google-research / google-research/google-research

Incorrect lower/upper bounds for exponential decay in TaskSet code

Open
#2,736 0 comments 0 reactions 0 assignees View on GitHub
Dominant language
Jupyter Notebook
Stars
38.8k
Forks
8.5k
PR merge metrics
No merged PRs in 30d

Description

Hi,

Shouldn't this line:
https://github.com/google-research/google-research/blob/05e3a6881c67b537722588e1d7c612b6c2fa8282/task_set/optimizers/adam8p.py#L214

be:
```python
"exponential_decay": utils.sample_log_float(rng, 1e-6, 1e-3),
```

since based on the [definition of `sample_log_float`](https://github.com/google-research/google-research/blob/05e3a6881c67b537722588e1d7c612b6c2fa8282/task_set/optimizers/utils.py#L21) the lower bound should come first?

Contributor guide

Open the contributing guide

Research direction

Open task_set/optimizers/adam8p.py around line 214 and read sample_log_float in task_set/optimizers/utils.py around line 21. Verify the helper's bound ordering and update the exponential_decay arguments if they are reversed; done means the lower bound precedes the upper bound.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Bug
Difficulty
1/5
Estimated time
Under an hour
Activity status
Stale
Clarity
Clearly specified
Newbie friendliness
45/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.