The meaning of C_loss and M_loss

[9p15p]

OpenDriveLab/DriveLM#51

LLaMA-Adapter/llama_adapter_v2_multimodal7b/llama/llama_adapter.py

Line 185 in 521a09d

return c_loss, c_loss

LLaMA-Adapter/llama_adapter_v2_multimodal7b/engine_finetune.py

Line 39 in 521a09d

loss = c_loss + m_loss * 0

It seems that C_loss and M_loss are the same variable.
what exactly do C_loss and M_loss mean? and why do we distinguish between them?

Thank you.

[gaopengpjlab]

C_loss and M_loss stands for cross entropy loss and masked prediction loss, respectively. M_loss is never used in the current version of code.