*   Extend scatter operations to work with a scalar update parameter.
    *   Move cuDNN RNN ops to core for use in TensorFlow codebase only.
    *   Add `float64` support for `Conv2d`, `Conv2dBackpropInput`, and
        `Conv2dBackpropFilter`.
    *   Add `float64` support for `AvgPool`/`AvgPoolGrad`.
    *   Make graph name scope thread local so that they work correctly in
        multi-threaded environments.

    * This is not really a bug in FastAPI but in Swagger UI, nevertheless pinning the version will work while a solution is found on the [Swagger UI side](https://github.com/swagger-api/swagger-ui/issues/9337).

### Docs

reducing modulo m each time. for i >= 0 { out.shiftIn(x.limbs[i], m) i-- } return out } // ExpandFor ensures x has the right size to work with operations modulo m. // // The announced size of x must be smaller than or equal to that of m. func (x *Nat) ExpandFor(m *Modulus) *Nat { return x.expand(len(m.nat.limbs)) } // resetFor ensures out has the right size to work with operations modulo m. // // out is zeroed and may start at any size. func (out *Nat) resetFor(m *Modulus) *Nat { return out.reset(len(m.nat.limbs))...

reducing modulo m each time. for i >= 0 { out.shiftIn(x.limbs[i], m) i-- } return out } // ExpandFor ensures x has the right size to work with operations modulo m. // // The announced size of x must be smaller than or equal to that of m. func (x *Nat) ExpandFor(m *Modulus) *Nat { return x.expand(len(m.nat.limbs)) } // resetFor ensures out has the right size to work with operations modulo m. // // out is zeroed and may start at any size. func (out *Nat) resetFor(m *Modulus) *Nat { return out.reset(len(m.nat.limbs))...