husnu's picture
Update README.md
4dce54e
|
raw
history blame
1.34 kB
---
tags:
- generated_from_trainer
datasets:
- squad
model-index:
- name: ft_electra-small-turkish-uncased-discriminator_lr-2e-1_epochs-1
results: []
---
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
This model is a fine-tuned version of [loodos/electra-small-turkish-uncased-discriminator](https://huggingface.co/loodos/electra-small-turkish-uncased-discriminator) on the squad dataset.
It achieves the following results on the evaluation set:
- Loss: 5.9506
## 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: 0.2
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 1
### Training results
| Training Loss | Epoch | Step | Validation Loss |
|:-------------:|:-----:|:----:|:---------------:|
| 5.951 | 1.0 | 5818 | 5.9506 |
### Framework versions
- Transformers 4.15.0
- Pytorch 1.10.0+cu111
- Datasets 1.17.0
- Tokenizers 0.10.3