In [1]:

    

import numpy
import matplotlib.pyplot as plt
import pickle as pkl
import numpy as np
%matplotlib inline


    

In [23]:

    

losses_v=pkl.load(open('losses/lossv.pkl','rb'))


    

In [45]:

    

loss=pkl.load(open('losses/3s_18b_30l_1024h_19d_512z_512zm_defdef4.pkl','rb'))


    

In [24]:

    

losses_e=pkl.load(open('losses/losse.pkl','rb'))


    

In [40]:

    

losses=pkl.load(open('losses/loss.pkl','rb'))


    

In [7]:

    

ctc=pkl.load(open('losses/1s_500b_30l_1024h_19d_512z_512zm_ctcclipvarvaedef.pkl','rb'))


    

In [111]:

    

len(losses_e)


    

    
Out[111]:





62075

In [29]:

    

total_batch=(158195/(128/2))


    

In [30]:

    

max([x/float(total_batch) for x in range(len(losses_e))])


    

    
Out[30]:





25.121003642250102

In [31]:

    

min(losses_e)


    

    
Out[31]:





2.2984068

In [32]:

    

def moving_average(a, n=15) :
    ret = np.cumsum(a, dtype=float)
    ret[n:] = ret[n:] - ret[:-n]
    return ret[n - 1:] / n


    

In [33]:

    

smooth_loss_e=moving_average(losses_e)


    

In [34]:

    

smooth_loss_v=moving_average(losses_v)


    

In [38]:

    

smooth_loss=moving_average(losses)


    

In [39]:

    

plt.plot([x/float(total_batch) for x in range(len(smooth_loss_e))],smooth_loss_e,label='W2V Embeddings')
plt.plot([x/float(total_batch) for x in range(len(smooth_loss_v))],smooth_loss_v,label='E2E Training')
plt.plot([x*2/float(total_batch) for x in range(len(smooth_loss))],smooth_loss,label='VAE2Def Embeddings')
plt.title('Flickr-30k Caption Reconstruction Loss')
plt.xlabel('epochs')
plt.ylabel('Softmax Cross Entropy Loss')
plt.legend()


    

    
Out[39]:





<matplotlib.legend.Legend at 0x7f6538a05650>






    

In [123]:

    

smooth_ctc=moving_average(ctc)


    

In [126]:

    

plt.plot(ctc)


    

    
Out[126]:





[<matplotlib.lines.Line2D at 0x7ff84bb88d50>]






    

In [125]:

    

ctc


    

    
Out[125]:





[-0.68619465827941895,
 -0.5740787422223268,
 -0.46343051827569476,
 -0.44442209447493575,
 -0.44236837467397266,
 -0.44231022432656375]

In [42]:

    

smooth_loss=moving_average(loss,3)


    

In [43]:

    

plt.plot(smooth_loss)


    

    
Out[43]:





[<matplotlib.lines.Line2D at 0x7f6537daded0>]






    

In [46]:

    

plt.plot(loss)


    

    
Out[46]:





[<matplotlib.lines.Line2D at 0x7f653887e5d0>]






    

In [ ]: