speller-t5-9

This model is a fine-tuned version of sberbank-ai/ruT5-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1614
• Rouge1: 14.9554
• Rouge2: 8.3333
• Rougel: 14.9554
• Rougelsum: 14.9554
• Gen Len: 42.8661

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: 5e-05
• 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
• num_epochs: 1
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
1.0873	0.04	500	0.5259	13.7946	7.1429	13.8393	13.8393	40.7946
0.6932	0.07	1000	0.3914	14.0625	8.3333	14.0625	14.0625	43.5357
0.5471	0.11	1500	0.3349	13.9633	7.9507	13.8641	13.9633	45.0089
0.5566	0.14	2000	0.2954	14.0625	8.3333	14.0625	14.0625	43.1429
0.4985	0.18	2500	0.2802	14.9554	8.3333	14.9554	14.9554	44.125
0.5175	0.22	3000	0.2631	14.9554	8.3333	14.9554	14.9554	44.4286
0.4377	0.25	3500	0.2431	14.9554	8.3333	14.9554	14.9554	42.5893
0.4356	0.29	4000	0.2315	14.9554	8.3333	14.9554	14.9554	42.9286
0.4052	0.32	4500	0.2258	14.9554	8.3333	14.9554	14.9554	43.2232
0.3888	0.36	5000	0.2179	14.9554	8.3333	14.9554	14.9554	42.6607
0.3731	0.39	5500	0.2063	14.9554	8.3333	14.9554	14.9554	42.9196
0.436	0.43	6000	0.2075	14.9554	8.3333	14.9554	14.9554	42.7589
0.42	0.47	6500	0.1993	14.9554	8.3333	14.9554	14.9554	42.5446
0.378	0.5	7000	0.2036	14.9554	8.3333	14.9554	14.9554	43.0179
0.3431	0.54	7500	0.1914	14.9554	8.3333	14.9554	14.9554	42.6875
0.3574	0.57	8000	0.1852	14.9554	8.3333	14.9554	14.9554	42.7321
0.302	0.61	8500	0.1900	14.9554	8.3333	14.9554	14.9554	42.7946
0.3081	0.65	9000	0.1807	14.9554	8.3333	14.9554	14.9554	42.7054
0.3266	0.68	9500	0.1755	14.9554	8.3333	14.9554	14.9554	42.5714
0.3834	0.72	10000	0.1726	14.9554	8.3333	14.9554	14.9554	42.8482
0.2802	0.75	10500	0.1736	14.9554	8.3333	14.9554	14.9554	42.8036
0.3013	0.79	11000	0.1675	14.9554	8.3333	14.9554	14.9554	42.7054
0.3404	0.83	11500	0.1630	14.9554	8.3333	14.9554	14.9554	42.6786
0.2945	0.86	12000	0.1627	14.9554	8.3333	14.9554	14.9554	42.6607
0.2819	0.9	12500	0.1633	14.9554	8.3333	14.9554	14.9554	42.7321
0.3028	0.93	13000	0.1597	14.9554	8.3333	14.9554	14.9554	42.6429
0.3138	0.97	13500	0.1614	14.9554	8.3333	14.9554	14.9554	42.8661

Framework versions

• Transformers 4.26.0
• Pytorch 1.13.1+cu116
• Datasets 2.9.0
• Tokenizers 0.13.2