Yuqing Yang

Certainly! You have the flexibility to modify the dataset, data_collator, and compute_metrics according to your requirements. However, if your task does not involve multiple-choice question answering or generation, you'll also...

你好，使用如下配置和命令，是可以在4张3090上训练LLaMA-13B的。需要注意的是，该`batch_size`开得过大，并不能优化出WIC数据集上的最优结果。 ``` # model model_name_or_path: '/remote-home/share/llama_hf/13B' # data dataset_name: 'wic' refresh: false data_tag: 'base' train_on_inputs: false data_max_length: 1024 # training # trainer tag: 'lomo' output_dir: 'outputs' overwrite_output_dir: true deepspeed: 'config/ds_config.json'...