Can we use this code for CPU?

songmzhang / DSKD

Repo for Paper "Dual-Space Knowledge Distillation for Large Language Models".

29 stars 3 forks source link

Can we use this code for CPU? #6

Closed survivebycoding closed 1 month ago

survivebycoding commented 1 month ago

Can we change this code GPUS=(0) export CUDA_VISIBLE_DEVICES=$(IFS=,; echo "${GPUS[*]}")

to CUDA_VISIBLE_DEVICES="" to use CPUs?

songmzhang commented 1 month ago

It may not directly support CPU since we implement the training process with DDP. You may need to modify the distillation.py by yourself to train models on the CPU.