Closed Dlinuigh closed 4 months ago
length_map defines the maximum length of generated tokens. It depends on the dataset you are using. The max_length of input texts is 1024 (in the updated version)
May I inquire if your latest CPT version can support the summarization of input texts with a length of 1,024 characters?
in finetune/run_gen.py, defines in a dict,
but I want to process some sentences with max_length 512.