|
|
--- |
|
|
license: afl-3.0 |
|
|
datasets: |
|
|
- WillHeld/hinglish_top |
|
|
language: |
|
|
- en |
|
|
- hi |
|
|
metrics: |
|
|
- accuracy |
|
|
library_name: transformers |
|
|
pipeline_tag: fill-mask |
|
|
--- |
|
|
|
|
|
### SRDberta |
|
|
|
|
|
This is a BERT model trained for Masked Language Modeling for Hinglish Data. |
|
|
|
|
|
Hinglish is a term used to describe the hybrid language spoken in India, which combines elements of Hindi and English. It is commonly used in informal conversations and in media such as Bollywood films |
|
|
|
|
|
### Dataset |
|
|
Hinglish-Top [Dataset](https://huggingface.co/datasets/WillHeld/hinglish_top) columns |
|
|
- en_query |
|
|
- cs_query |
|
|
- en_parse |
|
|
- cs_parse |
|
|
- domain |
|
|
|
|
|
### Training |
|
|
|Epoch|Loss| |
|
|
|:--:|:--:| |
|
|
|1 |0.0485| |
|
|
|2 |0.00837| |
|
|
|3 |0.00812| |
|
|
|4 |0.0029| |
|
|
|5 |0.014| |
|
|
|6 |0.00748| |
|
|
|7 |0.0041| |
|
|
|8 |0.00543| |
|
|
|9 |0.00304| |
|
|
|10 |0.000574| |
|
|
|
|
|
### Inference |
|
|
```python |
|
|
from transformers import AutoTokenizer, AutoModelForMaskedLM, pipeline |
|
|
|
|
|
tokenizer = AutoTokenizer.from_pretrained("SRDdev/SRDBerta") |
|
|
|
|
|
model = AutoModelForMaskedLM.from_pretrained("SRDdev/SRDBerta") |
|
|
|
|
|
fill = pipeline('fill-mask', model='SRDberta', tokenizer='SRDberta') |
|
|
``` |
|
|
```python |
|
|
fill_mask = fill.tokenizer.mask_token |
|
|
fill(f'Aap {fill_mask} ho?') |
|
|
``` |
|
|
|
|
|
### Citation |
|
|
Author: @[SRDdev](https://huggingface.co/SRDdev) |
|
|
``` |
|
|
Name : Shreyas Dixit |
|
|
framework : Pytorch |
|
|
Year: Jan 2023 |
|
|
Pipeline : fill-mask |
|
|
Github : https://github.com/SRDdev |
|
|
LinkedIn : https://www.linkedin.com/in/srddev/ |
|
|
``` |
|
|
|
