Update README.md

# Whisper Small Serbian (sr) — fine-tuned

This is a fine-tuned version of `openai/whisper-small` for Serbian ASR (transcription).
It is optimized for **Serbian Latin** output and tuned for better WER on Serbian speech.

## Highlights
- Base model: `openai/whisper-small`
- Task: transcription (`task="transcribe"`)
- Language: Serbian (`language="sr"`)
- Output: Serbian Latin (text normalization applied in evaluation)
- Decoding: greedy (`num_beams=1`)

README.md

@@ -1,3 +1,7 @@
----
-license: apache-2.0
----
+---
+license: apache-2.0
+language:
+- sr
+base_model:
+- openai/whisper-small
+---