[ASR] support wav2vec2-zh cli, test=asr (#2697)

* support wav2vec2-zh cli, test=asr

* support wav2vec2-zh cli, test=asr

* support wav2vec2-zh cli, test=asr

* support wav2vec2-zh cli, test=asr

paddlespeech/cli/ssl/infer.py

@@ -25,6 +25,7 @@ import librosa
 import numpy as np
 import paddle
 import soundfile
+from paddlenlp.transformers import AutoTokenizer
 from yacs.config import CfgNode
 
 from ..executor import BaseExecutor
@@ -50,7 +51,7 @@ class SSLExecutor(BaseExecutor):
         self.parser.add_argument(
             '--model',
             type=str,
-            default='wav2vec2ASR_librispeech',
+            default=None,
             choices=[
                 tag[:tag.index('-')]
                 for tag in self.task_resource.pretrained_models.keys()
@@ -123,7 +124,7 @@ class SSLExecutor(BaseExecutor):
             help='Increase logger verbosity of current task.')
 
     def _init_from_path(self,
-                        model_type: str='wav2vec2ASR_librispeech',
+                        model_type: str=None,
                         task: str='asr',
                         lang: str='en',
                         sample_rate: int=16000,
@@ -134,6 +135,18 @@ class SSLExecutor(BaseExecutor):
         Init model and other resources from a specific path.
         """
         logger.debug("start to init the model")
+
+        if model_type is None:
+            if lang == 'en':
+                model_type = 'wav2vec2ASR_librispeech'
+            elif lang == 'zh':
+                model_type = 'wav2vec2ASR_aishell1'
+            else:
+                logger.error(
+                    "invalid lang, please input --lang en or --lang zh")
+            logger.debug(
+                "Model type had not been specified, default {} was used.".
+                format(model_type))
         # default max_len: unit:second
         self.max_len = 50
         if hasattr(self, 'model'):
@@ -167,9 +180,13 @@ class SSLExecutor(BaseExecutor):
         self.config.merge_from_file(self.cfg_path)
         if task == 'asr':
             with UpdateConfig(self.config):
-                self.text_feature = TextFeaturizer(
-                    unit_type=self.config.unit_type,
-                    vocab=self.config.vocab_filepath)
+                if lang == 'en':
+                    self.text_feature = TextFeaturizer(
+                        unit_type=self.config.unit_type,
+                        vocab=self.config.vocab_filepath)
+                elif lang == 'zh':
+                    self.text_feature = AutoTokenizer.from_pretrained(
+                        self.config.tokenizer)
                 self.config.decode.decoding_method = decode_method
             model_name = model_type[:model_type.rindex(
                 '_')]  # model_type: {model_name}_{dataset}
@@ -253,7 +270,8 @@ class SSLExecutor(BaseExecutor):
                     audio,
                     text_feature=self.text_feature,
                     decoding_method=cfg.decoding_method,
-                    beam_size=cfg.beam_size)
+                    beam_size=cfg.beam_size,
+                    tokenizer=getattr(self.config, 'tokenizer', None))
                 self._outputs["result"] = result_transcripts[0][0]
             except Exception as e:
                 logger.exception(e)
@@ -413,7 +431,7 @@ class SSLExecutor(BaseExecutor):
     @stats_wrapper
     def __call__(self,
                  audio_file: os.PathLike,
-                 model: str='wav2vec2ASR_librispeech',
+                 model: str=None,
                  task: str='asr',
                  lang: str='en',
                  sample_rate: int=16000,

paddlespeech/resource/pretrained_models.py

@@ -70,6 +70,38 @@ ssl_dynamic_pretrained_models = {
             'exp/wav2vec2ASR/checkpoints/avg_1.pdparams',
         },
     },
+    "wav2vec2-zh-16k": {
+        '1.3': {
+            'url':
+            'https://paddlespeech.bj.bcebos.com/s2t/aishell/asr3/wav2vec2-large-wenetspeech-self_ckpt_1.3.0.model.tar.gz',
+            'md5':
+            '00ea4975c05d1bb58181205674052fe1',
+            'cfg_path':
+            'model.yaml',
+            'ckpt_path':
+            'chinese-wav2vec2-large',
+            'model':
+            'chinese-wav2vec2-large.pdparams',
+            'params':
+            'chinese-wav2vec2-large.pdparams',
+        },
+    },
+    "wav2vec2ASR_aishell1-zh-16k": {
+        '1.3': {
+            'url':
+            'https://paddlespeech.bj.bcebos.com/s2t/aishell/asr3/wav2vec2ASR-large-aishell1_ckpt_1.3.0.model.tar.gz',
+            'md5':
+            'ac8fa0a6345e6a7535f6fabb5e59e218',
+            'cfg_path':
+            'model.yaml',
+            'ckpt_path':
+            'exp/wav2vec2ASR/checkpoints/avg_1',
+            'model':
+            'exp/wav2vec2ASR/checkpoints/avg_1.pdparams',
+            'params':
+            'exp/wav2vec2ASR/checkpoints/avg_1.pdparams',
+        },
+    },
 }
 
