Skip to content

Latest commit

 

History

History
63 lines (36 loc) · 4.11 KB

Queries.md

File metadata and controls

63 lines (36 loc) · 4.11 KB
Raw

Preprocessing --> ML1M, Yahoo, Pinterest Done

sbatch --job-name=LXRpreproc --mail-user=[email protected] --time=12:00:00 --mem=64G ~/jobs/single-node-gpu.job "python code/data_processing.py"

Recommender Training

sbatch --job-name=XXXX --mail-user=[email protected] --time=12:00:00 --mem=64G ~/jobs/single-node-gpu.job "python code/LXR_training.py --trial 0 --learning_rate 0.004 --lambda_neg XXX --lambda_pos XXX --alpha XXX"

sbatch --job-name=Pos --mail-user=[email protected] --time=12:00:00 --mem=64G ~/jobs/single-node-gpu.job "python code/LXR_training.py --trial 0 --learning_rate 0.004 --lambda_neg 1 --lambda_pos 35 --alpha 1"

sbatch --job-name=Neg --mail-user=[email protected] --time=12:00:00 --mem=64G ~/jobs/single-node-gpu.job "python code/LXR_training.py --trial 0 --learning_rate 0.004 --lambda_neg 35 --lambda_pos 1 --alpha 1"

sbatch --job-name=Equal --mail-user=[email protected] --time=12:00:00 --mem=64G ~/jobs/single-node-gpu.job "python code/LXR_training.py --trial 0 --learning_rate 0.004 --lambda_neg 35 --lambda_pos 35 --alpha 1"

sbatch --job-name=Alpha --mail-user=[email protected] --time=12:00:00 --mem=64G ~/jobs/single-node-gpu.job "python code/LXR_training.py --trial 0 --learning_rate 0.004 --lambda_neg 8.3 --lambda_pos 35 --alpha 20"

LXR Training

sbatch --job-name=LXR --mail-user=[email protected] --time=12:00:00 --mem=64G ~/jobs/single-node-gpu.job "python code/LXR_training.py --directory Sel2/VAE_ML1M_1_12_256.pt --whereToSave LXR2/12"

--> Directory: Recommender Checkpoint --> Ml1M/Recommenders/VAE_ML1M_1_12_256.pt --> whereToSave: Where to dump the LXR checkpoint --> ML1M/LXR/number --> model: VAE, MLP --> Data: ML1M. Pinterest, Yahoo

Metrics

python code/metrics.py --directoryRec Sel1/VAE_ML1M_2_0_64.pt --directoryLXR LXR1/0/LXR_ML1M_VAE_0_24_128_35_7.pt --SizeLXR 128

For xp size (best Rec & LXR) (Sparsity)

python code/metricsXpSize.py --directoryRec Sel1/VAE_ML1M_2_18_64.pt --directoryLXR LXR1/18/LXR_ML1M_VAE_1_24_128_35_7.pt --SizeLXR 128 --sizeXP 10

1 sbatch --job-name=LXR --mail-user=[email protected] --time=12:00:00 --mem=64G ~/jobs/single-node-gpu.job "python code/metricsXpSize.py --directoryRec Sel1/VAE_ML1M_2_18_64.pt --directoryLXR LXR1/18/LXR_ML1M_VAE_1_24_128_35_7.pt --SizeLXR 128 --sizeXP 1"

Pinterest Yahoo

sbatch --job-name=LXR --mail-user=[email protected] --time=12:00:00 --mem=64G ~/jobs/single-node-gpu.job "python code/metricsXpSize.py --dataset Pinterest --sizeXP 1"

For Rec TopK (Consistency)

python code/metricsTopK.py --directoryRec Sel1/VAE_ML1M_2_18_64.pt --directoryLXR LXR1/18/LXR_ML1M_VAE_1_24_128_35_7.pt --SizeLXR 128 --sizeXP 10 --lentopk 4

2 sbatch --job-name=LXR --mail-user=[email protected] --time=12:00:00 --mem=64G ~/jobs/single-node-gpu.job "python code/metricsTopK.py --directoryRec Sel2/VAE_ML1M_1_2_256.pt --directoryLXR LXR2/2/LXR_ML1M_VAE_0_39_128_35_7.pt --SizeLXR 128 --sizeXP 10 --lentopk 1"

Fur 7,1 ToDo 3

4

sbatch --job-name=LXR --mail-user=[email protected] --time=12:00:00 --mem=64G ~/jobs/single-node-gpu.job "python code/metricsTopK.py --directoryRec Sel1/VAE_ML1M_2_1_64.pt --directoryLXR LXR1/1/LXR_ML1M_VAE_1_24_128_35_7.pt --SizeLXR 128 --sizeXP 10 --lentopk 4"

commands

srun --nodes=1 --nodelist=gc7 --ntasks-per-node=1 --time=12:00:00 --pty bash -i source ../CFX-REC/cfx-recVenv/bin/activate