Project Not Compiling

[liaq192]

when i try to compile using python vsum_train.py --dataset datasets/eccv16_dataset_tvsum_google_pool5.h5 --max-epochs 60 --hidden-dim 256

It gives following error

python vsum_train.py --dataset datasets/eccv16_dataset_tvsum_google_pool5.h5 --max-epochs 60 --hidden-dim 256

WARNING (theano.tensor.blas): Using NumPy C-API based implementation for BLAS functions. [29/06/2018 08:46:03] model options: {'disp_freq': 1, 'optimizer': 'adam', 'regularizer': 'L2', 'U_init': 'normal', 'base_lr': 1e-05, 'train_dataset_path': 'datasets/eccv16_dataset_tvsum_google_pool5.h5', 'alpha': 0.01, 'max_epochs': 60, 'W_init': 'normal', 'weight_decay': 1e-05, 'distant_sim_thre': 20, 'ignore_distant_sim': False, 'decay_stepsize': -1, 'model_file': None, 'decay_rate': 0.1, 'hidden_dim': 256, 'input_dim': 1024, 'n_episodes': 5} [29/06/2018 08:46:03] initializing net model Traceback (most recent call last): File "vsum_train.py", line 158, in train_dataset_path=args.dataset) File "vsum_train.py", line 56, in train net = reinforceRNN(model_options) File "/home/ssong/campTest1/vsumm-reinforce/model_reinforceRNN.py", line 40, in init init_state=None, init_memory=None, go_backwards=False File "/home/ssong/campTest1/vsumm-reinforce/theano_nets.py", line 342, in init self.output = self.step(self.state_below) File "/home/ssong/campTest1/vsumm-reinforce/theano_nets.py", line 396, in step go_backwards=self.go_backwards File "/home/ssong/.local/lib/python2.7/site-packages/theano/scan_module/scan.py", line 1077, in scan scan_outs = local_op(*scan_inputs) File "/home/ssong/.local/lib/python2.7/site-packages/theano/gof/op.py", line 615, in call node = self.make_node(*inputs, **kwargs) File "/home/ssong/.local/lib/python2.7/site-packages/theano/scan_module/scan_op.py", line 546, in make_node inner_sitsot_out.type.dtype)) ValueError: When compiling the inner function of scan the following error has been encountered: The initial state (outputs_info in scan nomenclature) of variable IncSubtensor{Set;:int64:}.0 (argument number 1) has dtype float32, while the result of the inner function (fn) has dtype float64. This can happen if the inner function of scan results in an upcast or downcast.

using following python version Python 2.7.12

[KaiyangZhou]

It says the dtype is supposed to be float32 but found float64. You need to change the global config to floatX = float32.

[liaq192]

python vsum_train.py --dataset datasets/eccv16_dataset_tvsum_google_pool5.h5 --max-epochs 60 --hidden-dim 256 WARNING (theano.tensor.blas): Using NumPy C-API based implementation for BLAS functions. [29/06/2018 02:57:38] model options: {'disp_freq': 1, 'optimizer': 'adam', 'regularizer': 'L2', 'U_init': 'normal', 'base_lr': 1e-05, 'train_dataset_path': 'datasets/eccv16_dataset_tvsum_google_pool5.h5', 'alpha': 0.01, 'max_epochs': 60, 'W_init': 'normal', 'weight_decay': 1e-05, 'distant_sim_thre': 20, 'ignore_distant_sim': False, 'decay_stepsize': -1, 'model_file': None, 'decay_rate': 0.1, 'hidden_dim': 256, 'input_dim': 1024, 'n_episodes': 5} [29/06/2018 02:57:38] initializing net model Traceback (most recent call last): File "vsum_train.py", line 158, in train_dataset_path=args.dataset) File "vsum_train.py", line 56, in train net = reinforceRNN(model_options) File "/home/ssong/campTest1/vsumm-reinforce/model_reinforceRNN.py", line 110, in init grads = [T.grad(cost=cost, wrt=p) for p in self.params] File "/home/ssong/.local/lib/python2.7/site-packages/theano/gradient.py", line 611, in grad rval[i].type.why_null) theano.gradient.NullTypeGradError: tensor.grad encountered a NaN. This variable is Null because the grad method for input 0 (Reshape{2}.0) of the UndefinedGrad op is mathematically undefined.

[KaiyangZhou]

you need to use theano 0.9.

i suggest using the pytorch version, which requires less efforts in configuration.

[liaq192]

Hi , Can u add some config details in readme about the packages u have used and version that can help me and other users to configure this project easily . ( it may be obvious for some users but for someone like me would be very helpful) .

[XChuanLee]

Hi! I met the same problem, have you solved it? using following version: python: 2.7.12 theano:0.9.0