test_script.py

"""CONFIG"""
#!/usr/bin/env python3
from transformers import RobertaConfig
config = RobertaConfig.from_pretrained("roberta-large")
config.save_pretrained("./")

"""TOKENIZER"""
#!/usr/bin/env python3
from datasets import load_dataset
from tokenizers import ByteLevelBPETokenizer
# load dataset
dataset = load_dataset("large_spanish_corpus")
# Instantiate tokenizer
tokenizer = ByteLevelBPETokenizer()
def batch_iterator(batch_size=1000):
    for i in range(0, len(dataset), batch_size):
        yield dataset[i: i + batch_size]["text"]
# Customized training
tokenizer.train_from_iterator(batch_iterator(), vocab_size=50265, min_frequency=2, special_tokens=[
    "<s>",
    "<pad>",
    "</s>",
    "<unk>",
    "<mask>",
])
# Save files to disk
tokenizer.save("./tokenizer.json")

"""TOKENIZER"""
#!/usr/bin/env bash
./run_mlm_flax.py \
    --output_dir="./" \
    --model_type="roberta" \
    --config_name="./" \
    --tokenizer_name="./" \
    --dataset_name="large_spanish_corpus" \
    --dataset_config_name \ # I think this would be empty
    --max_seq_length="128" \
    --per_device_train_batch_size="4" \
    --per_device_eval_batch_size="4" \
    --learning_rate="3e-4" \
    --warmup_steps="1000" \
    --overwrite_output_dir \
    --num_train_epochs="8" \
    --push_to_hub