bibyutatsu / fastjtnnpy3 Goto Github PK

This error occurred while I was running the code，I couldn't find the solution on the Internet. I hope I successfully installed Theano(because I'm not sure)：
UserWarning: size_average and reduce args will be deprecated, please use reduction='sum' instead.
warnings.warn(warning.format(ret))
Initializing network
WARNING (theano.gof.compilelock): Overriding existing lock by dead process '17108' (I am process '11780')
Traceback (most recent call last):
File "F:/Study/YJS/FastJTNNpy3-master/fast_bo/run_bo.py", line 248, in
random_seed)
File "F:/Study/YJS/FastJTNNpy3-master/fast_bo/run_bo.py", line 112, in main_bo
learning_rate=0.001)
File "F:\Study\YJS\FastJTNNpy3-master\fast_jtnn\sparse_gp.py", line 209, in train_via_ADAM
e = self.getEnergy()
File "F:\Study\YJS\FastJTNNpy3-master\fast_jtnn\sparse_gp.py", line 99, in getEnergy
return self.sparse_gp.getContributionToEnergy()[ 0, 0 ]
File "F:\Study\YJS\FastJTNNpy3-master\fast_jtnn\sparse_gp_theano_internal.py", line 191, in getContributionToEnergy
logZpost = self.getLogNormalizerPosterior()
File "F:\Study\YJS\FastJTNNpy3-master\fast_jtnn\sparse_gp_theano_internal.py", line 180, in getLogNormalizerPosterior
return casting(0.5 * self.n_inducing_points * np.log(2 * np.pi)) + casting(0.5) * T.nlinalg.LogDetPSD()(self.covPosterior) +
AttributeError: module 'theano.tensor.nlinalg' has no attribute 'LogDetPSD'

when I run preprocess.py on custom data, I meet a problem.

Reason: 'RecursionError('maximum recursion depth exceeded while pickling an object')'
  File "/y/home/src/jtnnvae/fast_molvae/preprocess.py", line 56, in convert
    all_data = pool.map(tensorize, data)
  File "/y/home/src/jtnnvae/fast_molvae/preprocess.py", line 91, in <module>
    convert(opts.train_path, pool, num_splits, opts.output_path)

Any advice? Many thanks.

from sample import load_model
model = load_model('/content/vocab.txt','/content/model.epoch-19')

where it is saying JTNNVAE is not defined even it is defined in sample.py
Let me know the solution for that

I am trying to load the model but every time an error comes tensors are not on the same device CPU and GPU.

Namespace(anneal_iter=40000, anneal_rate=0.9, batch_size=32, beta=0.0, clip_norm=50.0, depthG=3, depthT=20, epoch=20, hidden_size=450, kl_anneal_iter=2000, latent_size=56, load_epoch=0, lr=0.001, max_beta=1.0, print_iter=50, save_dir='vae_model/', save_iter=5000, step_beta=0.002, train='moses-processed', vocab='../data/vocab.txt', warmup=40000)
Traceback (most recent call last):
File "vae_train.py", line 166, in
args.save_iter)
File "vae_train.py", line 47, in main__vae__train
model = JTNNVAE(vocab, int(hidden_size), int(latent_size), int(depthT), int(depthG)).cuda()
TypeError: init() missing 1 required positional argument: 'depthG'

Can anyone help with this?