From 650be26d9637f37f25aec37ff4bd73472347e49d Mon Sep 17 00:00:00 2001
From: Franck Dary <franck.dary@lis-lab.fr>
Date: Wed, 8 Apr 2020 22:56:31 +0200
Subject: [PATCH] updated machines
---
UD_any/morpho/machine.rm | 25 ++++++++++++++++++++++++-
UD_any/parser/machine.rm | 25 ++++++++++++++++++++++++-
UD_any/tagger/machine.rm | 25 ++++++++++++++++++++++++-
UD_any/tokemorpho_incr/machine.rm | 25 ++++++++++++++++++++++++-
UD_any/tokemorpho_seq/machine.rm | 25 ++++++++++++++++++++++++-
UD_any/tokenizer/machine.rm | 25 ++++++++++++++++++++++++-
6 files changed, 144 insertions(+), 6 deletions(-)
diff --git a/UD_any/morpho/machine.rm b/UD_any/morpho/machine.rm
index 61f4a84..19000fa 100644
--- a/UD_any/morpho/machine.rm
+++ b/UD_any/morpho/machine.rm
@@ -1,5 +1,28 @@
Name : Feats Machine
-Classifier : tokenizer LSTM(-1,{-3,-2,-1,0,1,2},{},{FORM,UPOS},{-1,0},{},{ID,FORM,FEATS},{1,10,10},-1,-1) {data/morpho_parts.ts}
+Classifier : morpho
+{
+ Transitions : {data/morpho_parts.ts}
+ Network type : LSTM
+ Unknown value threshold : -1
+ Buffer context : {-3 -2 -1 0 1 2}
+ Stack context : {}
+ Columns : {FORM UPOS}
+ Focused buffer : {-1 0}
+ Focused stack : {}
+ Focused columns : {ID FORM FEATS}
+ Max nb elements : {1 10 10}
+ Raw input left window : -1
+ Raw input right window : -1
+ Embeddings size : 256
+ Hidden size : 8192
+ Context LSTM size : 1024
+ Focused LSTM size : 256
+ Rawinput LSTM size : 32
+ Split trans LSTM size : 256
+ Num layers : 2
+ BiLSTM : true
+ LSTM dropout : 0.3
+}
Predictions : FEATS
Strategy : sequential
morpho morpho NOTHING 1
diff --git a/UD_any/parser/machine.rm b/UD_any/parser/machine.rm
index 5d8e94c..a28b2ee 100644
--- a/UD_any/parser/machine.rm
+++ b/UD_any/parser/machine.rm
@@ -1,5 +1,28 @@
Name : Parser Machine
-Classifier : tagger RLT(15,15,4) data/parser.ts
+Classifier : parser
+{
+ Transitions : {data/parser.ts}
+ Network type : LSTM
+ Unknown value threshold : -1
+ Buffer context : {-3 -2 -1 0 1 2}
+ Stack context : {}
+ Columns : {FORM UPOS}
+ Focused buffer : {-1 0}
+ Focused stack : {-1 0}
+ Focused columns : {ID FORM FEATS}
+ Max nb elements : {1 10 10}
+ Raw input left window : -1
+ Raw input right window : -1
+ Embeddings size : 256
+ Hidden size : 8192
+ Context LSTM size : 1024
+ Focused LSTM size : 256
+ Rawinput LSTM size : 32
+ Split trans LSTM size : 256
+ Num layers : 2
+ BiLSTM : true
+ LSTM dropout : 0.3
+}
Predictions : HEAD DEPREL EOS
Strategy : sequential
parser parser SHIFT 1
diff --git a/UD_any/tagger/machine.rm b/UD_any/tagger/machine.rm
index ac8ee82..389942e 100644
--- a/UD_any/tagger/machine.rm
+++ b/UD_any/tagger/machine.rm
@@ -1,5 +1,28 @@
Name : Tagger Machine
-Classifier : tagger LSTM(-1,{-3,-2,-1,0,1,2},{},{FORM,UPOS},{-1,0},{},{ID,FORM},{1,10},-1,-1) {data/tagger.ts}
+Classifier : tagger
+{
+ Transitions : {data/tagger.ts}
+ Network type : LSTM
+ Unknown value threshold : -1
+ Buffer context : {-3 -2 -1 0 1 2}
+ Stack context : {}
+ Columns : {FORM UPOS}
+ Focused buffer : {-1 0}
+ Focused stack : {}
+ Focused columns : {ID FORM}
+ Max nb elements : {1 10}
+ Raw input left window : -1
+ Raw input right window : -1
+ Embeddings size : 256
+ Hidden size : 8192
+ Context LSTM size : 1024
+ Focused LSTM size : 256
+ Rawinput LSTM size : 32
+ Split trans LSTM size : 256
