cpt找不到验证集

[nuochenpku]

Reminder

• I have read the README and searched the existing issues.

System Info

它ransformers==4.36.1

Reproduction

model

model_name_or_path:models/Qwen2-7B

method

stage: pt
do_train: true
finetuning_type: full
deepspeed: examples/deepspeed/ds_z3_config.json # choices: [ds_z0_config.json, ds_z2_config.json, ds_z3_config.json]

dataset

dataset: pretrain/cpt/saves/qwen2-7b/cpt_v1/
eval_dataset: openmath_test,clrs_test
cutoff_len: 4096
template: qwen
overwrite_cache: true
preprocessing_num_workers: 64
tokenized_path: pretrain/cpt/saves/qwen2-7b/cpt_v1/

output

output_dir: output/saves/qwen2-7b/cpt_v1
logging_steps: 10
save_steps: 2000
plot_loss: true
overwrite_output_dir: true

train

per_device_train_batch_size: 4
gradient_accumulation_steps: 8
learning_rate: 5.0e-6
num_train_epochs: 3.0
lr_scheduler_type: cosine
warmup_ratio: 0.1
bf16: true
ddp_timeout: 180000000

eval

per_device_eval_batch_size: 4
eval_strategy: steps
eval_steps: 500

Expected behavior

上述命令，在运行时报错： File "/opt/conda/lib/python3.10/site-packages/transformers/trainer.py", line 419, in init
[rank4]: raise ValueError(
[rank4]: ValueError: You have set args.eval_strategy to steps but you didn't pass an eval_dataset to Trainer. Either set args.eval_strategy to no or pass an eval_dataset.

可是在sh文件中已经指定了eval_dataset

Others

No response

[Geministudents]

有bug，tokenized_path把这个参数去掉

[nuochenpku]

可是我tokenized——path已经是我事先tokenized 过的，直接去掉，也会自动load我之前tokenized id么？

[Geministudents]

那个只是针对大数据集进行个预处理，你不加那个参数就是在训练的开始前处理一下，不影响的，反正我刚才出现这个问题的时候就把这个参数去除就行了

[hiyouga]

预处理时候也需要指定 eval_dataset