Update app.py

app.py

@@ -51,7 +51,7 @@ torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
 use_flash_attention_2 = is_flash_attn_2_available()
 
 model = AutoModelForSpeechSeq2Seq.from_pretrained(
-    "openai/whisper-small", torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True, use_flash_attention_2=use_flash_attention_2
+    "models/kanyekuthi/AfriSpeech-whisper-tiny", torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True, use_flash_attention_2=use_flash_attention_2
 )
 distilled_model = AutoModelForSpeechSeq2Seq.from_pretrained(
     "models/kanyekuthi/AfriSpeech-whisper-tiny", torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True, use_flash_attention_2=use_flash_attention_2
@@ -62,7 +62,7 @@ if not use_flash_attention_2:
     model = model.to_bettertransformer()
     distilled_model = distilled_model.to_bettertransformer()
 
-processor = AutoProcessor.from_pretrained("openai/whisper-small")
+processor = AutoProcessor.from_pretrained("models/kanyekuthi/AfriSpeech-whisper-tiny")
 
 model.to(device)
 distilled_model.to(device)
@@ -205,10 +205,10 @@ torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
 use_flash_attention_2 = is_flash_attn_2_available()
 
 model = AutoModelForSpeechSeq2Seq.from_pretrained(
-    "openai/whisper-small", torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True, use_flash_attention_2=use_flash_attention_2
+    "models/kanyekuthi/AfriSpeech-whisper-tiny", torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True, use_flash_attention_2=use_flash_attention_2
 )
 distilled_model = AutoModelForSpeechSeq2Seq.from_pretrained(
-    "distil-whisper/distil-small", torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True, use_flash_attention_2=use_flash_attention_2
+    "models/kanyekuthi/AfriSpeech-whisper-tiny", torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True, use_flash_attention_2=use_flash_attention_2
 )
 
 if not use_flash_attention_2:
@@ -216,7 +216,7 @@ if not use_flash_attention_2:
     model = model.to_bettertransformer()
     distilled_model = distilled_model.to_bettertransformer()
 
-processor = AutoProcessor.from_pretrained("openai/whisper-small")
+processor = AutoProcessor.from_pretrained("models/kanyekuthi/AfriSpeech-whisper-tiny")
 
 model.to(device)
 distilled_model.to(device)