@@ -20,14 +20,14 @@ def __init__(self, emission_weights, transition_weights,
2020 self .w = transition_weights [2 :, :]
2121
2222 self .track = np .zeros (
23- (seq_start_positions [- 1 ], self .tag_num ), dtype = "int32 " )
23+ (seq_start_positions [- 1 ], self .tag_num ), dtype = "int64 " )
2424 self .decoded_path = np .zeros (
25- (seq_start_positions [- 1 ], 1 ), dtype = "int32 " )
25+ (seq_start_positions [- 1 ], 1 ), dtype = "int64 " )
2626
2727 def _decode_one_sequence (self , decoded_path , x ):
2828 seq_len , tag_num = x .shape
2929 alpha = np .zeros ((seq_len , tag_num ), dtype = "float64" )
30- track = np .zeros ((seq_len , tag_num ), dtype = "int32 " )
30+ track = np .zeros ((seq_len , tag_num ), dtype = "int64 " )
3131
3232 for i in range (tag_num ):
3333 alpha [0 , i ] = self .a [i ] + x [0 , i ]
@@ -125,10 +125,10 @@ def setUp(self):
125125 axis = 0 )
126126
127127 labels = np .random .randint (
128- low = 0 , high = TAG_NUM , size = (lod [- 1 ][- 1 ], 1 ), dtype = "int32 " )
128+ low = 0 , high = TAG_NUM , size = (lod [- 1 ][- 1 ], 1 ), dtype = "int64 " )
129129 predicted_labels = np .ones (
130- (lod [- 1 ][- 1 ], 1 ), dtype = "int32 " ) * (TAG_NUM - 1 )
131- expected_output = (labels == predicted_labels ).astype ("int32 " )
130+ (lod [- 1 ][- 1 ], 1 ), dtype = "int64 " ) * (TAG_NUM - 1 )
131+ expected_output = (labels == predicted_labels ).astype ("int64 " )
132132
133133 self .inputs = {
134134 "Emission" : (emission , lod ),
0 commit comments