Extend 2Dim embedding bag benchmarking to include 3Dim benchmarks (#64647)
authorEddie Ren <edwardren@fb.com>
Fri, 10 Sep 2021 19:31:27 +0000 (12:31 -0700)
committerFacebook GitHub Bot <facebook-github-bot@users.noreply.github.com>
Fri, 10 Sep 2021 23:49:02 +0000 (16:49 -0700)
commit3fbb49e75dc3e7d1d5b14e8a18e983ef9e04ae0b
tree71dab9868bc778b30c1f511bfa922bcb1cd3c03f
parent227aafd1d93d23c3f85564c94f03e91a0fff8fa0
Extend 2Dim embedding bag benchmarking to include 3Dim benchmarks (#64647)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64647

Add support for benchmarking of 8 bit quantizations of N-D batched embeddings. Currently only works for 3Dim embeddings and still requires thought on ramping up from 3Dim to NDim.

Test Plan: ```buck run //caffe2/benchmarks/operator_benchmark/pt:qembedding_pack_test```

Reviewed By: jingsh

Differential Revision: D30770085

fbshipit-source-id: 26659020f3458991592065a05366bde0f060494e
benchmarks/operator_benchmark/pt/qembedding_pack_test.py