Fix typos (#1143)

Found via `codespell -S .mypy_cache,yarn.lock,*.json,*.ipynb -L
rouge,nam,vie`

model/reward/instructor/TODO.md

@@ -1,6 +1,6 @@
 Some other reward features we can use
 
-0. Finish classifcation feature
+0. Finish classification feature
 
 1. Summaries from human feedback
 

model/reward/instructor/rank_datasets.py

@@ -14,7 +14,7 @@
     [] support additional negative samples generated from other models.
 
         For example we can use galactica-125m to generate a TLDR and assume it was
-        inferior than the human perference one
+        inferior than the human preference one
 
 
 """