  early_stopping: 0/10 0.91191
+stage 102/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.91it/s val_accuracy: 0.911 val_word_accuracy: 0.629  early_stopping: 1/10 0.91191
+stage 103/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.99it/s val_accuracy: 0.911 val_word_accuracy: 0.639  early_stopping: 2/10 0.91191
+stage 104/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.94it/s val_accuracy: 0.912 val_word_accuracy: 0.631  early_stopping: 0/10 0.91221
+stage 105/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.99it/s val_accuracy: 0.911 val_word_accuracy: 0.641  early_stopping: 1/10 0.91221
+stage 106/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.03it/s val_accuracy: 0.916 val_word_accuracy: 0.65  early_stopping: 0/10 0.91627
+stage 107/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.91it/s val_accuracy: 0.916 val_word_accuracy: 0.652  early_stopping: 1/10 0.91627
+stage 108/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.99it/s val_accuracy: 0.913 val_word_accuracy: 0.639  early_stopping: 2/10 0.91627
+stage 109/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.01it/s val_accuracy: 0.914 val_word_accuracy: 0.638  early_stopping: 3/10 0.91627
+stage 110/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.99it/s val_accuracy: 0.915 val_word_accuracy: 0.649  early_stopping: 4/10 0.91627
+stage 111/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.76it/s val_accuracy: 0.916 val_word_accuracy: 0.652  early_stopping: 5/10 0.91627
+stage 112/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.01it/s val_accuracy: 0.917 val_word_accuracy: 0.648  early_stopping: 0/10 0.91687
+stage 113/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.98it/s val_accuracy: 0.919 val_word_accuracy: 0.644  early_stopping: 0/10 0.91882
+stage 114/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.93it/s val_accuracy: 0.916 val_word_accuracy: 0.649  early_stopping: 1/10 0.91882
+stage 115/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.93it/s val_accuracy: 0.914 val_word_accuracy: 0.637  early_stopping: 2/10 0.91882
+stage 116/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.86it/s val_accuracy: 0.914 val_word_accuracy: 0.654  early_stopping: 3/10 0.91882
+stage 117/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.92it/s val_accuracy: 0.92 val_word_accuracy: 0.657  early_stopping: 0/10 0.91957
+stage 118/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.89it/s val_accuracy: 0.917 val_word_accuracy: 0.647  early_stopping: 1/10 0.91957
+stage 119/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.95it/s val_accuracy: 0.916 val_word_accuracy: 0.648  early_stopping: 2/10 0.91957
+stage 120/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.92it/s val_accuracy: 0.919 val_word_accuracy: 0.663  early_stopping: 3/10 0.91957
+stage 121/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.96it/s val_accuracy: 0.918 val_word_accuracy: 0.663  early_stopping: 4/10 0.91957
+stage 122/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.90it/s val_accuracy: 0.918 val_word_accuracy: 0.652  early_stopping: 5/10 0.91957
+stage 123/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.07it/s val_accuracy: 0.914 val_word_accuracy: 0.646  early_stopping: 6/10 0.91957
+stage 124/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 12.83it/s val_accuracy: 0.919 val_word_accuracy: 0.668  early_stopping: 7/10 0.91957
+stage 125/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.01it/s val_accuracy: 0.919 val_word_accuracy: 0.662  early_stopping: 8/10 0.91957
+stage 126/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.02it/s val_accuracy: 0.918 val_word_accuracy: 0.665  early_stopping: 9/10 0.91957
+stage 127/∞ ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 281/281 0:00:22 • 0:00:00 13.01it/s val_accuracy: 0.919 val_word_accuracy: 0.659  early_stopping: 10/10 0.91957
+Moving best model /home/ROCQ/almanach/achague/peraire/peraire-ground-truth/scratch/peraire_D_scratch_117.mlmodel (0.9195730686187744) to /home/ROCQ/almanach/achague/peraire/peraire-ground-truth/scratch/peraire_D_scratch_best.mlmodel
diff --git a/test.files b/test.files
new file mode 100644
index 0000000..35bb478
--- /dev/null
+++ b/test.files
@@ -0,0 +1,8 @@
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/test/test_B.12.japon3_0048.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/test/test_B.18.chine4_0003.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/test/test_B.3.pologne-urss1_0007.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/test/test_B.5.urss3_0045.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/test/test_D.3.eurasie-5-6_0018.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/test/test_D.4.eurasie-7-8_0077.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/test/test_D.7.eurasie-13-14_0056.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/test/test_D.8.eurasie-15_0062.xml
diff --git a/train.files b/train.files
new file mode 100644
index 0000000..ef82350
--- /dev/null
+++ b/train.files
@@ -0,0 +1,67 @@
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.1.intro-eurasie_0005.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.10.chine2_0020.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.11.japon2_0043.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.12.japon3_0014.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.13.japon3b_0048.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.14.japon4_0005.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.15.japon5_0009.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.16.chine2b_0014.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.17.chine3_0003.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.18.chine4_0024.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.19.chine5_0035.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.1b.europe-centrale_0005.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.2.europe-orientale_0007.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.2.europe-orientale_0029.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.20.chine6_0051.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.21.chine7_0022.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.22.indochine2_0025.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.23.indochine3_0018.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.24.cambodge_0037.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.25.siam_0003.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.26.malais_0048.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.27.java1_0006.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.28.java2_0017.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.29.fin_0025.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.3.pologne-urss1_0026.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.30.tdm_0014.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.30.tdm_0023.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.4.urss2_0017.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.5.urss3_0003.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.6.urss4_0050.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.7.urss5_0019.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.8.urss6_0049.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.9.chine1_0052.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.1.notes-1-et-2_0010.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.1.notes-1-et-2_0011.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.1.notes-1-et-2_0012.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.1.notes-1-et-2_0013.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.2.eurasie-3-4_0016.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.2.eurasie-3-4_0017.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.2.eurasie-3-4_0018.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.2.eurasie-3-4_0019.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.2.eurasie-3-4_0020.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.3.eurasie-5-6_0017.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.3.eurasie-5-6_0019.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.3.eurasie-5-6_0020.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.4.eurasie-7-8_0078.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.4.eurasie-7-8_0079.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.4.eurasie-7-8_0080.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.5.eurasie-9-10_0058.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.5.eurasie-9-10_0059.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.5.eurasie-9-10_0060.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.5.eurasie-9-10_0061.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.6.eurasie-11-12_0004.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.6.eurasie-11-12_0005.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.6.eurasie-11-12_0006.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.6.eurasie-11-12_0007.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.7.eurasie-13-14_0055.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.7.eurasie-13-14_0057.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.7.eurasie-13-14_0058.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.8.eurasie-15_0061.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.8.eurasie-15_0063.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.8.eurasie-15_0064.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.9.eurasie-16-17_0013.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.9.eurasie-16-17_0014.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.9.eurasie-16-17_0058.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.9.eurasie-16-17_0059.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.9.eurasie-16-17_0100.xml
diff --git a/train.sh b/train.sh
new file mode 100755
index 0000000..d2152fa
--- /dev/null
+++ b/train.sh
@@ -0,0 +1,126 @@
+#!/bin/bash
+
+source $PWD/../anaconda3/etc/profile.d/conda.sh # for conda envs
+
+echo "Stage 1 of 4: Identifying the datasets"
+
+#playground=$(pwd)/playground
+echo "Building playground if necessary"
+#mkdir -p $playground
+
+# test if the playground directory exists
+#if [ ! -d "$playground" ]; then
+#    echo "The playground directory does not exist. Please run step 1 of the workflow first."
+#    exit 1
+#fi
+
+ls -d $PWD/data/train/*.xml > "$PWD/train.files"
+ls -d $PWD/data/test/*.xml > "$PWD/test.files"
+
+ls -d $PWD/data/train/D*.xml > "$PWD/train_D.files"
+ls -d $PWD/data/train/B*.xml > "$PWD/train_B.files"
+
+echo "Stage 1 of 4: Identifying the datasets - DONE"
+
+# -------------------------------------------
+
+echo "Stage 2 of 4: Compiling the datasets into arrow files"
+echo "This may take a while."
+
+n_workers_comp=8
+n_workers_train=16
+batch_size=4 # the greater the number, the more general the models will be
+
+echo "Activating Kraken conda environment"
+conda activate kraken_latest
+
+# if "kraken --version" does not work, stop the script
+if ! kraken --version; then
+    echo "Couldn't find kraken. Please check that the conda environment is activated."
+    exit 1
+fi
+
+echo "Working with $(kraken --version)"
+
+echo "Compiling the datasets into arrow files"
+
+echo "Compiling the train and dev files into arrow files"
+#cat $PWD/train.files | xargs ketos compile --workers $n_workers_comp -o "$PWD/train.arrow" --random-split 0.9 0.1 0.0 --save-splits -f xml
+cat $PWD/train_B.files | xargs ketos compile --workers $n_workers_comp -o "$PWD/train_B.arrow" --random-split 0.9 0.1 0.0 --save-splits -f xml
+cat $PWD/train_D.files | xargs ketos compile --workers $n_workers_comp -o "$PWD/train_D.arrow" --random-split 0.9 0.1 0.0 --save-splits -f xml
+
+echo "Compiling the test files into arrow files"
+echo "skipping it for now"
+#cat $PWD/test.files | xargs ketos compile --workers $n_workers_comp -o "$PWD/test.arrow" --random-split 0.0 0.0 1.0 --save-splits -f xml
+
+#for file in $PWD/*.files
+#do
+#    base=$(basename $file .files)
+#    #cat $file | xargs ketos compile --workers $n_workers -o "$playground/$base.arrow" --random-split .9 .1 .0 --save-splits -f alto 
+#    if [[ $base != *"test" ]]
+#    then
+#        echo "Compiling the train and dev files into arrow files"
+#        cat $file | xargs ketos compile --workers $n_workers -o "$PWD/$base.arrow" --random-split .9 .1 .0 --save-splits -f alto 
+#    fi
+#    if [[ $base == *"test" ]]
+#    then
+#        echo "Compiling the test files into arrow files"
+#        cat $file | xargs ketos compile --workers $n_workers -o "$PWD/$base.arrow" --random-split .0 .0 1.0 --save-splits -f alto 
+#    fi
+#done
+
+#echo "Making a list of dev and train arrow files for each datasets" # cf. issue #397 of kraken
+#ls -d $PWD/*.arrow > "arrow.files"
+
+echo "Stage 2 of 4: Compiling the datasets into arrow files - DONE"
+
+# -------------------------------------------
+
+echo "Stage 3 of 4: Training Kraken models"
+echo "This may take a while."
+
+
+#n_workers=16
+#n_epochs=50 # 4 for test purposes, 50 for real
+#batch_size=2 # the greater the number, the more general the models will be
+#seeds="42 123 456 789 1011"
+
+
+echo "Start training the models"
+
+#cat $file | xargs ketos -s $seed train -o $playground/$base/models/$seed-$base -f binary --device cuda:0 -u NFC -s '[1,120,0,1 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 S1(1x0)1,3 Lbx200 Do0.1,2 Lbx200 Do.1,2 Lbx200 Do]' -r 0.0001 -B $batch_size --augment --workers $n_workers --epochs $n_epochs -q dumb --logger tensorboard --log-dir $playground/$base > "$playground/$base/logs/$seed-$base.log"
+#ls "$PWD/train.arrow" | xargs ketos train -o $PWD/peraire2_ft_MMCFR -f binary --device cuda:0 -u NFC -s '[1,120,0,1 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 S1(1x0)1,3 Lbx200 Do0.1,2 Lbx200 Do.1,2 Lbx200 Do]' -r 0.0001 -B $batch_size --augment --workers $n_workers > "$PWD/kraken.log"
+
+#ls train.arrow | xargs ketos train -f binary --load HTR-United-Manu_McFrench.mlmodel -o $PWD/models/peraire2_ft_MMCFR --resize both --device cuda:0 -u NFC -s '[1,120,0,1 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 S1(1x0)1,3 Lbx200 Do0.1,2 Lbx200 Do.1,2 Lbx200 Do]' -r 0.0001 -B $batch_size --augment --workers $n_workers_train --min-epochs 15 > "$PWD/kraken.log"
