dfdx Files

Breaking changes

• [Breaking] Adding Tensor::try_realize, and Tensor::realize no longer returns Result by @coreylowman in https://github.com/coreylowman/dfdx/pull/758
• [Breaking] ReshapeTo::reshape_like and ReshapeTo::try_reshape_like now panic instead of returning option by @coreylowman in https://github.com/coreylowman/dfdx/pull/766
• [Breaking] Adding dilation/groups to Conv2D. Adding dilation to Pool2D by @coreylowman in https://github.com/coreylowman/dfdx/pull/767
• [Breaking] Use gemm for matmul. Removes support for matrixmultiply & MKL by @coreylowman in https://github.com/coreylowman/dfdx/pull/776
• [Breaking] Moving storage GAT to trait level generic. Split DeviceStorage into multiple traits by @coreylowman in https://github.com/coreylowman/dfdx/pull/782
• [Breaking] Adding dilation/groups to ConvTranspose2D by @coreylowman in https://github.com/coreylowman/dfdx/pull/783

What's Changed

• Adding f16 as Dtype by @coreylowman in https://github.com/coreylowman/dfdx/pull/696
• Adding example by @sirandreww in https://github.com/coreylowman/dfdx/pull/740
• Adds TryConcatAlong to support Concat along any axis by @coreylowman in https://github.com/coreylowman/dfdx/pull/750
• Changed CUDA_ARCH in compatibility.cuh by @jafioti in https://github.com/coreylowman/dfdx/pull/752
• Allow broadcast_like to accept tensors OR shapes by @VasanthakumarV in https://github.com/coreylowman/dfdx/pull/751
• Removing rerun build.rs for output destination by @coreylowman in https://github.com/coreylowman/dfdx/pull/754
• Fixing compatibility for compute cap 70-75 by @coreylowman in https://github.com/coreylowman/dfdx/pull/757
• Adds TriangleTensor and CmpKernel traits to Device bound by @coreylowman in https://github.com/coreylowman/dfdx/pull/760
• Using Bernoulli distribution in dropout - makes dropout reproducible across dtypes by @coreylowman in https://github.com/coreylowman/dfdx/pull/761
• Fixes bug with f16 mean where number of elements reduced was f16::INF by @coreylowman in https://github.com/coreylowman/dfdx/pull/763
• Placeholder f16 gemm speedups by @coreylowman in https://github.com/coreylowman/dfdx/pull/765
• MultiHeadAttention 3d impl now broadcasts to 4d instead of duplicating logic by @coreylowman in https://github.com/coreylowman/dfdx/pull/768
• Moving cudarc?/f16 behind f16 feature by @coreylowman in https://github.com/coreylowman/dfdx/pull/774
• impl Clone for Adam, SGD, RMSprop by @coreylowman in https://github.com/coreylowman/dfdx/pull/775
• Properly setting read_dst for gemm in forward/backward pass by @coreylowman in https://github.com/coreylowman/dfdx/pull/777
• Adds rayon dependency. Using gemm::Parallelism::Rayon(rayon::current_num_threads()) by @coreylowman in https://github.com/coreylowman/dfdx/pull/778
• Add LogSoftmax by @kurnevsky in https://github.com/coreylowman/dfdx/pull/769
• Moving some tests off nightly. Adding docs to conv2d op by @coreylowman in https://github.com/coreylowman/dfdx/pull/779
• Adding better error messages if nvidia-smi/nvcc are not found by @coreylowman in https://github.com/coreylowman/dfdx/pull/784
• Using for loop with gridDim.x * blockDim.x as increment by @coreylowman in https://github.com/coreylowman/dfdx/pull/787
• Removing __hmax and __hmin compat functions by @coreylowman in https://github.com/coreylowman/dfdx/pull/788
• Uses grid striding in fill_with by @coreylowman in https://github.com/coreylowman/dfdx/pull/790
• Exposed NumpyDType publicly by @jafioti in https://github.com/coreylowman/dfdx/pull/791
• Fixing weight shape for grouped Conv2D by @coreylowman in https://github.com/coreylowman/dfdx/pull/797
• Bump half/cudarc versions by @coreylowman in https://github.com/coreylowman/dfdx/pull/805
• Using Groups in conv weight init by @coreylowman in https://github.com/coreylowman/dfdx/pull/806
• Add scalar support to TensorCollection by @nkoppel in https://github.com/coreylowman/dfdx/pull/799

New Contributors

• @sirandreww made their first contribution in https://github.com/coreylowman/dfdx/pull/740
• @kurnevsky made their first contribution in https://github.com/coreylowman/dfdx/pull/769

Full Changelog: https://github.com/coreylowman/dfdx/compare/v0.11.2...v0.12.0