I cant generate images after update. CUDA error: an illegal instruction was encountered

### Expected Behavior

I shouldnt be getting CUDA error.

### Actual Behavior

I am not able to use comfyui after the last update, it was working fine yesterday.

### Steps to Reproduce

It is not about specific wfs but here it is.
[Flux.json](https://github.com/user-attachments/files/17828571/Flux.json)

### Debug Logs

```powershell
(venv) ubuntu@129-146-162-177:~/cemil-test/ComfyUI$ python main.py --listen 0.0.0.0 --port 8334 --disable-all-custom-nodes
Total VRAM 40326 MB, total RAM 221449 MB
pytorch version: 2.4.1+cu124
Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA A100-SXM4-40GB : cudaMallocAsync
Using pytorch cross attention
[Prompt Server] web root: /home/ubuntu/cemil-test/ComfyUI/web
Adding extra search path loras /home/ubuntu/cemil-test/ai-toolkit/output
/home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/kornia/feature/lightglue.py:44: FutureWarning: `torch.cuda.amp.custom_fwd(args...)` is deprecated. Please use `torch.amp.custom_fwd(args..., device_type='cuda')` instead.
@torch.cuda.amp.custom_fwd(cast_inputs=torch.float32)
Skipping loading of custom nodes
Starting server

To see the GUI go to: http://0.0.0.0:8334
got prompt
model weight dtype torch.float16, manual cast: None
model_type EPS
Using pytorch attention in VAE
Using pytorch attention in VAE
/home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/transformers/tokenization_utils_base.py:1601: FutureWarning: `clean_up_tokenization_spaces` was not set. It will be set to `True` by default. This behavior will be depracted in transformers v4.45, and will be then set to `False` by default. For more details check this issue: https://github.com/huggingface/transformers/issues/31884
warnings.warn(
Requested to load SD1ClipModel
Loading 1 new model
loaded completely 0.0 235.84423828125 True
Requested to load BaseModel
Loading 1 new model
loaded completely 0.0 1639.406135559082 True
20%|█████████████████████████████████ | 4/20 [00:00<00:02, 5.86it/s]terminate called after throwing an instance of 'c10::Error'
what(): CUDA error: misaligned address
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

Exception raised from c10_cuda_check_implementation at ../c10/cuda/CUDAException.cpp:43 (most recent call first):
frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x96 (0x7bdf43f0ef86 in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libc10.so)
frame #1: c10::detail::torchCheckFail(char const*, char const*, unsigned int, std::string const&) + 0x64 (0x7bdf43ebdd10 in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libc10.so)
frame #2: c10::cuda::c10_cuda_check_implementation(int, char const*, char const*, int, bool) + 0x118 (0x7bdf43fe9f08 in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libc10_cuda.so)
frame #3: + 0x587d0 (0x7bdf43fef7d0 in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libc10_cuda.so)
frame #4: + 0x5a4f4 (0x7bdf43ff14f4 in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libc10_cuda.so)
frame #5: + 0x5db920 (0x7bdf41bdb920 in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libtorch_python.so)
frame #6: + 0x6abdf (0x7bdf43ef2bdf in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libc10.so)
frame #7: c10::TensorImpl::~TensorImpl() + 0x21b (0x7bdf43eebc3b in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libc10.so)
frame #8: c10::TensorImpl::~TensorImpl() + 0x9 (0x7bdf43eebde9 in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libc10.so)
frame #9: + 0x11543d7 (0x7bdf2a5543d7 in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libtorch_cpu.so)
frame #10: + 0x2e0927b (0x7bdf2c20927b in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libtorch_cpu.so)
frame #11: at::_ops::_to_copy::redispatch(c10::DispatchKeySet, at::Tensor const&, std::optional, std::optional, std::optional, std::optional, bool, std::optional) + 0xf5 (0x7bdf2b911f15 in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libtorch_cpu.so)
frame #12: + 0x2c47d23 (0x7bdf2c047d23 in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libtorch_cpu.so)
frame #13: at::_ops::_to_copy::call(at::Tensor const&, std::optional, std::optional, std::optional, std::optional, bool, std::optional) + 0x1eb (0x7bdf2b9a197b in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libtorch_cpu.so)
frame #14: at::native::to(at::Tensor const&, std::optional, std::optional, std::optional, std::optional, bool, bool, std::optional) + 0x11d (0x7bdf2b3ca2dd in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libtorch_cpu.so)
frame #15: + 0x300d631 (0x7bdf2c40d631 in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libtorch_cpu.so)
frame #16: at::_ops::to_dtype_layout::redispatch(c10::DispatchKeySet, at::Tensor const&, std::optional, std::optional, std::optional, std::optional, bool, bool, std::optional) + 0x114 (0x7bdf2bad7e24 in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libtorch_cpu.so)
