Marialab/finetuned-whisper-medium-1000-step

Finetuned Whisper medium for darija speech translation

This model is a fine-tuned version of openai/whisper-medium on the Darija-C dataset. It achieves the following results on the evaluation set:

• Loss: 0.0000
• Bleu: 0.7440

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: 1e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• 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: 100
• training_steps: 1000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu
2.3191	12.5	50	0.3022	0.4084
0.0359	25.0	100	0.0001	0.7440
0.0001	37.5	150	0.0001	0.7440
0.0001	50.0	200	0.0000	0.7440
0.0	62.5	250	0.0000	0.7440
0.0	75.0	300	0.0000	0.7440
0.0	87.5	350	0.0000	0.7440
0.0	100.0	400	0.0000	0.7440
0.0	112.5	450	0.0000	0.7440
0.0	125.0	500	0.0000	0.7440
0.0	137.5	550	0.0000	0.7440
0.0	150.0	600	0.0000	0.7440
0.0	162.5	650	0.0000	0.7440
0.0	175.0	700	0.0000	0.7440
0.0	187.5	750	0.0000	0.7440
0.0	200.0	800	0.0000	0.7440
0.0	212.5	850	0.0000	0.7440
0.0	225.0	900	0.0000	0.7440
0.0	237.5	950	0.0000	0.7440
0.0	250.0	1000	0.0000	0.7440

Framework versions

• Transformers 4.47.0
• Pytorch 2.5.1+cu121
• Datasets 2.19.2
• Tokenizers 0.21.0