This model is a fine-tuned version of bert-base-uncased on the squad dataset. It achieves the following results on the evaluation set:
Most base model weights were frozen leaving only to finetune the last layer (qa outputs) and 3 last layers of the encoder.
Achieved EM: 76.77388836329234, F1: 85.41893520501723
The following hyperparameters were used during training:
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 1.2944 | 1.0 | 44262 | 1.3432 |
| 1.0152 | 2.0 | 88524 | 1.3450 |
| 1.0062 | 3.0 | 132786 | 1.4571 |
pytorch_model.bin 436 MBtraining_args.bin 3 kB