#!/bin/bash
HF_DATASETS_IN_MEMORY_MAX_SIZE=200000000
MODEL_DIR=.
TOKENIZER_DIR=tokenizer
./run_t5_mlm_flax.py \
    --output_dir="${MODEL_DIR}/checkpoints" \
    --model_type="t5" \
    --config_name="t5-small" \
    --tokenizer_name="${TOKENIZER_DIR}" \
    --dataset_name="mc4" \
    --dataset_config_name="bn" \
    --max_seq_length="512" \
    --per_device_train_batch_size="128" \
    --per_device_eval_batch_size="128" \
    --learning_rate="1e-3" \
    --weight_decay="0.001" \
    --warmup_steps="500" \
    --overwrite_output_dir \
    --num_train_epochs="10" \
    --logging_steps="1" \
    --save_steps="1" \
    --eval_steps="1" \
    --push_to_hub