bohraanuj23/testing

MarathiLORA_test

This model is a fine-tuned version of openai/whisper-small on the Common Voice 11.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3212

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: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 50
• training_steps: 3000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
1.1553	0.2037	100	0.7732
0.6893	0.4073	200	0.6536
0.5571	0.6110	300	0.4662
0.438	0.8147	400	0.4299
0.3976	1.0183	500	0.4117
0.3582	1.2220	600	0.3969
0.3537	1.4257	700	0.3871
0.3379	1.6293	800	0.3749
0.3182	1.8330	900	0.3629
0.3198	2.0367	1000	0.3572
0.2771	2.2403	1100	0.3505
0.2886	2.4440	1200	0.3482
0.2707	2.6477	1300	0.3457
0.2777	2.8513	1400	0.3406
0.2615	3.0550	1500	0.3337
0.2361	3.2587	1600	0.3320
0.2436	3.4623	1700	0.3319
0.2375	3.6660	1800	0.3284
0.2321	3.8697	1900	0.3282
0.224	4.0733	2000	0.3255
0.211	4.2770	2100	0.3270
0.203	4.4807	2200	0.3257
0.2074	4.6843	2300	0.3227
0.2185	4.8880	2400	0.3234
0.2117	5.0916	2500	0.3232
0.1913	5.2953	2600	0.3228
0.1921	5.4990	2700	0.3219
0.1888	5.7026	2800	0.3217
0.1925	5.9063	2900	0.3216
0.1955	6.1100	3000	0.3212

Framework versions

• PEFT 0.12.1.dev0
• Transformers 4.44.0
• Pytorch 2.4.0
• Datasets 2.21.0
• Tokenizers 0.19.1