From af4d201a7eeb3c6a0b51b71f040851d2a68c9f3e Mon Sep 17 00:00:00 2001
From: Jonas Ehrenstein <jonas@seekwhens.com>
Date: Mon, 5 Jul 2021 12:06:19 +0200
Subject: [PATCH 1/2] Fix NER script:

'ALL_MODELS' is not needed and 'pretrained_config_archive_map' is deprecated.

Error in predictions for label-ends leads to labels snaking the next one without accounting for 'O'-labels between them.
---
 models/ner.py | 22 +++++++++++++++++-----
 1 file changed, 17 insertions(+), 5 deletions(-)

diff --git a/models/ner.py b/models/ner.py
index 36f9c3f..801a9f9 100644
--- a/models/ner.py
+++ b/models/ner.py
@@ -29,10 +29,10 @@
 
 logger = logging.getLogger(__name__)
 
-
-ALL_MODELS = sum(
-    [list(conf.pretrained_config_archive_map.keys()) for conf in (BertConfig, RobertaConfig, DistilBertConfig)],
-    [])
+# 'pretrained_config_archive_map' is deprecated and variable not actually in use
+# ALL_MODELS = sum(
+#     [list(conf.pretrained_config_archive_map.keys()) for conf in (BertConfig, RobertaConfig, DistilBertConfig)],
+#     [])
 
 MODEL_CLASSES = {
     'bert': (BertConfig, BertForTokenClassification, BertTokenizer),
@@ -407,14 +407,23 @@ def predict(self, tasks, **kwargs):
 
                 result = []
 
+                # bool for when 'O'-label is not appended in 'result'
+                skipped = False
+
                 for label, group in groupby(zip(preds, starts, scores), key=lambda i: re.sub('^(B-|I-)', '', i[0])):
                     _, group_start, _ = list(group)[0]
                     if len(result) > 0:
                         if group_start == 0:
                             result.pop(-1)
-                        else:
+                        # when 'O' is skipped when appending 'result', 'end' of previous 'result' sould not be changed, until new label is appended
+                        elif not skipped:
                             result[-1]['value']['end'] = group_start - 1
+                            # remove incorrect predictions, where 'end is smaller than 'start'
+                            if result[-1]['value']['end'] is not None and result[-1]['value']['end'] < result[-1]['value']['start']:
+                                result.pop(-1)
+
                     if label != 'O':
+                        skipped = False
                         result.append({
                             'from_name': from_name,
                             'to_name': to_name,
@@ -426,6 +435,9 @@ def predict(self, tasks, **kwargs):
                                 'text': '...'
                             }
                         })
+                    else:
+                        skipped = True
+
                 if result and result[-1]['value']['end'] is None:
                     result[-1]['value']['end'] = len(string)
                 results.append({

From 0a80880a4586720f7288ca6506b13e02373480d7 Mon Sep 17 00:00:00 2001
From: Jonas Ehrenstein <jonas@seekwhens.com>
Date: Tue, 13 Jul 2021 13:29:34 +0200
Subject: [PATCH 2/2] learnwhens v0.1

---
 .gitignore | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/.gitignore b/.gitignore
index 8bc1d55..f5936ee 100644
--- a/.gitignore
+++ b/.gitignore
@@ -101,3 +101,6 @@ venv.bak/
 
 # mypy
 .mypy_cache/
+
+# Models
+/predictwhens
\ No newline at end of file