|
17 | 17 | kernel = CategoricalHyperparam(['model','shallow_model','kernel'],["rbf","sigmoid","linear","poly"]) |
18 | 18 | xg_learning_rate = ContinuousHyperparam(['model','shallow_model','learning_rate'],0,1) |
19 | 19 | scale_pos_weight = CategoricalHyperparam(['model','shallow_model','scale_pos_weight'],[1,10.0,100.0]) |
20 | | -num_samples = CategoricalHyperparam(['model','shallow_model','num_samples'],[10000,100000,1000000,1e10]) |
| 20 | +num_samples = CategoricalHyperparam(['model','shallow_model','num_samples'],[10000,100000,1000000,1e7]) |
21 | 21 | tunables = [shallow_model,n_estimators,max_depth,C,kernel,xg_learning_rate,scale_pos_weight,num_samples] #target |
22 | 22 |
|
23 | 23 | #for DL |
|
28 | 28 | #target = CategoricalHyperparam(['target'],['maxhinge','hinge']) |
29 | 29 | #batch_size = CategoricalHyperparam(['training','batch_size'],[256,128,32,64]) |
30 | 30 | #dropout_prob = CategoricalHyperparam(['model','dropout_prob'],[0.1,0.3,0.5]) |
31 | | -# tunables = [lr,lr_decay,fac] #target |
| 31 | +#tunables = [lr,lr_decay,fac] #target |
32 | 32 |
|
33 | 33 |
|
34 | 34 | run_directory = "/tigress/{}/hyperparams/".format(getpass.getuser()) |
35 | 35 | template_path = os.environ['PWD'] #"/home/{}/plasma-python/examples/".format(getpass.getuser()) |
36 | 36 | conf_name = "conf.yaml" |
37 | 37 | num_nodes = 1 |
38 | | -num_trials = 10 |
| 38 | +num_trials = 100 |
39 | 39 |
|
40 | 40 | def generate_conf_file(tunables,template_path = "../",save_path = "./",conf_name="conf.yaml"): |
41 | 41 | assert(template_path != save_path) |
|
0 commit comments