@ -18,8 +18,9 @@ import sys
from paddle . trainer_config_helpers import * from paddle . trainer_config_helpers import *
#file paths #file paths
word_dict_file = ' ./data/src.dict ' word_dict_file = ' ./data/wordDict.txt '
label_dict_file = ' ./data/tgt.dict ' label_dict_file = ' ./data/targetDict.txt '
predicate_file = ' ./data/verbDict.txt '
train_list_file = ' ./data/train.list ' train_list_file = ' ./data/train.list '
test_list_file = ' ./data/test.list ' test_list_file = ' ./data/test.list '
@ -30,8 +31,10 @@ if not is_predict:
#load dictionaries
#load dictionaries
word_dict = dict ( )
word_dict = dict ( )
label_dict = dict ( )
label_dict = dict ( )
predicate_dict = dict ( )
with open ( word_dict_file , ' r ' ) as f_word , \
with open ( word_dict_file , ' r ' ) as f_word , \
open ( label_dict_file , ' r ' ) as f_label :
open ( label_dict_file , ' r ' ) as f_label , \
open ( predicate_file , ' r ' ) as f_pre :
for i , line in enumerate ( f_word ) :
for i , line in enumerate ( f_word ) :
w = line . strip ( )
w = line . strip ( )
word_dict [ w ] = i
word_dict [ w ] = i
@ -40,6 +43,11 @@ if not is_predict:
w = line . strip ( )
w = line . strip ( )
label_dict [ w ] = i
label_dict [ w ] = i
for i , line in enumerate ( f_pre ) :
w = line . strip ( )
predicate_dict [ w ] = i
if is_test :
if is_test :
train_list_file = None
train_list_file = None
@ -50,91 +58,157 @@ if not is_predict:
module = ' dataprovider ' ,
module = ' dataprovider ' ,
obj = ' process ' ,
obj = ' process ' ,
args = { ' word_dict ' : word_dict ,
args = { ' word_dict ' : word_dict ,
' label_dict ' : label_dict } )
' label_dict ' : label_dict ,
' predicate_dict ' : predicate_dict } )
word_dict_len = len ( word_dict )
word_dict_len = len ( word_dict )
label_dict_len = len ( label_dict )
label_dict_len = len ( label_dict )
pred_len = len ( predicate_dict )
else : else :
word_dict_len = get_config_arg ( ' dict_len ' , int )
word_dict_len = get_config_arg ( ' dict_len ' , int )
label_dict_len = get_config_arg ( ' label_len ' , int )
label_dict_len = get_config_arg ( ' label_len ' , int )
pred_len = get_config_arg ( ' pred_len ' , int )
############################## Hyper-parameters ##################################
mark_dict_len = 2 mark_dict_len = 2
word_dim = 32 word_dim = 32
mark_dim = 5 mark_dim = 5
hidden_dim = 8 hidden_dim = 5 12
depth = 8 depth = 8
emb_lr = 1e-2
fc_lr = 1e-2
lstm_lr = 2e-2
########################### Optimizer #######################################
settings ( settings (
batch_size = 150 ,
batch_size = 150 ,
learning_method = AdamOptimizer( ) ,
learning_method = MomentumOptimizer( momentum = 0 ) ,
learning_rate = 1e-3 ,
learning_rate = 2e-2 ,
regularization = L2Regularization ( 8e-4 ) ,
regularization = L2Regularization ( 8e-4 ) ,
gradient_clipping_threshold = 25 )
is_async = False ,
model_average = ModelAverage ( average_window = 0.5 ,
max_average_window = 10000 ) ,
)
#6 features
####################################### network ##############################
#8 features and 1 target
word = data_layer ( name = ' word_data ' , size = word_dict_len ) word = data_layer ( name = ' word_data ' , size = word_dict_len )
predicate = data_layer ( name = ' verb_data ' , size = word_dict_len ) predicate = data_layer ( name = ' verb_data ' , size = pred_len )
ctx_n2 = data_layer ( name = ' ctx_n2_data ' , size = word_dict_len )
ctx_n1 = data_layer ( name = ' ctx_n1_data ' , size = word_dict_len ) ctx_n1 = data_layer ( name = ' ctx_n1_data ' , size = word_dict_len )
ctx_0 = data_layer ( name = ' ctx_0_data ' , size = word_dict_len ) ctx_0 = data_layer ( name = ' ctx_0_data ' , size = word_dict_len )
ctx_p1 = data_layer ( name = ' ctx_p1_data ' , size = word_dict_len ) ctx_p1 = data_layer ( name = ' ctx_p1_data ' , size = word_dict_len )
ctx_p2 = data_layer ( name = ' ctx_p2_data ' , size = word_dict_len )
mark = data_layer ( name = ' mark_data ' , size = mark_dict_len ) mark = data_layer ( name = ' mark_data ' , size = mark_dict_len )
if not is_predict : if not is_predict :
target = data_layer ( name = ' target ' , size = label_dict_len )
target = data_layer ( name = ' target ' , size = label_dict_len )
ptt = ParameterAttribute ( name = ' src_emb ' , learning_rate = emb_lr )
layer_attr = ExtraLayerAttribute ( drop_rate = 0.5 )
fc_para_attr = ParameterAttribute ( learning_rate = fc_lr )
lstm_para_attr = ParameterAttribute ( initial_std = 0. , learning_rate = lstm_lr )
para_attr = [ fc_para_attr , lstm_para_attr ]
