Migrate model card from transformers-repo

Read announcement at https://discuss.huggingface.co/t/announcement-all-model-cards-will-be-migrated-to-hf-co-model-repos/2755
Original file history: https://github.com/huggingface/transformers/commits/master/model_cards/NeuML/bert-small-cord19-squad2/README.md

README.md

@@ -0,0 +1,26 @@
+# BERT-Small CORD-19 fine-tuned on SQuAD 2.0
+
+[bert-small-cord19 model](https://huggingface.co/NeuML/bert-small-cord19) fine-tuned on SQuAD 2.0
+
+## Building the model
+
+```bash
+python run_squad.py
+    --model_type bert
+    --model_name_or_path bert-small-cord19
+    --do_train
+    --do_eval
+    --do_lower_case
+    --version_2_with_negative
+    --train_file train-v2.0.json
+    --predict_file dev-v2.0.json
+    --per_gpu_train_batch_size 8
+    --learning_rate 3e-5
+    --num_train_epochs 3.0
+    --max_seq_length 384
+    --doc_stride 128
+    --output_dir bert-small-cord19-squad2
+    --save_steps 0
+    --threads 8
+    --overwrite_cache
+    --overwrite_output_dir