# Copyright 2022 Cerebras Systems.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
Script that generates a dataset in HDF5 format for GPT Models.
"""
import logging
import os
import sys
from multiprocessing import cpu_count
from pathlib import Path
sys.path.append(os.path.join(os.path.dirname(__file__), "../../../"))
from cerebras.modelzoo.common.utils.utils import check_and_create_output_dirs
from cerebras.modelzoo.data_preparation.nlp.hdf5_preprocessing.utils import (
dump_args,
dump_result,
get_files,
get_params,
get_verification_args,
multimodal_add_image_patch_start_idx,
process_dataset,
verify_saved_hdf5_files_mp,
)
from cerebras.modelzoo.data_preparation.nlp.hdf5_preprocessing.hdf5_dataset_preprocessors import ( # noqa
LMDataPreprocessor,
SummarizationPreprocessor,
FIMDataPreprocessor,
VSLLMDataPreprocessor,
VSLSummarizationPreprocessor,
LlavaPhaseOnePreprocessor,
LlavaPhaseTwoPreprocessor,
)
# Custom preprocessors
from cerebras.modelzoo.data_preparation.nlp.hdf5_preprocessing.hdf5_curation_corpus_preprocessor import ( # noqa
CurationCorpusPreprocessor,
)
from cerebras.modelzoo.data_preparation.nlp.hdf5_preprocessing.hdf5_nlg_preprocessor import ( # noqa
NLGPreprocessor,
)
logging.basicConfig()
logger = logging.getLogger(__file__)
logger.setLevel(logging.INFO)
[docs]def main():
"""Main function for execution."""
params = get_params(desc="Create HDF5 dataset for language models")
output_dir = params["setup"].get("output_dir", "./data_dir/")
if not params["processing"].get("resume_from_checkpoint", False):
check_and_create_output_dirs(output_dir, filetype="h5")
logger.info(f"\nWriting data to {output_dir}.")
json_params_file = os.path.join(output_dir, "data_params.json")
dump_args(params, json_params_file)
metadata_files = params["setup"].pop("metadata_files", None)
if metadata_files:
metadata_files = metadata_files.split(",")
input_dir = params["setup"].pop("input_dir", None)
input_files = get_files(input_dir=input_dir, metadata_files=metadata_files)
processes = params["setup"].pop("processes", 0)
if processes == 0:
processes = cpu_count()
ds_processor = params["setup"].pop(
"dataset_processor", "LMDataPreprocessor"
)
module_name = params["setup"].pop("module", None)
dataset_processor = getattr(sys.modules[__name__], ds_processor)(params)
unused_setup_params = [
key for key in params["setup"].keys() if key != "output_dir"
]
if unused_setup_params:
logger.warning(
"The following setup params are unused: "
+ ", ".join(unused_setup_params)
)
unused_dataset_params = [key for key in params["dataset"].keys()]
if unused_dataset_params:
logger.warning(
"The following dataset params are unused: "
+ ", ".join(unused_dataset_params)
)
## Set this to avoid the warning - The current process just got forked. Disabling parallelism to avoid deadlocks...
# To disable this warning, please explicitly set TOKENIZERS_PARALLELISM=(true | false)
os.environ["TOKENIZERS_PARALLELISM"] = "false"
results = process_dataset(input_files, dataset_processor, processes)
vocab_size = dataset_processor.get_vocab_size()
logger.info(
f"\nFinished writing data to {output_dir}."
f" Runtime arguments and outputs can be found at {json_params_file}."
)
logger.info(f"Verifying the converted dataset at: {output_dir}")
output_files = list(Path(output_dir).glob("*.h5"))
verification_args = get_verification_args(
processes, dataset_processor
) # for verify_saved_hdf5_files_mp
dataset_stats = verify_saved_hdf5_files_mp(
output_files, verification_args, vocab_size
)
logger.info("Done verifying the converted dataset.")
dump_result(
results,
dataset_stats,
json_params_file,
dataset_processor.eos_id,
dataset_processor.pad_id,
vocab_size,
)
multimodal_add_image_patch_start_idx(
json_params_file,
dataset_processor,
)
if __name__ == "__main__":
main()