Fix using std::any_cast instead of absl::any_cast

Inconsistencies exist in Lite repo using std::variant/absl::variant and std::any_cast/absl::any_cast. When abseil is built with ABSL_USES_STD_ANY (Linux), everything works even with this bug. But if not (on Windows build), this uses std::any_cast on variable created by absl::variant, causing a throwing of exception at runtime.

Fix using std::any_cast instead of absl::any_cast

Inconsistencies exist in Lite repo using std::variant/absl::variant and std::any_cast/absl::any_cast. When abseil is built with ABSL_USES_STD_ANY (Linux), everything works even with this bug. But if not (on Windows build), this uses std::any_cast on variable created by absl::variant, causing a throwing of exception at runtime.

Hello all, And thank you for all your work. Any news on this, or should we stay in python 3.10 for tflite-runtime?

Hi @ramizouari ,

The script for ./configure can be found here.

If you are interested then go through the source code and analyse the behaviour and may let us know if you have any pointers for this behaviour.

Thanks!

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

Fix using std::any_cast instead of absl::any_cast

Inconsistencies exist in Lite repo using std::variant/absl::variant and std::any_cast/absl::any_cast. When abseil is built with ABSL_USES_STD_ANY (Linux), everything works even with this bug. But if not (on Windows build), this uses std::any_cast on variable created by absl::variant, causing a throwing of exception at runtime.

Hello @pkgoogle, The model was attached in the first comment here: https://drive.google.com/file/d/1jeKm7EesBZqi_lgPrCSq_HwPapX54OlL/view?usp=sharing I would have zipped and sent it but the last time I did that it wiped the entire MedtaData. Here is the training script I used: tflite_maker.zip

• Name of dataset: <name>
• URL of dataset: <url>
• License of dataset: <license type>
• Short description of dataset and use case(s): <description>

Folks who would also like to see this dataset in tensorflow/datasets, please thumbs-up so the developers can know which requests to prioritize.

And if you'd like to contribute the dataset (thank you!), see our guide to adding a dataset.

Hi, thank you for reaching out.

The imdb_reviews dataset is already available in TFDS: https://www.tensorflow.org/datasets/catalog/imdb_reviews

You should be able to use it by simply:

import tensorflow_datasets as tfds

ds = tfds.load('imdb_reviews', split='train')

• Name of dataset: <name>
• URL of dataset: <url>
• License of dataset: <license type>
• Short description of dataset and use case(s): <description>

Folks who would also like to see this dataset in tensorflow/datasets, please thumbs-up so the developers can know which requests to prioritize.

And if you'd like to contribute the dataset (thank you!), see our guide to adding a dataset.

Hi, thank you for reaching out! Please note that the imdb_reviews dataset is already available in TFDS: https://www.tensorflow.org/datasets/catalog/imdb_reviews

This PR separates the graph scheduling & evaluation into the core and implementation modules respectively.

Other relevant changes:

• implemented backends.py library to store available backends (only pandas for the time being)
• added .csv to PandasEvent parsing
• implemented core_mapping.py library to map core operators to their respective pandas implementations

Temporian has moved to its own repo, closing this, see https://github.com/google/temporian

This PR has added ROCm for xla_call_module_test.

/cc @cheshire

@tilakrayal are you talking about something like this ?

import tensorflow as tf
import tensorflow_datasets as tfds
# Set TPU address (if using Google Cloud TPU)
tpu_address = 'grpc://<tpu-ip-address>'
tpu_resolver = tf.distribute.cluster_resolver.TPUClusterResolver(tpu=tpu_address)
tf.config.experimental_connect_to_cluster(tpu_resolver)
tf.tpu.experimental.initialize_tpu_system(tpu_resolver)
data_dir = 'gs://your-bucket/data'  # GCS bucket path for data
model_dir = 'gs://your-bucket/model'  # GCS bucket path for model
builder = tfds.builder('your_dataset_name', data_dir=data_dir)
builder.download_and_prepare()

# Define the training and validation datasets
train_dataset = builder.as_dataset(split='train', shuffle_files=True)
val_dataset = builder.as_dataset(split='validation', shuffle_files=True)

# Enable batching and shuffling
train_dataset = train_dataset.shuffle(1024).batch(batch_size)
val_dataset = val_dataset.batch(batch_size)
strategy = tf.distribute.experimental.TPUStrategy(tpu_resolver)
with strategy.scope():
    # Define and compile your model
    model = ...
    model.compile(...)
callbacks = [
    tf.keras.callbacks.ModelCheckpoint(
        filepath=model_dir,  # Save the model to the GCS bucket
        save_best_only=True,
        save_weights_only=True
    ),
    # Other callbacks...
]

# Train the model
model.fit(train_dataset, validation_data=val_dataset, callbacks=callbacks, ...)

How can I access "gs://your-bucket/data" ? I mean how can I get my bucket address ? (what will be the "gs://your-bucket/" ?)