v1.0.3

Release 1.0 is the first with a stable public API.

What's Changed in r1.0

• Overall
  • Use with incompatible Keras v3 raises a clear error.
    • As of release 1.0.3, the error refers to the new Keras version guide and explains how to get Keras v2 with TF2.16+ via TF_USE_LEGACY_KERAS=1.
    • Releases 1.0.0 to 1.0.2 had a pip package requirement for TF <2.16 but could be made to work the same way.
  • Minimum supported TF/Keras version moved to >=2.12.
  • Importing the library no longer leaks private module names.
  • All parts of the GraphSchema protobuf are now exposed undertfgnn.proto.*.
  • Model saving now clearly distinguishes export to inference (pure TF, fully supported) from misc ways of saving for model reuse.
  • Numerous small bug fixes.
• Subgraph sampling: major upgrade
  • New and unified sampler for in-memory and beam-based subgraph sampling.
  • Module tfgnn.experimental.in_memory is removed in favor of the new sampler.
  • New console script tfgnn_sampler replaces the old tfgnn_graph_sampler.
• GraphTensor
  • Most tfgnn.* functions on GraphTensor now work in Keras' Functional API, including the factory methods GraphTensor.from_pieces(...) etc.
  • New static checks for GraphTensor field shapes, opt out with tfgnn.disable_graph_tensor_validation().
  • New runtime checks for GraphTensor field shapes, sizes and index ranges, opt in with tfgnn.enable_graph_tensor_validation_at_runtime().
  • GraphTensor maintains .row_splits_dtype separately from .indices_dtype.
  • The GraphSchema and the I/O functions for tf.Example now support all non-quantized, non-complex floating-point and integer types as well as bool and string.
  • Added convenience wrapper tfgnn.pool_neighbors_to_node().
  • Misc fixes to tfgnn.random_graph_tensor(), now respects component boundaries.
• Runner
  • New tasks for link prediction and node classification/regression based on structured readout.
  • Now comes with API docs.
• Models collection
  • models/contrastive_losses gets multiple extensions, including a triplet loss and API docs.
  • models/multi_head_attention replaces sigmoid with elu+1 in trained scaling.
  • Bug fixes for mixed precision.

Full Changelog: v0.6.1...v1.0.0

What's Changed in v1.0.3 over v1.0.2

• Support TF2.16+ via TF_USE_LEGACY_KERAS=1: updated setup.py, docs and error messages (605b552)

Full Changelog: v1.0.2...v1.0.3

What's Changed in v1.0.2 over v1.0.1

• Bugfixes for use with TF 2.14 and 2.15 in case tf_keras is installed but not used as tf.keras (ffa453f) (e1d9210).

Full Changelog: v1.0.1...v1.0.2

What's Changed in v1.0.1 over v1.0.0

• Bugfix for regression tasks runner.GraphMean*Error: the reduce_type is again passed through correctly (19c10f2).

Full Changelog: v1.0.0...v1.0.1

v1.0.3rc0

Release 1.0 is the first with a stable public API.

What's Changed in r1.0

• Overall
  • Use with incompatible Keras v3 raises a clear error.
    • As of release 1.0.3, the error refers to the new Keras version guide and explains how to get Keras v2 with TF2.16+ via TF_USE_LEGACY_KERAS=1.
    • Releases 1.0.0 to 1.0.2 had a pip package requirement for TF <2.16 but could be made to work the same way.
  • Minimum supported TF/Keras version moved to >=2.12.
  • Importing the library no longer leaks private module names.
  • All parts of the GraphSchema protobuf are now exposed undertfgnn.proto.*.
  • Model saving now clearly distinguishes export to inference (pure TF, fully supported) from misc ways of saving for model reuse.
  • Numerous small bug fixes.
• Subgraph sampling: major upgrade
  • New and unified sampler for in-memory and beam-based subgraph sampling.
  • Module tfgnn.experimental.in_memory is removed in favor of the new sampler.
  • New console script tfgnn_sampler replaces the old tfgnn_graph_sampler.
• GraphTensor
  • Most tfgnn.* functions on GraphTensor now work in Keras' Functional API, including the factory methods GraphTensor.from_pieces(...) etc.
  • New static checks for GraphTensor field shapes, opt out with tfgnn.disable_graph_tensor_validation().
  • New runtime checks for GraphTensor field shapes, sizes and index ranges, opt in with tfgnn.enable_graph_tensor_validation_at_runtime().
  • GraphTensor maintains .row_splits_dtype separately from .indices_dtype.
  • The GraphSchema and the I/O functions for tf.Example now support all non-quantized, non-complex floating-point and integer types as well as bool and string.
  • Added convenience wrapper tfgnn.pool_neighbors_to_node().
  • Misc fixes to tfgnn.random_graph_tensor(), now respects component boundaries.
• Runner
  • New tasks for link prediction and node classification/regression based on structured readout.
  • Now comes with API docs.
• Models collection
  • models/contrastive_losses gets multiple extensions, including a triplet loss and API docs.
  • models/multi_head_attention replaces sigmoid with elu+1 in trained scaling.
  • Bug fixes for mixed precision.

