continue training (tamasheq only feature extractor)

README.md

@@ -15,8 +15,8 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model was trained from scratch on the None dataset.
 It achieves the following results on the evaluation set:
-- Loss: 1.5169
-- Wer: 0.8575
+- Loss: 1.3689
+- Wer: 0.8342
 
 ## Model description
 
@@ -36,11 +36,11 @@ More information needed
 
 The following hyperparameters were used during training:
 - learning_rate: 3e-05
-- train_batch_size: 16
+- train_batch_size: 8
 - eval_batch_size: 8
 - seed: 42
 - gradient_accumulation_steps: 2
-- total_train_batch_size: 32
+- total_train_batch_size: 16
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: linear
 - lr_scheduler_warmup_steps: 500
@@ -50,12 +50,12 @@ The following hyperparameters were used during training:
 
 | Training Loss | Epoch | Step | Validation Loss | Wer    |
 |:-------------:|:-----:|:----:|:---------------:|:------:|
-| 8.0502        | 15.79 | 300  | 3.0170          | 0.9972 |
-| 2.2958        | 31.58 | 600  | 1.2753          | 0.8575 |
-| 0.5807        | 47.37 | 900  | 1.2622          | 0.8547 |
-| 0.3897        | 63.16 | 1200 | 1.3396          | 0.8603 |
-| 0.311         | 78.95 | 1500 | 1.4590          | 0.8743 |
-| 0.2819        | 94.74 | 1800 | 1.5169          | 0.8575 |
+| 7.4856        | 5.71  | 300  | 2.9956          | 0.9974 |
+| 2.3903        | 11.43 | 600  | 1.2600          | 0.8816 |
+| 0.9577        | 17.14 | 900  | 1.1878          | 0.8342 |
+| 0.7051        | 22.86 | 1200 | 1.1907          | 0.8053 |
+| 0.5821        | 28.57 | 1500 | 1.2621          | 0.8316 |
+| 0.5037        | 34.29 | 1800 | 1.3689          | 0.8342 |
 
 
 ### Framework versions

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:28ed9fb28eea050068b246d10d18d70fdd64fa6d7d7b2fc90c061dd0e74e09b8
+oid sha256:b38d7dd7b1a594c232bef815eb7488de6a8f01d655b7c936d9b61e40f4d6f452
 size 1262078125