Merge remote-tracking branch 'upstream/master'

uma-pi1 · Jul 9, 2020 · 4e06713 · 4e06713
2 parents 162741c + d918c48
commit 4e06713
Show file tree

Hide file tree

Showing 2 changed files with 93 additions and 22 deletions.
diff --git a/data/download_all.sh b/data/download_all.sh
@@ -11,7 +11,7 @@ fi
 if [ ! -d "$BASEDIR/toy" ]; then
  echo Downloading toy
  cd $BASEDIR
- curl -O https:https://www.uni-mannheim.de/media/Einrichtungen/dws/pi1/kge_datasets/toy.tar.gz
+ curl -O http:https://web.informatik.uni-mannheim.de/pi1/kge-datasets/toy.tar.gz
  tar xvf toy.tar.gz
 else
  echo toy already present
@@ -27,7 +27,7 @@ fi
 if [ ! -d "$BASEDIR/fb15k" ]; then
  echo Downloading fb15k
  cd $BASEDIR
- curl -O https:https://www.uni-mannheim.de/media/Einrichtungen/dws/pi1/kge_datasets/fb15k.tar.gz
+ curl -O http:https://web.informatik.uni-mannheim.de/pi1/kge-datasets/fb15k.tar.gz
  tar xvf fb15k.tar.gz
  cd fb15k
  case "$(uname -s)" in
@@ -56,7 +56,7 @@ fi
 if [ ! -d "$BASEDIR/fb15k-237" ]; then
  echo Downloading fb15k-237
  cd $BASEDIR
- curl -O https:https://www.uni-mannheim.de/media/Einrichtungen/dws/pi1/kge_datasets/fb15k-237.tar.gz
+ curl -O http:https://web.informatik.uni-mannheim.de/pi1/kge-datasets/fb15k-237.tar.gz
  tar xvf fb15k-237.tar.gz
 else
  echo fb15k-237 already present
@@ -71,7 +71,7 @@ fi
 if [ ! -d "$BASEDIR/wn18" ]; then
  echo Downloading wn18
  cd $BASEDIR
- curl -O https:https://www.uni-mannheim.de/media/Einrichtungen/dws/pi1/kge_datasets/wn18.tar.gz
+ curl -O http:https://web.informatik.uni-mannheim.de/pi1/kge-datasets/wn18.tar.gz
  tar xvf wn18.tar.gz
  cd wn18
  case "$(uname -s)" in
@@ -100,7 +100,7 @@ fi
 if [ ! -d "$BASEDIR/wnrr" ]; then
  echo Downloading wnrr
  cd $BASEDIR
- curl -O https:https://www.uni-mannheim.de/media/Einrichtungen/dws/pi1/kge_datasets/wnrr.tar.gz
+ curl -O http:https://web.informatik.uni-mannheim.de/pi1/kge-datasets/wnrr.tar.gz
  tar xvf wnrr.tar.gz
 else
  echo wnrr already present
@@ -116,7 +116,7 @@ fi
 if [ ! -d "$BASEDIR/dbpedia50" ]; then
  echo Downloading dbpedia50
  cd $BASEDIR
- curl -O https:https://www.uni-mannheim.de/media/Einrichtungen/dws/pi1/kge_datasets/dbpedia50.tar.gz
+ curl -O http:https://web.informatik.uni-mannheim.de/pi1/kge-datasets/dbpedia50.tar.gz
  tar xvf dbpedia50.tar.gz
 else
  echo dbpedia50 already present
@@ -131,7 +131,7 @@ fi
 if [ ! -d "$BASEDIR/dbpedia500" ]; then
  echo Downloading dbpedia500
  cd $BASEDIR
- curl -O https:https://www.uni-mannheim.de/media/Einrichtungen/dws/pi1/kge_datasets/dbpedia500.tar.gz
+ curl -O http:https://web.informatik.uni-mannheim.de/pi1/kge-datasets/dbpedia500.tar.gz
  tar xvf dbpedia500.tar.gz
 else
  echo dbpedia500 already present
@@ -146,7 +146,7 @@ fi
 if [ ! -d "$BASEDIR/db100k" ]; then
  echo Downloading db100k
  cd $BASEDIR
- curl -O https:https://www.uni-mannheim.de/media/Einrichtungen/dws/pi1/kge_datasets/db100k.tar.gz
+ curl -O http:https://web.informatik.uni-mannheim.de/pi1/kge-datasets/db100k.tar.gz
  tar xvf db100k.tar.gz
  cd db100k
  case "$(uname -s)" in
@@ -175,7 +175,7 @@ fi
 if [ ! -d "$BASEDIR/yago3-10" ]; then
  echo Downloading yago3-10
  cd $BASEDIR
- curl -O https:https://www.uni-mannheim.de/media/Einrichtungen/dws/pi1/kge_datasets/yago3-10.tar.gz
+ curl -O http:https://web.informatik.uni-mannheim.de/pi1/kge-datasets/yago3-10.tar.gz
  tar xvf yago3-10.tar.gz
 else
  echo yago3-10 already present
@@ -190,7 +190,7 @@ fi
 if [ ! -d "$BASEDIR/wikidata5m" ]; then
  echo Downloading wikidata5m
  cd $BASEDIR
- curl -O https:https://www.uni-mannheim.de/media/Einrichtungen/dws/pi1/kge_datasets/wikidata5m.tar.gz
+ curl -O http:https://web.informatik.uni-mannheim.de/pi1/kge-datasets/wikidata5m.tar.gz
  tar xvf wikidata5m.tar.gz
 else
  echo wikidata5m already present
@@ -201,3 +201,47 @@ else
  echo wikidata5m already prepared
 fi
 
