search

QueuQ / CGLB

Other
49 stars 14 forks source link

ValueError from GEM on Tox21 #22

Open ghost opened 1 year ago

ghost commented 1 year ago

Thanks for your effort to keep this benchmark better!

I have continuously kept facing errors from GEM on Tox21 in GCGL as below:

.... EarlyStopping counter: 41 out of 10

error constraints are inconsistent, no solution Traceback (most recent call last): File "train.py", line 118, in AP, AF, acc_matrix = main(args,valid=True) File "/home/python/CGLB/GCGL/pipeline_org.py", line 375, in pipeline_multi_label life_model_ins.observe(train_loader, loss_criterion, tid, args) File "/home/python/CGLB/GCGL/Baselines/gem_model.py", line 122, in observe self.grads.index_select(1, indx), self.margin) File "/home/python/CGLB/GCGL/Baselines/gem_utils.py", line 66, in project2cone2 v = quadprog.solve_qp(P, q, G, h)[0] File "quadprog/quadprog.pyx", line 102, in quadprog.solve_qp ValueError: constraints are inconsistent, no solution

Since the error occured during quadratic programming, the error might not be due to the original code. However, I've never got the result but only error while I implementing GEM on Tox21. Have you ever faced the same or similar case of error?

ghost commented 1 year ago

Furthermore, another error by GEM on SIDER dataset occurred as below:

best params is {memory_strength_0.5_n_memories_100}, best AP is 0.6157 ----------Now in testing-------- Processing dgl graphs from scratch... Processing molecule 1000/1427

****[0, [0]] Warning: Only one class 0.0 present in y_true for a task. ROC AUC score is not defined in that case. T00 0.9821|T01 0.6163|T02 0.4373|T03 0.5536|T04 0.3704|T05 0.5103|T06 0.3935|T07 0.4642|T08 0.6771|T09 0.4217|T10 0.4634|T11 0.4933|T12 0.4384|T13 0.5725|T14 0.5148|T15 0.6018|T16 0.3849|T17 0.6000|T18 0.3954|T19 0.5164|T20 0.4329|T21 0.6351|T22 0.5980|T23 0.3839|T24 0.4687|T25 0.4883|Traceback (most recent call last): File "train.py", line 156, in AP_test, AF_test, acc_matrix_test = main(args, valid=False) File "/home/python/CGLB/GCGL/pipeline_org.py", line 389, in pipeline_multi_label score_matrix[tid] = test_func(args, model, test_loader, tid) File "/home/python/CGLB/GCGL/pipeline_org.py", line 159, in eval_all_learnt_task_multi_label score = test_score[t] IndexError: list index out of range