wav2vec2 demo update: support different optimizer and lr_schedular, align mdoel, update input type, test=asr

examples/librispeech/asr3/RESULTS.md

@@ -5,4 +5,4 @@ train: Epoch 1, 1*V100-32G, batchsize: 6
 
 | Model | Params | Config | Augmentation| Test set | Decode method | WER |  
 | --- | --- | --- | --- | --- | --- | --- |
-| wav2vec2ASR | 302.86 M | conf/wav2vec2ASR.yaml | spec_aug | test-clean | greedy search | 0.019 |  
+| wav2vec2ASR | 302.86 M | conf/wav2vec2ASR.yaml | spec_aug | test-clean | greedy search | 0.018906 |  

examples/librispeech/asr3/conf/wav2vec2ASR.yaml

@@ -108,7 +108,7 @@ audio_augment:  # for raw audio
 ###########################################
 n_epoch: 1
 accum_grad: 1
-global_grad_clip: 3.0
+global_grad_clip: 5.0
 model_optim: adadelta
 model_optim_conf:
   lr: 0.9

paddlespeech/resource/pretrained_models.py

@@ -59,7 +59,7 @@ ssl_dynamic_pretrained_models = {
             'url':
             'https://paddlespeech.bj.bcebos.com/s2t/librispeech/asr3/wav2vec2ASR-large-960h-librispeech_ckpt_1.3.1.model.tar.gz',
             'md5':
-            '71ee652cdae60fc289a467291017b51c',
+            'cbe28d6c78f3dd2e189968402381f454',
             'cfg_path':
             'model.yaml',
             'ckpt_path':