frame #17: + 0x2c47e5e (0x7bdf2c047e5e in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libtorch_cpu.so)
frame #18: at::_ops::to_dtype_layout::call(at::Tensor const&, std::optional, std::optional, std::optional, std::optional, bool, bool, std::optional) + 0x200 (0x7bdf2bb50d10 in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libtorch_cpu.so)
frame #19: + 0x57023b (0x7bdf41b7023b in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libtorch_python.so)
frame #20: + 0x5c4057 (0x7bdf41bc4057 in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libtorch_python.so)

Aborted

(venv) ubuntu@129-146-162-177:~/cemil-test/ComfyUI$ python main.py --listen 0.0.0.0 --port 8334 --disable-all-custom-nodes
Total VRAM 40326 MB, total RAM 221449 MB
pytorch version: 2.4.1+cu124
Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA A100-SXM4-40GB : cudaMallocAsync
Using pytorch cross attention
[Prompt Server] web root: /home/ubuntu/cemil-test/ComfyUI/web
Adding extra search path loras /home/ubuntu/cemil-test/ai-toolkit/output
/home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/kornia/feature/lightglue.py:44: FutureWarning: `torch.cuda.amp.custom_fwd(args...)` is deprecated. Please use `torch.amp.custom_fwd(args..., device_type='cuda')` instead.
@torch.cuda.amp.custom_fwd(cast_inputs=torch.float32)
Skipping loading of custom nodes
Starting server

To see the GUI go to: http://0.0.0.0:8334
got prompt
Using pytorch attention in VAE
Using pytorch attention in VAE
/home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/transformers/tokenization_utils_base.py:1601: FutureWarning: `clean_up_tokenization_spaces` was not set. It will be set to `True` by default. This behavior will be depracted in transformers v4.45, and will be then set to `False` by default. For more details check this issue: https://github.com/huggingface/transformers/issues/31884
warnings.warn(
clip missing: ['text_projection.weight']
Requested to load FluxClipModel_
Loading 1 new model
loaded completely 0.0 9319.23095703125 True
model weight dtype torch.bfloat16, manual cast: None
model_type FLUX
Requested to load Flux
Loading 1 new model
loaded completely 0.0 22700.097778320312 True
0%| | 0/4 [00:00 + 0x587d0 (0x7084091cf7d0 in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libc10_cuda.so)
frame #4: + 0x5a4f4 (0x7084091d14f4 in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libc10_cuda.so)
frame #5: + 0x5db920 (0x708406ddb920 in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libtorch_python.so)
frame #6: + 0x6abdf (0x7084090d2bdf in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libc10.so)
frame #7: c10::TensorImpl::~TensorImpl() + 0x21b (0x7084090cbc3b in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libc10.so)
frame #8: c10::TensorImpl::~TensorImpl() + 0x9 (0x7084090cbde9 in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libc10.so)
frame #9: + 0x891b68 (0x708407091b68 in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libtorch_python.so)
frame #10: THPVariable_subclass_dealloc(_object*) + 0x2c6 (0x708407091eb6 in /home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/torch/lib/libtorch_python.so)

Aborted

(venv) ubuntu@129-146-162-177:~/cemil-test/ComfyUI$ python main.py --listen 0.0.0.0 --port 8334 --disable-all-custom-nodes --disable-cuda-malloc
Total VRAM 40326 MB, total RAM 221449 MB
pytorch version: 2.4.1+cu124
Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA A100-SXM4-40GB : native
Using pytorch cross attention
[Prompt Server] web root: /home/ubuntu/cemil-test/ComfyUI/web
Adding extra search path loras /home/ubuntu/cemil-test/ai-toolkit/output
/home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/kornia/feature/lightglue.py:44: FutureWarning: `torch.cuda.amp.custom_fwd(args...)` is deprecated. Please use `torch.amp.custom_fwd(args..., device_type='cuda')` instead.
@torch.cuda.amp.custom_fwd(cast_inputs=torch.float32)
Skipping loading of custom nodes
Starting server

To see the GUI go to: http://0.0.0.0:8334
got prompt
Using pytorch attention in VAE
Using pytorch attention in VAE
/home/ubuntu/cemil-test/ComfyUI/venv/lib/python3.10/site-packages/transformers/tokenization_utils_base.py:1601: FutureWarning: `clean_up_tokenization_spaces` was not set. It will be set to `True` by default. This behavior will be depracted in transformers v4.45, and will be then set to `False` by default. For more details check this issue: https://github.com/huggingface/transformers/issues/31884
warnings.warn(
clip missing: ['text_projection.weight']
Requested to load FluxClipModel_
Loading 1 new model
loaded completely 0.0 9319.23095703125 True
model weight dtype torch.bfloat16, manual cast: None
model_type FLUX
Requested to load Flux
Loading 1 new model
loaded completely 0.0 22700.097778320312 True
0%| | 0/4 [00:00