README.md

Instructions for the paper

---

1. 😊 init

   • mkdir best_model predicts

   • mkdir ./data/ED/comet

   • download comet checkpoint to comet directory.

   • wget http://nlp.stanford.edu/data/glove.6B.zip to ./vectors .

---

result:

---

2. 🔥 code params explanation

params	instruction
model	model type available contains trans,mult,empdg,mime,moel,kemp,cem,emf
code_check	store_strue for fast check program runnable in ur machine
devices	value passed to os['CUDA_VISIBLE_DEVICE']
mode	train_only,train_and_test,test_only indicates run partial or whole process
max_epoch	max epochs to train model
emotion_emb_type	origin,coarse,contrastive indicates different emotion embedings,details see paper
batch_size	batch size for train,valid,test

3. 🐶 run experiment

   • nohup python train.py --model trans --mode train_and_test --batch_size 32 --max_epoch 128 --devices 0 >trans.log& trans training
   • nohup python train.py --model mult --mode train_and_test --batch_size 32 --max_epoch 128 --devices 1 >mult.log& mult training
   • nohup python train.py --model empdg --mode train_and_test --batch_size 32 --max_epoch 128 --devices 2 >empdg.log& empdg training
   • nohup python train.py --model mime --mode train_and_test --batch_size 32 --max_epoch 128 --devices 3 >mime.log& mime training
   • nohup python train.py --model moel --mode train_and_test --batch_size 32 --max_epoch 128 --devices 4 >moel.log& moel training
   • nohup python train.py --model cem --mode train_and_test --batch_size 32 --max_epoch 128 --devices 5 >cem.log& cem training
   • nohup python train.py --model kemp --mode train_and_test --batch_size 32 --max_epoch 128 --devices 6 >kemp.log& kemp training

4. 🔎 todo list

   • run all models
   • add more params to control
   • run on other datasets