word_embedding = embedding_layer ( size = word_dim , input = word , param_attr = ptt ) default_std = 1 / math . sqrt ( hidden_dim ) / 3.0
predicate_embedding = embedding_layer (
size = word_dim , input = predicate , param_attr = ptt )
emb_para = ParameterAttribute ( name = ' emb ' , initial_std = 0. , learning_rate = 0. )
ctx_n1_embedding = embedding_layer ( size = word_dim , input = ctx_n1 , param_attr = ptt ) std_0 = ParameterAttribute ( initial_std = 0. )
ctx_0_embedding = embedding_layer ( size = word_dim , input = ctx_0 , param_attr = ptt ) std_default = ParameterAttribute ( initial_std = default_std )
ctx_p1_embedding = embedding_layer ( size = word_dim , input = ctx_p1 , param_attr = ptt )
mark_embedding = embedding_layer ( size = mark_dim , input = mark ) predicate_embedding = embedding_layer ( size = word_dim , input = predicate , param_attr = ParameterAttribute ( name = ' vemb ' , initial_std = default_std ) )
mark_embedding = embedding_layer ( name = ' word_ctx-in_embedding ' , size = mark_dim , input = mark , param_attr = std_0 )
word_input = [ word , ctx_n2 , ctx_n1 , ctx_0 , ctx_p1 , ctx_p2 ]
emb_layers = [ embedding_layer ( size = word_dim , input = x , param_attr = emb_para ) for x in word_input ]
emb_layers . append ( predicate_embedding )
emb_layers . append ( mark_embedding )
hidden_0 = mixed_layer ( hidden_0 = mixed_layer (
name = ' hidden0 ' ,
size = hidden_dim ,
size = hidden_dim ,
input = [
bias_attr = std_default ,
full_matrix_projection ( input = word_embedding ) ,
input = [ full_matrix_projection ( input = emb , param_attr = std_default ) for emb in emb_layers ] )
full_matrix_projection ( input = predicate_embedding ) ,
full_matrix_projection ( input = ctx_n1_embedding ) ,
full_matrix_projection ( input = ctx_0_embedding ) ,
full_matrix_projection ( input = ctx_p1_embedding ) ,
full_matrix_projection ( input = mark_embedding ) ,
] )
lstm_0 = lstmemory ( input = hidden_0 , layer_attr = layer_attr ) mix_hidden_lr = 1e-3
lstm_para_attr = ParameterAttribute ( initial_std = 0.0 , learning_rate = 1.0 )
hidden_para_attr = ParameterAttribute ( initial_std = default_std , learning_rate = mix_hidden_lr )
lstm_0 = lstmemory ( name = ' lstm0 ' ,
input = hidden_0 ,
act = ReluActivation ( ) ,
gate_act = SigmoidActivation ( ) ,
state_act = SigmoidActivation ( ) ,
bias_attr = std_0 ,
param_attr = lstm_para_attr )
#stack L-LSTM and R-LSTM with direct edges #stack L-LSTM and R-LSTM with direct edges
input_tmp = [ hidden_0 , lstm_0 ] input_tmp = [ hidden_0 , lstm_0 ]
for i in range ( 1 , depth ) : for i in range ( 1 , depth ) :
fc = fc_layer ( input = input_tmp , size = hidden_dim , param_attr = para_attr )
mix_hidden = mixed_layer ( name = ' hidden ' + str ( i ) ,
size = hidden_dim ,
bias_attr = std_default ,
input = [ full_matrix_projection ( input = input_tmp [ 0 ] , param_attr = hidden_para_attr ) ,
full_matrix_projection ( input = input_tmp [ 1 ] , param_attr = lstm_para_attr )
]
)
lstm = lstmemory ( name = ' lstm ' + str ( i ) ,
input = mix_hidden ,
act = ReluActivation ( ) ,
gate_act = SigmoidActivation ( ) ,
state_act = SigmoidActivation ( ) ,
reverse = ( ( i % 2 ) == 1 ) ,
bias_attr = std_0 ,
param_attr = lstm_para_attr )
input_tmp = [ mix_hidden , lstm ]
feature_out = mixed_layer ( name = ' output ' ,
size = label_dict_len ,
bias_attr = std_default ,
input = [ full_matrix_projection ( input = input_tmp [ 0 ] , param_attr = hidden_para_attr ) ,
full_matrix_projection ( input = input_tmp [ 1 ] , param_attr = lstm_para_attr )
] ,
)
lstm = lstmemory (
input = fc ,
act = ReluActivation ( ) ,
reverse = ( i % 2 ) == 1 ,
layer_attr = layer_attr )
input_tmp = [ fc , lstm ]
prob = fc_layer (
input = input_tmp ,
size = label_dict_len ,
act = SoftmaxActivation ( ) ,
param_attr = para_attr )
if not is_predict : if not is_predict :
cls = classification_cost ( input = prob , label = target )
crf_l = crf_layer ( name = ' crf ' ,
outputs ( cls )
size = label_dict_len ,
input = feature_out ,
label = target ,
param_attr = ParameterAttribute ( name = ' crfw ' , initial_std = default_std , learning_rate = mix_hidden_lr )
)
crf_dec_l = crf_decoding_layer ( name = ' crf_dec_l ' ,
size = label_dict_len ,
input = feature_out ,
label = target ,
param_attr = ParameterAttribute ( name = ' crfw ' )
)
eval = sum_evaluator ( input = crf_dec_l )
outputs ( crf_l )
else : else :
outputs ( prob )
crf_dec_l = crf_decoding_layer ( name = ' crf_dec_l ' ,
size = label_dict_len ,
input = feature_out ,
param_attr = ParameterAttribute ( name = ' crfw ' )
)
outputs ( crf_dec_l )
Luo Tao
9 years ago
