Edit model card

t5-small-pointer-top_v2

This model is a fine-tuned version of google/mt5-small on the top_v2 dataset. It achieves the following results on the evaluation set:

  • Loss: 0.0306
  • Exact Match: 0.8264

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
1.9316 0.82 200 0.4566 0.0084
0.3713 1.65 400 0.1473 0.1230
0.1747 2.47 600 0.0788 0.1984
0.1104 3.29 800 0.0568 0.2149
0.0842 4.12 1000 0.0473 0.2217
0.0694 4.94 1200 0.0426 0.2260
0.0603 5.76 1400 0.0383 0.2279
0.0534 6.58 1600 0.0367 0.2281
0.0477 7.41 1800 0.0347 0.2301
0.0441 8.23 2000 0.0334 0.2314
0.0413 9.05 2200 0.0323 0.2315
0.0387 9.88 2400 0.0316 0.2316
0.0366 10.7 2600 0.0311 0.2324
0.0358 11.52 2800 0.0307 0.2324
0.0343 12.35 3000 0.0306 0.2327

Framework versions

  • Transformers 4.25.1
  • Pytorch 1.13.0+cu117
  • Datasets 2.7.1
  • Tokenizers 0.13.2
Downloads last month
1
Hosted inference API
This model can be loaded on the Inference API on-demand.