SQUAD 2.0 upgrade - Deep Learning (NLP)
Budget: $750 – $1,500 USD
Currently we have a model based on open-source: https://github.com/kamalkraj/ALBERT-TF2.0
We need your help to upgrade it:
1. using TF 2.0 DeBerta V2 xLarge/xxLarge, instead of Albert TF 2.0 model
(https://huggingface.co/docs/transformers/v4.16.2/en/model_doc/deberta-v2#transformers.TFDebertaV2Model) (or other published TF Deberta model)
we expect SQUAD 2.0 accuracy to be very close to published Deberta data (xxLarge: 92.2/89.7 F1/EM)
2. If successful, export the above TF model for serving, including data pipeline layer.
We will only pay for result (very close to target), not just effort/time, please only apply if you have prior SQUAD 2.0 experience.
We will pay for any additional training cost incurred, or we can provide you GPU server (48GB) free.
We are open for alternative solutions/suggestions if you might have.
We need your help to upgrade it:
1. using TF 2.0 DeBerta V2 xLarge/xxLarge, instead of Albert TF 2.0 model
(https://huggingface.co/docs/transformers/v4.16.2/en/model_doc/deberta-v2#transformers.TFDebertaV2Model) (or other published TF Deberta model)
we expect SQUAD 2.0 accuracy to be very close to published Deberta data (xxLarge: 92.2/89.7 F1/EM)
2. If successful, export the above TF model for serving, including data pipeline layer.
We will only pay for result (very close to target), not just effort/time, please only apply if you have prior SQUAD 2.0 experience.
We will pay for any additional training cost incurred, or we can provide you GPU server (48GB) free.
We are open for alternative solutions/suggestions if you might have.