pre-train accuracy for fbwq_half

[lihuiliullh]

May I know how much the accuracy is when using kge trains fbwq_half ?
Also, when you use kgc to train fbwq_half, did you only use train.txt?

[lihuiliullh]

I used the pre-trained embedding on KGQA, the accuracy is very low. How did you learn the embedding use kge? @apoorvumang

[apoorvumang]

Can you please elaborate on what experiment you did, what commands you used?

[lihuiliullh]

I try to use kge to get the pre-train embedding of fbwq_half, and use the pre-trained embedding in multi-hop answering on fbwq_half. @apoorvumang

[lihuiliullh]

But the pre-train accuracy is very low when using kge. If I use that embedding, the kgqa accuracy is also much lower than the result in the paper. So, I want to know, how do you use kge to pre-train the model. The config file I use is listed below.

complex:
entity_embedder:
dropout: 0.44299429655075073
regularize_weight: 7.830760727899156e-12
relation_embedder:
dropout: -0.4746062345802784
regularize_weight: 1.182876478423781e-10
dataset:
name: fbwq_half
eval:
batch_size: 200
chunk_size: 25000
num_workers: 2
import:

• complex
  lookup_embedder:
  dim: 400
  initialize_args:
  normal_:
  mean: 0.0
  std: 5.8970567449527816e-05
  regularize_args:
  p: 1
  weighted: true
  sparse: true
  model: complex
  negative_sampling:
  implementation: batch
  num_samples:
  o: 7851
  s: 2176
  shared: true
  with_replacement: false
  train:
  auto_correct: true
  batch_size: 1024
  loss_arg: .nan
  lr_scheduler: ReduceLROnPlateau
  lr_scheduler_args:
  factor: 0.95
  mode: max
  patience: 1
  threshold: 0.0001
  max_epochs: 200
  num_workers: 8
  optimizer_args:
  lr: 0.6560544891789137
  type: negative_sampling
  valid:
  early_stopping:
  min_threshold:
  epochs: 10
  metric_value: 0.1
  patience: 10

[apoorvumang]

Thanks for the details, let me try it with this config and I'll get back to you

[lihuiliullh]

Can you share your config file to me? so I can get the same performance in the paper @apoorvumang