Full Changelog: v0.6.1...v1.0.0

What's Changed in v1.0.3 over v1.0.2

• Support TF2.16+ via TF_USE_LEGACY_KERAS=1: updated setup.py, docs and error messages (605b552)

Full Changelog: v1.0.2...v1.0.3rc0

What's Changed in v1.0.2 over v1.0.1

• Bugfixes for use with TF 2.14 and 2.15 in case tf_keras is installed but not used as tf.keras (ffa453f) (e1d9210).

Full Changelog: v1.0.1...v1.0.2

What's Changed in v1.0.1 over v1.0.0

• Bugfix for regression tasks runner.GraphMean*Error: the reduce_type is again passed through correctly (19c10f2).

Full Changelog: v1.0.0...v1.0.1

v1.0.2

Release 1.0 is the first with a stable public API.

What's Changed in r1.0

• Overall
  • Supported TF/Keras versions moved to >=2.12,<2.16, incompatible Keras v3 raises a clear error.
  • Importing the library no longer leaks private module names.
  • All parts of the GraphSchema protobuf are now exposed undertfgnn.proto.*.
  • Model saving now clearly distinguishes export to inference (pure TF, fully supported) from misc ways of saving for model reuse.
  • Numerous small bug fixes.
• Subgraph sampling: major upgrade
  • New and unified sampler for in-memory and beam-based subgraph sampling.
  • Module tfgnn.experimental.in_memory is removed in favor of the new sampler.
  • New console script tfgnn_sampler replaces the old tfgnn_graph_sampler.
• GraphTensor
  • Most tfgnn.* functions on GraphTensor now work in Keras' Functional API, including the factory methods GraphTensor.from_pieces(...) etc.
  • New static checks for GraphTensor field shapes, opt out with tfgnn.disable_graph_tensor_validation().
  • New runtime checks for GraphTensor field shapes, sizes and index ranges, opt in with tfgnn.enable_graph_tensor_validation_at_runtime().
  • GraphTensor maintains .row_splits_dtype separately from .indices_dtype.
  • The GraphSchema and the I/O functions for tf.Example now support all non-quantized, non-complex floating-point and integer types as well as bool and string.
  • Added convenience wrapper tfgnn.pool_neighbors_to_node().
  • Misc fixes to tfgnn.random_graph_tensor(), now respects component boundaries.
• Runner
  • New tasks for link prediction and node classification/regression based on structured readout.
  • Now comes with API docs.
• Models collection
  • models/contrastive_losses gets multiple extensions, including a triplet loss and API docs.
  • models/multi_head_attention replaces sigmoid with elu+1 in trained scaling.
  • Bug fixes for mixed precision.

Full Changelog: v0.6.1...v1.0.0

What's Changed in v1.0.2 over v1.0.1

• Bugfixes for use with TF 2.14 and 2.15 in case tf_keras is installed but not used as tf.keras (ffa453f) (e1d9210).

Full Changelog: v1.0.1...v1.0.2

What's Changed in v1.0.1 over v1.0.0

• Bugfix for regression tasks runner.GraphMean*Error: the reduce_type is again passed through correctly (19c10f2).

Full Changelog: v1.0.0...v1.0.1

v1.0.2rc1

Release 1.0 is the first with a stable public API.

