t5-base-vanilla-mtop

This model is a fine-tuned version of google/mt5-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2080
• Exact Match: 0.6394

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: 0.001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 64
• total_train_batch_size: 512
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• training_steps: 3000

Training results

Training Loss	Epoch	Step	Validation Loss	Exact Match
1.0516	6.65	200	0.1173	0.5875
0.0541	13.33	400	0.1130	0.6331
0.0468	19.98	600	0.1290	0.6036
0.0241	26.65	800	0.1306	0.6273
0.0125	33.33	1000	0.1425	0.6291
0.0077	39.98	1200	0.1518	0.6345
0.0054	46.65	1400	0.1643	0.6362
0.004	53.33	1600	0.1718	0.6362
0.0033	59.98	1800	0.1803	0.6336
0.0026	66.65	2000	0.1808	0.6394
0.0021	73.33	2200	0.1915	0.6371
0.0017	79.98	2400	0.1919	0.6403
0.0013	86.65	2600	0.2024	0.6358
0.0011	93.33	2800	0.2049	0.6353
0.0008	99.98	3000	0.2080	0.6394

Framework versions

• Transformers 4.24.0
• Pytorch 1.13.0+cu117
• Datasets 2.7.0
• Tokenizers 0.13.2