#965 add hydra plugin Pull request opened by samhita-alla Adds

flyteplugins-hydra

, a Hydra integration for Flyte that lets users compose Hydra configs and launch Flyte tasks across three entry points: •

hydra/launcher=flyte

for existing

@hydra.main

workflows •

hydra_run

/

hydra_sweep

Python SDK helpers •

flyte hydra run

for Flyte CLI usage without requiring a

@hydra.main

entry point What this plugin supports • Single-run and multirun execution for Flyte tasks using Hydra-composed configs • Remote execution by default, with local execution available via: •

hydra.launcher.mode=local

•

mode="local"

•

flyte hydra run --local

• Fire-and-forget remote submission via: •

hydra.launcher.wait=false

•

wait=False

•

--no-wait

• Basic grid sweeps and custom Hydra sweepers such as Optuna • YAML configs and structured

ConfigStore

configs • App-level overrides using the task’s actual

DictConfig

parameter name • Hydra namespace overrides through

--hydra-override

• Entry-task

task.override(...)

support through configurable task environment config, defaulting to

task_env

• Child task override patterns documented for users to apply inside their own pipeline code uv run flyte hydra run --multirun --config-path conf --config-name training \ --project flytetester --domain production \ --name my-sweep \ train.py pipeline --dataset s3://my-bucket/imagenet \ --hydra-override "hydra/sweeper=optuna" \ --hydra-override "hydra.sweeper.n_trials=20" \ --hydra-override "hydra.sweeper.n_jobs=4" \ --cfg "optimizer.lr=interval(1e-4,1e-1)" [2026-04-13 173310,043][HYDRA] setup plugin alembic.autogenerate.schemas [2026-04-13 173310,043][HYDRA] setup plugin alembic.autogenerate.tables [2026-04-13 173310,043][HYDRA] setup plugin alembic.autogenerate.types [2026-04-13 173310,043][HYDRA] setup plugin alembic.autogenerate.constraints [2026-04-13 173310,043][HYDRA] setup plugin alembic.autogenerate.defaults [2026-04-13 173310,043][HYDRA] setup plugin alembic.autogenerate.comments [I 2026-04-13 173311,040] A new study created in memory with name: no-name-266c9173-7396-4cc6-965e-9ba2eceff429 [2026-04-13 173311,040][HYDRA] Study name: no-name-266c9173-7396-4cc6-965e-9ba2eceff429 [2026-04-13 173311,040][HYDRA] Storage: None [2026-04-13 173311,040][HYDRA] Sampler: TPESampler [2026-04-13 173311,040][HYDRA] Directions: ['minimize'] [2026-04-13 173311,042][HYDRA] #0 : optimizer.lr=0.01819662332129959 > Building 1 image... ... Flyte run submitted [0]: ... [2026-04-13 173315,413][flyteplugins.hydra._launcher][INFO] - #1 : optimizer.lr=0.047139913742000106 Flyte run submitted [1]: ... [2026-04-13 173316,063][flyteplugins.hydra._launcher][INFO] - #2 : optimizer.lr=0.08064613359196803 ... [2026-04-13 173340,335][flyteplugins.hydra._launcher][INFO] - Waiting for 4 Flyte runs to complete. [2026-04-13 173340,904][hydra_plugins.hydra_optuna_sweeper._impl][INFO] - Best parameters: {'optimizer.lr': 0.01819662332129959} [2026-04-13 173340,904][hydra_plugins.hydra_optuna_sweeper._impl][INFO] - Best value: 0.5779719250354236 python train.py --multirun \ hydra/launcher=flyte hydra.launcher.mode=remote \ optimizer.lr=0.001,0.01,0.1 training.epochs=10,20 A detailed description is available in the plugin README. flyteorg/flyte-sdk