Update app.py to change model loading to use float16 for improved performance, and remove dynamic quantization. Update README.md to change the emoji and adjust color settings for better visual consistency.

README.md

@@ -1,8 +1,8 @@
 ---
 title: Transcribed Urdu
-emoji: ☪️
+emoji: 🎙️
 colorFrom: indigo
-colorTo: blue
+colorTo: indigo
 sdk: gradio
 sdk_version: 5.35.0
 app_file: app.py
@@ -14,6 +14,3 @@ short_description: The most accurate Urdu speech recognition app.
 # whisper-large-v3-turbo-urdu
 
 This model is a fine-tuned version of [openai/whisper-large-v3-turbo](https://huggingface.co/openai/whisper-large-v3-turbo) on the common_voice_17_0 dataset.
-It achieves the following results on the evaluation set:
-- Loss: 0.4630
-- Wer: 0.3826

app.py

@@ -27,11 +27,9 @@ model_id = "kingabzpro/whisper-large-v3-turbo-urdu"
 # Load in fp32 and quantize to int8
 model = AutoModelForSpeechSeq2Seq.from_pretrained(
     model_id,
-    torch_dtype=torch.float32,
+    torch_dtype=torch.float16,
     use_safetensors=True,
 )
-model.eval()
-model = torch.quantization.quantize_dynamic(model, {torch.nn.Linear}, dtype=torch.qint8)
 
 processor = AutoProcessor.from_pretrained(model_id)