t5-small-pointer-adv-mtop

This model is a fine-tuned version of google/mt5-small on the mtop dataset. It achieves the following results on the evaluation set:

• Loss: 0.1341
• Exact Match: 0.5817

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
2.1628	1.09	200	0.7205	0.0022
1.1208	2.17	400	0.6393	0.0013
0.8675	3.26	600	0.5905	0.0027
1.8729	4.35	800	0.5726	0.0031
3.5417	5.43	1000	0.5371	0.0067
0.9087	6.52	1200	0.3512	0.1119
1.2224	7.61	1400	0.2739	0.1911
0.7597	8.69	1600	0.2151	0.3016
0.6981	9.78	1800	0.1736	0.3749
0.4779	10.87	2000	0.1548	0.4166
0.4397	11.96	2200	0.1377	0.4510
0.4101	13.04	2400	0.1480	0.4197
0.3323	14.13	2600	0.1396	0.4398
0.2565	15.22	2800	0.1351	0.4523
0.2108	16.3	3000	0.1341	0.4541

Framework versions

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