update readme

README.md

@@ -1,12 +1,24 @@
-This is a dummy model that can be used for testing.
+This is a dummy model that can be used for testing.  It should always give random results (i.e. `{"label": "negative", "score": 0.5}`).
 
-It was created as follows in a python shell
+It was created as follows:
+
+1. Create a vocab.txt file (in /tmp/vocab.txt in this example).
+
+```
+[UNK]
+[SEP]
+[PAD]
+[CLS]
+[MASK]
+```
+
+2. Open a python shell and run the following
 
 ```python
 import transformers
 config = transformers.DistilBertConfig(vocab_size=5, n_layers=1, n_heads=1, dim=1, hidden_dim=4 * 1, num_labels=2, id2label={0: "negative", 1: "positive"}, label2id={"negative": 0, "positive": 1})
 model = transformers.DistilBertForSequenceClassification(config)
-tokenizer = transformers.DistilBertTokenizer("/tmp/empty_vocab.txt", model_max_length=512)
+tokenizer = transformers.DistilBertTokenizer("/tmp/vocab.txt", model_max_length=512)
 config.save_pretrained(".")
 model.save_pretrained(".")
 tokenizer.save_pretrained(".")