Snowpark ML FileSystem and FileSet — Deprecated¶

Installation¶

The FileSystem and FileSet APIs are part of the Snowpark ML Python package, snowflake-ml-python. See Using Snowflake ML Locally for installation instructions.

Creating and Using a File System¶

Creating a Snowpark ML file system requires either a Snowflake Connector for Python Connection object or a Snowpark Python Session. See Connecting to Snowflake for instructions.

After you have either a connection or a session, you can create a Snowpark ML SFFileSystem instance through which you can access data in your internal stage.

If you have a Snowflake Connector for Python connection, pass it as the sf_connection argument:

import fsspec
from snowflake.ml.fileset import sfcfs

sf_fs1 = sfcfs.SFFileSystem(sf_connection=sf_connection)

If you have a Snowpark Python session, pass it as the snowpark_session argument:

import fsspec
from snowflake.ml.fileset import sfcfs

sf_fs2 = sfcfs.SFFileSystem(snowpark_session=sp_session)

SFFileSystem inherits many features from fsspec.FileSystem, such as local caching of files. You can enable this and other features by instantiating a Snowflake file system through the fsspec.filesystem factory function, passing target_protocol="sfc" to use the Snowflake FileSystem implementation:

local_cache_path = "/tmp/sf_files/"
cached_fs = fsspec.filesystem("cached", target_protocol="sfc",
                    target_options={"sf_connection": sf_connection,
                                    "cache_types": "bytes",
                                    "block_size": 32 * 2**20},
                    cache_storage=local_cache_path)

The Snowflake file system supports most read-only methods defined for a fsspec FileSystem, including find, info, isdir, isfile, and exists.

print(*cached_fs.ls("@ML_DATASETS.public.my_models/sales_predict/"), end='\n')

path = '@ML_DATASETS.public.my_models/test/data_7_7_3.snappy.parquet'

with sf_fs1.open(path, mode='rb') as f:
    print(f.read(16))

import pyarrow.parquet as pq

table = pq.read_table(path, filesystem=sf_fs1)
table.take([1, 3])

path = "sfc://@ML_DATASETS.public.my_models/dataset.csv.gz"

with cached_fs.open(path, mode='rb', sf_connection=sf_connection) as f:
    g = gzip.GzipFile(fileobj=f)
    for i in range(3):
        print(g.readline())

Creating and Using a FileSet¶

A Snowflake FileSet represents an immutable snapshot of the result of a SQL query in the form of files in an internal server-side encrypted stage. These files can be accessed through a FileSystem to feed data to tools such as PyTorch and TensorFlow so that you can train models at scale and within your existing data governance model. To create a FileSet, use the FileSet.make method.

You need a Snowflake Python connection or a Snowpark session to create a FileSet. See Connecting to Snowflake for instructions. You must also provide the path to an existing internal server-side encrypted stage, or a subdirectory under such a stage, where the FileSet will be stored.

To create a FileSet from a Snowpark DataFrame, construct a DataFrame and pass it to FileSet.make as snowpark_dataframe; do not call the DataFrame’s collect method:

# Snowpark Python equivalent of "SELECT * FROM MYDATA LIMIT 5000000"
df = snowpark_session.table('mydata').limit(5000000)
fileset_df = fileset.FileSet.make(
    target_stage_loc="@ML_DATASETS.public.my_models/",
    name="from_dataframe",
    snowpark_dataframe=df,
    shuffle=True,
)

To create a FileSet using a Snowflake Connector for Python connection, pass the connection to Fileset.make as sf_connection, and pass the SQL query as query:

fileset_sf = fileset.FileSet.make(
    target_stage_loc="@ML_DATASETS.public.my_models/",
    name="from_connector",
    sf_connection=sf_connection,
    query="SELECT * FROM MYDATA LIMIT 5000000",
    shuffle=True,           # see later section about shuffling
)

Use the files method to get a list of the files in the FileSet:

print(*fileset_df.files())

For information about feeding the data in the FileSet to PyTorch or TensorFlow, see Feeding a FileSet to PyTorch or Feeding a FileSet to TensorFlow, respectively.

Feeding a FileSet to PyTorch¶

From a Snowflake FileSet, you can get a PyTorch DataPipe, which can be passed to a PyTorch DataLoader. The DataLoader iterates over the FileSet data and yields batched PyTorch tensors. Create the DataPipe using the FileSet’s to_torch_datapipe method, and then pass the DataPipe to PyTorch’s DataLoader:

from torch.utils.data import DataLoader

# See later sections about shuffling and batching
pipe = fileset_df.to_torch_datapipe(
    batch_size=4,
    shuffle=True,
    drop_last_batch=True)

for batch in DataLoader(pipe, batch_size=None, num_workers=0):
    print(batch)
    break

Feeding a FileSet to TensorFlow¶

You can get a TensorFlow Dataset from a Snowflake FileSet using the FileSet’s to_tf_dataset method:

import tensorflow as tf

# See following sections about shuffling and batching
ds = fileset_df.to_tf_dataset(
    batch_size=4,
    shuffle=True,
    drop_last_batch=True)

for batch in ds:
    print(batch)
    break

Iterating over the Dataset yields batched tensors.

Shuffling Data in FileSets¶

It is often valuable to shuffle the training data to avoid overfitting and other issues. For a discussion of the value of shuffling, see Why should the data be shuffled for machine learning tasks?

If your query does not already shuffle your data sufficiently, a FileSet can shuffle data at two points:

• When the FileSet is created by using FileSet.make.

  All rows in your query are shuffled before they are written to the FileSet. This is a high-quality global shuffle and can be expensive with large datasets. Therefore, it is performed only once, when materializing the FileSet. Pass shuffle=True as a keyword argument to FileSet.make.

• When you create a PyTorch DataPipe or a TensorFlow Dataset from a FileSet.

  At this point, the order of the files in the FileSet is randomized, as is the order of the rows in each file. This can be considered an “approximate” global shuffle. It is of lower quality than a true global shuffle, but it is much less expensive. To shuffle at this stage, pass shuffle=True as a keyword argument to the FileSet’s to_torch_datapipe or to_tf_dataset method.

For best results, shuffle twice: when creating the FileSet and when feeding the data to PyTorch or TensorFlow.

Batching Data in FileSets¶

FileSets have a batching feature that works the same as the batching functionality in PyTorch and TensorFlow but is more efficient. Snowflake recommends that you use the batch_size parameter in the FileSet’s to_torch_datapipe and to_tf_dataset methods instead of having PyTorch or TensorFlow do the batching. With PyTorch, to disable its batching functionality, you must explicitly pass batch_size=None when instantiating DataLoader.

You can also drop the last batch if it is incomplete by passing drop_last_batch=True to to_torch_datapipe or to to_tf_dataset.