diff --git a/UD_any/tokeparser_base/machine.rm b/UD_any/tokeparser_base/machine.rm
index 9560d0bf5cdf163a26207f29cf89dffc0572eced..39ec5b5d0b3bf47acd3e0ad88f659447adda6b01 100644
--- a/UD_any/tokeparser_base/machine.rm
+++ b/UD_any/tokeparser_base/machine.rm
@@ -1,36 +1,22 @@
Name : Tokenizer, Tagger, Morpho and Parser Machine
Classifier : tokeparser
{
- Transitions : {tokenizer,data/tokenizer.ts tagger,data/tagger.ts morpho,data/morpho_parts.ts parser,data/parser.ts}
- Network type : LSTM
- Unknown value threshold : 1
- Buffer context : {-3 -2 -1}
- Stack context : {2 1 0}
- Columns : {FORM UPOS}
- Focused buffer : {-1 0}
- Focused stack : {2 1 0}
- Focused columns : {ID EOS FORM FEATS DEPREL}
- Max nb elements : {1 1 10 10 1}
- Raw input left window : 5
- Raw input right window : 5
- Embeddings size : 128
- MLP : {2048 0.3}
- Context LSTM size : 512
- Focused LSTM size : 256
- Rawinput LSTM size : 32
- Split trans LSTM size : 256
- Num layers : 3
- BiLSTM : true
- LSTM dropout : 0.1
- Total input dropout : 0.3
- Embeddings dropout : 0.3
- Dropout 2d : false
- Tree embedding columns : {DEPREL}
- Tree embedding buffer : {-1}
- Tree embedding stack : {0 1 2}
- Tree embedding nb : {6}
- Tree embedding size : 128
- Optimizer : Adam {0.0005 0.9 0.999 0.00000001 0.00001 true}
+ Transitions : {tokenizer,data/tokenizer.ts tagger,data/tagger.ts morpho,data/morpho_parts.ts parser,data/parser.ts}
+ Network type : Modular
+ StateName : States{tokenizer tagger morpho parser} Out{128}
+ Context : Buffer{-3 -2 -1} Stack{2 1 0} Columns{FORM UPOS} GRU{1 1 0.0 1} In{128} Out{128}
+ Focused : Column{ID} NbElem{1} Buffer{0} Stack{0} GRU{1 1 0.0 1} In{128} Out{128}
+ Focused : Column{EOS} NbElem{1} Buffer{-1} Stack{} GRU{1 1 0.0 1} In{128} Out{128}
+ Focused : Column{FORM} NbElem{10} Buffer{-1 0} Stack{2 1 0} GRU{1 1 0.0 1} In{128} Out{128}
+ Focused : Column{FEATS} NbElem{10} Buffer{-1 0} Stack{2 1 0} GRU{1 1 0.0 1} In{128} Out{128}
+ Focused : Column{DEPREL} NbElem{1} Buffer{} Stack{2 1 0} GRU{1 1 0.0 1} In{128} Out{128}
+ RawInput : Left{5} Right{5} GRU{1 1 0.0 1} In{64} Out{64}
+ SplitTrans : GRU{1 1 0.0 1} In{128} Out{128}
+ DepthLayerTree : Columns{DEPREL} Buffer{} Stack{2 1 0} LayerSizes{3} GRU{1 1 0.0 1} In{128} Out{128}
+ InputDropout : 0.4
+ MLP : {2048 0.3 512 0.1}
+ End
+ Optimizer : Adam {0.0005 0.9 0.999 0.00000001 0.00001 true}
}
Splitwords : data/splitwords.ts
Predictions : ID FORM UPOS FEATS HEAD DEPREL EOS
diff --git a/UD_any/tokeparser_incr/machine.rm b/UD_any/tokeparser_incr/machine.rm
index fa36bfe1ea8359dade6ddf473483c5dcaf4f365f..4405538c6b8432f34fbd8aaa0d213715ef1f7383 100644
--- a/UD_any/tokeparser_incr/machine.rm
+++ b/UD_any/tokeparser_incr/machine.rm
@@ -1,36 +1,22 @@
Name : Tokenizer, Tagger, Morpho and Parser Machine
Classifier : tokeparser
{
- Transitions : {tokenizer,data/tokenizer.ts tagger,data/tagger.ts morpho,data/morpho_parts.ts parser,data/parser.ts}
- Network type : LSTM
- Unknown value threshold : 1
- Buffer context : {-3 -2 -1}
- Stack context : {2 1 0}
- Columns : {FORM UPOS}
- Focused buffer : {-1 0}
- Focused stack : {2 1 0}
- Focused columns : {ID EOS FORM FEATS DEPREL}
- Max nb elements : {1 1 10 10 1}
- Raw input left window : 5
- Raw input right window : 5
- Embeddings size : 128
- MLP : {2048 0.3}
- Context LSTM size : 512
- Focused LSTM size : 256
- Rawinput LSTM size : 32
- Split trans LSTM size : 256
- Num layers : 3
- BiLSTM : true
- LSTM dropout : 0.1
- Total input dropout : 0.3
- Embeddings dropout : 0.3
- Dropout 2d : false
- Tree embedding columns : {DEPREL}
- Tree embedding buffer : {-1}
- Tree embedding stack : {0 1 2}
- Tree embedding nb : {6}
- Tree embedding size : 128
