End of training

README.md

@@ -20,7 +20,7 @@ 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 the Common Voice 16.1 dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.9023
+- Loss: 0.9134
 
 ## Model description
 
@@ -48,16 +48,15 @@ The following hyperparameters were used during training:
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: linear
 - lr_scheduler_warmup_ratio: 0.1
-- num_epochs: 5
+- num_epochs: 2
 - mixed_precision_training: Native AMP
 
 ### Training results
 
-| Training Loss | Epoch | Step  | Validation Loss |
-|:-------------:|:-----:|:-----:|:---------------:|
-| 0.9794        | 1.0   | 2500  | 1.0233          |
-| 0.7059        | 3.0   | 7500  | 0.8989          |
-| 0.5047        | 5.0   | 12500 | 0.9023          |
+| Training Loss | Epoch | Step | Validation Loss |
+|:-------------:|:-----:|:----:|:---------------:|
+| 0.9769        | 1.0   | 2500 | 1.0105          |
+| 0.8011        | 2.0   | 5000 | 0.9134          |
 
 
 ### Framework versions

runs/Aug22_17-58-54_4c8bbe3eacd7/events.out.tfevents.1724349534.4c8bbe3eacd7.26.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ba6041b0d07cb1d91e1439302925424c40bb0dcd59da0a245beb43ae8d73feca
-size 47895
+oid sha256:7e3490971440b98a725b3b46d60f44bbdef153ba805641f1e9003fd3713cdaef
+size 48520