v0.4.0

PyCallChainRules v0.4.0

Diff since v0.3.2

• Add integration and example usage with Lux.jl and jax.

Merged pull requests:

• Add training examples for mixing Flux and Torch layers (#22) (@rejuvyesh)
• use BatchNorm removal function from functorch (#25) (@rejuvyesh)
• CompatHelper: bump compat for Functors to 0.3, (keep existing compat) (#26) (@github-actions[bot])
• Add Lux.jl integration with Jax (#27) (@rejuvyesh)

v0.3.2

PyCallChainRules v0.3.2

Diff since v0.3.1

• Fix Adapt usage for CUDA array views.

Merged pull requests:

• Add example scripts for training with flux (#20) (@rejuvyesh)
• Add DiffEqFlux training on GPU with pytorch (#23) (@rejuvyesh)

v0.3.1

PyCallChainRules v0.3.1

Diff since v0.3.0

• Fix Adapt usage for array views.

Merged pull requests:

• Do type piracy in a single place (#17) (@rejuvyesh)
• Add an example for DiffEqFlux training with pytorch (#21) (@rejuvyesh)

v0.3.0

PyCallChainRules v0.3.0

Diff since v0.2.1

• Support multiple output python functions

Merged pull requests:

• Support multi-output PyTorch/Jax functions (#16) (@rejuvyesh)

v0.2.1

PyCallChainRules v0.2.1

Diff since v0.2.0

• Fix Adapt usage.

Closed issues:

• Support PythonCall.jl? (#11)
• Example for CPU fails (#13)

Merged pull requests:

• fix typo in readme (#12) (@terasakisatoshi)
• Fix Adapt usage (#15) (@rejuvyesh)

v0.2.0

PyCallChainRules v0.2.0

Diff since v0.1.1

• Use dlpack (via DLPack.jl) for zero-copy tensor sharing with python
• Enable optional CUDA support

Merged pull requests:

• Use dlpack for array interop (#10) (@rejuvyesh)

v0.1.1

PyCallChainRules v0.1.1

Diff since v0.1.0

• Fix support for pytorch modules with buffers.
• Add benchmark suite

v0.1.0

PyCallChainRules v0.1.0

Merged pull requests:

• CompatHelper: add new compat entry for Zygote at version 0.6, (keep existing compat) (#1) (@github-actions[bot])
• CompatHelper: add new compat entry for ChainRulesCore at version 1, (keep existing compat) (#2) (@github-actions[bot])