More README

README.md

@@ -39,6 +39,29 @@ generation that supports 18 programming languages. It takes the OpenAI
 translate them  to other languages. It is easy to add support for new languages 
 and benchmarks.
 
+## Subsets
+
+For most purposes, you should use the variations called *SRCDATA-LANG*, where
+*SRCDATA* is either "humaneval" or "mbpp" and *LANG* is one of the supported
+languages. We use the canonical file extension for each language to identify
+the language, e.g., "py" for Python, "cpp" for C++, "lua" for Lua, and so on.
+
+We also provide a few other variations:
+
+- *SRCDATA-LANG-keep* is the same as *SRCDATA-LANG*, but the text of the prompt
+  is totally unchanged. If the original prompt had Python doctests, they remain
+  as Python instead of being translated to *LANG*. If the original prompt had 
+  Python-specific terminology, e.g., "list", it remains "list", instead of 
+  being translated, e.g., to "vector" for C++.
+
+- *SRCDATA-LANG-transform* transforms the doctests to *LANG* but leaves
+  the natural language text of the prompt unchanged.
+
+- *SRCDATA-LANG-removed* removes the doctests from the prompt.
+
+Note that MBPP does not have any doctests, so the "removed" and "transform"
+variations are not available for MBPP.
+
 ## Example
 
 The following script uses the Salesforce/codegen model to generate Lua
@@ -72,7 +95,7 @@ for problem in problems["test"]:
         return_tensors="pt",
     ).input_ids.cuda()
     generated_ids = model.generate(
-        input_ids, max_length=256, pad_token_id=tokenizer.eos_token_id + 2
+        input_ids, max_length=512, pad_token_id=tokenizer.eos_token_id + 2
     )
     truncated_string = stop_at_stop_token(tokenizer.decode(generated_ids[0]), problem)
     filename = problem["name"] + "." + LANG