Merge pull request #1 from anishthite/master

Add coqa extraction

README.md

+# Evaluation Harness for Large Language Models
+
+### Currently based on evaluations of GPT3 as mentioned in https://arxiv.org/pdf/2005.14165.pdf
+
+## Summary (need to convert from google docs at some point):
+https://docs.google.com/document/d/177dwJpH8GHebISXYZSn4NL98sXdCtQMH82b7O5F7jmw/edit?usp=sharing

datasets/coqa/evaluateCoQA.py

+
+from ...base import Dataset
+import os
+import json
+class CoQA(Dataset):
+    def training_docs(self):
+        pass
+
+    def validation_docs(self):
+        pass
+    
+    def test_docs(self):
+        pass
+    
+    def fewshot_examples(self, k):
+        traindocs = list(self.training_docs())
+        random.seed(123)
+        random.shuffle(traindocs)
+
+        return traindocs[:k]
+    
+    def fewshot_description(self):
+        pass
+
+    def doc_to_text(self, doc, include_target=True):
+        json.load(open(doc))
+
+        
+