Update app.py

app.py

@@ -13,7 +13,7 @@ from bambara_utils import BambaraWhisperTokenizer
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
 # Define the model checkpoint and language
-# model_checkpoint = "oza75/whisper-bambara-asr-002"
+# model_checkpoint = "oza75/whisper-bambara-asr-002" # first model
 # revision = "831cd15ed74a554caac9f304cf50dc773841ba1b"
 # model_checkpoint = "oza75/whisper-bambara-asr-005"
 # revision = "6a92cd0f19985d12739c2f6864607627115e015d" # first good checkpoint for bambara
@@ -27,12 +27,15 @@ device = "cuda" if torch.cuda.is_available() else "cpu"
 #revision = "96535debb4ce0b7af7c9c186d09d088825f63840"
 #revision = "4549778c08f29ed2e033cc9a497a187488b6bf56"
 
-model_checkpoint = "oza75/bm-whisper-02"
-revision = "06e81aa0214f6d07d3d787b367e3e8357b171549"
+# model_checkpoint = "oza75/bm-whisper-02"
+# revision = "06e81aa0214f6d07d3d787b367e3e8357b171549"
 
 # language = "bambara"
-language = "icelandic" # we use icelandic as the model was trained to replace the icelandic with bambara.
+# language = "icelandic" # we use icelandic as the model was trained to replace the icelandic with bambara.
 
+model_checkpoint = "oza75/bm-whisper-from-swa-02"
+revision = None
+language = "swahili"
 
 # Load the custom tokenizer designed for Bambara and the ASR model
 #tokenizer = BambaraWhisperTokenizer.from_pretrained(model_checkpoint, language=language, device=device)