提交 76501c82 编写于 作者: L Luo Tao

remove ProtoData, update related files

上级 56ec40ad
因为 它太大了无法显示 source diff 。你可以改为 查看blob
...@@ -15,12 +15,7 @@ ...@@ -15,12 +15,7 @@
from paddle.trainer_config_helpers import * from paddle.trainer_config_helpers import *
TrainData(ProtoData( TrainData(SimpleData(
files = "dummy_list",
constant_slots = [1.0],
async_load_data = True))
TestData(SimpleData(
files = "trainer/tests/sample_filelist.txt", files = "trainer/tests/sample_filelist.txt",
feat_dim = 3, feat_dim = 3,
context_len = 0, context_len = 0,
......
...@@ -1116,35 +1116,6 @@ def PyData(files=None, ...@@ -1116,35 +1116,6 @@ def PyData(files=None,
return data_config return data_config
@config_func
def ProtoData(files=None,
type=None,
file_group_queue_capacity=None,
load_file_count=None,
constant_slots=None,
load_thread_num=None,
**xargs):
data_config = create_data_config_proto(**xargs)
if type is None:
data_config.type = 'proto'
else:
data_config.type = type
data_config.files = files
# When type="proto_group", one data provider contains at most
# load_file_count files, and there are at most
# (queue_capacity + load_thread_num + 1) data providers in memory
if file_group_queue_capacity is not None:
data_config.file_group_conf.queue_capacity = file_group_queue_capacity
if load_file_count is not None:
data_config.file_group_conf.load_file_count = load_file_count
if load_thread_num is not None:
data_config.file_group_conf.load_thread_num = load_thread_num
if constant_slots:
data_config.constant_slots.extend(constant_slots)
return data_config
#real data for training is actually provided by "sub_data" data providers. #real data for training is actually provided by "sub_data" data providers.
@config_func @config_func
def MultiData(sub_data=[]): def MultiData(sub_data=[]):
...@@ -2714,7 +2685,7 @@ Usage: ...@@ -2714,7 +2685,7 @@ Usage:
max_sort_size = -1, inputs = ["output", "score"]) max_sort_size = -1, inputs = ["output", "score"])
Input data: Samples of the same query should be loaded as a sequence, Input data: Samples of the same query should be loaded as a sequence,
by ProtoDataProvider or PyDataProvider etc.. User should provide by PyDataProvider etc.. User should provide
scores for each sample. The score slot should be the 2nd scores for each sample. The score slot should be the 2nd
input of lambdaRank layer. input of lambdaRank layer.
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册