yandex / faster-rnnlm

Faster Recurrent Neural Network Language Modeling Toolkit with Noise Contrastive Estimation and Hierarchical Softmax
Other
561 stars 138 forks source link

error by training data #31

Open Thanh-Binh opened 8 years ago

Thanh-Binh commented 8 years ago

dear all, I am trying to train using dataset from rnnlm.org, but have errors as follow: ./rnnlm -rnnlm binh_model -train /home/binh/DeepLearning/simple-examples/data/ptb.char.train.txt -valid /home/binh/DeepLearning/simple-examples/data/ptb.char.valid.txt -hidden 128 -hidden-type gru -nce 20 -alpha 0.01 Read the vocabulary: 8 words Restoring existing nnet Constructing RNN: layer_size=128, layer_type=gru, layer_count=1, maxent_hash_size=0, maxent_order=0, vocab_size=8, use_nce=1 Constructing NCE: layer_size=128, maxent_hash_size=0, cuda=1, ln(Z)=9.000000 Constructed UnigramNoiseGenerator: power=1.000, mincells(param)=5.000, mincells(real)=9090909 Initial entropy (bits) valid: -nan Epoch 1 lr: 1.00e-02/1.00e-01 progress: 99.79% 49385.93 Kwords/sec entropy (bits) valid: -nan elapsed: 0.1s+0.0s Awful: Nnet rejected Epoch 2 lr: 5.00e-03/5.00e-02 progress: 99.79% 56978.07 Kwords/sec entropy (bits) valid: -nan elapsed: 0.1s+0.0s Awful: Nnet rejected Epoch 3 lr: 2.50e-03/2.50e-02 progress: 99.79% 54694.62 Kwords/sec entropy (bits) valid: -nan elapsed: 0.1s+0.0s Awful: Nnet rejected

Could anyone help me? Best thanks