VERSION 1 WILL NOT RUN on a blackwell rtx 5060 ti

[LIQUIDMIND111]

I TRIED everything, and nothing.... not even with CUDA 12.6 or CUDA 12.8

any hints?

[Redtash1]

Are you using Windows or Linux?

[LIQUIDMIND111]

Are you using Windows or Linux?

windows my friend. i did a LOT of uninstalling an older version and installing a NEW git pull FRESH, and nothing, i used GEMINI and nothing...... it will load, and all, but when click GENERTE i get this issue:

2026-02-07 19:46:46.693 | WARNING  | acestep.pipeline_ace_step:call:1480 - Checkpoint not loaded, loading checkpoint...
2026-02-07 19:46:46.693 | INFO     | acestep.pipeline_ace_step:get_checkpoint_path:178 - Download models from Hugging Face: ACE-Step/ACE-Step-v1-3.5B, cache to: C:\Users\eddie.cache/ace-step/checkpoints
Fetching 14 files: 100%|███████████████████████████████████████████████████████████████████████| 14/14 [00:00<?, ?it/s]
Traceback (most recent call last):
  File "E:\ACE-STEP=1\ACE-Step\venv\lib\site-packages\gradio\queueing.py", line 536, in process_events
    response = await route_utils.call_process_api(
  File "E:\ACE-STEP=1\ACE-Step\venv\lib\site-packages\gradio\route_utils.py", line 322, in call_process_api
    output = await app.get_blocks().process_api(
  File "E:\ACE-STEP=1\ACE-Step\venv\lib\site-packages\gradio\blocks.py", line 1935, in process_api
    result = await self.call_function(
  File "E:\ACE-STEP=1\ACE-Step\venv\lib\site-packages\gradio\blocks.py", line 1520, in call_function
    prediction = await anyio.to_thread.run_sync(  # type: ignore
  File "E:\ACE-STEP=1\ACE-Step\venv\lib\site-packages\anyio\to_thread.py", line 63, in run_sync
    return await get_async_backend().run_sync_in_worker_thread(
  File "E:\ACE-STEP=1\ACE-Step\venv\lib\site-packages\anyio_backends_asyncio.py", line 2502, in run_sync_in_worker_thread
    return await future
  File "E:\ACE-STEP=1\ACE-Step\venv\lib\site-packages\anyio_backends_asyncio.py", line 986, in run
    result = context.run(func, *args)
  File "E:\ACE-STEP=1\ACE-Step\venv\lib\site-packages\gradio\utils.py", line 826, in wrapper
    response = f(*args, **kwargs)
  File "E:\ACE-STEP=1\ACE-Step\acestep\pipeline_ace_step.py", line 1484, in call
    self.load_checkpoint(self.checkpoint_dir)
  File "E:\ACE-STEP=1\ACE-Step\acestep\pipeline_ace_step.py", line 199, in load_checkpoint
    self.ace_step_transformer.to(self.device).eval().to(self.dtype)
  File "E:\ACE-STEP=1\ACE-Step\venv\lib\site-packages\diffusers\models\modeling_utils.py", line 1435, in to
    return super().to(*args, **kwargs)
  File "E:\ACE-STEP=1\ACE-Step\venv\lib\site-packages\torch\nn\modules\module.py", line 1381, in to
    return self._apply(convert)
  File "E:\ACE-STEP=1\ACE-Step\venv\lib\site-packages\torch\nn\modules\module.py", line 933, in _apply
    module._apply(fn)
  File "E:\ACE-STEP=1\ACE-Step\venv\lib\site-packages\torch\nn\modules\module.py", line 1031, in _apply
    self._buffers[key] = fn(buf)
  File "E:\ACE-STEP=1\ACE-Step\venv\lib\site-packages\torch\nn\modules\module.py", line 1367, in convert
    return t.to(
torch.AcceleratorError: CUDA error: no kernel image is available for execution on the device
Search for cudaErrorNoKernelImageForDevice' in https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__TYPES.html for more information. 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.

[LIQUIDMIND111]

Are you using Windows or Linux?

E:\ACE-STEP=1\ACE-Step>python -c "import torch; print(f'CUDA Version: {torch.version.cuda}'); print(f'Compatible with sm_120: {torch.cuda.get_device_capability(0) >= (12,0) if torch.cuda.is_available() else False}')"

CUDA Version: 12.8

Compatible with sm_120: True

[LIQUIDMIND111]

Turn off compile.

ok where is that?

[LIQUIDMIND111]

In the tab where you load the models at the bottom.

i dont see anything here in settings - basic or advanced.... no tab

[LIQUIDMIND111]

In the tab where you load the models at the bottom.

i even added to the arguments

python -m acestep.gui --port 7865 --torch_compile False --bf16 True --device_id 0

and still error..... when generating....

[LIQUIDMIND111]

Sorry your on v1, I thought you were using v1.5 which is much better. https://github.com/ace-step/ACE-Step-1.5

ah, no, i got it running,,, i just wanted to use again the older version...... indeed, i have made many audio with 1.5 already.... thanks

[LIQUIDMIND111]

Sorry your on v1, I thought you were using v1.5 which is much better. https://github.com/ace-step/ACE-Step-1.5

i even change the codes of the TIERS so i can make an 8-minute-long song with only 16GB..... lol