Home

Most part of the project would need to be completed by the end of March.

Questions

• Standard LoRA: what training configuration. (one LoRA adaptor for each dataset? or one LoRA adaptor for each type of task? or one LoRA adaptor for all tasks?)
• Using LoRA MoE: what training configuration.

Todo

Datasets

• Confirm XBRL extraction pipeline (what are the subtasks) and finish creating the dataset.

Finetuning

• Finetune XBRL extraction.

Benchmark

• FNXL