Skip to content

Commit

Permalink
remove ProtoData, update related files
Browse files Browse the repository at this point in the history
  • Loading branch information
luotao1 committed Nov 20, 2017
1 parent 56ec40a commit 76501c8
Show file tree
Hide file tree
Showing 6 changed files with 2 additions and 252 deletions.
Binary file removed paddle/trainer/tests/compare_sparse_data
Binary file not shown.
214 changes: 0 additions & 214 deletions paddle/trainer/tests/data_bin_part

This file was deleted.

7 changes: 1 addition & 6 deletions paddle/trainer/tests/test_config.conf
Original file line number Diff line number Diff line change
Expand Up @@ -15,12 +15,7 @@

from paddle.trainer_config_helpers import *

TrainData(ProtoData(
files = "dummy_list",
constant_slots = [1.0],
async_load_data = True))

TestData(SimpleData(
TrainData(SimpleData(
files = "trainer/tests/sample_filelist.txt",
feat_dim = 3,
context_len = 0,
Expand Down
1 change: 0 additions & 1 deletion paddle/trainer/tests/train.list

This file was deleted.

1 change: 0 additions & 1 deletion paddle/trainer/tests/train_sparse.list

This file was deleted.

31 changes: 1 addition & 30 deletions python/paddle/trainer/config_parser.py
Original file line number Diff line number Diff line change
Expand Up @@ -1116,35 +1116,6 @@ def get_path(module):
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.
@config_func
def MultiData(sub_data=[]):
Expand Down Expand Up @@ -2714,7 +2685,7 @@ def __init__(self, name, num_classes, inputs, device=None, bias=True):
max_sort_size = -1, inputs = ["output", "score"])
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
input of lambdaRank layer.
Expand Down

0 comments on commit 76501c8

Please sign in to comment.