Execution¶

Maturity: verified

Execution separates the application that performs training from the executor that provisions a worker, starts it, and manages the provider lifecycle.

Single-process execution¶

The current application runs one Python process on either a CPU or an NVIDIA CUDA GPU. The shipped Gemini Enterprise Agent Platform preset requests one replica and one NVIDIA L4 GPU. Provider-managed execution is not the same thing as distributed training.

Application responsibility¶

The application resolves configuration and mounted gs:// paths, reads the canonical dataset, constructs the selected program, performs bounded optimizer steps, and publishes one attempt’s evidence.

Executor responsibility¶

The caller or provider adapter owns the mounted Cloud Storage namespace, container lifecycle, source and image provenance inputs, job submission, monitoring, retry policy, cancellation, and provider cleanup.

The committed Gemini Enterprise Agent Platform submission script validates the launch YAML, full 40-digit source commit, immutable image digest, safe run and attempt IDs, and paired resume inputs. It uploads the exact launch YAML with create-only semantics and submits a CustomJob. It does not wait for success.

Distributed training limitations¶

There is no DistributedDataParallel, FullyShardedDataParallel, multi-process launcher, rank-aware sampler, collective metric reduction, sharded checkpoint, or elastic restart contract. Documentation and launch configuration must not imply otherwise.