Best Cloud GPU Providers for Kohya_ss SDXL Trainer

Kohya_ss is the most popular open-source trainer for custom character, style, and concept LoRAs for Stable Diffusion XL, SD 1.5, and Flux.

Verified 1-Click Cloud Launchers for Kohya_ss SDXL Trainer

Click to launch directly with pre-installed CUDA drivers, PyTorch, and template weights

RunPodInstant Launch

Kohya_ss SD-Scripts Template

From$0.44/hr

Launch

Vast.aiInstant Launch

Vast Kohya SDXL Training Image

From$0.22/hr

Launch

Direct Docker Run Command

Standard Container Startup

docker run --gpus all -d -p 7860:7860 \
  -v /workspace/dataset:/dataset \
  -v /workspace/output:/output \
  ashleyk/kohya-ss-gui:latest

Deployment Best Practices & Gotchas:

Use AdaFactor or AdamW8bit optimizer to train SDXL LoRAs within 24GB VRAM.
Enable xFormers or FlashAttention to double step throughput.
Store output models in network persistent volumes to protect against spot pod eviction.

Recommended GPU Hardware for Kohya_ss SDXL Trainer

To achieve reliable throughput without running into CUDA Out of Memory (OOM) errors, we recommend deploying on one of the following verified GPU models:

RTX 4090 (24GB)RTX 3090 (24GB)A100 (80GB)