carlot/whisper-base-mixedv1.3

README.md

@@ -15,8 +15,8 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model was trained from scratch on the audiofolder dataset.
 It achieves the following results on the evaluation set:
-- Loss: 3.5988
-- Cer: 127.5927
+- Loss: 0.8720
+- Cer: 32.0507
 
 ## Model description
 
@@ -35,10 +35,12 @@ More information needed
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
-- learning_rate: 0.001
+- learning_rate: 0.0001
 - train_batch_size: 2
 - eval_batch_size: 8
 - seed: 42
+- gradient_accumulation_steps: 4
+- total_train_batch_size: 8
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: linear
 - lr_scheduler_warmup_steps: 500
@@ -47,12 +49,12 @@ The following hyperparameters were used during training:
 
 ### Training results
 
-| Training Loss | Epoch | Step | Validation Loss | Cer      |
-|:-------------:|:-----:|:----:|:---------------:|:--------:|
-| 1.0341        | 8.55  | 1000 | 2.0062          | 197.7475 |
-| 0.2818        | 17.09 | 2000 | 2.2179          | 117.1281 |
-| 0.0934        | 25.64 | 3000 | 3.0194          | 185.1244 |
-| 0.0115        | 34.19 | 4000 | 3.5988          | 127.5927 |
+| Training Loss | Epoch  | Step | Validation Loss | Cer     |
+|:-------------:|:------:|:----:|:---------------:|:-------:|
+| 0.0156        | 34.19  | 1000 | 0.8428          | 26.7480 |
+| 0.0004        | 68.38  | 2000 | 0.8710          | 28.1558 |
+| 0.0004        | 102.56 | 3000 | 0.8693          | 31.2529 |
+| 0.0002        | 136.75 | 4000 | 0.8720          | 32.0507 |
 
 
 ### Framework versions

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:144a63a051a1be3ae375445f9b13f90bc4ca0d6b6165698fd7ab65f826685eaa
+oid sha256:030a707ffbd000b1288727bc6fa45ec97b59b18f506ed93547bd92ad6d6ad39b
 size 290403936