- Optimizer : Adam {0.0005 0.9 0.999 0.00000001 0.00001 true}
+ Transitions : {tokenizer,data/tokenizer.ts tagger,data/tagger.ts morpho,data/morpho_parts.ts parser,data/parser.ts}
+ Network type : Modular
+ StateName : States{tokenizer tagger morpho parser} Out{128}
+ Context : Buffer{-3 -2 -1} Stack{2 1 0} Columns{FORM UPOS} GRU{1 1 0.0 1} In{128} Out{128}
+ Focused : Column{ID} NbElem{1} Buffer{0} Stack{0} GRU{1 1 0.0 1} In{128} Out{128}
+ Focused : Column{EOS} NbElem{1} Buffer{-1} Stack{} GRU{1 1 0.0 1} In{128} Out{128}
+ Focused : Column{FORM} NbElem{10} Buffer{-1 0} Stack{2 1 0} GRU{1 1 0.0 1} In{128} Out{128}
+ Focused : Column{FEATS} NbElem{10} Buffer{-1 0} Stack{2 1 0} GRU{1 1 0.0 1} In{128} Out{128}
+ Focused : Column{DEPREL} NbElem{1} Buffer{} Stack{2 1 0} GRU{1 1 0.0 1} In{128} Out{128}
+ RawInput : Left{5} Right{5} GRU{1 1 0.0 1} In{64} Out{64}
+ SplitTrans : GRU{1 1 0.0 1} In{128} Out{128}
+ DepthLayerTree : Columns{DEPREL} Buffer{} Stack{2 1 0} LayerSizes{3} GRU{1 1 0.0 1} In{128} Out{128}
+ InputDropout : 0.4
+ MLP : {2048 0.3 512 0.1}
+ End
+ Optimizer : Adam {0.0005 0.9 0.999 0.00000001 0.00001 true}
}
Splitwords : data/splitwords.ts
Predictions : ID FORM UPOS FEATS HEAD DEPREL EOS
diff --git a/UD_any/tokeparser_seq/machine.rm b/UD_any/tokeparser_seq/machine.rm
index a79fb52144488c19ebfd66840e90261bfc17a237..a7a9bea134856be7ba3e285eeae50dc779b32a98 100644
--- a/UD_any/tokeparser_seq/machine.rm
+++ b/UD_any/tokeparser_seq/machine.rm
@@ -1,36 +1,22 @@
Name : Tokenizer, Tagger, Morpho and Parser Machine
Classifier : tokeparser
{
- Transitions : {tokenizer,data/tokenizer.ts tagger,data/tagger.ts morpho,data/morpho_parts.ts parser,data/parser.ts}
- Network type : LSTM
- Unknown value threshold : 1
- Buffer context : {-3 -2 -1 1 2 3}
- Stack context : {2 1 0}
- Columns : {FORM UPOS}
- Focused buffer : {-1 0 1 2}
- Focused stack : {2 1 0}
- Focused columns : {ID EOS FORM FEATS DEPREL}
- Max nb elements : {1 1 10 10 1}
- Raw input left window : 5
- Raw input right window : 5
- Embeddings size : 128
- MLP : {2048 0.3}
- Context LSTM size : 512
- Focused LSTM size : 256
- Rawinput LSTM size : 32
- Split trans LSTM size : 256
- Num layers : 3
- BiLSTM : true
- LSTM dropout : 0.1
- Total input dropout : 0.3
- Embeddings dropout : 0.3
- Dropout 2d : false
- Tree embedding columns : {DEPREL}
- Tree embedding buffer : {-1}
- Tree embedding stack : {0 1 2}
- Tree embedding nb : {6}
- Tree embedding size : 128
- Optimizer : Adam {0.0005 0.9 0.999 0.00000001 0.00001 true}
+ Transitions : {tokenizer,data/tokenizer.ts tagger,data/tagger.ts morpho,data/morpho_parts.ts parser,data/parser.ts}
+ Network type : Modular
+ StateName : States{tokenizer tagger morpho parser} Out{128}
+ Context : Buffer{-3 -2 -1 1 2} Stack{2 1 0} Columns{FORM UPOS} GRU{1 1 0.0 1} In{128} Out{128}
+ Focused : Column{ID} NbElem{1} Buffer{0} Stack{0} GRU{1 1 0.0 1} In{128} Out{128}
+ Focused : Column{EOS} NbElem{1} Buffer{-1} Stack{} GRU{1 1 0.0 1} In{128} Out{128}
+ Focused : Column{FORM} NbElem{10} Buffer{-1 0 1} Stack{2 1 0} GRU{1 1 0.0 1} In{128} Out{128}
+ Focused : Column{FEATS} NbElem{10} Buffer{-1 0 1} Stack{2 1 0} GRU{1 1 0.0 1} In{128} Out{128}
+ Focused : Column{DEPREL} NbElem{1} Buffer{} Stack{2 1 0} GRU{1 1 0.0 1} In{128} Out{128}
+ RawInput : Left{5} Right{5} GRU{1 1 0.0 1} In{64} Out{64}
+ SplitTrans : GRU{1 1 0.0 1} In{128} Out{128}
+ DepthLayerTree : Columns{DEPREL} Buffer{} Stack{2 1 0} LayerSizes{3} GRU{1 1 0.0 1} In{128} Out{128}
+ InputDropout : 0.4
+ MLP : {2048 0.3 512 0.1}
+ End
+ Optimizer : Adam {0.0005 0.9 0.999 0.00000001 0.00001 true}
}
Splitwords : data/splitwords.ts
Predictions : ID FORM UPOS FEATS HEAD DEPREL EOS