End of training

README.md

@@ -19,10 +19,10 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on an unknown dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.0333
-- Bleu: 50.39
-- Chrf: 88.17
-- Wer: 35.9658
+- Loss: 0.0332
+- Bleu: 50.76
+- Chrf: 88.11
+- Wer: 36.0650
 
 ## Model description
 
@@ -54,9 +54,9 @@ The following hyperparameters were used during training:
 
 | Training Loss | Epoch | Step  | Validation Loss | Bleu  | Chrf  | Wer     |
 |:-------------:|:-----:|:-----:|:---------------:|:-----:|:-----:|:-------:|
-| 0.0383        | 1.0   | 5323  | 0.0390          | 42.06 | 85.56 | 44.6237 |
-| 0.0071        | 2.0   | 10646 | 0.0335          | 48.32 | 87.59 | 38.2148 |
-| 0.0555        | 3.0   | 15969 | 0.0333          | 50.39 | 88.17 | 35.9658 |
+| 0.0408        | 1.0   | 5323  | 0.0388          | 42.82 | 86.05 | 42.9847 |
+| 0.0095        | 2.0   | 10646 | 0.0335          | 48.25 | 87.64 | 37.9428 |
+| 0.0364        | 3.0   | 15969 | 0.0332          | 50.76 | 88.11 | 36.0650 |
 
 
 ### Framework versions

generation_config.json

@@ -160,7 +160,7 @@
     "<|yo|>": 50325,
     "<|zh|>": 50260
   },
-  "language": "be",
+  "language": "en",
   "max_initial_timestamp_index": 50,
   "max_length": 448,
   "no_timestamps_token_id": 50363,