[Docs] De-flake doctests

[bveeramani]

Why are these changes needed?

:book: Doctest (CPU) fails 25% of runs due to a few flaky tests. This PR deflakes those tests.

---

This test is flaky because the order of columns is non-deterministic.

______________________ [doctest] working-with-images.rst _______________________

  | 111
  | 112             ds = (
  | 113                 ray.data.read_tfrecords(
  | 114                     "s3://anonymous@air-example-data/cifar-10/tfrecords"
  | 115                 )
  | 116                 .map(decode_bytes)
  | 117             )
  | 118
  | 119             print(ds.schema())
  | 120
  | Differences (unified diff with -expected +actual):
  | @@ -1,4 +1,4 @@
  | Column  Type
  | ------  ----
  | +label   int64
  | image   numpy.ndarray(shape=(32, 32, 3), dtype=uint8)
  | -label   int64

---

This test is flaky because, in rare occasions, all outputs from the previous code block actually show up in the previous testoutput. (Usually, a few of the outputs show up in this testoutput).

___________________________ [doctest] async_api.rst ____________________________
  | 062     (AsyncActor pid=40293) finished
  | 063
  | 064 .. testcode::
  | 065     :hide:
  | 066
  | 067     # NOTE: The outputs from the previous code block can show up in subsequent tests.
  | 068     # To prevent flakiness, we wait for the async calls finish.
  | 069     import time
  | 070     time.sleep(3)
  | 071
  | Expected:
  | ...
  | Got nothing

---

These tests are flaky because the outputs contain extra logs if the dataset isn't cached.

______________________ [doctest] working-with-pytorch.rst ______________________
  | 353
  | 354     import torchvision
  | 355     import ray
  | 356
  | 357     mnist = torchvision.datasets.MNIST(root="/tmp/", download=True)
  | 358     ds = ray.data.from_torch(mnist)
  | 359
  | 360     # The data for each item of the torch dataset is under the "item" key.
  | 361     print(ds.schema())
  | 362
  | Differences (unified diff with -expected +actual):
  | @@ -1,2 +1,18 @@
  | +Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz
  | +Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz to /tmp/MNIST/raw/train-images-idx3-ubyte.gz
  | +Extracting /tmp/MNIST/raw/train-images-idx3-ubyte.gz to /tmp/MNIST/raw
  | +<BLANKLINE>
  | +Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz
  | +Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz to /tmp/MNIST/raw/train-labels-idx1-ubyte.gz
  | +Extracting /tmp/MNIST/raw/train-labels-idx1-ubyte.gz to /tmp/MNIST/raw
  | +<BLANKLINE>
  | +Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz
  | +Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz to /tmp/MNIST/raw/t10k-images-idx3-ubyte.gz
  | +Extracting /tmp/MNIST/raw/t10k-images-idx3-ubyte.gz to /tmp/MNIST/raw
  | +<BLANKLINE>
  | +Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz
  | +Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz to /tmp/MNIST/raw/t10k-labels-idx1-ubyte.gz
  | +Extracting /tmp/MNIST/raw/t10k-labels-idx1-ubyte.gz to /tmp/MNIST/raw
  | +<BLANKLINE>
  | Column  Type
  | ------  ----

__________________________ [doctest] loading-data.rst __________________________
  | 613         .. testcode::
  | 614
  | 615             import ray
  | 616             import tensorflow_datasets as tfds
  | 617
  | 618             tf_ds, _ = tfds.load("cifar10", split=["train", "test"])
  | 619             ds = ray.data.from_tf(tf_ds)
  | 620
  | 621             print(ds)
  | 622
  | Differences (unified diff with -expected +actual):
  | @@ -1,4 +1,6 @@
  | +Downloading and preparing dataset 162.17 MiB (download: 162.17 MiB, generated: 132.40 MiB, total: 294.58 MiB) to /root/tensorflow_datasets/cifar10/3.0.2...
  | +Dataset cifar10 downloaded and prepared to /root/tensorflow_datasets/cifar10/3.0.2. Subsequent calls will reuse this data.
  | MaterializedDataset(
  | -   num_blocks=...,
  | +   num_blocks=200,
  | num_rows=50000,
  | schema={

Related issue number

Checks

• I've signed off every commit(by using the -s flag, i.e., git commit -s) in this PR.
• I've run scripts/format.sh to lint the changes in this PR.
• I've included any doc changes needed for https://docs.ray.io/en/master/.
  • I've added any new APIs to the API Reference. For example, if I added a
    method in Tune, I've added it in doc/source/tune/api/ under the
    corresponding .rst file.
• I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
• Testing Strategy
  • Unit tests
  • Release tests
  • This PR is not tested :(

[amogkam]

what is this for?

[bveeramani]

Usually, there are outputs from the previous code block (which asynchronously prints out stuff), and this code block is intended to catch them. However, sometimes everything actually gets caught in the previous testoutput block. In this case, there are no outputs from this code block, and the test errors.

This print ensures there are always some outputs from this code block.

[amogkam]

data changes lgtm

[angelinalg]

Suggested change

	Datasets containing torch tensors can be saved to files, like parquet or numpy.
	You can save Datasets containing torch Tensors to files, like parquet or NumPy.