Downloads · 30 days
66
9% of all-time downloads
bert-base-uncased-pretrained-mlm-coqa-stories
alistvt/bert-base-uncased-pretrained-mlm-coqa-stories
bert-base-uncased-pretrained-mlm-coqa-stories is a fill-mask model from alistvt. Use it when you need the model to fill a missing word. It is set up for transformers.
should probably proofread and complete it, then remove this comment. --
All-time downloads
776
Public
Repo size
1.3 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.bin438 MB · 100%
At a glance
- Task
- Fill-Mask
- Library
- transformers
- Model type
- bert
- Access
- Public
- Created
- Mar 2, 2022
- Updated
- Jan 21, 2022
- SHA
- 9168ae7f
Try a prompt
Model card
From the Hugging Face model README
<!-- 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. -->
bert-base-uncased-pretrained-mlm-coqa-stories
This model was trained from scratch on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.8310
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: 2e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 3.0
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 2.0573 | 1.0 | 2479 | 1.8805 |
| 1.9517 | 2.0 | 4958 | 1.8377 |
| 1.9048 | 3.0 | 7437 | 1.8310 |
Framework versions
- Transformers 4.15.0
- Pytorch 1.10.0+cu111
- Datasets 1.17.0
- Tokenizers 0.10.3
- Task
- Fill-Mask
- Library
- transformers
- Type
- bert
- Created
- Mar 2, 2022
- Updated
- Jan 21, 2022