new tasks

Author: Damien Sileo

src/tasksource/mtasks.py

@@ -103,9 +103,6 @@ def concatenate_configs(dataset):
 xstory = MultipleChoice(constant(''),choices=["text_right_ending","text_wrong_ending"],labels=constant(0),
     **all("juletxara/xstory_cloze"))
 
-anthropic_rlhf = MultipleChoice(constant(''), ['chosen','rejected'], constant(0),
-    dataset_name="Anthropic/hh-rlhf")
-
 xglue_ner = TokenClassification("words","ner", dataset_name="xglue",config_name="ner")
 xglue_pos = TokenClassification("words","pos", dataset_name="xglue",config_name="pos")
 

src/tasksource/tasks.py

@@ -984,4 +984,9 @@ def _udep_post_process(ds):
 sharc_classification = Classification("snippet", lambda x:f'{x["scenario"]}\n{x["question"]}' ,labels="answer",dataset_name='sharc_modified',config_name='mod',
     pre_process = lambda ds:ds.filter(lambda x:x['answer'] in {"Yes","No","Irrelevant"}))
 
-conceptrules_v2 = Classification("context", "text", "label", dataset_name="tasksource/conceptrules_v2")
+conceptrules_v2 = Classification("context", "text", "label", dataset_name="tasksource/conceptrules_v2")
+
+scidtb = Classification("unit1_txt","unit2_txt","label", dataset_name="metaeval/disrpt",config_name='eng.dep.scidtb')
+
+chunking = TokenClassification("tokens","chunk_tags", dataset_name="conll2000")
+few_nerd=TokenClassification("tokens","fine_ner_tags",dataset_name="DFKI-SLT/few-nerd",config_name='supervised')