the prompt is too long for CodeGen model

[LionLin2003]

Hello, great work! When i try to reproduce the results of RepoCoder, there is an issue:
When i run the file codegen_inference.py using prompt file rg-one-gram-ws-20-ss-2.jsonl, it comes the error the current text generation call will exceed the model's predefined maximum length (2048).
I noticed that you limited the retrieval context with 1k for CodeGen model. But i think the original prompt in the file api_level_completion_1k_context_codegen.test.jsonl is too long, and it is directly concatenated to the additional retrieval context. So the total length of the input can be long as 1900+, leading to this error. How should i solve this problem? Thanks!