This commit is contained in:
@@ -0,0 +1,25 @@
|
||||
CUDA_VISIBLE_DEVICES=0,1,2,3 \
|
||||
NPROC_PER_NODE=4 \
|
||||
swift sft \
|
||||
--model Qwen/Qwen3-Reranker-0.6B \
|
||||
--task_type generative_reranker \
|
||||
--loss_type pointwise_reranker \
|
||||
--tuner_type full \
|
||||
--dataset MTEB/scidocs-reranking \
|
||||
--load_from_cache_file true \
|
||||
--split_dataset_ratio 0.05 \
|
||||
--eval_strategy steps \
|
||||
--output_dir output \
|
||||
--eval_steps 100 \
|
||||
--num_train_epochs 1 \
|
||||
--save_steps 200 \
|
||||
--per_device_train_batch_size 2 \
|
||||
--per_device_eval_batch_size 2 \
|
||||
--gradient_accumulation_steps 8 \
|
||||
--dataset_num_proc 8 \
|
||||
--learning_rate 6e-6 \
|
||||
--label_names labels \
|
||||
--dataloader_drop_last true \
|
||||
--sequence_parallel_size 4 \
|
||||
--padding_free true \
|
||||
--attn_impl flash_attn
|
||||
Reference in New Issue
Block a user