Improved measurements and tables pipelines

.gitignore

@@ -53,6 +53,9 @@ _build/
 *.tar.gz
 *.tsv
 *.ann
+*.spacy
+*.RRF
+*.pkl
 
 # Editors
 .idea
@@ -71,3 +74,6 @@ _build/
 docs/reference
 docs/changelog.md
 docs/contributing.md
+
+# Logs
+*.log

NER_model/configs/config.cfg

@@ -0,0 +1,185 @@
+[paths]
+train = "corpus/train.spacy"
+dev = "corpus/dev.spacy"
+vectors = null
+init_tok2vec = null
+bert = "../../word-embedding/finetuning-camembert-2021-07-29"
+
+[system]
+gpu_allocator = "pytorch"
+seed = 0
+
+[nlp]
+lang = "eds"
+pipeline = ["tok2vec","ner","qualifier"]
+batch_size = 1
+disabled = []
+before_creation = null
+after_creation = null
+after_pipeline_creation = null
+
+[components]
+
+[components.ner]
+factory = "nested_ner"
+scorer = {"@scorers": "eds.nested_ner_scorer.v1"}
+ent_labels = ["DISO","Constantes","BIO_comp","Chemical_and_drugs","dosage","BIO","strength","form","SECTION_antecedent","SECTION_motif","SECTION_histoire","SECTION_examen_clinique","SECTION_examen_complementaire","SECTION_mode_de_vie","SECTION_traitement_entree","SECTION_antecedent_familiaux","SECTION_traitement_sortie","SECTION_conclusion"]
+
+[components.qualifier]
+factory = "eds.span_qualifier"
+
+[components.qualifier.candidate_getter]
+@misc = "eds.candidate_span_qualifier_getter"
+on_ents = true
+on_span_groups = false
+
+qualifiers = ["_.Allergie", "_.Action","_.Certainty","_.Temporality","_.Negation","_.Family"]
+label_constraints = {"_.Allergie":["Chemical_and_drugs"],"_.Action":["Chemical_and_drugs"],"_.Certainty":["Chemical_and_drugs"],"_.Temporality":["Chemical_and_drugs"],"_.Negation":["Chemical_and_drugs"],"_.Family":["Chemical_and_drugs"]}
+
+[components.qualifier.model]
+@architectures = "eds.span_multi_classifier.v1"
+projection_mode = "dot"
+pooler_mode = "max"
+
+[components.qualifier.model.tok2vec]
+@architectures = "spacy.Tok2Vec.v2"
+
+[components.qualifier.model.tok2vec.embed]
+@architectures = "spacy-transformers.TransformerListener.v1"
+pooling = {"@layers":"reduce_mean.v1"}
+
+[components.qualifier.model.tok2vec.encode]
+@architectures = "spacy.MishWindowEncoder.v2"
+width = 768
+window_size = 1
+depth = 2
+
+[components.clean-entities]
+factory = "clean-entities"
+
+[components.ner.model]
+@architectures = "eds.stack_crf_ner_model.v1"
+mode = "joint"
+
+[components.ner.model.tok2vec]
+@architectures = "spacy.Tok2Vec.v2"
+
+[components.ner.model.tok2vec.embed]
+@architectures = "spacy-transformers.TransformerListener.v1"
+pooling = {"@layers":"reduce_mean.v1"}
+
+[components.ner.model.tok2vec.encode]
+@architectures = "spacy.MishWindowEncoder.v2"
+width = 768
+window_size = 1
+depth = 2
+
+[components.tok2vec]
+factory = "transformer"
+max_batch_items = 4096
+set_extra_annotations = {"@annotation_setters":"spacy-transformers.null_annotation_setter.v1"}
+
+[components.tok2vec.model]
+@architectures = "spacy-transformers.TransformerModel.v3"
+name = ${paths.bert}
+# name = "camembert-base"
+mixed_precision = true
+
+[components.tok2vec.model.get_spans]
+@span_getters = "spacy-transformers.strided_spans.v1"
+window = 128
+stride = 64
+
+[components.tok2vec.model.grad_scaler_config]
+
+[components.tok2vec.model.tokenizer_config]
+use_fast = true
+
+[components.tok2vec.model.transformer_config]
+
+[corpora]
+
+[corpora.train]
+@readers = "eds.Corpus.v1"
+path = ${paths.train}
+max_length = 384
+gold_preproc = false
+limit = 0
+augmenter = null
+seed = ${system.seed}
+shuffle = true
+filter_expr = null
+#filter_expr = "any(ent._.event_type is not None for ent in doc.ents)"
+
+[corpora.dev]
+@readers = "eds.Corpus.v1"
+path = ${paths.dev}
+max_length = 0
+gold_preproc = false
+limit = 0
+augmenter = null
+
+[training]
+accumulate_gradient = 1
+dev_corpus = "corpora.dev"
+train_corpus = "corpora.train"
+seed = ${system.seed}
+gpu_allocator = ${system.gpu_allocator}
+dropout = 0.1
+patience = 1500
+max_epochs = -1
+max_steps = 20000
+eval_frequency = 100
+frozen_components = []
+annotating_components = []
+before_to_disk = null
+
+[training.batcher]
+@batchers = "spacy.batch_by_padded.v1"
+size = 3072
+discard_oversize = true
+buffer = 256
+
+[training.logger]
+@loggers = "spacy.ConsoleLogger.v1"
+progress_bar = true
+#@loggers = "DVCLive.v1"
+
+[training.optimizer]
+@optimizers = "Adam.v1"
+beta1 = 0.9
+beta2 = 0.999
+L2_is_weight_decay = true
+L2 = 0.01
+grad_clip = 0.0
+use_averages = false
+eps = 0.000001
+
+[training.optimizer.learn_rate]
+@schedules = "warmup_linear.v1"
+warmup_steps = 250
+total_steps = ${training.max_steps}
+initial_rate = 0.0001
+
+[training.score_weights]
+ents_f = 0.8
+ents_p = 0.0
+ents_r = 0.0
+qual_f = 0.2
+speed = 0.0
+ents_per_type = null
+qual_per_type = null
+
+[pretraining]
+
+[initialize]
+vectors = ${paths.vectors}
+init_tok2vec = ${paths.init_tok2vec}
+vocab_data = null
+lookups = null
+before_init = null
+after_init = null
+
+[initialize.components]
+
+[initialize.tokenizer]