byt5-small-wikipron-eng-latn-multi-broad-p2g

This model is a fine-tuned version of google/byt5-small on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1238
• Per: 0.2052
• Gen Len: 8.4891

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: 0.0002
• train_batch_size: 128
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20.0

Training results

Training Loss	Epoch	Step	Validation Loss	Per	Gen Len
2.0082	1.0	1177	0.4061	0.6392	8.2917
0.4295	2.0	2354	0.2953	0.5242	8.3425
0.3179	3.0	3531	0.2338	0.4552	8.4024
0.255	4.0	4708	0.2011	0.4038	8.4287
0.2131	5.0	5885	0.1753	0.3669	8.4356
0.1813	6.0	7062	0.1567	0.3341	8.4336
0.157	7.0	8239	0.1459	0.3098	8.4546
0.1368	8.0	9416	0.1349	0.2859	8.4531
0.1202	9.0	10593	0.1302	0.2663	8.4621
0.1067	10.0	11770	0.1240	0.2514	8.4701
0.0946	11.0	12947	0.1203	0.2415	8.4734
0.0857	12.0	14124	0.1180	0.2347	8.4782
0.0779	13.0	15301	0.1187	0.226	8.4827
0.0709	14.0	16478	0.1180	0.2211	8.4781
0.0646	15.0	17655	0.1176	0.2147	8.4856
0.0602	16.0	18832	0.1178	0.2129	8.4858
0.0563	17.0	20009	0.1200	0.2113	8.4844
0.0532	18.0	21186	0.1218	0.2069	8.4907
0.0501	19.0	22363	0.1228	0.2057	8.4891
0.0486	20.0	23540	0.1238	0.2052	8.4891

Framework versions

• Transformers 4.28.1
• Pytorch 2.0.0+cu117
• Datasets 2.11.1.dev0
• Tokenizers 0.13.2