You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi, as written in the README, these values are MSE and you will have to take the square root.
Also, as I note in the README, you should consult the tables of optimal hyperparameters in the paper to achieve the best performance. For example, for this dataset, you should do the pretraining for max. 700 epochs with a batch size of 128, not 32. I have now added this value for this dataset explicitly in the README.
Finally, there is definitely variance when running experiments, and thus you would generally have to run several iterations, but in expectation get something like MSE = 2870.
Dear Author,
I am running your commands and find that the pretraining process seems good while the finetuning is weird. The pretraining loss is just 0.140160306.
The commands I run are
CUDA_VISIBLE_DEVICES=4 python src/main.py --output_dir experiments --comment "pretraining through imputation" --name BeijingPM25Quality_pretrained --records_file Imputation_records.xls --data_dir BeijingPM25Quality --data_class tsra --pattern TRAIN --val_ratio 0.2 --epochs 700 --lr 0.001 --optimizer RAdam --batch_size 32 --pos_encoding learnable --d_model 128
CUDA_VISIBLE_DEVICES=1 python src/main.py --output_dir experiments --comment "finetune for regression" --name BeijingPM25Quality_finetuned --records_file Regression_records.xls --data_dir BeijingPM25Quality --data_class tsra --pattern TRAIN --val_pattern TEST --epochs 200 --lr 0.001 --optimizer RAdam --pos_encoding learnable --d_model 128 --load_model /home/xzhoubi/paperreading/mvts_transformer/experiments/BeijingPM25Quality_pretrained_2022-07-19_10-27-28_tlB/checkpoints/model_best.pth --task regression --change_output --batch_size 128
Can you please help check it?
The text was updated successfully, but these errors were encountered: