-
Notifications
You must be signed in to change notification settings - Fork 3
/
rotate-freebase-distributed-graph-cut-LB-2@2.yaml
111 lines (110 loc) · 2.11 KB
/
rotate-freebase-distributed-graph-cut-LB-2@2.yaml
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
dataset:
files:
test_small:
filename: test_small.del
type: triples
test_tiny:
filename: test_tiny.del
type: triples
test_tiny_tiny:
filename: test_tiny_tiny.del
type: triples
valid_small:
filename: valid_small.del
type: triples
valid_tiny:
filename: valid_tiny.del
type: triples
valid_tiny_tiny:
filename: valid_tiny_tiny.del
type: triples
name: freebase
distributed_model:
base_model:
type: rotate
entity_ranking:
chunk_size: 15000
eval:
batch_size: 45
num_workers: 4
split: test_small
import:
- distributed_model
- rotate
job:
device_pool:
- cuda:0
- cuda:1
distributed:
parameter_server: lapse
entity_localize: true
gloo_socket_ifname: bond0
num_eval_workers: 4
num_machines: 2
num_partitions: 8
num_workers: 8
num_workers_machine: 4
partition_type: graph-cut
lookup_embedder:
dim: 128
initialize: xavier_uniform_
initialize_args:
normal_:
mean: 0.0
std: 0.7247714657764338
uniform_:
a: -0.7264893817037996
xavier_normal_:
gain: 1.0
xavier_uniform_:
gain: 1.0
regularize_args:
weighted: true
sparse: true
model: distributed_model
modules:
- kge.model
- kge.job
- kge.model.embedder
negative_sampling:
combined: true
implementation: batch
num_samples:
o: 400
s: 191
sampling_type: pooled
shared: true
shared_type: naive
rotate:
entity_embedder:
dropout: -0.4202541569247842
regularize_weight: 0.0016039710378808378
relation_embedder:
dropout: -0.4545119274407625
regularize_weight: 9.709504921691681e-20
train:
auto_correct: true
batch_size: 1024
loss_arg: .nan
lr_scheduler: ReduceLROnPlateau
lr_scheduler_args:
factor: 0.95
mode: max
patience: 9
threshold: 0.0001
max_epochs: 10
num_workers: 4
optimizer:
default:
args:
lr: 0.16755594652229405
type: dist_rowadagrad
type: distributed_negative_sampling
valid:
early_stopping:
patience: 10
threshold:
epochs: 50
metric_value: 0.1
every: 10
split: valid_tiny_tiny