zhpinkman/Advanced_NLP_course_HW2

This assignment is about using the infrastructure you built in HW1, now applying a non-linear classifier (specifically, a one-layer feed-forward network) to the problem. The key is that we want you to implement the model by hand, without using neural network or machine learning frameworks. Here is a nice post on why such exercise of implementing your own forward and backward passes is useful. After you do this, you should compare your implementation to a PyTorch implementation (that you should also write) and to the results you got in HW1.

Advanced_NLP_course_HW2

This assignment is about using the infrastructure you built in HW1, now applying a non-linear classifier (specifically, a one-layer feed-forward network) to the problem. The key is that we want you to implement the model by hand, without using neural network or machine learning frameworks. Here is a nice post on why such exercise of implementing your own forward and backward passes is useful. After you do this, you should compare your implementation to a PyTorch implementation (that you should also write) and to the results you got in HW1.

get the embedding files from link.

All the experiments, training and evaluation scripts are in pipeline.sh, torch_pipeline.sh.

There are a couple of notes that you have to consider before running the models

The embedding files should contain the UNK token also included as a separate token
Also if you wanted to run the models on a slurm based cluster you can use the scripts `cluster_pipeline.sh`, `cluster_torch_pipeline.sh`