README.md

metadata

library_name: transformers
language:
  - ar
license: apache-2.0
base_model: openai/whisper-base
tags:
  - generated_from_trainer
metrics:
  - wer
model-index:
  - name: Whisper base AR - YA
    results: []

Whisper base AR - YA

This model is a fine-tuned version of openai/whisper-base on the quran-ayat-speech-to-text dataset. It achieves the following results on the evaluation set:

• Loss: 0.0479
• Wer: 0.2408

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0001
• train_batch_size: 8
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• 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_steps: 500
• num_epochs: 25
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0498	0.4651	100	0.0378	0.3465
0.0397	0.9302	200	0.0313	0.2965
0.0235	1.3953	300	0.0323	0.2712
0.0342	1.8605	400	0.0337	0.2919
0.0151	2.3256	500	0.0428	0.2779
0.0206	2.7907	600	0.0382	0.3221
0.0115	3.2558	700	0.0414	0.3213
0.0126	3.7209	800	0.0435	0.3636
0.0075	4.1860	900	0.0425	0.2909
0.0071	4.6512	1000	0.0439	0.3039
0.0049	5.1163	1100	0.0442	0.2731
0.0047	5.5814	1200	0.0443	0.3161
0.0047	6.0465	1300	0.0452	0.2933
0.0027	6.5116	1400	0.0459	0.2898
0.004	6.9767	1500	0.0456	0.2771
0.0017	7.4419	1600	0.0463	0.2482
0.0018	7.9070	1700	0.0458	0.2818
0.0016	8.3721	1800	0.0464	0.2523
0.0014	8.8372	1900	0.0460	0.2621
0.0012	9.3023	2000	0.0474	0.2606
0.0009	9.7674	2100	0.0473	0.2575
0.0008	10.2326	2200	0.0479	0.2723
0.0008	10.6977	2300	0.0469	0.2710
0.0007	11.1628	2400	0.0488	0.2773
0.0008	11.6279	2500	0.0481	0.2507
0.0006	12.0930	2600	0.0481	0.2632
0.0007	12.5581	2700	0.0496	0.2586
0.0003	13.0233	2800	0.0477	0.2646
0.0003	13.4884	2900	0.0484	0.2455
0.0003	13.9535	3000	0.0476	0.2496
0.0003	14.4186	3100	0.0485	0.2487
0.0001	14.8837	3200	0.0483	0.2474
0.0002	15.3488	3300	0.0484	0.2468
0.0002	15.8140	3400	0.0484	0.2469
0.0001	16.2791	3500	0.0485	0.2448
0.0	16.7442	3600	0.0479	0.2408
0.0001	17.2093	3700	0.0481	0.2554
0.0	17.6744	3800	0.0482	0.2505
0.0	18.1395	3900	0.0482	0.2501
0.0	18.6047	4000	0.0483	0.2480
0.0	19.0698	4100	0.0484	0.2431
0.0	19.5349	4200	0.0485	0.2470
0.0	20.0	4300	0.0486	0.2447
0.0	20.4651	4400	0.0486	0.2461
0.0	20.9302	4500	0.0487	0.2539
0.0	21.3953	4600	0.0488	0.2482
0.0	21.8605	4700	0.0488	0.2505
0.0	22.3256	4800	0.0488	0.2482
0.0	22.7907	4900	0.0489	0.2487
0.0	23.2558	5000	0.0489	0.2494
0.0	23.7209	5100	0.0489	0.2525
0.0	24.1860	5200	0.0489	0.2479
0.0	24.6512	5300	0.0490	0.2440

Framework versions

• Transformers 4.46.3
• Pytorch 2.4.0
• Datasets 3.1.0
• Tokenizers 0.20.3