vakyansh-wav2vec2-malayalam-mlm-8-audio-abuse-feature

This model is a fine-tuned version of Harveenchadha/vakyansh-wav2vec2-malayalam-mlm-8 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5558
• Accuracy: 0.8118
• Macro F1-score: 0.7576

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: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Macro F1-score
6.7625	0.77	10	6.7667	0.0	0.0
6.6935	1.54	20	6.5968	0.2527	0.0191
6.5445	2.31	30	6.3912	0.6909	0.4086
6.338	3.08	40	6.0995	0.6909	0.4086
6.1044	3.85	50	5.7554	0.6909	0.4086
5.755	4.62	60	5.4099	0.6909	0.4086
5.4695	5.38	70	5.1276	0.6909	0.4086
5.1723	6.15	80	4.8795	0.6909	0.4086
4.9648	6.92	90	4.6629	0.6909	0.4086
4.7414	7.69	100	4.4481	0.6909	0.4086
4.5793	8.46	110	4.2398	0.6909	0.4086
4.4455	9.23	120	4.0339	0.6909	0.4086
4.2287	10.0	130	3.8247	0.6909	0.4086
3.9367	10.77	140	3.6164	0.6909	0.4086
3.7916	11.54	150	3.4090	0.6909	0.4086
3.6112	12.31	160	3.2043	0.6909	0.4086
3.408	13.08	170	3.0023	0.6909	0.4086
3.1359	13.85	180	2.8029	0.6909	0.4086
2.9607	14.62	190	2.6125	0.6909	0.4086
2.83	15.38	200	2.4336	0.6909	0.4086
2.4853	16.15	210	2.2649	0.6909	0.4086
2.3841	16.92	220	2.1059	0.6909	0.4086
2.2296	17.69	230	1.9583	0.6909	0.4086
1.9631	18.46	240	1.8302	0.6909	0.4086
2.0456	19.23	250	1.7146	0.6909	0.4086
1.8406	20.0	260	1.6100	0.6909	0.4086
1.7127	20.77	270	1.5130	0.6909	0.4086
1.5241	21.54	280	1.4264	0.6909	0.4086
1.4366	22.31	290	1.3458	0.6909	0.4086
1.4368	23.08	300	1.2710	0.6909	0.4086
1.2664	23.85	310	1.2024	0.6909	0.4086
1.2681	24.62	320	1.1391	0.6909	0.4086
1.1518	25.38	330	1.0791	0.6909	0.4086
1.0681	26.15	340	1.0221	0.6909	0.4086
1.014	26.92	350	0.9679	0.6909	0.4086
0.9918	27.69	360	0.9197	0.6909	0.4086
1.0046	28.46	370	0.8839	0.6909	0.4086
0.9582	29.23	380	0.8422	0.6909	0.4086
0.927	30.0	390	0.8017	0.6909	0.4086
0.8853	30.77	400	0.7666	0.6909	0.4086
0.7872	31.54	410	0.7353	0.6909	0.4086
0.7773	32.31	420	0.7032	0.6909	0.4086
0.7163	33.08	430	0.6929	0.6909	0.4086
0.7686	33.85	440	0.6617	0.6909	0.4086
0.7504	34.62	450	0.6623	0.6909	0.4086
0.7491	35.38	460	0.6333	0.6909	0.4086
0.6688	36.15	470	0.6115	0.6962	0.4348
0.6785	36.92	480	0.5968	0.6909	0.4086
0.6511	37.69	490	0.5879	0.6909	0.4086
0.5906	38.46	500	0.5855	0.8253	0.7679
0.6	39.23	510	0.5837	0.8065	0.7299
0.604	40.0	520	0.5683	0.8226	0.7699
0.6269	40.77	530	0.5697	0.8065	0.7362
0.5643	41.54	540	0.5628	0.8199	0.7687
0.6269	42.31	550	0.5650	0.8145	0.7570
0.5965	43.08	560	0.5666	0.8065	0.7473
0.5578	43.85	570	0.5683	0.8065	0.7401
0.5571	44.62	580	0.5607	0.8172	0.7690
0.5511	45.38	590	0.5566	0.8145	0.7618
0.5404	46.15	600	0.5587	0.8091	0.7482
0.5708	46.92	610	0.5541	0.8172	0.7660
0.62	47.69	620	0.5524	0.8145	0.7618
0.6095	48.46	630	0.5573	0.8065	0.7438
0.5282	49.23	640	0.5559	0.8145	0.7586
0.5307	50.0	650	0.5558	0.8118	0.7576

Framework versions

• Transformers 4.33.0
• Pytorch 2.0.0
• Datasets 2.1.0
• Tokenizers 0.13.3