def attention_map(self, text):
"""
Text to visualze attention map for.
"""
# encode the string
d = self.input_vocab.string_to_int(text)
print('d: ', d)
# get the output sequence
predicted_text = run_example(self.pred_model, self.input_vocab, self.output_vocab, text)
print('predicted_text: ', predicted_text)
text_ = list(text) + ['<eot>'] + ['<unk>'] * self.input_vocab.padding
# get the lengths of the string
input_length = len(text)+1
output_length = predicted_text.index('<eot>')+1
# get the activation map
activation_map = np.squeeze(self.proba_model.predict(np.array([d])))[0:output_length, 0:input_length]
print('activation_map: ', activation_map)
# [[1.04707105e-05 1.22802967e-05 8.08871482e-06 2.06340337e-05
# 9.13377789e-06 8.17141245e-06 2.89358250e-05 1.30348863e-05
# 3.70874773e-06 1.70587246e-05 7.16923250e-06 4.97975234e-05
# 4.53671564e-05 2.57728461e-05 2.45305255e-05 3.59793594e-05
# 1.75800902e-04 3.21106811e-04 2.58878747e-04 9.57598037e-04]
# [2.88392557e-03 2.04139692e-03 8.94600758e-04 1.82232610e-03
# ...
# import seaborn as sns
plt.clf()
f = plt.figure(figsize=(8, 8.5))
ax = f.add_subplot(1, 1, 1)
# add image
i = ax.imshow(activation_map, interpolation='nearest', cmap='gray') #weight값을 회색으로 표시...
# add colorbar
cbaxes = f.add_axes([0.2, 0, 0.6, 0.03])
cbar = f.colorbar(i, cax=cbaxes, orientation='horizontal')
cbar.ax.set_xlabel('Probability', labelpad=2)
# add labels
ax.set_yticks(range(output_length))
ax.set_yticklabels(predicted_text[:output_length])
ax.set_xticks(range(input_length))
ax.set_xticklabels(text_[:input_length], rotation=45)
ax.set_xlabel('Input Sequence')
ax.set_ylabel('Output Sequence')
# add grid and legend
ax.grid()
# ax.legend(loc='best')
f.savefig(os.path.join(HERE, 'attention_maps', text.replace('/', '')+'.pdf'), bbox_inches='tight')
f.show()
def attention_map(self, text):
"""
Text to visualze attention map for.
"""
# encode the string
d = self.input_vocab.string_to_int(text)
# get the output sequence
predicted_text = run_example(self.pred_model, self.input_vocab,
self.output_vocab, text)
text_ = list(text) + ['<eot>'] + ['<unk>'] * self.input_vocab.padding
# get the lengths of the string
input_length = len(text) + 1
output_length = predicted_text.index('<eot>') + 1
# get the activation map
activation_map = np.squeeze(self.proba_model.predict(np.array(
[d])))[0:output_length, 0:input_length]
# import seaborn as sns
plt.clf()
f = plt.figure(figsize=(8, 8.5))
ax = f.add_subplot(1, 1, 1)
# add image
i = ax.imshow(activation_map, interpolation='nearest', cmap='gray')
# add colorbar
cbaxes = f.add_axes([0.2, 0, 0.6, 0.03])
cbar = f.colorbar(i, cax=cbaxes, orientation='horizontal')
cbar.ax.set_xlabel('Probability', labelpad=2)
# add labels
ax.set_yticks(range(output_length))
ax.set_yticklabels(predicted_text[:output_length])
ax.set_xticks(range(input_length))
ax.set_xticklabels(text_[:input_length], rotation=45)
ax.set_xlabel('Input Sequence')
ax.set_ylabel('Output Sequence')
# add grid and legend
ax.grid()
# ax.legend(loc='best')
f.savefig(os.path.join(HERE, 'attention_maps',
text.replace('/', '') + '.pdf'),
bbox_inches='tight')
f.show()
def attention_map(self, text):
"""
Text to visualze attention map for.
"""
# encode the string
d = self.input_vocab.string_to_int(text)
# get the output sequence
predicted_text = run_example(
self.pred_model, self.input_vocab, self.output_vocab, text)
text_ = list(text) + ['<eot>'] + ['<unk>'] * self.input_vocab.padding
# get the lengths of the string
input_length = len(text)+1
output_length = predicted_text.index('<eot>')+1
# get the activation map
activation_map = np.squeeze(self.proba_model.predict(np.array([d])))[
0:output_length, 0:input_length]
# import seaborn as sns
plt.clf()
f = plt.figure(figsize=(8, 8.5))
ax = f.add_subplot(1, 1, 1)
# add image
i = ax.imshow(activation_map, interpolation='nearest', cmap='gray')
# add colorbar
cbaxes = f.add_axes([0.2, 0, 0.6, 0.03])
cbar = f.colorbar(i, cax=cbaxes, orientation='horizontal')
cbar.ax.set_xlabel('Probability', labelpad=2)
# add labels
ax.set_yticks(range(output_length))
ax.set_yticklabels(predicted_text[:output_length])
ax.set_xticks(range(input_length))
ax.set_xticklabels(text_[:input_length], rotation=45)
ax.set_xlabel('Input Sequence')
ax.set_ylabel('Output Sequence')
# add grid and legend
ax.grid()
# ax.legend(loc='best')
f.savefig(os.path.join(HERE, 'attention_maps', text.replace('/', '')+'.pdf'), bbox_inches='tight')
f.show()