+#ls train_B.arrow | xargs ketos train -f binary --load HTR-United-Manu_McFrench.mlmodel -o $PWD/models/peraireB_ft_MMCFR --resize both --device cuda:0 -u NFC -s '[1,120,0,1 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 S1(1x0)1,3 Lbx200 Do0.1,2 Lbx200 Do.1,2 Lbx200 Do]' -r 0.0001 -B $batch_size --augment --workers $n_workers_train --min-epochs 15 > "$PWD/kraken_B.log"
+#ls train_D.arrow | xargs ketos train -f binary --load HTR-United-Manu_McFrench.mlmodel -o $PWD/models/peraireD_ft_MMCFR --resize both --device cuda:0 -u NFC -s '[1,120,0,1 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 S1(1x0)1,3 Lbx200 Do0.1,2 Lbx200 Do.1,2 Lbx200 Do]' -r 0.0001 -B $batch_size --augment --workers $n_workers_train --min-epochs 15 > "$PWD/kraken_D.log"
+
+# !!! J'ai fait les entraînements la première en mode "both" au lieu de "add": en "both", on ne garde que les caractères vus dans le new train set
+ls train.arrow | xargs ketos train -f binary -o $PWD/scratch/peraire_BD_scratch --device cuda:0 -u NFC -s '[1,120,0,1 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 S1(1x0)1,3 Lbx200 Do0.1,2 Lbx200 Do.1,2 Lbx200 Do]' -r 0.0001 -B 4 --augment --workers 16 --min-epochs 15 > "$PWD/kraken_BD_scratch.log"
+ls train_B.arrow | xargs ketos train -f binary -o $PWD/scratch/peraire_B_scratch --device cuda:1 -u NFC -s '[1,120,0,1 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 S1(1x0)1,3 Lbx200 Do0.1,2 Lbx200 Do.1,2 Lbx200 Do]' -r 0.0001 -B 4 --augment --workers 16 --min-epochs 15 > "$PWD/kraken_B_scratch.log"
+ls train_D.arrow | xargs ketos train -f binary -o $PWD/scratch/peraire_D_scratch --device cuda:1 -u NFC -s '[1,120,0,1 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 S1(1x0)1,3 Lbx200 Do0.1,2 Lbx200 Do.1,2 Lbx200 Do]' -r 0.0001 -B 4 --augment --workers 16 --min-epochs 15 > "$PWD/kraken_D_scratch.log"
+
+ls train.arrow | xargs ketos train -f binary --load HTR-United-Manu_McFrench.mlmodel -o $PWD/finetuned/peraire_BD_add_ --resize add --device cuda:0 -u NFC -s '[1,120,0,1 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 S1(1x0)1,3 Lbx200 Do0.1,2 Lbx200 Do.1,2 Lbx200 Do]' -r 0.0001 -B 4 --augment --workers 16 --min-epochs 15 > "$PWD/kraken_BD_add.log"
+ls train_B.arrow | xargs ketos train -f binary --load HTR-United-Manu_McFrench.mlmodel -o $PWD/finetuned/peraire_B_add --resize add --device cuda:0 -u NFC -s '[1,120,0,1 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 S1(1x0)1,3 Lbx200 Do0.1,2 Lbx200 Do.1,2 Lbx200 Do]' -r 0.0001 -B 4 --augment --workers 16 --min-epochs 15 > "$PWD/kraken_B_add.log"
+ls train_D.arrow | xargs ketos train -f binary --load HTR-United-Manu_McFrench.mlmodel -o $PWD/finetuned/peraireD_D_add --resize add --device cuda:0 -u NFC -s '[1,120,0,1 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,13,32 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 Mp2,2 Cr3,9,64 Do0.1,2 S1(1x0)1,3 Lbx200 Do0.1,2 Lbx200 Do.1,2 Lbx200 Do]' -r 0.0001 -B 4 --augment --workers 16 --min-epochs 15 > "$PWD/kraken_D_add.log"
+
+conda deactivate
+
+echo "Stage 3 of 4: Training Kraken models - DONE"
+
+# -------------------------------------------
+
+echo "Stage 4 of 4: Testing the models"
+echo "Do it manually!"
