wehub-resource-sync
23 lines
612 B
Bash
23 lines
612 B
Bash
# loss_scale all to train all tokens
|
|
# use loss_type loss_scale
|
|
# This is just an example
|
|
CUDA_VISIBLE_DEVICES=0 \
|
|
swift sft \
|
|
--model Qwen/Qwen2.5-7B-Instruct \
|
|
--tuner_type lora \
|
|
--dataset 'swift/self-cognition#1000' \
|
|
--num_train_epochs 1 \
|
|
--per_device_train_batch_size 1 \
|
|
--learning_rate 1e-4 \
|
|
--lora_rank 8 \
|
|
--lora_alpha 32 \
|
|
--gradient_accumulation_steps 16 \
|
|
--eval_steps 100 \
|
|
--save_steps 100 \
|
|
--save_total_limit 2 \
|
|
--logging_steps 5 \
|
|
--model_author swift \
|
|
--model_name swift-robot \
|
|
--loss_scale all \
|
|
--loss_type loss_scale
|