Update README.md

README.md

@@ -1,8 +1,16 @@
-model: `facebook/bart-large`
-max position embeddings: 2048
+```
+{
+  model: facebook/bart-large,
+  max_position_embeddings: 2048,
+  learning_rate: 3e-4,
+  no_repeat_ngram_size: 0
+  num_steps: 520000
+}
+```
 dataset: RISC -> ARM cloze training data
-LR: 3e-4, 520k steps
 
+
+  
 Inference:
-beam size 20, top-100, gets 34 / 45 of the Project Euler test set.
-(ensure that `no_repeat_ngram_size`=0)
+
+beam size 20, top-100, gets 34 / 45 of the Project Euler test set.