+
+#for model in $PWD/*.mlmodel
+#do
+#    basemodel=$(basename $model .mlmodel)
+#    echo "Testing $model"
+#    #ketos test -m $model -e "$playground/test_files.txt" -f alto --device cuda:0 -u NFC -B $batch_size --workers $n_workers > "$folder/tests/model_$basemodel.test"
+#    ls "$PWD/test.arrow" | xargs ketos test -m $model -f binary --device cuda:0 -u NFC -B $batch_size --workers $n_workers > "$PWD/model_$basemodel.test"
+#done
+
+echo "Execution finished!"
+
+
diff --git a/train_B.files b/train_B.files
new file mode 100644
index 0000000..ff3e58e
--- /dev/null
+++ b/train_B.files
@@ -0,0 +1,33 @@
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.1.intro-eurasie_0005.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.10.chine2_0020.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.11.japon2_0043.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.12.japon3_0014.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.13.japon3b_0048.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.14.japon4_0005.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.15.japon5_0009.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.16.chine2b_0014.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.17.chine3_0003.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.18.chine4_0024.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.19.chine5_0035.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.1b.europe-centrale_0005.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.2.europe-orientale_0007.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.2.europe-orientale_0029.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.20.chine6_0051.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.21.chine7_0022.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.22.indochine2_0025.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.23.indochine3_0018.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.24.cambodge_0037.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.25.siam_0003.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.26.malais_0048.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.27.java1_0006.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.28.java2_0017.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.29.fin_0025.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.3.pologne-urss1_0026.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.30.tdm_0014.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.30.tdm_0023.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.4.urss2_0017.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.5.urss3_0003.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.6.urss4_0050.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.7.urss5_0019.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.8.urss6_0049.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/B.9.chine1_0052.xml
diff --git a/train_D.files b/train_D.files
new file mode 100644
index 0000000..6e17741
--- /dev/null
+++ b/train_D.files
@@ -0,0 +1,34 @@
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.1.notes-1-et-2_0010.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.1.notes-1-et-2_0011.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.1.notes-1-et-2_0012.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.1.notes-1-et-2_0013.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.2.eurasie-3-4_0016.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.2.eurasie-3-4_0017.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.2.eurasie-3-4_0018.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.2.eurasie-3-4_0019.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.2.eurasie-3-4_0020.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.3.eurasie-5-6_0017.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.3.eurasie-5-6_0019.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.3.eurasie-5-6_0020.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.4.eurasie-7-8_0078.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.4.eurasie-7-8_0079.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.4.eurasie-7-8_0080.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.5.eurasie-9-10_0058.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.5.eurasie-9-10_0059.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.5.eurasie-9-10_0060.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.5.eurasie-9-10_0061.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.6.eurasie-11-12_0004.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.6.eurasie-11-12_0005.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.6.eurasie-11-12_0006.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.6.eurasie-11-12_0007.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.7.eurasie-13-14_0055.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.7.eurasie-13-14_0057.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.7.eurasie-13-14_0058.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.8.eurasie-15_0061.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.8.eurasie-15_0063.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.8.eurasie-15_0064.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.9.eurasie-16-17_0013.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.9.eurasie-16-17_0014.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.9.eurasie-16-17_0058.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.9.eurasie-16-17_0059.xml
+/home/ROCQ/almanach/achague/peraire/peraire-ground-truth/data/train/D.9.eurasie-16-17_0100.xml