GLM-130B/tasks/language-modeling/pile.yaml

name: "Pile"
type: "lm"
module: "tasks.language-modeling.pile.Pile"
path: "pile"
file-pattern:
  test: "**/test_tokenized.jsonl"
#  validation: "**/val_tokenized.jsonl"

generation-length: 1024
use_task_mask: true