Changes to random utils #625

priyakasimbeg · 2024-02-02T22:59:28Z

Update function signatures for _split and _fold_in to accurately reflect SeedType inputs and outputs.
Change _fold_in to use data for sampling new seed instead of returning data.
Force PyTorch function SeedType outputs to be 1-dim as opposed to 2-dim which is typical for Jax PRNGKeys
Update run_workloads.py to use newly added randint for getting workload seeds.

github-actions · 2024-02-02T22:59:38Z

MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅

algorithmic_efficiency/random_utils.py

priyakasimbeg added 2 commits February 2, 2024 01:46

fix fold_in in pytorch

c794339

random utils fixes

247dcb0

priyakasimbeg changed the base branch from main to dev February 2, 2024 23:00

priyakasimbeg added 16 commits February 2, 2024 23:06

remove indexing from rngs in pytorch workloads

759c90d

formatting

6baacd7

formatting

368e348

fix seed shapes

4be495f

fix dataset seed

5495d72

fix rng utils

355ebe0

fix

43c381a

remove unused types from random_utils

f81e877

fix overflow error in jax sampling

3494f16

use bits instead of randint

665049a

add numpy bits function

d0d3e3e

change seed method call for mnist dataset

e4476aa

add documentation

9ebdca7

add fold in method

f888a99

reformatting

4b30a6e

add type signature to _fold_in

43b27b2

georgedahl previously approved these changes Feb 7, 2024

View reviewed changes

algorithmic_efficiency/random_utils.py Outdated Show resolved Hide resolved

change _bits

1fc4ce2

priyakasimbeg dismissed georgedahl’s stale review via 1fc4ce2 February 7, 2024 00:22

priyakasimbeg marked this pull request as ready for review February 7, 2024 00:24

priyakasimbeg requested a review from a team as a code owner February 7, 2024 00:24

priyakasimbeg merged commit 3421276 into dev Feb 7, 2024
31 checks passed

github-actions bot locked and limited conversation to collaborators Feb 7, 2024

Provide feedback