Pre-train:
python run.py --datasets=dbpedia,atis,slurp,snips --model_name=google/flan-t5-large --output_dir=outputs/pre-training --max_iterations=400000
Fine-tune:
python run.py --datasets=gpt-emails --model_name=outputs/pre-training/checkpoints/latest --output_dir=outputs/fine-tuning --max_iterations=10000
Test:
python run.py --datasets=gpt-emails --model_name=outputs/fine-tuning/checkpoints/best --output_dir=outputs/testing --no_train --test