Downloads · 30 days
4
8% of all-time downloads
dariurs/eduro-bert-romanian-fillblank-v3
eduro-bert-romanian-fillblank-v3 is a fill-mask model from dariurs. Use it when you need the model to fill a missing word. The card lists the license as cc-by-sa-4.0.
Fine-tuned Romanian BERT for academic fill-in-the-blank question generation, trained on the EduRO-FillBlank dataset (72,541 examples from Romanian Wikipedia).
Downloads · 30 days
4
8% of all-time downloads
All-time downloads
48
Public
Parameters
124M
498 MB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors498 MB · 100%
From the Hugging Face model README
Fine-tuned Romanian BERT for academic fill-in-the-blank question generation, trained on the EduRO-FillBlank dataset (72,541 examples from Romanian Wikipedia).
| Parameter | Value |
|---|---|
| Base model | dumitrescustefan/bert-base-romanian-cased-v1 |
| Dataset | EduRO-FillBlank v2 (72,541 examples) |
| Epochs | 10 |
| Learning Rate | 2e-5 |
| Batch Size | 8 |
| MLM Probability | 15% |
| Hardware | NVIDIA A100 |
| Metric | Score |
|---|---|
| Acc@1 | 34.82% |
| Acc@5 | 56.65% |
| MRR | 0.4482 |
| Metric | Score |
|---|---|
| Acc@1 | 32.5% |
| Acc@5 | 55.2% |
| MRR | 0.410 |
from transformers import pipeline
pipe = pipeline("fill-mask",
model="dariurs/eduro-bert-romanian-fillblank-v3")
result = pipe("Fotosinteza este procesul [MASK] prin care plantele produc energie.")
print(result)
[to be added after paper publication]