 # ---------------------------------

paddlespeech/s2t/models/wav2vec2/modules/modeling_wav2vec2.py

@@ -1173,10 +1173,6 @@ class Wav2Vec2ConfigPure():
         self.proj_codevector_dim = config.proj_codevector_dim
         self.diversity_loss_weight = config.diversity_loss_weight
 
-        # ctc loss
-        self.ctc_loss_reduction = config.ctc_loss_reduction
-        self.ctc_zero_infinity = config.ctc_zero_infinity
-
         # adapter
         self.add_adapter = config.add_adapter
         self.adapter_kernel_size = config.adapter_kernel_size

paddlespeech/s2t/models/wav2vec2/wav2vec2_ASR.py

@@ -76,28 +76,66 @@ class Wav2vec2ASR(nn.Layer):
                feats: paddle.Tensor,
                text_feature: Dict[str, int],
                decoding_method: str,
-               beam_size: int):
+               beam_size: int,
+               tokenizer: str=None):
         batch_size = feats.shape[0]
 
         if decoding_method == 'ctc_prefix_beam_search' and batch_size > 1:
-            logger.error(
-                f'decoding mode {decoding_method} must be running with batch_size == 1'
+            raise ValueError(
+                f"decoding mode {decoding_method} must be running with batch_size == 1"
             )
-            logger.error(f"current batch_size is {batch_size}")
-            sys.exit(1)
 
         if decoding_method == 'ctc_greedy_search':
-            hyps = self.ctc_greedy_search(feats)
-            res = [text_feature.defeaturize(hyp) for hyp in hyps]
-            res_tokenids = [hyp for hyp in hyps]
+            if tokenizer is None:
+                hyps = self.ctc_greedy_search(feats)
+                res = [text_feature.defeaturize(hyp) for hyp in hyps]
+                res_tokenids = [hyp for hyp in hyps]
+            else:
+                hyps = self.ctc_greedy_search(feats)
+                res = []
+                res_tokenids = []
+                for sequence in hyps:
+                    # Decode token terms to words
+                    predicted_tokens = text_feature.convert_ids_to_tokens(
+                        sequence)
+                    tmp_res = []
+                    tmp_res_tokenids = []
+                    for c in predicted_tokens:
+                        if c == "[CLS]":
+                            continue
+                        elif c == "[SEP]" or c == "[PAD]":
+                            break
+                        else:
+                            tmp_res.append(c)
+                            tmp_res_tokenids.append(text_feature.vocab[c])
+                    res.append(''.join(tmp_res))
+                    res_tokenids.append(tmp_res_tokenids)
         # ctc_prefix_beam_search and attention_rescoring only return one
         # result in List[int], change it to List[List[int]] for compatible
         # with other batch decoding mode
         elif decoding_method == 'ctc_prefix_beam_search':
             assert feats.shape[0] == 1
-            hyp = self.ctc_prefix_beam_search(feats, beam_size)
-            res = [text_feature.defeaturize(hyp)]
-            res_tokenids = [hyp]
+            if tokenizer is None:
+                hyp = self.ctc_prefix_beam_search(feats, beam_size)
+                res = [text_feature.defeaturize(hyp)]
+                res_tokenids = [hyp]
+            else:
+                hyp = self.ctc_prefix_beam_search(feats, beam_size)
+                res = []
+                res_tokenids = []
+                predicted_tokens = text_feature.convert_ids_to_tokens(hyp)
+                tmp_res = []
+                tmp_res_tokenids = []
+                for c in predicted_tokens:
+                    if c == "[CLS]":
+                        continue
+                    elif c == "[SEP]" or c == "[PAD]":
+                        break
+                    else:
+                        tmp_res.append(c)
+                        tmp_res_tokenids.append(text_feature.vocab[c])
+                res.append(''.join(tmp_res))
+                res_tokenids.append(tmp_res_tokenids)
         else:
             raise ValueError(
                 f"wav2vec2 not support decoding method: {decoding_method}")