updated readme

README.md

@@ -83,7 +83,7 @@ Audio should be readable by `soundfile`/`librosa`; sample rate is normalized to
 
 ### 1) Fine-tune Whisper-small
 ```bash
-python scripts/finetune_whisper_small.py   --csv /absolute/path/to/data_for_whisper.csv   --out_dir models/whisper-small-finetuned   --batch 4   --num_workers 4
+python scripts/finetune_whisper_small.py   --csv /absolute/path/to/sample_stt.csv   --out_dir models/whisper-small-finetuned   --batch 4   --num_workers 4
 ```
 
 - The trainer uses **lazy transforms** to avoid OOM with large datasets (100k+ files ok).
@@ -96,7 +96,7 @@ ct2-transformers-converter   --model models/whisper-small-finetuned   --output_d
 
 ### 2) Train the text classifier
 ```bash
-python scripts/train_text_classifier.py   --csv /absolute/path/to/data_for_text.csv   --out models/text_cls.joblib   --word_ngrams 1,2   --char_ngrams 3,6
+python scripts/train_text_classifier.py   --csv /absolute/path/to/sample_text_labels.csv   --out models/text_cls.joblib   --word_ngrams 1,2   --char_ngrams 3,6
 ```
 The script prints precision/recall/F1 on a held-out split and writes `models/text_cls.joblib`.