Memory error with last commits

[tester4488]

i'm on commit 5b7f9d8

here i can set:

and al works.

with the last commits if i chose acestep-v15-base Initialize 5Hz LM only disable, and i got an error: prompt too long no memory available.

RTX 4060 8GB vram 32 GB ram windows

[coderabbitai]

⚠️ Possible Duplicate Issue(s)

• Latest origin/main has increased VRAM usage #384

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!

[ChuxiJ]

related PR

#570

#567

[ChuxiJ]

The current optimal configuration is optimized based on Turbo.
If you are using the base model, you need to check if the batch size is greater than 1.
The base model requires CFG, which will consume more VRAM.

[tester4488]

The current optimal configuration is optimized based on Turbo. If you are using the base model, you need to check if the batch size is greater than 1. The base model requires CFG, which will consume more VRAM.

batch size is 1

[ChuxiJ]

What's your test input? could you please check this #577 if solving your issue?

[tester4488]

last git commit de8438b

batch 1

2026-02-15 11:06:41.373 | INFO | acestep.llm_inference:_load_model_context:3855 - Loading LLM to auto 2026-02-15 11:06:41.475 | ERROR | acestep.llm_inference:generate_from_formatted_prompt:2281 - Error in generate_from_formatted_prompt: RuntimeError: Expected one of cpu, cuda, ipu, xpu, mkldnn, opengl, opencl, ideep, hip, ve, fpga, maia, xla, lazy, vulkan, mps, meta, hpu, mtia, privateuseone device type at start of device string: auto Traceback (most recent call last): File "E:\ace\ACE-Step-1.5\acestep\llm_inference.py", line 2254, in generate_from_formatted_prompt output_text = self._run_pt( ^^^^^^^^^^^^^ File "E:\ace\ACE-Step-1.5\acestep\llm_inference.py", line 1052, in _run_pt return self._run_pt_single( ^^^^^^^^^^^^^^^^^^^^ File "E:\ace\ACE-Step-1.5\acestep\llm_inference.py", line 858, in _run_pt_single with self._load_model_context(): File "C:\Users\mc\AppData\Roaming\uv\python\cpython-3.11.14-windows-x86_64-none\Lib\contextlib.py", line 137, in __enter__ return next(self.gen) ^^^^^^^^^^^^^^ File "E:\ace\ACE-Step-1.5\acestep\llm_inference.py", line 3858, in _load_model_context model.to(self.device).to(self.dtype) ^^^^^^^^^^^^^^^^^^^^^ File "E:\ace\ACE-Step-1.5\.venv\Lib\site-packages\transformers\modeling_utils.py", line 4343, in to return super().to(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "E:\ace\ACE-Step-1.5\.venv\Lib\site-packages\torch\nn\modules\module.py", line 1314, in to device, dtype, non_blocking, convert_to_format = torch._C._nn._parse_to( ^^^^^^^^^^^^^^^^^^^^^^^ RuntimeError: Expected one of cpu, cuda, ipu, xpu, mkldnn, opengl, opencl, ideep, hip, ve, fpga, maia, xla, lazy, vulkan, mps, meta, hpu, mtia, privateuseone device type at start of device string: auto

[tester4488]

example from random

[tester4488]

status dont show nothing

[ChuxiJ]

I am asking you to test #577

not the last git commit de8438b

[tester4488]

turbo and 2 batch works

[tester4488]

I am asking you to test #577

not the last git commit de8438b

i have to change
gpu_config.py
and
handler.py?

[ChuxiJ]

try:
git pull
uv sync
MAX_CUDA_VRAM=8 uv run acestep