complex-yago-graph.yaml

complex:
  class_name: ComplEx
  entity_embedder:
    dropout: 0.213007293546981
    regularize_weight: 4.239683948309551e-06
    type: lookup_embedder
  relation_embedder:
    dropout: -0.16925297383059523
    regularize_weight: 2.205945531766492e-16
    type: lookup_embedder
dataset:
  name: yago3-10
entity_ranking:
  chunk_size: 50000
eval:
  batch_size: 128
import:
- complex
job:
  type: train
lookup_embedder:
  class_name: LookupEmbedder
  dim: 128
  dropout: 0.0
  initialize: xavier_uniform_
  initialize_args:
    normal_:
      mean: 0.0
      std: 5.6549962925003294e-05
    uniform_:
      a: -0.8255589171977215
    xavier_normal_:
      gain: 1.0
    xavier_uniform_:
      gain: 1.0
  normalize:
    p: -1.0
  regularize: lp
  regularize_args:
    p: 2
    weighted: true
  sparse: true
model: complex
negative_sampling:
  class_name: TrainingJobNegativeSampling
  implementation: batch
  num_samples:
    o: 6425
    p: 0
    s: 3870
  sampling_type: uniform
  shared: true
  shared_type: default
  with_replacement: false
train:
  auto_correct: true
  batch_size: 1024
  loss: kl
  loss_arg: 1.0
  max_epochs: 400
  num_workers: 0
  optimizer:
    default:
      args:
        lr: 0.30874798537148224
      type: Adagrad
  type: negative_sampling
valid:
  early_stopping:
    patience: 50
    threshold:
      epochs: 10
      metric_value: 0.05
  every: 5