# via
    #   -r ci/official/requirements_updater/requirements.in
    #   h5py
    #   jax
    #   keras-nightly
    #   ml-dtypes
    #   opt-einsum
    #   scipy
    #   tb-nightly
nvidia-cublas-cu12==12.5.3.2 \
    --hash=sha256:4960f3dc5f39699acadf76fa6d94b10a2a00f2956c2c442efa299fb22b0748f3 \

    # via
    #   -r ci/official/requirements_updater/requirements.in
    #   h5py
    #   jax
    #   keras-nightly
    #   ml-dtypes
    #   opt-einsum
    #   scipy
    #   tb-nightly
nvidia-cublas-cu12==12.5.3.2 \
    --hash=sha256:4960f3dc5f39699acadf76fa6d94b10a2a00f2956c2c442efa299fb22b0748f3 \

    # via
    #   -r ci/official/requirements_updater/requirements.in
    #   h5py
    #   jax
    #   keras-nightly
    #   ml-dtypes
    #   opt-einsum
    #   scipy
    #   tb-nightly
nvidia-cublas-cu12==12.5.3.2 \
    --hash=sha256:4960f3dc5f39699acadf76fa6d94b10a2a00f2956c2c442efa299fb22b0748f3 \

    # via
    #   -r ci/official/requirements_updater/requirements.in
    #   h5py
    #   jax
    #   keras-nightly
    #   ml-dtypes
    #   opt-einsum
    #   scipy
    #   tb-nightly
nvidia-cublas-cu12==12.5.3.2 \
    --hash=sha256:4960f3dc5f39699acadf76fa6d94b10a2a00f2956c2c442efa299fb22b0748f3 \

    # via
    #   -r ci/official/requirements_updater/requirements.in
    #   h5py
    #   jax
    #   keras-nightly
    #   ml-dtypes
    #   opt-einsum
    #   scipy
    #   tb-nightly
nvidia-cublas-cu12==12.5.3.2 \
    --hash=sha256:4960f3dc5f39699acadf76fa6d94b10a2a00f2956c2c442efa299fb22b0748f3 \

    # via
    #   -r ci/official/requirements_updater/requirements.in
    #   h5py
    #   jax
    #   keras-nightly
    #   ml-dtypes
    #   opt-einsum
    #   scipy
    #   tb-nightly
nvidia-cublas-cu12==12.5.3.2 \
    --hash=sha256:4960f3dc5f39699acadf76fa6d94b10a2a00f2956c2c442efa299fb22b0748f3 \

    # via
    #   -r ci/official/requirements_updater/requirements.in
    #   h5py
    #   jax
    #   keras-nightly
    #   ml-dtypes
    #   opt-einsum
    #   scipy
    #   tb-nightly
nvidia-cublas-cu12==12.5.3.2 \
    --hash=sha256:4960f3dc5f39699acadf76fa6d94b10a2a00f2956c2c442efa299fb22b0748f3 \

    # via
    #   -r ci/official/requirements_updater/requirements.in
    #   h5py
    #   jax
    #   keras-nightly
    #   ml-dtypes
    #   opt-einsum
    #   scipy
    #   tb-nightly
nvidia-cublas-cu12==12.5.3.2 \
    --hash=sha256:4960f3dc5f39699acadf76fa6d94b10a2a00f2956c2c442efa299fb22b0748f3 \

   *       example both {@code Enum<?>} and {@code Enum<? extends Enum<?>>} canonicalize to {@code
   *       Enum<? extends Enum<E>}.
   *   <li>{@code canonicalize(t)} produces a "literal" supertype of t. For example: {@code Enum<?
   *       extends Enum<?>>} canonicalizes to {@code Enum<?>}, which is a supertype (if we disregard
   *       the upper bound is implicitly an Enum too).

   * <p><b>Note:</b> if elements are non-null and won't be added or removed after this point, use
   * {@link ImmutableSet#of()} or {@link ImmutableSet#copyOf(Object[])} instead. If {@code E} is an
   * {@link Enum} type, use {@link EnumSet#of(Enum, Enum[])} instead. Otherwise, strongly consider
   * using a {@code LinkedHashSet} instead, at the cost of increased memory footprint, to get
   * deterministic iteration behavior.
   *