asr

app.py

@@ -8,7 +8,7 @@ mms_transcribe = gr.Interface(
     inputs=[
         gr.Audio(),
         gr.Dropdown(
-            choices=[model for model in asr.models_info] + ["Compare All Models"],
+            choices=[model for model in asr.models_info],
             label="Select a Model for ASR",
             value="ixxan/wav2vec2-large-mms-1b-uyghur-latin",
             interactive=True
@@ -54,6 +54,7 @@ tabbed_interface = gr.TabbedInterface(
 )
 
 with gr.Blocks() as demo:
+    gr.Markdown("Comparision of STT and TTS models for Uyghur language.")
     tabbed_interface.render()
 
 if __name__ == "__main__":

asr.py

@@ -19,6 +19,12 @@ models_info = {
         "ctc_model": False,
         "arabic_script": False
     },
+    "Meta-MMS-Uyghur": {
+        "processor": AutoProcessor.from_pretrained("facebook/mms-1b-all", target_lang='uig-script_arabic'),
+        "model": AutoModelForCTC.from_pretrained("facebook/mms-1b-all", target_lang='uig-script_arabic', ignore_mismatched_sizes=True),
+        "ctc_model": True,
+        "arabic_script": True
+    },
     "ixxan/whisper-small-thugy20": {
         "processor": AutoProcessor.from_pretrained("ixxan/whisper-small-thugy20"),
         "model": AutoModelForSpeechSeq2Seq.from_pretrained("ixxan/whisper-small-thugy20"),
@@ -31,12 +37,6 @@ models_info = {
         "ctc_model": False,
         "arabic_script": False
     },
-    "Meta-MMS": {
-        "processor": AutoProcessor.from_pretrained("facebook/mms-1b-all", target_lang='uig-script_arabic'),
-        "model": AutoModelForCTC.from_pretrained("facebook/mms-1b-all", target_lang='uig-script_arabic', ignore_mismatched_sizes=True),
-        "ctc_model": True,
-        "arabic_script": True
-    },
     "ixxan/wav2vec2-large-mms-1b-uyghur-latin": {
         "processor": Wav2Vec2Processor.from_pretrained("ixxan/wav2vec2-large-mms-1b-uyghur-latin", target_lang='uig-script_latin'),
         "model": Wav2Vec2ForCTC.from_pretrained("ixxan/wav2vec2-large-mms-1b-uyghur-latin", target_lang='uig-script_latin'),