--allfolds KeyError

[SanFran-Me]

when I use --allfolds flag

KeyError: 'train'

[Dadiao-shuai]

I also encountered this error when I use --allfolds to train on a single types file 'train_all.types' with
python3 train.py -m default2018.model --allfolds -p train_all -d /root/gnina_docker --weights crossdock_default2018.caffemodel --dynamic

  File "train.py", line 479, in train_and_test_model
    idx = test_idxs[key]
KeyError: 'train'

[dkoes]

You can skip the case where the key isn't in test_idxs. This is happening because the train and test set are the same so the train set isn't put in test_idxs.

if key not in test_idxs: continue

[JonasLi-19]

Hello professor, it might still need some other codes modified when running --allfolds option:

Iteration 1000
Train time: 208.816554
Traceback (most recent call last):
  File "train.py", line 934, in <module>
    results = train_and_test_model(args, train_test_files[i], outname, cont)
  File "train.py", line 541, in train_and_test_model
    test_net, n_tests = test_nets[key]
KeyError: 'train'

[JonasLi-19]

--allfolds still causes error:

Iteration 1000
Train time: 208.816554
Traceback (most recent call last):
  File "train.py", line 934, in <module>
    results = train_and_test_model(args, train_test_files[i], outname, cont)
  File "train.py", line 541, in train_and_test_model
    test_net, n_tests = test_nets[key]
KeyError: 'train'

so can I directly add : if key not in test_idxs: continue

        #evaluate train set
        start = time.time()
        if use_reduced and (not last_test or args.skip_full):
            key = 'reduced_train'
        else:
            key = 'train'
        test_net, n_tests = test_nets[key]
        freemem()

[JonasLi-19]

I tried directly add : if key not in test_idxs: continue

        #evaluate train set
        start = time.time()
        if use_reduced and (not last_test or args.skip_full):
            key = 'reduced_train'
        else:
            key = 'train'
        test_net, n_tests = test_nets[key]
        freemem()

it avoid the error.

But I find the result .out file cannot be generated! #62

[Dadiao-shuai]

even I follow the modifications, I find --allfolds cannot generate .out files, too! @JonasLi-19

You can skip the case where the key isn't in test_idxs. This is happening because the train and test set are the same so the train set isn't put in test_idxs.

if key not in test_idxs: continue