HiDream E1 isn't loading and crash

### Expected Behavior

hidream-e1 works

### Actual Behavior

I updated comfy to latest version, and download the bf16 hidream-e1 checkpoint. When the load diffusion model nodes runs, it just crashes the comfy without any error message.
I have two rtx 5090 cards and 96GB RAM. And the hidream-l1 fp8 workflow works well for me.
If someone has it working with 3090, why will not 5090?
I try fp8 or default option in load diffusion node, nothing is different.
It seems to crash between clip and diffusion loading.

### Steps to Reproduce

run a hidream-e1 workflow

### Debug Logs

```powershell
got prompt
Using xformers attention in VAE
Using xformers attention in VAE
VAE load device: cuda:0, offload device: cpu, dtype: torch.bfloat16
Requested to load AutoencodingEngine
loaded completely 28255.125 159.87335777282715 True
Using scaled fp8: fp8 matrix mult: False, scale input: False
CLIP/text encoder model load device: cuda:0, offload device: cpu, current: cpu, dtype: torch.float16
Requested to load HiDreamTEModel_
loaded completely 29149.926642227172 18303.292846679688 True

and comfyui crash...
```

### Other

_No response_