optional tokenizer and fix some doc. (#3042)

examples/librispeech/asr3/run.sh

@@ -6,7 +6,7 @@ set -e
 
 gpus=0
 stage=0
-stop_stage=0
+stop_stage=4
 conf_path=conf/wav2vec2ASR.yaml
 ips=            #xx.xx.xx.xx,xx.xx.xx.xx
 decode_conf_path=conf/tuning/decode.yaml

paddlespeech/s2t/exps/wav2vec2/bin/test_wav.py

@@ -34,8 +34,9 @@ class Wav2vec2Infer():
         self.args = args
         self.config = config
         self.audio_file = args.audio_file
+        self.tokenizer = config.get("tokenizer", None)
 
-        if self.config.tokenizer:
+        if self.tokenizer:
             self.text_feature = AutoTokenizer.from_pretrained(
                 self.config.tokenizer)
         else:
@@ -72,7 +73,7 @@ class Wav2vec2Infer():
                 text_feature=self.text_feature,
                 decoding_method=decode_config.decoding_method,
                 beam_size=decode_config.beam_size,
-                tokenizer=self.config.tokenizer, )
+                tokenizer=self.tokenizer, )
             rsl = result_transcripts[0]
             utt = Path(self.audio_file).name
             logger.info(f"hyp: {utt} {rsl}")