Search results

Add a mode argument to ensure in generalize, users are forced to provide a train/val split.

Status: Open.

#144 In gepa-ai/gepa;

· LakshyAAAgrawal opened on Dec 5, 2025

Advice on integrating GEPA for CUDA kernels

Status: Open.

#143 In gepa-ai/gepa;

· siddagra opened on Nov 24, 2025

Run multiple rollouts per task

Status: Open.

#139 In gepa-ai/gepa;

· jjwang01 opened on Nov 17, 2025

Add Docs (MkDocs + ReadTheDocs) with auto deployment.

Status: Open.

#138 In gepa-ai/gepa;

· LakshyAAAgrawal opened on Nov 17, 2025

log_metrics for MLFlow raise excemption for string value

Status: Open.

#137 In gepa-ai/gepa;

· BnnaFish opened on Nov 12, 2025

Ability to provide N candidates instead of just 1 seed

Status: Open.

#134 In gepa-ai/gepa;

· LakshyAAAgrawal opened on Nov 11, 2025

Feature Request: Mid-Training k-Fold Cross-Validation in GEPA

Status: Open.

#133 In gepa-ai/gepa;

· nickarafyllis opened on Nov 10, 2025

GEPA reflection minibatch size is akin to learning_rate. We should have a LR schedule strategy support that can modify it dynamically

Status: Open.

#132 In gepa-ai/gepa;

· LakshyAAAgrawal opened on Nov 10, 2025

Support for pairwise comparison

Status: Open.

#128 In gepa-ai/gepa;

· LakshyAAAgrawal opened on Nov 5, 2025

With val_evaluation_policy, there shouldn't be a distinction between trainset and valset

Status: Open.

#125 In gepa-ai/gepa;

· LakshyAAAgrawal opened on Oct 26, 2025

Dealing with VLM

Status: Open.

#121 In gepa-ai/gepa;

· UlrickBL opened on Oct 22, 2025

Once the score reaches 80%, many minibatches are discarded due to getting perfect scores

Status: Open.

#117 In gepa-ai/gepa;

· LakshyAAAgrawal opened on Oct 19, 2025