Merge pull request #222 from megagonlabs/feature/update_config

update model config files

config/ja_ginza.meta.json

@@ -1,7 +1,7 @@
 {
   "lang":"ja",
   "name":"ginza",
-  "version":"5.0.0",
+  "version":"5.1.0",
   "description":"Japanese multi-task CNN trained on UD-Japanese BCCWJ r2.8 + GSK2014-A(2019). Assigns word2vec token vectors. Components: tok2vec, parser, ner, morphologizer, atteribute_ruler, compound_splitter, bunsetu_recognizer.",
   "author":"Megagon Labs Tokyo.",
   "email":"ginza@megagon.ai",
@@ -57,8 +57,8 @@
     "attribute_ruler"
   ],
   "requirements":[
-    "sudachipy>=0.5.2,<0.6.0",
-    "sudachidict_core>=20210608",
-    "ginza>=5.0.0,<5.1.0"
+    "sudachipy>=0.5.4,<0.7.0",
+    "sudachidict_core>=20210802",
+    "ginza>=5.1.0,<5.2.0"
   ]
 }

config/ja_ginza_electra.analysis.cfg

@@ -25,6 +25,7 @@ split_mode = "C"
 
 [components.attribute_ruler]
 factory = "attribute_ruler"
+scorer = {"@scorers":"spacy.attribute_ruler_scorer.v1"}
 validate = false
 
 [components.bunsetu_recognizer]
@@ -37,6 +38,9 @@ split_mode = null
 
 [components.morphologizer]
 factory = "morphologizer"
+extend = true
+overwrite = true
+scorer = {"@scorers":"spacy.morphologizer_scorer.v1"}
 
 [components.morphologizer.model]
 @architectures = "spacy.Tagger.v1"
@@ -52,6 +56,7 @@ upstream = "*"
 factory = "ner"
 incorrect_spans_key = null
 moves = null
+scorer = {"@scorers":"spacy.ner_scorer.v1"}
 update_with_oracle_cut_size = 100
 
 [components.ner.model]
@@ -74,6 +79,7 @@ factory = "parser"
 learn_tokens = false
 min_action_freq = 30
 moves = null
+scorer = {"@scorers":"spacy.parser_scorer.v1"}
 update_with_oracle_cut_size = 100
 
 [components.parser.model]
@@ -97,14 +103,17 @@ max_batch_items = 4096
 set_extra_annotations = {"@annotation_setters":"spacy-transformers.null_annotation_setter.v1"}
 
 [components.transformer.model]
-@architectures = "ginza-transformers.TransformerModel.v1"
-name = "megagonlabs/transformers-ud-japanese-electra-base-ginza"
+@architectures = "spacy-transformers.TransformerModel.v3"
+name = "megagonlabs/transformers-ud-japanese-electra-base-ginza-510"
+mixed_precision = false
 
 [components.transformer.model.get_spans]
 @span_getters = "spacy-transformers.strided_spans.v1"
 window = 128
 stride = 96
 
+[components.transformer.model.grad_scaler_config]
+
 [components.transformer.model.tokenizer_config]
 use_fast = false
 tokenizer_class = "sudachitra.tokenization_electra_sudachipy.ElectraSudachipyTokenizer"
@@ -119,6 +128,8 @@ word_form_type = "dictionary_and_surface"
 split_mode = "A"
 dict_type = "core"
 
+[components.transformer.model.transformer_config]
+
 [corpora]
 
 [corpora.dev]
@@ -149,8 +160,8 @@ max_epochs = 0
 max_steps = 50000
 eval_frequency = 200
 frozen_components = []
-before_to_disk = null
 annotating_components = []
+before_to_disk = null
 
 [training.batcher]
 @batchers = "spacy.batch_by_padded.v1"

config/ja_ginza_electra.meta.json

@@ -1,7 +1,7 @@
 {
   "lang":"ja",
   "name":"ginza_electra",
-  "version":"5.0.0",
+  "version":"5.1.0",
   "description":"Japanese multi-task CNN trained on UD-Japanese BCCWJ r2.8 + GSK2014-A(2019) + transformers-ud-japanese-electra--base. Components: transformer, parser, atteribute_ruler, ner, morphologizer, compound_splitter, bunsetu_recognizer.",
   "author":"Megagon Labs Tokyo.",
   "email":"ginza@megagon.ai",
@@ -64,10 +64,10 @@
     "attribute_ruler"
   ],
   "requirements":[
-    "sudachipy>=0.5.2,<0.6.0",
-    "sudachidict_core>=20210608",
+    "sudachipy>=0.5.4,<0.7.0",
+    "sudachidict_core>=20210802",
     "sudachitra>=0.1.5,<0.2.0",
-    "ginza-transformers>=0.3.1,<1.0.0",
-    "ginza>=5.0.0,<5.1.0"
+    "ginza-transformers>=0.4.0,<0.5.0",
+    "ginza>=5.1.0,<5.2.0"
   ]
 }

setup.py

@@ -19,7 +19,7 @@
     install_requires=[
         "spacy>=3.2.0,<3.3.0",
         "plac>=1.3.3",
-        "SudachiPy>=0.6.0,<0.7.0",
+        "SudachiPy>=0.5.4,<0.7.0",
         "SudachiDict-core>=20210802",
     ],
     setup_requires=["pytest-runner"],