How should token_max be configured in mapreduce chain?

[vr140]

See:
https://github.com/hwchase17/langchain/blob/master/langchain/chains/combine_documents/map_reduce.py

with default token_max: int = 3000

1. Am I correct in that if after mapreduce, the response is > 3000 tokens, the response is collapsed further using the collapse_llm or collapse_prompt?

2. Curious why default is set to 3000?

3. How do we override this when using mapreduce chains for say Q&A?

[ShantanuNair]

Like so :)

res = await chain(inputs={'input_documents': texts, 'token_max': 12000}, return_only_outputs=True)