removed GPU inference

app.py

@@ -12,7 +12,6 @@ import torch
 import gradio as gr
 from transformers import Wav2Vec2Tokenizer, Wav2Vec2ForCTC
 nltk.download("punkt")
-torch_device = 'cuda'
 
 
 # In[ ]:
@@ -23,7 +22,7 @@ model_name = "facebook/wav2vec2-base-960h"
 
 #model_name = "facebook/wav2vec2-large-xlsr-53"
 tokenizer = Wav2Vec2Tokenizer.from_pretrained(model_name)#omdel_name
-model = Wav2Vec2ForCTC.from_pretrained(model_name).to(torch_device)
+model = Wav2Vec2ForCTC.from_pretrained(model_name)
 
 
 # In[ ]:
@@ -96,7 +95,7 @@ def asr_transcript_long(input_file,tokenizer=tokenizer, model=model ):
         if sample_rate !=16000:
             speech = librosa.resample(speech, sample_rate,16000)
         input_values = tokenizer(speech, return_tensors="pt").input_values
-        logits = model(input_values.to(torch_device)).logits
+        logits = model(input_values).logits
 
         predicted_ids = torch.argmax(logits, dim=-1)
         transcription = tokenizer.decode(predicted_ids[0])