distilbert-base-uncased__subj__train-8-6

This model is a fine-tuned version of distilbert-base-uncased on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6075
• Accuracy: 0.7485

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: 4
• eval_batch_size: 4
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.7163	1.0	3	0.6923	0.5
0.6648	2.0	6	0.6838	0.5
0.6329	3.0	9	0.6747	0.75
0.5836	4.0	12	0.6693	0.5
0.5287	5.0	15	0.6670	0.25
0.4585	6.0	18	0.6517	0.5
0.415	7.0	21	0.6290	0.5
0.3353	8.0	24	0.6019	0.5
0.2841	9.0	27	0.5613	0.75
0.2203	10.0	30	0.5222	1.0
0.1743	11.0	33	0.4769	1.0
0.1444	12.0	36	0.4597	1.0
0.1079	13.0	39	0.4462	1.0
0.0891	14.0	42	0.4216	1.0
0.0704	15.0	45	0.3880	1.0
0.0505	16.0	48	0.3663	1.0
0.0428	17.0	51	0.3536	1.0
0.0356	18.0	54	0.3490	1.0
0.0283	19.0	57	0.3531	1.0
0.025	20.0	60	0.3595	1.0
0.0239	21.0	63	0.3594	1.0
0.0202	22.0	66	0.3521	1.0
0.0168	23.0	69	0.3475	1.0
0.0159	24.0	72	0.3458	1.0
0.0164	25.0	75	0.3409	1.0
0.0132	26.0	78	0.3360	1.0
0.0137	27.0	81	0.3302	1.0
0.0112	28.0	84	0.3235	1.0
0.0113	29.0	87	0.3178	1.0
0.0111	30.0	90	0.3159	1.0
0.0113	31.0	93	0.3108	1.0
0.0107	32.0	96	0.3101	1.0
0.0101	33.0	99	0.3100	1.0
0.0083	34.0	102	0.3110	1.0
0.0092	35.0	105	0.3117	1.0
0.0102	36.0	108	0.3104	1.0
0.0086	37.0	111	0.3086	1.0
0.0092	38.0	114	0.3047	1.0
0.0072	39.0	117	0.3024	1.0
0.0079	40.0	120	0.3014	1.0
0.0079	41.0	123	0.2983	1.0
0.0091	42.0	126	0.2948	1.0
0.0077	43.0	129	0.2915	1.0
0.0085	44.0	132	0.2890	1.0
0.009	45.0	135	0.2870	1.0
0.0073	46.0	138	0.2856	1.0
0.0073	47.0	141	0.2844	1.0
0.0076	48.0	144	0.2841	1.0
0.0065	49.0	147	0.2836	1.0
0.0081	50.0	150	0.2835	1.0

Framework versions

• Transformers 4.15.0
• Pytorch 1.10.2+cu102
• Datasets 1.18.2
• Tokenizers 0.10.3