ComfyUI Wan2.2 i2v workflow may no longer keep models cached in RAM
Fun-Class3451 · reddit · 2026-07-22
A ComfyUI user reports that the new Wan2.2 i2v standard workflow does not seem to keep models cached in RAM between runs, unlike the older KJ Workflow with WanVideoHelper Nodes.
- After upgrading to int8 quantized models, the system only uses about 10 GB of RAM during the run.
- The user says that in the previous workflow, models stayed cached in memory, which avoided reloading from SSD on every run or model switch.
- They ask whether this is expected behavior or if a startup flag is needed.
- The post is essentially a local inference workflow troubleshooting note for Wan2.2 in ComfyUI.
More from Infra
- AWS tutorial: fleet-wide edge CV model lifecycle with SageMaker SDK v3, ~$0.07 per run — AWS ML Blog · 2026-10-27
- Qwen3.8-27B hits 40 tok/s on Jetson Thor with INT4 + MTP, accuracy beats FP16 — No-Dragonfly6246 · 2026-10-27
- Celesto launches CI runners nearly 90% cheaper than GitHub Actions, with agent sandboxes — aniketmaurya · 2026-10-27
- Supabase founders on rebuilding the backend for a billion AI-built apps — Latent Space · 2026-10-27
- AW-0 runs at 230ms on a single RTX 5090, 2.4x faster than autoregressive code planning — trevordarrell · 2026-10-27
- Site tracks US city council topics: data centers rank only behind pickleball courts — zakkohane · 2026-10-27