80(DX), X11 PXOR X11, X5 MOVOU 96(DX), X11 PXOR X11, X6 MOVOU 112(DX), X11 PXOR X11, X7 MOVOU X0, (SI) MOVOU X1, 16(SI) MOVOU X2, 32(SI) MOVOU X3, 48(SI) MOVOU X4, 64(SI) MOVOU X5, 80(SI) MOVOU X6, 96(SI) MOVOU X7, 112(SI) LEAQ 128(SI), SI LEAQ 128(DX), DX JMP gcmAesDecOctetsLoop gcmAesDecEndOctets: SUBQ $0x07, R10 gcmAesDecSingles: MOVOU 16(AX), X1 MOVOU 32(AX), X2 MOVOU 48(AX), X3 MOVOU 64(AX), X4 MOVOU 80(AX), X5 MOVOU 96(AX), X6 MOVOU 112(AX), X7 MOVOU 224(DI), X13 gcmAesDecSinglesLoop: CMPQ R9,...

pkg debug/dwarf, const AttrStrOffsetsBase Attr
pkg debug/dwarf, const AttrStringLengthBitSize = 111
pkg debug/dwarf, const AttrStringLengthBitSize Attr
pkg debug/dwarf, const AttrStringLengthByteSize = 112
pkg debug/dwarf, const AttrStringLengthByteSize Attr
pkg debug/dwarf, const AttrThreadsScaled = 98
pkg debug/dwarf, const AttrThreadsScaled Attr
pkg debug/dwarf, const ClassAddrPtr = 15

* Make Swagger UI and ReDoc URLs parameterizable, allowing to host and serve local versions of them and have offline docs. PR [#112](https://github.com/tiangolo/fastapi/pull/112) by [@euri10](https://github.com/euri10).

## 0.22.0

* Add support for `dependencies` parameter:

        `tf.data.Dataset.sample_from_datasets` and deprecate the experimental
        endpoint.
    *   Added `TF_GPU_ALLOCATOR=cuda_malloc_async` that use cudaMallocAsync from
        CUDA 11.2. This could become the default in the future.
*   TF SavedModel:
    *   Custom gradients are now saved by default. See
        `tf.saved_model.SaveOptions` to disable this.