Add & refine WER evaluator for w2v

codes/models/asr/w2v_wrapper.py

@@ -11,6 +11,11 @@ from trainer.networks import register_model
 from utils.util import opt_get
 
 
+def only_letters(string):
+    allowlist = set(' ABCDEFGHIJKLMNOPQRSTUVWXYZ\'')
+    return ''.join(filter(allowlist.__contains__, string.upper()))
+
+
 class Wav2VecWrapper(nn.Module):
     """
     Basic wrapper class that makes Wav2Vec2 usable by DLAS.
@@ -77,8 +82,8 @@ class Wav2VecWrapper(nn.Module):
             pred_strings = []
             for last_labels, last_pred in zip(self.last_labels, self.last_pred):
                 last_labels[last_labels == -100] = 0
-                label_strings.extend([sequence_to_text(lbl) for lbl in last_labels])
-                pred_strings.extend([sequence_to_text(self.decode_ctc(pred)) for pred in last_pred])
+                label_strings.extend([only_letters(sequence_to_text(lbl)) for lbl in last_labels])
+                pred_strings.extend([only_letters(sequence_to_text(self.decode_ctc(pred))) for pred in last_pred])
             wer = wer_metric.compute(predictions=pred_strings, references=label_strings)
             res['wer'] = wer
             print(f"Sample prediction: {pred_strings[0]} <=> {label_strings[0]}")
@@ -88,7 +93,7 @@ class Wav2VecWrapper(nn.Module):
         audio_norm = (audio - audio.mean()) / torch.sqrt(audio.var() + 1e-7)
         logits = self.w2v(input_values=audio_norm.squeeze(1)).logits
         pred = logits.argmax(dim=-1)
-        return self.decode_ctc(pred)
+        return [self.decode_ctc(p) for p in pred]
 
 
 @register_model
@@ -97,6 +102,7 @@ def register_wav2vec2_finetune(opt_net, opt):
 
 
 if __name__ == '__main__':
+    print(only_letters("Hello, world!"))
     w2v = Wav2VecWrapper(basis_model='facebook/wav2vec2-large-960h', freeze_transformer=True)
     loss = w2v(torch.randn(2,1,50000), torch.randint(0,40,(2,70)), torch.tensor([20000, 30000]), torch.tensor([35, 50]))
     w2v.get_debug_values(0,"")

codes/trainer/eval/eval_wer.py

@@ -0,0 +1,48 @@
+from copy import deepcopy
+
+from datasets import load_metric
+
+import torch
+import trainer.eval.evaluator as evaluator
+from data import create_dataset, create_dataloader
+from models.asr.w2v_wrapper import only_letters
+from models.tacotron2.text import sequence_to_text
+
+
+class WerEvaluator(evaluator.Evaluator):
+    """
+    Evaluator that produces the WER for a speech recognition model on a test set.
+    """
+    def __init__(self, model, opt_eval, env):
+        super().__init__(model, opt_eval, env, uses_all_ddp=False)
+        self.clip_key = opt_eval['clip_key']
+        self.clip_lengths_key = opt_eval['clip_lengths_key']
+        self.text_seq_key = opt_eval['text_seq_key']
+        self.text_seq_lengths_key = opt_eval['text_seq_lengths_key']
+        self.wer_metric = load_metric('wer')
+
+    def perform_eval(self):
+        val_opt = deepcopy(self.env['opt']['datasets']['val'])
+        val_opt['batch_size'] = 1  # This is important to ensure no padding.
+        val_dataset, collate_fn = create_dataset(val_opt, return_collate=True)
+        val_loader = create_dataloader(val_dataset, val_opt, self.env['opt'], None, collate_fn=collate_fn)
+        model = self.model.module if hasattr(self.model, 'module') else self.model  # Unwrap DDP models
+        model.eval()
+        with torch.no_grad():
+            preds = []
+            reals = []
+            for batch in val_loader:
+                clip = batch[self.clip_key]
+                assert clip.shape[0] == 1
+                clip_len = batch[self.clip_lengths_key][0]
+                clip = clip[:, :, :clip_len].cuda()
+                pred_seq = model.inference(clip)
+                preds.append(only_letters(sequence_to_text(pred_seq[0])))
+                real_seq = batch[self.text_seq_key]
+                real_seq_len = batch[self.text_seq_lengths_key][0]
+                real_seq = real_seq[:, :real_seq_len]
+                reals.append(only_letters(sequence_to_text(real_seq[0])))
+        wer = self.wer_metric.compute(predictions=preds, references=reals)
+        model.train()
+        return {'eval_wer': wer}
+