Seq2Seq with Attention + BERT - Prediction Problem
Budget: $30 – $250 USD
There are 4 columns in our dataset. These are *'sequence1'* *'sequence_2'* *'sequence1_index'*, and *'sequence2_index'*. My project is about bioinformatic. The data used is protein sequences.
I have 10000 rows. Input shape (10000, 1024). Output shape is going to be (10000, 20, 18). So each prediction is a 20x18 matrix.
Preparing training and the testing dataset were done by me. I developed the Encoder-Decoder deep learning model, but the model couldn't learn it.
Things to do in the project:
-Encoder-Decoder model
-Encoder-Decoder with Attention Mechanism
-BERT model
Prediction needs to be made using the above models. It needs to run with high accuracy. Pytorch or Tensorflow can be used. I prefer Pytorch.
NOTE IMPORTANT: IF YOU DON'T UNDERSTAND THE CODES BELOW OR DO NOT HAVE EXPERIENCE, PLEASE DO NOT BID ON THE PROJECT. DON'T WASTE YOUR TIME AND MINE.
1) https://github.com/Kalwing/BahdanauAttention/blob/master/models.py
2) https://github.com/wanfeather/transformer/blob/56ca3907524580db21e2e213d3232ea17d906480/model.py
I have 10000 rows. Input shape (10000, 1024). Output shape is going to be (10000, 20, 18). So each prediction is a 20x18 matrix.
Preparing training and the testing dataset were done by me. I developed the Encoder-Decoder deep learning model, but the model couldn't learn it.
Things to do in the project:
-Encoder-Decoder model
-Encoder-Decoder with Attention Mechanism
-BERT model
Prediction needs to be made using the above models. It needs to run with high accuracy. Pytorch or Tensorflow can be used. I prefer Pytorch.
NOTE IMPORTANT: IF YOU DON'T UNDERSTAND THE CODES BELOW OR DO NOT HAVE EXPERIENCE, PLEASE DO NOT BID ON THE PROJECT. DON'T WASTE YOUR TIME AND MINE.
1) https://github.com/Kalwing/BahdanauAttention/blob/master/models.py
2) https://github.com/wanfeather/transformer/blob/56ca3907524580db21e2e213d3232ea17d906480/model.py