added eval section

README.md

@@ -107,4 +107,13 @@ if __name__ == "__main__":
 - Knowledge Distillation
 - Onnx format weights
     - The weights are converted in the Onnx format (in order to optimize CPU & GPU Performance)
-    - As tested on an AMD Instinct MI100 GPU - sub 10ms inference per 704ms audio chunk
+    - As tested on an AMD Instinct MI100 GPU - sub 10ms inference per 704ms audio chunk
+
+# Evaluation
+
+Accuracy at 0.95 with 8120 samples tested.
+
+classes | precision | recall | f1-score | support |
+---|---|---|---|---|
+eos | 0.94 | 0.95 | 0.95 | 4060 |
+not_eos | 0.95 | 0.94 | 0.95 | 4060 |