Replaces the custom MLIR-based TFLite conversion logic in LiteRTExporter with direct use of the standard TFLiteConverter. Also improves input signature handling for tf.function tracing and updates imports accordingly.

Moved imports of get_input_signature and make_tf_tensor_spec inside functions in saved_model.py to prevent circular imports. Updated EXPORT_FORMATS in export_utils.py to use string references instead of direct imports.

Adds max_sequence_length parameter to input signature generation for sequence models, bounding sequence length for transformer-like architectures. Improves LiteRTExporter with heuristics for complex models, fallback conversion via SavedModel for large models, and exposes max_sequence_length in Model export options. Updates documentation accordingly.

Adds logic to dynamically reduce max_sequence_length for large vocabulary models in export_utils.py to prevent tensor size overflow. In lite_rt_exporter.py, introduces checks and workarounds for models with _DictWrapper issues, and applies memory optimizations for large models during TFLite conversion. These changes improve export reliability and prevent memory errors for models such as Gemma, Llama, and similar architectures.

Removed custom trackable object logic from LiteRTExporter and now save the model directly, simplifying the export process. Also streamlined vocabulary size checks in export_utils to prevent tensor size overflow, removing verbose warnings and redundant comments.

Refactors the TFLite conversion logic in lite_rt_exporter.py to attempt direct conversion first and only fall back to SavedModel if necessary, improving robustness and clarity. Adds a new lite_rt_exporter_simple.py file with a streamlined LiteRTExporter class for direct TFLite export, bypassing complex MLIR conversion paths.

…port

Refactors export_utils and lite_rt_exporter to better detect large vocabulary and Keras-Hub models, applying safer sequence length limits and more robust TFLite conversion paths. Adds heuristics for model type detection, ensures memory safety, and improves handling of TensorFlow introspection issues during export.

Working well with keras

Eliminates the logic for bounding sequence length in model export utilities and related code paths. The max_sequence_length parameter and associated shape bounding for large vocabulary models are removed from export_utils.py and lite_rt_exporter.py. Updates model export documentation accordingly. Adds a comprehensive test script for Keras Hub LiteRT export, verifying numerical accuracy between original and exported models.

Refactors and improves the LiteRT exporter for better input signature inference, error handling, and verbose output. Updates and expands the test suite to cover multi-input, multi-output, and error scenarios, and ensures the Model.export() API is used consistently. Also improves docstrings and formatting for clarity and maintainability.

Replaces direct imports and availability checks for LiteRT with the LazyModule-based `litert` utility from `module_utils`. Updates tests and exporter logic to require '.tflite' filepaths and consistently use the new import pattern, improving modularity and error handling.

…port

Simplifies and standardizes input signature generation for Functional, Sequential, and subclassed models to always use a flexible batch dimension (None) for export. Refactors Sequential input shape inference to remove hardcoded fallbacks and ensure dynamic batching support. Cleans up and clarifies error handling and structure of input signature utilities.

Renamed litert_exporter.py and its test to litert.py and litert_test.py, respectively. Updated all relevant imports to use the new module name. Refactored LitertExporter to simplify model building logic and removed redundant input shape inference for Sequential models. Also cleaned up export_utils.py by removing the export format registry and related functions.

Moved imports of get_input_signature and make_tf_tensor_spec to the module level in saved_model.py for consistency and efficiency. In litert.py, imports remain local to preserve lazy loading where needed.

LiteRT export and related tests now import TensorFlow conditionally, allowing Keras to function without TensorFlow installed. If TensorFlow is unavailable, LiteRT export is disabled and tests are skipped, with informative messages and error handling added to guide users.