Feature req - parallel batching across multiple GPUs

Nobody has claimed this yet.

I get that splitting a single gen across multiple GPUs is tough, and there's at least one still-open issue regarding this. But what about using multiple GPUs in parallel, and just letting each do its own gen based on the same prompt/settings? There would be no need to coordinate between GPUs, just have each load the same model and run gens based on the same prompt, settings (different seeds, obviously), etc at the same time - separate, but in parallel.

I have three 3090s, and I was considering doing this manually by running multiple instances of forge (one per GPU). But I think it shouldn't be too terribly difficult to have forge do this automatically - assuming the user has similar-enough GPUs.

The issue names no files, tests, or entry points. Start by mapping Forge’s model-loading and generation orchestration paths, then determine how separate GPU workers could receive identical prompts and settings with independent seeds. Done means Forge can launch and coordinate independent generations across multiple compatible GPUs without requiring manual instances.

Written by the indexing model from the issue text.