manuel-delverme

import torch
import tensorboardX
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torch.autograd import Variable
import torch.utils.data

writer = tensorboardX.SummaryWriter()

manuel-delverme

    def train_network(samples, neural_network, nr_epochs=10, batch_size=64):
        optimizer = optim.Adam(neural_network.parameters())
        neural_network.train()

        for epoch_nr in range(nr_epochs):
            sample_ids = np.random.shuffle(range(len(samples)))

            for start in range(0, len(samples) // batch_size, batch_size):
                mini_batch = samples[sample_ids[start: start + batch_size]]
                boards, pis, vs = zip(*mini_batch)

manuel-delverme

    environment = environments.GoEnvironment(board_size=19)
    player_mcts = mcts.MCTS(
        environment,
        networks.NeuralNetwork(board_size=environment.getStateSize(), action_size=environment.getActionSize()),
    )

    training_samples = collections.deque(maxlen=opt.training_samples_buffer_size)
    
    for iteration_number in range(opt.num_iters):

manuel-delverme

score:	19.46759259259259	options: nr:4	names:7 10 25 27
score:	11.435185185185185	options: nr:5	names:7 10 25 27 28
score:	25.23611111111111	options: nr:6	names:6 7 10 25 27 28
score:	18.35648148148148	options: nr:7	names:6 7 9 10 25 27 28
score:	23.541666666666668	options: nr:8	names:6 7 9 10 24 25 27 28
score:	23.38888888888889	options: nr:9	names:6 7 9 10 13 24 25 27 28
score:	8.62037037037037	options: nr:10	names:6 7 9 10 13 16 24 25 27 28
score:	24.101851851851848	options: nr:11	names:6 7 9 10 13 16 24 25 27 28 31
score:	18.96759259259259	options: nr:12	names:6 7 9 10 13 16 24 25 27 28 31 34
score:	17.833333333333332	options: nr:13	names:6 7 9 10 12 13 16 24 25 27 28 31 34

manuel-delverme

/home/awok/Projects/supervised_reward/env_reward/bin/python /home/awok/Projects/supervised_reward/main.py
(2_w,4mirr1)-aCMA-ES (mu_w=1.5,w_1=80%) in dimension 18 (seed=237288, Fri Jan  5 12:17:27 2018)
score:	7940.826388888889	options: 8	4 5 10 11 15 16 17 23
score:	5839.784722222223	options: 10	21 22 23 26 27 28 29 33 34 35
score:	8771.63888888889	options: 4	4 5 10 11
score:	-1107.7361111111113	options: 14	0 1 2 3 4 6 7 8 9 10 13 14 15 16
[7940.826388888889, 5839.784722222223, 8771.63888888889, -1107.7361111111113]
best [-0.21687281 -0.29423262  0.10809115 -0.3457722  -0.18912326  0.17178892   0.14703262  0.94997003 -0.18883859 -0.82346577  0.50633336 -0.17325047   0.37087813  0.63369408  0.07967291 -0.47341161 -0.68896583 -0.4226999 ] fitness -8771.63888888889
score:	11841.0625	options: 3	29 34 35
score:	11929.47222222222	options: 4	28 29 34 35

manuel-delverme

import timeit
setup = 'import numpy as np; a=np.random.randn(10)'
reshape = timeit.Timer('a.reshape(-1, 10)', setup=setup)
transp = timeit.Timer('a.transpose()', setup=setup)
T = timeit.Timer('a.T', setup=setup)
print("reshape", reshape.timeit(number=int(1e6)))
print("transp", transp.timeit(number=int(1e6)))
print("T", T.timeit(number=int(1e6)))

print("reshape", reshape.timeit(number=int(1e6)))

manuel-delverme

manuel-delverme

manuel-delverme

/usr/bin/python3.5 /home/awok/Documents/sapienza/s1/nlp/hw2/src/homework2.py model ../ ../resources
Using Theano backend.
WARNING (theano.sandbox.cuda): The cuda backend is deprecated and will be removed in the next release (v0.10).  Please switch to the gpuarray backend. You can get more information about how to switch at this URL:
 https://github.com/Theano/Theano/wiki/Converting-to-the-new-gpu-back-end%28gpuarray%29

ERROR (theano.sandbox.cuda): nvcc compiler not found on $PATH. Check your nvcc installation and try again.
model_output_path model
homework_dir: ../

model output: model

manuel-delverme

/usr/bin/python3.5 /home/awok/Documents/sapienza/s1/nlp/hw2/src/homework2.py model ../ ../resources
Using Theano backend.
model_output_path model
homework_dir: ../

model output: model
homework dir: ../
src dir: ../src/
data dir: ../data/
WARNING (theano.sandbox.cuda): The cuda backend is deprecated and will be removed in the next release (v0.10).  Please switch to the gpuarray backend. You can get more information about how to switch at this URL: