Update README.md

README.md

@@ -129,11 +129,60 @@ The following hyperparameters were used during training:
 | 0.8238        | 11.88 | 1900 | 0.6735          | 0.3297 |
 | 0.7618        | 12.5  | 2000 | 0.6728          | 0.3286 |
 
-#### Choosing the best model
+#### Hyperparameter Tuning
 Several models with differet hyperparameters were trained. The following figures show the training process for three of them.
 ![wer](wandb-wer.png)
 ![loss](wandb-loss.png)
-As you can see this model performs better on evaluation set.
+'20_2000_1e-5_hp-mehrdad' is the current model and it's hyperparameter are:
+```python
+model = Wav2Vec2ForCTC.from_pretrained(
+    model_name_or_path if not last_checkpoint else last_checkpoint,
+    # hp-mehrdad: Hyperparams of 'm3hrdadfi/wav2vec2-large-xlsr-persian-v3'
+    attention_dropout = 0.05316,
+    hidden_dropout    = 0.01941,
+    feat_proj_dropout = 0.01249,
+    mask_time_prob    = 0.04529,
+    layerdrop         = 0.01377,
+    ctc_loss_reduction = 'mean',
+    ctc_zero_infinity = True,
+)
+
+learning_rate = 1e-5
+```
+The hyperparameters of '19_2000_1e-5_hp-base' are:
+```python
+model = Wav2Vec2ForCTC.from_pretrained(
+    model_name_or_path if not last_checkpoint else last_checkpoint,
+    # hp-base: Hyperparams simmilar to ('facebook/wav2vec2-large-xlsr-53' or 'facebook/wav2vec2-xls-r-300m')
+    attention_dropout = 0.1,
+    hidden_dropout    = 0.1,
+    feat_proj_dropout = 0.1,
+    mask_time_prob    = 0.075,
+    layerdrop         = 0.1,
+    ctc_loss_reduction = 'mean',
+    ctc_zero_infinity = True,
+)
+
+learing_rate = 1e-5
+```
+
+And the hyperparameters of '22_2000_1e-5_hp-masoud' are:
+```python
+model = Wav2Vec2ForCTC.from_pretrained(
+    model_name_or_path if not last_checkpoint else last_checkpoint,
+    # hp-masoud: Hyperparams of 'masoudmzb/wav2vec2-xlsr-multilingual-53-fa'
+	attention_dropout = 0.2,
+	hidden_dropout    = 0.2,
+	feat_proj_dropout = 0.1,
+	mask_time_prob    = 0.2,
+	layerdrop         = 0.2,
+	ctc_loss_reduction = 'mean',
+	ctc_zero_infinity = True,
+)
+
+learning_rate = 1e-5
+```
+As you can see this model performs better with WER metric on validation(evaluation) set.
 
 #### Framework versions