Silent crash on KSampler

### Expected Behavior

I have low-end machine Ryzen 780M that borrows 12 GB of memory out of 32 GB availoable. I expect some image generation models will work.

### Actual Behavior

Every, even the simplest, model that I tried crash silently on KSampler stage without any error message.

### Steps to Reproduce

CPU spikes to 50%, GPU remains clear (I use radeontop to check)

RUN with the command
```sh
uv run main.py
```

### Debug Logs

```powershell
HSA_OVERRIDE_GFX_VERSION=11.0.0 uv run main.py --novram --cpu-vae
[START] Security scan
[DONE] Security scan
## ComfyUI-Manager: installing dependencies done.
** ComfyUI startup time: 2025-04-19 09:48:10.603
** Platform: Linux
** Python version: 3.12.9 (main, Feb 12 2025, 14:50:50) [Clang 19.1.6 ]
** Python executable: /home/webui/ComfyUI/.venv/bin/python3
** ComfyUI Path: /home/webui/ComfyUI
** ComfyUI Base Folder Path: /home/webui/ComfyUI
** User directory: /home/webui/ComfyUI/user
** ComfyUI-Manager config path: /home/webui/ComfyUI/user/default/ComfyUI-Manager/config.ini
** Log path: /home/webui/ComfyUI/user/comfyui.log

Prestartup times for custom nodes:
0.7 seconds: /home/webui/ComfyUI/custom_nodes/comfyui-manager

Checkpoint files will always be loaded safely.
Total VRAM 11872 MB, total RAM 20480 MB
pytorch version: 2.5.1+rocm6.2
AMD arch: gfx1100
Set vram state to: NO_VRAM
Device: cuda:0 AMD Radeon Graphics : native
Using sub quadratic optimization for attention, if you have memory or speed issues try using: --use-split-cross-attention
Python version: 3.12.9 (main, Feb 12 2025, 14:50:50) [Clang 19.1.6 ]
ComfyUI version: 0.3.29
ComfyUI frontend version: 1.16.8
[Prompt Server] web root: /home/webui/ComfyUI/.venv/lib/python3.12/site-packages/comfyui_frontend_package/static
### Loading: ComfyUI-Manager (V3.31.10)
[ComfyUI-Manager] network_mode: public
### ComfyUI Revision: 3347 [93292bc4] *DETACHED | Released on '2025-04-17'
ComfyUI-GGUF: Partial torch compile only, consider updating pytorch

Import times for custom nodes:
0.0 seconds: /home/webui/ComfyUI/custom_nodes/websocket_image_save.py
0.0 seconds: /home/webui/ComfyUI/custom_nodes/ComfyUI-GGUF
0.0 seconds: /home/webui/ComfyUI/custom_nodes/comfyui-manager

Starting server

To see the GUI go to: http://127.0.0.1:8188
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/alter-list.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/model-list.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/github-stats.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/extension-node-map.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/custom-node-list.json
FETCH ComfyRegistry Data: 5/82
FETCH ComfyRegistry Data: 10/82
FETCH ComfyRegistry Data: 15/82
FETCH ComfyRegistry Data: 20/82
FETCH ComfyRegistry Data: 25/82
FETCH ComfyRegistry Data: 30/82
FETCH ComfyRegistry Data: 35/82
FETCH ComfyRegistry Data: 40/82
FETCH ComfyRegistry Data: 45/82
FETCH ComfyRegistry Data: 50/82
got prompt
model weight dtype torch.float16, manual cast: None
model_type EPS
Using split attention in VAE
Using split attention in VAE
FETCH ComfyRegistry Data: 55/82
VAE load device: cpu, offload device: cpu, dtype: torch.float32
Requested to load SD1ClipModel
loaded completely 9.5367431640625e+25 235.84423828125 True
CLIP/text encoder model load device: cpu, offload device: cpu, current: cpu, dtype: torch.float16
```

### Other

_No response_