End of training

README.md

@@ -1,4 +1,5 @@
 ---
+library_name: transformers
 language:
 - ar
 license: apache-2.0
@@ -6,11 +7,12 @@ base_model: openai/whisper-base
 tags:
 - generated_from_trainer
 datasets:
-- ymoslem/MediaSpeech
-- fixie-ai/common_voice_17_0
+- deepdml/Tunisian_MSA
 - UBC-NLP/Casablanca
+- fixie-ai/common_voice_17_0
 - google/fleurs
-- deepdml/Tunisian_MSA
+- pain/MASC
+- ymoslem/MediaSpeech
 metrics:
 - wer
 model-index:
@@ -21,12 +23,13 @@ model-index:
       type: automatic-speech-recognition
     dataset:
       name: Common Voice 17.0
-      type: ymoslem/MediaSpeech
+      type: deepdml/Tunisian_MSA
     metrics:
     - name: Wer
       type: wer
-      value: 45.035897247571654
+      value: 39.07383265088779
 ---
+
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
 should probably proofread and complete it, then remove this comment. -->
 
@@ -34,9 +37,9 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [openai/whisper-base](https://huggingface.co/openai/whisper-base) on the Common Voice 17.0 dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.5226
-- Wer: 45.0359
-- Cer: 14.7675
+- Loss: 0.4415
+- Wer: 39.0738
+- Cer: 12.5149
 
 ## Model description
 
@@ -59,41 +62,38 @@ The following hyperparameters were used during training:
 - train_batch_size: 64
 - eval_batch_size: 64
 - seed: 42
-- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
+- optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
 - lr_scheduler_type: linear
 - lr_scheduler_warmup_ratio: 0.04
-- training_steps: 8000
+- training_steps: 18000
 
 ### Training results
 
-| Training Loss | Epoch  | Step | Validation Loss | Wer     | Cer     |
-|:-------------:|:------:|:----:|:---------------:|:-------:|:-------:|
-| 0.6613        | 0.125  | 1000 | 0.6130          | 53.4217 | 18.1546 |
-| 0.3701        | 0.25   | 2000 | 0.5791          | 50.5885 | 17.1229 |
-| 0.1948        | 0.375  | 3000 | 0.5610          | 49.0241 | 16.4269 |
-| 0.1173        | 0.5    | 4000 | 0.5481          | 47.8526 | 16.0193 |
-| 0.089         | 0.625  | 5000 | 0.5337          | 47.0171 | 15.6262 |
-| 0.0606        | 0.75   | 6000 | 0.5255          | 45.5317 | 15.0271 |
-| 0.046         | 1.0442 | 7000 | 0.5235          | 45.1791 | 14.8553 |
-| 0.0458        | 1.1692 | 8000 | 0.5226          | 45.0359 | 14.7675 |
+| Training Loss | Epoch  | Step  | Validation Loss | Wer     | Cer     |
+|:-------------:|:------:|:-----:|:---------------:|:-------:|:-------:|
+| 0.7722        | 0.0556 | 1000  | 0.6243          | 54.6887 | 19.2051 |
+| 0.4875        | 0.1111 | 2000  | 0.5574          | 49.7126 | 16.7685 |
+| 0.2778        | 0.1667 | 3000  | 0.5361          | 47.8269 | 15.8165 |
+| 0.1987        | 0.2222 | 4000  | 0.5172          | 45.5776 | 14.8800 |
+| 0.1477        | 0.2778 | 5000  | 0.5015          | 45.5666 | 14.9556 |
+| 0.1382        | 0.3333 | 6000  | 0.4788          | 43.4751 | 14.1194 |
+| 0.0871        | 0.3889 | 7000  | 0.4689          | 42.8049 | 13.8911 |
+| 0.0769        | 0.4444 | 8000  | 0.4575          | 41.8777 | 13.6289 |
+| 0.082         | 0.5    | 9000  | 0.4542          | 41.3801 | 13.3983 |
+| 0.0873        | 0.5556 | 10000 | 0.4483          | 41.4700 | 13.3972 |
+| 0.0654        | 0.6111 | 11000 | 0.4399          | 40.4179 | 13.1265 |
+| 0.0537        | 0.6667 | 12000 | 0.4434          | 40.2380 | 12.9609 |
+| 0.0806        | 0.7222 | 13000 | 0.4402          | 40.1296 | 12.9885 |
+| 0.0678        | 0.7778 | 14000 | 0.4363          | 39.9589 | 12.8484 |
+| 0.064         | 0.8333 | 15000 | 0.4333          | 39.5476 | 12.7333 |
+| 0.0515        | 0.8889 | 16000 | 0.4360          | 38.8939 | 12.5456 |
+| 0.0725        | 0.9444 | 17000 | 0.4386          | 39.0793 | 12.6691 |
+| 0.0654        | 1.0    | 18000 | 0.4415          | 39.0738 | 12.5149 |
 
 
 ### Framework versions
 
-- Transformers 4.42.0.dev0
-- Pytorch 2.3.0+cu121
-- Datasets 2.19.1
-- Tokenizers 0.19.1
-
-## Citation
-
-Please cite the model using the following BibTeX entry:
-
-```bibtex
-@misc{deepdml/whisper-base-ar-mix-norm,
-      title={Fine-tuned Whisper base ASR model for speech recognition in Arabic},
-      author={Jimenez, David},
-      howpublished={\url{https://huggingface.co/deepdml/whisper-base-ar-mix-norm}},
-      year={2026}
-    }
-```
+- Transformers 4.48.0.dev0
+- Pytorch 2.5.1+cu121
+- Datasets 3.6.0
+- Tokenizers 0.21.0

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5c07bba6d9e0d56dc6f68a1e08ab779cacc3d283e8f990a7331b35a1763ff39b
+oid sha256:14f7110dac1328544e9058de3ce0473f1bb888ec39ec8ac8e7d2b5a59cce72d5
 size 290403936