Using fp8_e4m3fn_fast results in corrupted images with Qwen

### Custom Node Testing

- [x] I have tried disabling custom nodes and the issue persists (see [how to disable custom nodes](https://docs.comfy.org/troubleshooting/custom-node-issues#step-1%3A-test-with-all-custom-nodes-disabled) if you need help)

### Expected Behavior

Images of comparable quality to fp8_e4m3fn.

### Actual Behavior

An image that looks only 10% denoised. See image.

### Steps to Reproduce

Use the default Qwen workflow, switch to fp8_e4m3fn_fast in the Load Diffusion Model node, generate.

### Debug Logs

```powershell
No apparent problems in the log:

loaded completely 21074.77170448303 7263.013442993164 True
model weight dtype torch.float8_e4m3fn, manual cast: torch.bfloat16
model_type FLUX
Requested to load QwenImage
loaded completely 21143.983919891358 19483.948791503906 True
100%|██████████████████████████████████████████████████████████████████████████████████| 20/20 [00:39<00:00, 1.99s/it]
Requested to load WanVAE
0 models unloaded.
loaded partially 128.0 127.9998779296875 0
Prompt executed in 66.25 seconds
```

### Other

By the way, thank you for the day 1 Qwen support!!