| #! /usr/bin/env bash |
|
|
| python run_summarization_flax.py \ |
| --output_dir ./mt5-base-qgen \ |
| --model_name_or_path google/mt5-base \ |
| --tokenizer_name google/mt5-base \ |
| --pretokenized \ |
| --train_file asdf.csv \ |
| --text_column context \ |
| --summary_column question \ |
| --do_train \ |
| --do_eval \ |
| --num_train_epochs 4 \ |
| --learning_rate 5e-5 --warmup_steps 1000 \ |
| --dtype bfloat16 \ |
| --per_device_train_batch_size 4 \ |
| --per_device_eval_batch_size 4 \ |
| --overwrite_output_dir \ |
| --max_source_length 1024 \ |
| --max_target_length 64 \ |
| --push_to_hub |