+# kinship
+if [ ! -d "$BASEDIR/kinship" ]; then
+ echo Downloading kinship
+ cd $BASEDIR
+ curl -O https://web.informatik.uni-mannheim.de/pi1/kge-datasets/kinship.tar.gz
+ tar xvf kinship.tar.gz
+else
+ echo kinship already present
+fi
+if [ ! -f "$BASEDIR/kinship/dataset.yaml" ]; then
+ python preprocess.py kinship
+else
+ echo kinship already prepared
+fi
+
+# nations
+if [ ! -d "$BASEDIR/nations" ]; then
+ echo Downloading nations
+ cd $BASEDIR
+ curl -O https://web.informatik.uni-mannheim.de/pi1/kge-datasets/nations.tar.gz
+ tar xvf nations.tar.gz
+else
+ echo nations already present
+fi
+if [ ! -f "$BASEDIR/nations/dataset.yaml" ]; then
+ python preprocess.py nations
+else
+ echo nations already prepared
+fi
+
+# umls
+if [ ! -d "$BASEDIR/umls" ]; then
+ echo Downloading umls
+ cd $BASEDIR
+ curl -O https://web.informatik.uni-mannheim.de/pi1/kge-datasets/umls.tar.gz
+ tar xvf umls.tar.gz
+else
+ echo umls already present
+fi
+if [ ! -f "$BASEDIR/umls/dataset.yaml" ]; then
+ python preprocess.py umls
+else
+ echo umls already prepared
+fi
diff --git a/docs/examples/expanded-config-example.yaml b/docs/examples/expanded-config-example.yaml
@@ -1,11 +1,15 @@
 KvsAll:
  label_smoothing: 0.0
+ query_types:
+ _po: true
+ s_o: false
+ sp_: true
 ax_search:
- fixed_parameters: []
  num_sobol_trials: -1
  num_trials: 10
  parameter_constraints: []
  parameters: []
+ sobol_seed: 0
 complex:
  class_name: ComplEx
  entity_embedder:
@@ -39,13 +43,16 @@ dataset:
  valid:
  filename: valid.del
  type: triples
- name: fb15k-237
+ name: toy
  num_entities: -1
  num_relations: -1
-eval:
- batch_size: 100
+ pickle: true
+entity_ranking:
  chunk_size: -1
- data: valid
+ filter_splits:
+ - train
+ - valid
+ filter_with_test: true
  hits_at_k_s:
  - 1
  - 3
@@ -61,9 +68,13 @@ eval:
  argument_frequency: false
  head_and_tail: false
  relation_type: false
+ tie_handling: rounded_mean_rank
+eval:
+ batch_size: 100
  num_workers: 0
  pin_memory: false
- trace_level: example
+ split: valid
+ trace_level: epoch
  type: entity_ranking
 grid_search:
  parameters:
@@ -80,12 +91,16 @@ lookup_embedder:
  initialize: normal_
  initialize_args:
  +++: +++
+ normal_:
+ mean: 0.0
+ std: 0.1
  normalize:
  p: -1.0
  with_grad: false
  regularize: lp
  regularize_args:
  +++: +++
+ p: 2
  weighted: false
  regularize_weight: 8.0e-08
  round_dim_to: []
@@ -95,16 +110,23 @@ manual_search:
  run: true
 model: complex
 negative_sampling:
+ chunk_size: -1
  filtering:
+ implementation: fast_if_available
  o: false
  p: false
  s: false
- implementation: spo
+ split: ''
+ frequency:
+ smoothing: 1
+ implementation: triple
  num_samples:
  o: -1
  p: 0
  s: 3
  sampling_type: uniform
+ shared: false
+ with_replacement: true
 random_seed:
  numpy: -1
  python: -1
@@ -115,23 +137,27 @@ search:
  on_error: abort
  type: ax
 train:
+ abort_on_nan: true
  auto_correct: false
  batch_size: 100
  checkpoint:
  every: 5
  keep: 3
- loss: bce
- loss_arg: 1.0
- lr_scheduler: ConstantLRScheduler
+ loss: kl
+ loss_arg: .nan
+ lr_scheduler: ReduceLROnPlateau
  lr_scheduler_args:
  +++: +++
+ mode: max
+ patience: 4
  max_epochs: 20
  num_workers: 0
  optimizer: Adagrad
  optimizer_args:
  +++: +++
  lr: 0.2
  pin_memory: false
+ split: train
  trace_level: epoch
  type: KvsAll
  visualize_graph: false
@@ -144,7 +170,8 @@ valid:
  metric_value: 0.0
  patience: 5
  every: 5
- filter_with_test: false
- metric: mean_reciprocal_rank_filtered
+ metric: mean_reciprocal_rank_filtered_with_test
  metric_expr: float("nan")
+ split: valid
  trace_level: epoch
+verbose: true