#7376 Allow seeding the Orthogonal Initializer and use QR instead of gramSchmidt

[cosminc98]

Allow seeding the Orthogonal Initializer and add test. Use QR instead of gramSchmidt for consistency with the python implementation.

To see the logs from the Cloud Build CI, please join either our discussion or announcement mailing list.

---

This change is 

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[pyu10055]

Reviewable status: 0 of 1 approvals obtained (waiting on @cosminc98)

---

tfjs-layers/src/initializers.ts line 557 at r1 (raw file):

      for (const dim of shape.slice(0, -1)) {
        numRows *= dim;
      }

You can use the util.sizeFromShape(shape) method from tfjs-core

Code quote:

      let numRows = 1;
      for (const dim of shape.slice(0, -1)) {
        numRows *= dim;
      }

---

tfjs-layers/src/initializers.ts line 568 at r1 (raw file):

      const maxDim = Math.max(numCols, numRows);
      const minDim = Math.min(numCols, numRows);
      const flatShape = [maxDim, minDim];

Prefer removing the intermediate variables:
const flatShape = [Math.max(numCols, numRows), Math.min(numCols, numRows)];

Code quote:

      const maxDim = Math.max(numCols, numRows);
      const minDim = Math.min(numCols, numRows);
      const flatShape = [maxDim, minDim];

[pyu10055]

Reviewable status: complete! 1 of 1 approvals obtained

[mattsoulanille]

LGTM