app.py

# grab a dataset, prove we can save it
from datasets import load_dataset
raw_datasets = load_dataset("allocine")
raw_datasets.save_to_disk("my-arrow-datasets")

# load dataset from disk - prove we can reload it
from datasets import load_from_disk
arrow_datasets_reloaded = load_from_disk("my-arrow-datasets")
arrow_datasets_reloaded

#dataset_dict.save_to_disk("../data/wikipedia_rank_nocache")
raw_datasets.save_to_disk("../data/awacke1=allocine")

#prove the cache 
arrow_datasets_reloaded.cache_files

# prove we can save in CSV
for split, dataset in raw_datasets.items():
  dataset.to_csv(f"my-dataset-{split}.csv", index=None)
data_files = { 
  "train": "my-dataset-train.csv",
  "validation": "my-dataset-validation.csv",
  "test": "my-dataset-test.csv",
}
csv_datasets_reloaded = load_dataset("csv", data_files=data_files)
csv_datasets_reloaded

# prove we can save in JSON
for split, dataset in raw_datasets.items():
  dataset.to_json(f"my-dataset-{split}.jsonl")
json_data_files = { 
  "train": "my-dataset-train.jsonl",
  "validation": "my-dataset-validation.jsonl",
  "test": "my-dataset-test.jsonl",
}
json_datasets_reloaded = load_dataset("json", data_files=json_data_files)
json_datasets_reloaded 

# prove we can save in Parquet
for split, dataset in raw_datasets.items():
  dataset.to_parquet(f"my-dataset-{split}.parquet")
parquet_data_files = { 
  "train": "my-dataset-train.parquet",
  "validation": "my-dataset-validation.parquet",
  "test": "my-dataset-test.parquet",
}
parquet_datasets_reloaded = load_dataset("parquet", data_files=parquet_data_files)
parquet_datasets_reloaded 


# prove we can save and load public local dataset on huggingface spaces
raw_datasets.save_to_disk("awacke1/my-arrow-datasets")
arrow_datasets_reloaded = load_from_disk("awacke1/my-arrow-datasets")
thisworked="Yes really worked"
arrow_datasets_reloaded
thisworked

#awacke1_public_datasets = load_dataset("awacke1/my-arrow-datasets")
#awacke1_public_datasets