fengji.zhang

City University of Hong Kong Hong Kong, China PhD student at CityU. Interest includes Code Generation, Language Models. (avatar is created using DALLE2 using the prompt "Beyond The Horizon"

Results 2 comments of


                                            fengji.zhang

Absent new_lines and indentation in python data

Same question. linebreakers and indentation are really important to rebuild the syntax tree.

[Bug]: Qwen/Qwen2.5-1.5B-Instruct generates out of vocabulary tokens

Same issue, setting top-p to 0.99 still cannot totally avoid this problem