e-body>img{margin:10px}.timeline>div>.timeline-item>.timeline-body ol,.timeline>div>.timeline-item>.timeline-body ul,.timeline>div>.timeline-item>.timeline-body>dl{margin:0}.timeline>div>.timeline-item>.timeline-footer>a{color:#fff}.timeline>div>.fa,.timeline>div>.fab,.timeline>div>.far,.timeline>div>.fas,.timeline>div>.glyphicon,.timeline>div>.ion{background:#adb5bd;border-radius:50%;font-size:15px;height:30px;left:18px;line-height:30px;position:absolute;text-align:center;top:0;width:30px}.time...

    `self_adjoint_eig` or `self_adjoint_eigvals`.
*   Eliminated `batch_*` methods for most linear algebra and FFT ops and
    promoted the non-batch version of the ops to handle batches of matrices.
*   Tracing/timeline support for distributed runtime (no GPU profiler yet).
*   C API gives access to inferred shapes with `TF_GraphGetTensorNumDims` and
    `TF_GraphGetTensorShape`.
*   Shape functions for core ops have moved to C++ via