diff --git a/finetune/install-wsl-dep-and-train.sh b/finetune/install-wsl-dep-and-train.sh index 461edb3..4a2d25f 100644 --- a/finetune/install-wsl-dep-and-train.sh +++ b/finetune/install-wsl-dep-and-train.sh @@ -50,8 +50,9 @@ echo "loading $loadModel" modelInfo=$(python3 ./finetune/get_layer_and_embd.py $loadModel 5.2) echo $modelInfo if [[ $modelInfo =~ "--n_layer" ]]; then + sudo rm -rf /root/.cache/torch_extensions python3 ./finetune/lora/$modelInfo $@ --proj_dir lora-models --data_type binidx --lora \ - --lora_parts=att,ffn,time,ln --strategy deepspeed_stage_2 --accelerator gpu + --lora_parts=att,ffn,time,ln --strategy deepspeed_stage_2 --accelerator gpu --ds_bucket_mb 2 else echo "modelInfo is invalid" exit 1 diff --git a/finetune/requirements.txt b/finetune/requirements.txt index 2e218e8..09c962b 100644 --- a/finetune/requirements.txt +++ b/finetune/requirements.txt @@ -1,3 +1,3 @@ -torch==1.13.1 +torch==2.1.2 pytorch_lightning==1.9.5 -deepspeed==0.11.2 +deepspeed==0.12.6