[core] MatMul broadcasting.

[annxingyuan]

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

---

This change is 

[tafsiri]

hey @annxingyuan, looks like tests don't pass for this on node do you want to take a look at that before we review?

[annxingyuan]

yes @tafsiri sorry i thought it was flaking at first - removed reviewers and will re-add when ready

[pyu10055]

Reviewed 7 of 8 files at r1, 5 of 5 files at r2.
Reviewable status: 0 of 1 approvals obtained (waiting on @annxingyuan, @pyu10055, and @tafsiri)

---

tfjs-core/src/ops/mat_mul.ts, line 89 at r2 (raw file):

    const outShapeOuterDims =
        batchDimA > batchDimB ? $a.shape.slice(0, -2) : $b.shape.slice(0, -2);

can these be replaced with variables: outerDimsA and outerDimsB?

---

tfjs-node/src/run_tests.ts, line 84 at r2 (raw file):

  'maxPoolWithArgmax', 'rotate', 'flipLeftRight', 'unique',
  // libtensorflow does not yet support tf.matmul with broadcast
  'broadcast with unequal batch dims', 'broadcast with unequal ranks'

why tests for tfjs-node need to be skipped?

[annxingyuan]

Reviewable status: 0 of 1 approvals obtained (waiting on @pyu10055 and @tafsiri)

---

tfjs-core/src/ops/mat_mul.ts, line 89 at r2 (raw file):

Previously, pyu10055 (Ping Yu) wrote…

can these be replaced with variables: outerDimsA and outerDimsB?

Done

---

tfjs-node/src/run_tests.ts, line 84 at r2 (raw file):

Previously, pyu10055 (Ping Yu) wrote…

why tests for tfjs-node need to be skipped?

Broadcasting support was recently added in tensorflow 2.0 - libtensorflow does not yet support it.

[pyu10055]

Reviewable status: complete! 1 of 1 approvals obtained (waiting on @pyu10055 and @tafsiri)