Ax

Ax copied to clipboard

Warn when modelbridge generates repeat candidates

2

Summary: The aim was to do this as low level as applicable. Reviewed By: bletham Differential Revision: D52735217

danielcohenlive

Update all instances of MinimumPreferenceOccurances to directly use TransitionCriterion

1

Summary: We have replaced the more limited MinimumTrialsInStatus with the more flexible TransitionCriterion MinTrials. This diff updates MinimumPreferenceOccurances to directly inherit from its source in TransitionCriterion file In following diffs...

mgarrard

I have the following script where I'm simply minimizing / maximizing $\sum_i(x_i)$. It works for `botorch.acquisition.qExpectedImprovement`. However, for `botorch.acquisition.qMaxValueEntropy`, only maximize works. Minimize=true would also result in maximization. Reproduction: For...

rentruewang

Use scheduler and early stopping strategy

2

Summary: Scheduler now needs to be able to: - save trials with new data even if their status didn't update - catch extra fetching errors. it already handled most of...

danielcohenlive

`AxClient.get_best_trial()` with wrong index

6

I was working with the `Service API` in a constrained single-objective set and the `get_best_trial()` function returns me the parameters from a trial, but with the wrong index. This is...

leandrobbraga

add config option for inverse map in TrialAsTask

3

Summary: If the trial_level_map maps multiple trial indices to a single value, then it is convenient to be able to specify which trial index a given value should be mapped...

sdaulton

Submitit ask-tell tutorial

4

Adds a tutorial that shows how to use the AxClient to schedule jobs and tune parameters on a Slurm cluster via SubmitIt. Code changes: 1. The tutorial notebook 2. Linking...

mhavasi

Submitit integration

1

This pull requests adds a runner for the SubmitIt package (https://github.com/facebookincubator/submitit). With the SubmitIt package, we can use the Ax scheduler to schedule jobs on a SLURM cluster. The integration...

mhavasi

[QUESTION] MAB with Thompson Sampling not converging

3

I'm new to Ax and want to run Bandit with Thompson Sampling. I have a fixed number of arms with initial weights. When I tried the following code, it doesn't...

seabull

Question : modifications of compute_posterior_pareto_frontier

1

Hello, When using compute_posterior_pareto_frontier and adding outcome_constraints , the line : run = model.gen(1, model_gen_options=model_gen_options, optimization_config=oc) is giving us one point. If we add : run = model.gen(tests, model_gen_options=model_gen_options, optimization_config=oc)...

Kh-im