运行DGBO_GC-batch

[wkybucm]

崔老师您好，我在学习您的DGBO_GC方法，我采用的overfit_type是默认的opt。目前程序运行初始化参数，代码如下： def initialization(data,n,info): print ("-----------initialization-----------") xlist=np.linspace(0,len(data.candidates)-1,len(data.candidates)) xlist = xlist.tolist() xlistinit=[] if len(data.candidates)%n==0: delta=len(data.candidates)/n else: delta=len(data.candidates)/n+1 for i in range(n): xlistinit.append(random.sample(xlist[iint(delta):min((i+1)int(delta),len(data.candidates))],1)[0]) xlist=xlistinit xlist=np.array(xlist,dtype=np.integer) for i in range(n): print ("initialization [%s] ..."%(i)) y=evaluate(info,data.candidates[xlist[i]]) info.add(xlist[i],y) if info.overfittype=="opt": print ("init the setting of params on overfitting...") bo = BayesianOptimization(initsettingCV,{'dropout': (0,1), 'weightdecay': (1e-5,1e-1)},random_state=rng) bo.maximize(init_points=21, niter=30, acq='ei') print (bo.res['max']) 在上述代码中，运行到 bo = BayesianOptimization(initsettingCV,{'dropout': (0,1), 'weightdecay': (1e-5,1e-1)},random_state=rng)后进入了5折交叉验证，但是一直循环交叉验证的过程，结果如下： | 2 | -1.044 | 0.9985 | 0.01969 | [cv 1] [ 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19] [0 1 2 3] add a model [cv 2] [ 0 1 2 3 8 9 10 11 12 13 14 15 16 17 18 19] [4 5 6 7] add a model [cv 3] [ 0 1 2 3 4 5 6 7 12 13 14 15 16 17 18 19] [ 8 9 10 11] add a model [cv 4] [ 0 1 2 3 4 5 6 7 8 9 10 11 16 17 18 19] [12 13 14 15] add a model [cv 5] [ 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15] [16 17 18 19] add a model transition_matrix: (16, 257) train: features_comb shape is : (257, 68) train: constructing adj... Build DGCN_Comb2... Architecture: { gcn[1 2 3 4 5]->concat(0.8)->dnn[1 2 3 4 5] } heur: dropout=0.16990029785933825, weight_decay=0.047577751206738225 Start to train network. Epoch: 0001 train_loss= 45.74941 time= 1.25621s total_time= 1.25621s Epoch: 0101 train_loss= 33.50009 time= 0.00600s total_time= 1.88035s Train net Finished! acc= 1.3567712 transition_matrix: (16, 257) train: features_comb shape is : (257, 68) train: constructing adj... Build DGCN_Comb2... Architecture: { gcn[1 2 3 4 5]->concat(0.8)->dnn[1 2 3 4 5] } heur: dropout=0.16990029785933825, weight_decay=0.047577751206738225 Start to train network. Epoch: 0001 train_loss= 53.56653 time= 1.30929s total_time= 1.30929s Epoch: 0101 train_loss= 33.76947 time= 0.00700s total_time= 1.95644s Train net Finished! acc= 1.1174575 transition_matrix: (16, 257) train: features_comb shape is : (257, 68) train: constructing adj... Build DGCN_Comb2... Architecture: { gcn[1 2 3 4 5]->concat(0.8)->dnn[1 2 3 4 5] } heur: dropout=0.16990029785933825, weight_decay=0.047577751206738225 Start to train network. Epoch: 0001 train_loss= 51.28548 time= 1.23025s total_time= 1.23025s Epoch: 0101 train_loss= 33.82351 time= 0.00700s total_time= 1.86339s Train net Finished! acc= 0.5193358 transition_matrix: (16, 257) train: features_comb shape is : (257, 68) train: constructing adj... Build DGCN_Comb2... Architecture: { gcn[1 2 3 4 5]->concat(0.8)->dnn[1 2 3 4 5] } heur: dropout=0.16990029785933825, weight_decay=0.047577751206738225 Start to train network. Epoch: 0001 train_loss= 45.57104 time= 1.31298s total_time= 1.31298s Epoch: 0101 train_loss= 32.98197 time= 0.00700s total_time= 1.97312s Train net Finished! acc= 1.7546887 transition_matrix: (16, 257) train: features_comb shape is : (257, 68) train: constructing adj... Build DGCN_Comb2... Architecture: { gcn[1 2 3 4 5]->concat(0.8)->dnn[1 2 3 4 5] } heur: dropout=0.16990029785933825, weight_decay=0.047577751206738225 Start to train network. Epoch: 0001 train_loss= 47.50560 time= 1.25738s total_time= 1.25738s Epoch: 0101 train_loss= 33.11481 time= 0.00600s total_time= 1.89652s Train net Finished! acc= 1.9915786 上述过程一直循环，无法运行到 bo.maximize(init_points=21, n_iter=30, acq='ei')这一步。不知道问题出现在哪里。 另外发现当不debug，直接运行整个项目，当迭代次数为51之后会跳到bo.maximize(init_points=21, n_iter=30, acq='ei')这一步，但是会报错。错误如下：Traceback (most recent call last): File "E:/model/Scalable-and-Parallel-DGBO-master/Scalable-and-Parallel-DGBO-master/DGBO_GC-batch/DGBO.py", line 338, in BO_process(data,params,info) File "E:/model/Scalable-and-Parallel-DGBO-master/Scalable-and-Parallel-DGBO-master/DGBO_GC-batch/DGBO.py", line 210, in BO_process initialization(data,params.initn,info) File "E:/model/Scalable-and-Parallel-DGBO-master/Scalable-and-Parallel-DGBO-master/DGBO_GC-batch/DGBO.py", line 135, in initialization print (bo.res['max']) TypeError: list indices must be integers or slices, not str。 这个问题困扰我很久了，一直不知道原因，希望您百忙之中可以帮忙给解答一下。我将感激不尽，谢谢您~

[csjtx1021]

Hi，

Can you confirm that the installed bayesian-optimization package is version 1.0.0?

Best