+ Num layers : 2
+ BiLSTM : true
+ LSTM dropout : 0.3
+}
Predictions : UPOS
Strategy : sequential
tagger tagger 1
diff --git a/UD_any/tokemorpho_incr/machine.rm b/UD_any/tokemorpho_incr/machine.rm
index 45e91a4..0c9e187 100644
--- a/UD_any/tokemorpho_incr/machine.rm
+++ b/UD_any/tokemorpho_incr/machine.rm
@@ -1,5 +1,28 @@
Name : Tokenizer, Tagger and Morpho Machine
-Classifier : toketagger LSTM(-1,{-3,-2,-1},{},{FORM,UPOS},{-1,0},{},{ID,FORM,FEATS},{1,10,10},5,5) {data/tokenizer.ts data/tagger.ts data/morpho_parts.ts}
+Classifier : tokemorpho
+{
+ Transitions : {data/tokenizer.ts data/tagger.ts data/morpho_parts}
+ Network type : LSTM
+ Unknown value threshold : -1
+ Buffer context : {-3 -2 -1}
+ Stack context : {}
+ Columns : {FORM UPOS}
+ Focused buffer : {-1 0}
+ Focused stack : {}
+ Focused columns : {ID FORM FEATS}
+ Max nb elements : {1 10 10}
+ Raw input left window : 5
+ Raw input right window : 5
+ Embeddings size : 256
+ Hidden size : 8192
+ Context LSTM size : 1024
+ Focused LSTM size : 256
+ Rawinput LSTM size : 32
+ Split trans LSTM size : 256
+ Num layers : 2
+ BiLSTM : true
+ LSTM dropout : 0.3
+}
Splitwords : data/splitwords.ts
Predictions : ID FORM UPOS FEATS EOS
Strategy : incremental
diff --git a/UD_any/tokemorpho_seq/machine.rm b/UD_any/tokemorpho_seq/machine.rm
index 654d519..13715ac 100644
--- a/UD_any/tokemorpho_seq/machine.rm
+++ b/UD_any/tokemorpho_seq/machine.rm
@@ -1,5 +1,28 @@
Name : Tokenizer, Tagger and Morpho Machine
-Classifier : toketagger LSTM(-1,{-3,-2,-1,1,2},{},{FORM,UPOS},{-1,0},{},{ID,FORM,FEATS},{1,10,10},5,5) {data/tokenizer.ts data/tagger.ts data/morpho_parts.ts}
+Classifier : tokemorpho
+{
+ Transitions : {data/tokenizer.ts data/tagger.ts data/morpho_parts}
+ Network type : LSTM
+ Unknown value threshold : -1
+ Buffer context : {-3 -2 -1 1 2}
+ Stack context : {}
+ Columns : {FORM UPOS}
+ Focused buffer : {-1 0}
+ Focused stack : {}
+ Focused columns : {ID FORM FEATS}
+ Max nb elements : {1 10 10}
+ Raw input left window : 5
+ Raw input right window : 5
+ Embeddings size : 256
+ Hidden size : 8192
+ Context LSTM size : 1024
+ Focused LSTM size : 256
+ Rawinput LSTM size : 32
+ Split trans LSTM size : 256
+ Num layers : 2
+ BiLSTM : true
+ LSTM dropout : 0.3
+}
Splitwords : data/splitwords.ts
Predictions : ID FORM UPOS FEATS EOS
Strategy : sequential
diff --git a/UD_any/tokenizer/machine.rm b/UD_any/tokenizer/machine.rm
index 91c0331..afadef5 100644
--- a/UD_any/tokenizer/machine.rm
+++ b/UD_any/tokenizer/machine.rm
@@ -1,5 +1,28 @@
Name : Tokenizer Machine
-Classifier : tokenizer LSTM(-1,{-3,-2,-1},{},{FORM},{-1,0},{},{ID,FORM},{1,10},5,5) {data/tokenizer.ts}
+Classifier : tokenizer
+{
+ Transitions : {data/tokenizer.ts}
+ Network type : LSTM
+ Unknown value threshold : -1
+ Buffer context : {-3 -2 -1}
+ Stack context : {}
+ Columns : {FORM}
+ Focused buffer : {-1 0}
+ Focused stack : {}
+ Focused columns : {ID FORM}
+ Max nb elements : {1 10}
+ Raw input left window : 5
+ Raw input right window : 5
+ Embeddings size : 256
+ Hidden size : 8192
+ Context LSTM size : 1024
+ Focused LSTM size : 256
+ Rawinput LSTM size : 32
+ Split trans LSTM size : 256
+ Num layers : 2
+ BiLSTM : true
+ LSTM dropout : 0.3
+}
Splitwords : data/splitwords.ts
Predictions : ID FORM EOS
Strategy : sequential
--
GitLab