mirror of
https://github.com/modelscope/modelscope.git
synced 2025-12-22 02:59:24 +01:00
17 lines
640 B
Bash
17 lines
640 B
Bash
|
PYTHONPATH=. torchrun examples/pytorch/token_classification/finetune_token_classification.py \
|
||
|
|
--trainer 'nlp-base-trainer' \
|
||
|
|
--work_dir './tmp' \
|
||
|
|
--model 'damo/nlp_structbert_backbone_base_std' \
|
||
|
|
--dataset_name 'GeoGLUE' \
|
||
|
|
--subset_name 'GeoETA' \
|
||
|
|
--train_dataset_params 'first_sequence=tokens,label=ner_tags,sequence_length=128' \
|
||
|
|
--preprocessor 'token-cls-tokenizer' \
|
||
|
|
--preprocessor_padding 'max_length' \
|
||
|
|
--max_epochs 2 \
|
||
|
|
--per_device_train_batch_size 32 \
|
||
|
|
--lr 3e-5 \
|
||
|
|
--save_ckpt_strategy 'by_epoch' \
|
||
|
|
--logging_interval 1 \
|
||
|
|
--eval_strategy 'by_step' \
|
||
|
|
--eval_interval 20 \
|