TensorFlow
Data & MLAn Open Source Machine Learning Framework for Everyone
Release History
v2.21.0-rc0Breaking1 fix8 featuresRelease 2.21.0 removes support for Python 3.9 and introduces significant quantization and data type support enhancements in tf.lite, alongside JPEG XL support in tf.image.
v2.20.0Breaking1 featureTensorFlow 2.20.0 makes `tensorflow-io-gcs-filesystem` optional, adds `autotune.min_parallelism` to `tf.data.Options`, and deprecates `tf.lite` in favor of LiteRT.
v2.19.11 fix1 featureThis patch fixes the save_model.save function for serving embeddings and introduces SparseCore Reshard support.
v2.20.0-rc0Breaking1 featureTensorFlow 2.20.0 makes `tensorflow-io-gcs-filesystem` optional and deprecates `tf.lite` while adding `autotune.min_parallelism` to `tf.data.Options` for faster pipeline warm‑up.
v2.19.0Breaking1 fix1 featureTensorFlow 2.19 introduces breaking changes to LiteRT constants and deprecates `tf.lite.Interpreter`, adds bfloat16 support to `tfl.Cast`, and stops publishing `libtensorflow` packages.
v2.18.1Breaking1 fixThis patch updates curl for security, loosens the ml_dtypes version bound, and deprecates tf.lite.Interpreter in favor of ai_edge_litert.interpreter, while dropping TPU support until 2.19.0.
v2.19.0-rc0Breaking1 fix1 featureTensorFlow 2.19.0 adds bfloat16 support for the tfl.Cast op, deprecates tf.lite.Interpreter in favor of ai_edge_litert.interpreter, and changes two interpreter constants to const references, requiring code updates.
Common Errors
ResourceExhaustedError2 reportsResourceExhaustedError usually arises when TensorFlow tries to allocate memory beyond available resources (GPU or RAM). Reduce batch sizes, model complexity, or input data dimensions to decrease memory footprint, or move computation to a device with more memory if possible. Consider enabling memory growth using `tf.config.experimental.set_memory_growth` or using mixed precision training (`tf.keras.mixed_precision.Policy`) to further mitigate memory usage.
InvalidArgumentError2 reportsInvalidArgumentError in TensorFlow usually stems from providing function arguments with incorrect shapes, data types, or values that violate preconditions or constraints defined within the TensorFlow operation. To fix it, carefully inspect the error message for clues about the specific invalid argument, then validate the shapes, data types, and values of your input tensors before calling the TensorFlow operation using `tf.debugging.assert_*` functions (e.g., `tf.debugging.assert_rank`, `tf.debugging.assert_type`, `tf.debugging.assert_greater_equal`), and adjust them accordingly.
Related Data & ML Packages
🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.
Tensors and Dynamic neural networks in Python with strong GPU acceleration
scikit-learn: machine learning in Python
Flexible and powerful data analysis / manipulation library for Python, providing labeled data structures similar to R data.frame objects, statistical functions, and much more
Streamlit — A faster way to build and share data apps.
Build and share delightful machine learning apps, all in Python. 🌟 Star to support our work!
Subscribe to Updates
Get notified when new versions are released