What's Changed in r1.0

• Overall
  • Supported TF/Keras versions moved to >=2.12,<2.16, incompatible Keras v3 raises a clear error.
  • Importing the library no longer leaks private module names.
  • All parts of the GraphSchema protobuf are now exposed undertfgnn.proto.*.
  • Model saving now clearly distinguishes export to inference (pure TF, fully supported) from misc ways of saving for model reuse.
  • Numerous small bug fixes.
• Subgraph sampling: major upgrade
  • New and unified sampler for in-memory and beam-based subgraph sampling.
  • Module tfgnn.experimental.in_memory is removed in favor of the new sampler.
  • New console script tfgnn_sampler replaces the old tfgnn_graph_sampler.
• GraphTensor
  • Most tfgnn.* functions on GraphTensor now work in Keras' Functional API, including the factory methods GraphTensor.from_pieces(...) etc.
  • New static checks for GraphTensor field shapes, opt out with tfgnn.disable_graph_tensor_validation().
  • New runtime checks for GraphTensor field shapes, sizes and index ranges, opt in with tfgnn.enable_graph_tensor_validation_at_runtime().
  • GraphTensor maintains .row_splits_dtype separately from .indices_dtype.
  • The GraphSchema and the I/O functions for tf.Example now support all non-quantized, non-complex floating-point and integer types as well as bool and string.
  • Added convenience wrapper tfgnn.pool_neighbors_to_node().
  • Misc fixes to tfgnn.random_graph_tensor(), now respects component boundaries.
• Runner
  • New tasks for link prediction and node classification/regression based on structured readout.
  • Now comes with API docs.
• Models collection
  • models/contrastive_losses gets multiple extensions, including a triplet loss and API docs.
  • models/multi_head_attention replaces sigmoid with elu+1 in trained scaling.
  • Bug fixes for mixed precision.

Full Changelog: v0.6.1...v1.0.0

What's Changed in v1.0.2 over v1.0.1

• Bugfixes for use with TF 2.14 and 2.15 in case tf_keras is installed but not used as tf.keras (ffa453f) (e1d9210).

Full Changelog: v1.0.1...v1.0.2rc1

What's Changed in v1.0.1 over v1.0.0

• Bugfix for regression tasks runner.GraphMean*Error: the reduce_type is again passed through correctly (19c10f2).

Full Changelog: v1.0.0...v1.0.1

v1.0.2rc0

Release 1.0 is the first with a stable public API.

What's Changed in r1.0

• Overall
  • Supported TF/Keras versions moved to >=2.12,<2.16, incompatible Keras v3 raises a clear error.
  • Importing the library no longer leaks private module names.
  • All parts of the GraphSchema protobuf are now exposed undertfgnn.proto.*.
  • Model saving now clearly distinguishes export to inference (pure TF, fully supported) from misc ways of saving for model reuse.
  • Numerous small bug fixes.
• Subgraph sampling: major upgrade
  • New and unified sampler for in-memory and beam-based subgraph sampling.
  • Module tfgnn.experimental.in_memory is removed in favor of the new sampler.
  • New console script tfgnn_sampler replaces the old tfgnn_graph_sampler.
• GraphTensor
  • Most tfgnn.* functions on GraphTensor now work in Keras' Functional API, including the factory methods GraphTensor.from_pieces(...) etc.
  • New static checks for GraphTensor field shapes, opt out with tfgnn.disable_graph_tensor_validation().
  • New runtime checks for GraphTensor field shapes, sizes and index ranges, opt in with tfgnn.enable_graph_tensor_validation_at_runtime().
  • GraphTensor maintains .row_splits_dtype separately from .indices_dtype.
  • The GraphSchema and the I/O functions for tf.Example now support all non-quantized, non-complex floating-point and integer types as well as bool and string.
  • Added convenience wrapper tfgnn.pool_neighbors_to_node().
  • Misc fixes to tfgnn.random_graph_tensor(), now respects component boundaries.
• Runner
  • New tasks for link prediction and node classification/regression based on structured readout.
  • Now comes with API docs.
• Models collection
  • models/contrastive_losses gets multiple extensions, including a triplet loss and API docs.
  • models/multi_head_attention replaces sigmoid with elu+1 in trained scaling.
  • Bug fixes for mixed precision.

Full Changelog: v0.6.1...v1.0.0

What's Changed in v1.0.2 over v1.0.1

• Bugfix for use with TF 2.14 and 2.15 in case tf_keras is installed but not used as tf.keras (ffa453f).

Full Changelog: v1.0.1...v1.0.2rc0

What's Changed in v1.0.1 over v1.0.0

• Bugfix for regression tasks runner.GraphMean*Error: the reduce_type is again passed through correctly (19c10f2).

