Update app.py

app.py

@@ -15,34 +15,35 @@ import matplotlib.pyplot as plt
 
 import gradio as gr
 
-lang = 'multilingual'
-fs = 16000 
-tag = 'ftshijt/open_li52_asr_train_asr_raw_bpe7000_valid.acc.ave_10best' 
-
-d = ModelDownloader()
-speech2text = Speech2Text(
-    **d.download_and_unpack(tag),
-    device="cpu",
-    minlenratio=0.0,
-    maxlenratio=0.0,
-    ctc_weight=0.3,
-    beam_size=10,
-    batch_size=0,
-    nbest=1
-)
+
 
 def text_normalizer(text):
     text = text.upper()
     return text.translate(str.maketrans('', '', string.punctuation))
 
-def inference(audio):
+def inference(audio, model):
+  lang = 'multilingual'
+  fs = 16000 
+  tag = model
+    
+  d = ModelDownloader()
+  speech2text = Speech2Text(
+        **d.download_and_unpack(tag),
+        device="cpu",
+        minlenratio=0.0,
+        maxlenratio=0.0,
+        ctc_weight=0.3,
+        beam_size=10,
+        batch_size=0,
+        nbest=1
+  )
   speech, rate = librosa.load(audio.name, sr=16000)
   assert rate == fs, "mismatch in sampling rate"
   nbests = speech2text(speech)
   text, *_ = nbests[0]
   return f"ASR hypothesis: {text_normalizer(text)}"
   
-inputs = gr.inputs.Audio(label="Input Audio", type="file")
+inputs = [gr.inputs.Audio(label="Input Audio", type="file"),gradio.inputs.Dropdown(choices=["ftshijt/open_li52_asr_train_asr_raw_bpe7000_valid.acc.ave_10best","Shinji Watanabe/spgispeech_asr_train_asr_conformer6_n_fft512_hop_length256_raw_en_unnorm_bpe5000_valid.acc.ave"], type="value", default="ftshijt/open_li52_asr_train_asr_raw_bpe7000_valid.acc.ave_10best", label="model")]
 outputs =  gr.outputs.Textbox(label="Output Text")
 
 title = "ESPnet2-ASR"