Update MHSA-Word-Model/README.md

MHSA-Word-Model/README.md

@@ -7,7 +7,7 @@ CRNN with Transformer-style Multi-Head Self-Attention between BiLSTM layers 2 an
 - CNN backbone: 6 convolutional blocks (max 256 channels)
 - Recurrent block: 3 BiLSTM layers, hidden = 160 per direction
 - Decoding: Connectionist Temporal Classification (CTC)
-- Parameters: 4,047,576
+- Parameters: 4,044,625
 
 ## Test Performance (DASTNUS, seed 42)
 | Metric | Value |