Full Changelog: v1.0.0...v1.0.1

v1.0.1

Release 1.0 is the first with a stable public API.

What's Changed in r1.0

• Overall
  • Supported TF/Keras versions moved to >=2.12,<2.16, incompatible Keras v3 raises a clear error.
  • Importing the library no longer leaks private module names.
  • All parts of the GraphSchema protobuf are now exposed undertfgnn.proto.*.
  • Model saving now clearly distinguishes export to inference (pure TF, fully supported) from misc ways of saving for model reuse.
  • Numerous small bug fixes.
• Subgraph sampling: major upgrade
  • New and unified sampler for in-memory and beam-based subgraph sampling.
  • Module tfgnn.experimental.in_memory is removed in favor of the new sampler.
  • New console script tfgnn_sampler replaces the old tfgnn_graph_sampler.
• GraphTensor
  • Most tfgnn.* functions on GraphTensor now work in Keras' Functional API, including the factory methods GraphTensor.from_pieces(...) etc.
  • New static checks for GraphTensor field shapes, opt out with tfgnn.disable_graph_tensor_validation().
  • New runtime checks for GraphTensor field shapes, sizes and index ranges, opt in with tfgnn.enable_graph_tensor_validation_at_runtime().
  • GraphTensor maintains .row_splits_dtype separately from .indices_dtype.
  • The GraphSchema and the I/O functions for tf.Example now support all non-quantized, non-complex floating-point and integer types as well as bool and string.
  • Added convenience wrapper tfgnn.pool_neighbors_to_node().
  • Misc fixes to tfgnn.random_graph_tensor(), now respects component boundaries.
• Runner
  • New tasks for link prediction and node classification/regression based on structured readout.
  • Now comes with API docs.
• Models collection
  • models/contrastive_losses gets multiple extensions, including a triplet loss and API docs.
  • models/multi_head_attention replaces sigmoid with elu+1 in trained scaling.
  • Bug fixes for mixed precision.

Full Changelog: v0.6.1...v1.0.0

What's Changed in v1.0.1 over v1.0.0

• Bugfix for regression tasks runner.GraphMean*Error: the reduce_type is again passed through correctly
  (19c10f2).

Full Changelog: v1.0.0...v1.0.1

v1.0.0

First release with a stable public API.

What's Changed

• Overall
  • Supported TF/Keras versions moved to >=2.12,<2.16, incompatible Keras v3 raises a clear error.
  • Importing the library no longer leaks private module names.
  • All parts of the GraphSchema protobuf are now exposed undertfgnn.proto.*.
  • Model saving now clearly distinguishes export to inference (pure TF, fully supported) from misc ways of saving for model reuse.
  • Numerous small bug fixes.
• Subgraph sampling: major upgrade
  • New and unified sampler for in-memory and beam-based subgraph sampling.
  • Module tfgnn.experimental.in_memory is removed in favor of the new sampler.
  • New console script tfgnn_sampler replaces the old tfgnn_graph_sampler.
• GraphTensor
  • Most tfgnn.* functions on GraphTensor now work in Keras' Functional API, including the factory methods GraphTensor.from_pieces(...) etc.
  • New static checks for GraphTensor field shapes, opt out with tfgnn.disable_graph_tensor_validation().
  • New runtime checks for GraphTensor field shapes, sizes and index ranges, opt in with tfgnn.enable_graph_tensor_validation_at_runtime().
  • GraphTensor maintains .row_splits_dtype separately from .indices_dtype.
  • The GraphSchema and the I/O functions for tf.Example now support all non-quantized, non-complex floating-point and integer types as well as bool and string.
  • Added convenience wrapper tfgnn.pool_neighbors_to_node().
  • Misc fixes to tfgnn.random_graph_tensor(), now respects component boundaries.
• Runner
  • New tasks for link prediction and node classification/regression based on structured readout.
  • Now comes with API docs.
• Models collection
  • models/contrastive_losses gets multiple extensions, including a triplet loss and API docs.
  • models/multi_head_attention replaces sigmoid with elu+1 in trained scaling.
  • Bug fixes for mixed precision.

Full Changelog: v0.6.1...v1.0.0

v1.0.0rc0

Initial release candidate for v1.0.0.

v1.0.0.dev2

Early developmental release of tensorflow-gnn 1.0.0 code; docs still unfinished.

v0.6.1

• Running import tensorflow_gnn now checks if the version of tf.keras is compatible.