============ Initialized logger ============
add_traspose_rels: True
aggregate: pna
aug: False
batch_size: 12
constrained_neg_prob: 0.0
dataset: fb237_v3
dropout: 0.1
early_stop: 50
emb_dim: 32
enclosing_sub_graph: False
exp_dir: GNN/experiments/fb237_v3_mul
experiment_name: fb237_v3_mul
gpu: 5
hop: 3
l2: 0.0001
load_model: False
loss: bce
lr: 0.0005
main_dir: GNN
margin: 10
max_links: 1000000
message: mul
num_epochs: 5
num_gcn_layers: 6
num_neg_samples_per_link: 1
num_workers: 8
optimizer: Adam
residual: False
train_file: train
un_hop: 1
update: lstm
using_jk: False
valid_file: valid
============================================
No existing model found. Initializing new model..
Device: cuda
Input dim : 32, # Relations : 430
Total number of parameters: 205377
Starting training ...
====================================================================================================
Run Time Error! Reduce batch size. Current batch size: 8
No existing model found. Initializing new model..
Device: cuda
Input dim : 32, # Relations : 430
Total number of parameters: 205377
Starting training ...
====================================================================================================
Epoch 1 Training Performance:{'auc': 0.8341772556732012, 'auc_pr': 0.8189901406006872, 'acc': 0.7411319915489826} in 8316.186446666718 s 
Epoch 1 Validation Performance:{'auc': 0.898320001761662, 'auc_pr': 0.867934879602555, 'acc': 0.8220145852324522} in 554.3208560943604 s 
Epoch 1 Better models found w.r.t AUC-PR. Saved it!
Epoch 1 with loss: 0.5039523577407871, best validation AUC-PR: 0.867934879602555, weight_norm: 9.295869827270508 in 8870.551411628723 s 
====================================================================================================
Epoch 2 Training Performance:{'auc': 0.9231145255878063, 'auc_pr': 0.9068363376703447, 'acc': 0.8474924941621261} in 8256.728440523148 s 
Epoch 2 Validation Performance:{'auc': 0.922918766603873, 'auc_pr': 0.8963288446053442, 'acc': 0.8534639927073838} in 556.932502746582 s 
Epoch 2 Better models found w.r.t AUC-PR. Saved it!
Epoch 2 with loss: 0.3478653593947088, best validation AUC-PR: 0.8963288446053442, weight_norm: 7.6261677742004395 in 8813.70314359665 s 
====================================================================================================
Epoch 3 Training Performance:{'auc': 0.9462924781965597, 'auc_pr': 0.9319327349089601, 'acc': 0.8834093183587235} in 8205.199630022049 s 
Epoch 3 Validation Performance:{'auc': 0.9405255195864414, 'auc_pr': 0.924849718389984, 'acc': 0.8830902461257977} in 576.0604300498962 s 
Epoch 3 Better models found w.r.t AUC-PR. Saved it!
Epoch 3 with loss: 0.28412728416685823, best validation AUC-PR: 0.924849718389984, weight_norm: 6.748859405517578 in 8781.30405831337 s 
====================================================================================================
Epoch 4 Training Performance:{'auc': 0.9571967996186266, 'auc_pr': 0.9449507962209502, 'acc': 0.8990603802957856} in 8412.917364358902 s 
Epoch 4 Validation Performance:{'auc': 0.9495308743743814, 'auc_pr': 0.9395206197343771, 'acc': 0.8887876025524157} in 568.0880351066589 s 
Epoch 4 Better models found w.r.t AUC-PR. Saved it!
Epoch 4 with loss: 0.247626260508526, best validation AUC-PR: 0.9395206197343771, weight_norm: 6.2120161056518555 in 8981.051851272583 s 
====================================================================================================
Epoch 5 Training Performance:{'auc': 0.9651765070523425, 'auc_pr': 0.9553230280985872, 'acc': 0.9123206938730123} in 8061.344767808914 s 
Epoch 5 Validation Performance:{'auc': 0.9504696657578595, 'auc_pr': 0.9371106477013683, 'acc': 0.8974475843208751} in 538.4471004009247 s 
Epoch 5 with loss: 0.22080600044697113, best validation AUC-PR: 0.9395206197343771, weight_norm: 5.826440811157227 in 8599.821657419205 s 
====================================================================================================
