ACE Step audio generation slowdown caused by VRAM eviction triggered by deprecated API hooks and conflicting custom nodes
- Dominant language
- Python
- Stars
- 133k
- Forks
- 15.7k
- Avg merge
- 1d 7h
- Merged PRs (30d)
- 158
Description
### Custom Node Testing
- [x] I have tried disabling custom nodes and the issue persists (see [how to disable custom nodes](https://docs.comfy.org/troubleshooting/custom-node-issues#step-1%3A-test-with-all-custom-nodes-disabled) if you need help)
### Expected Behavior
## Expected Behavior
ACE Step should maintain stable VRAM allocation throughout audio generation, without being affected by unrelated UI‑patching or multi‑model nodes.
### Actual Behavior
## Observed Symptoms
### 1. Severe slowdown in ACE Step audio generation
- 90‑second songs take **10+ minutes**
- 4‑minute songs take **20 minutes**
### 2. Repeated VRAM unloading and reloading during generation
Example log excerpts:
```
Unloaded partially: 5585.50 MB freed
Unloaded partially: 4177.50 MB freed
loaded partially; 8676.58 MB usable, 8651.29 MB loaded, 476.00 MB offloaded
```
### 3. Deprecated API warnings at startup
```
[DEPRECATION WARNING] Detected import of deprecated legacy API: /scripts/ui.js
[DEPRECATION WARNING] Detected import of deprecated legacy API: /extensions/core/clipspace.js
[DEPRECATION WARNING] Detected import of deprecated legacy API: /extensions/core/widgetInputs.js
```
### 4. Non‑deterministic caching warnings from rgthree-comfy
```
[rgthree-comfy][Power Puter] cannot be cached b/c it's using a non-deterministic function call
```
### 5. Connection reset errors during LM sampling
```
ConnectionResetError: [WinError 10054] An existing connection was forcibly closed by the remote host
```
---
## Analysis
After isolating nodes and testing ACE Step in a reduced environment, the slowdown appears to be caused by **external custom nodes that patch ComfyUI’s UI or load additional heavy models**, resulting in:
- VRAM fragmentation
- forced model unload/reload cycles
- instability during ACE Step’s audio diffusion phase
Nodes that contributed to the issue include:
- `rgthree-comfy`
- `comfyui-workflow-prettier`
- `comfyui_fill-nodes`
- `comfyui-advanced-controlnet`
- `comfyui-animatediff-evolved`
- `comfyui-videohelpersuite`
- `comfyui-apex-artist`
- `comfyui-ollamagemini`
Disabling these nodes restored ACE Step’s normal performance and eliminated VRAM eviction during generation.
---
### Steps to Reproduce
## Analysis
After isolating nodes and testing ACE Step in a reduced environment, the slowdown appears to be caused by **external custom nodes that patch ComfyUI’s UI or load additional heavy models**, resulting in:
- VRAM fragmentation
- forced model unload/reload cycles
- instability during ACE Step’s audio diffusion phase
Nodes that contributed to the issue include:
- `rgthree-comfy`
- `comfyui-workflow-prettier`
- `comfyui_fill-nodes`
- `comfyui-advanced-controlnet`
- `comfyui-animatediff-evolved`
- `comfyui-videohelpersuite`
- `comfyui-apex-artist`
- `comfyui-ollamagemini`
Disabling these nodes restored ACE Step’s normal performance and eliminated VRAM eviction during generation.
### Debug Logs
```powershell
FETCH DATA from: C:\ComfyUI_windows_portable\python_embeded\Lib\site-packages\comfyui_manager\custom-node-list.json [DONE]
[WARNING] [DEPRECATION WARNING] Detected import of deprecated legacy API: /scripts/ui.js. This is likely caused by a custom node extension using outdated APIs. Please update your extensions or contact the extension author for an updated version.
[WARNING] [DEPRECATION WARNING] Detected import of deprecated legacy API: /extensions/core/clipspace.js. This is likely caused by a custom node extension using outdated APIs. Please update your extensions or contact the extension author for an updated version.
[WARNING] [DEPRECATION WARNING] Detected import of deprecated legacy API: /extensions/core/widgetInputs.js. This is likely caused by a custom node extension using outdated APIs. Please update your extensions or contact the extension author for an updated version.
Error fetching Gemini models: 400 API key not valid. Please pass a valid API key. [reason: "API_KEY_INVALID"
domain: "googleapis.com"
metadata {
key: "service"
value: "generativelanguage.googleapis.com"
}
, locale: "en-US"
message: "API key not valid. Please pass a valid API key."
]
Using default model list.
Error fetching Gemini models: 400 API key not valid. Please pass a valid API key. [reason: "API_KEY_INVALID"
domain: "googleapis.com"
metadata {
key: "service"
value: "generativelanguage.googleapis.com"
}
, locale: "en-US"
message: "API key not valid. Please pass a valid API key."
]
Using default model list.
Error fetching Gemini image models: 400 API key not valid. Please pass a valid API key. [reason: "API_KEY_INVALID"
domain: "googleapis.com"
metadata {
key: "service"
value: "generativelanguage.googleapis.com"
}
, locale: "en-US"
message: "API key not valid. Please pass a valid API key."
]
Using default image model list.
Error fetching Gemini image models: 400 API key not valid. Please pass a valid API key. [reason: "API_KEY_INVALID"
domain: "googleapis.com"
metadata {
key: "service"
value: "generativelanguage.googleapis.com"
}
, locale: "en-US"
message: "API key not valid. Please pass a valid API key."
]
Using default image model list.
[WARNING] [DEPRECATION WARNING] Detected import of deprecated legacy API: /scripts/ui/components/button.js. This is likely caused by a custom node extension using outdated APIs. Please update your extensions or contact the extension author for an updated version.
[INFO] HTTP Request: GET http://127.0.0.1:11434/api/tags "HTTP/1.1 200 OK"
[INFO] HTTP Request: GET http://127.0.0.1:11434/api/tags "HTTP/1.1 200 OK"
[INFO] HTTP Request: GET http://127.0.0.1:11434/api/tags "HTTP/1.1 200 OK"
[INFO] got prompt
[rgthree-comfy][Power Puter] Note, Power Puter (node #189) cannot be cached b/c it's using a non-deterministic function call. Matches function call for `random.int`.
[rgthree-comfy][Power Puter] Note, Power Puter (node #190) cannot be cached b/c it's using a non-deterministic function call. Matches function call for `random.int`.
[rgthree-comfy][Power Puter] Power Puter node #190 should update to use the `random.int` built-in instead of `random_int`.
[rgthree-comfy][Power Puter] Power Puter node #189 should update to use the `random.int` built-in instead of `random_int`.
[INFO] VAE load device: cuda:0, offload device: cpu, dtype: torch.float16
[INFO] FETCH ComfyRegistry Data [DONE]
[INFO] [ComfyUI-Manager] default cache updated: https://api.comfy.org/nodes
FETCH DATA from: C:\ComfyUI_windows_portable\ComfyUI\user\__manager\cache\1514988643_custom-node-list.json [DONE]
[INFO] [ComfyUI-Manager] All startup tasks have been completed.
[INFO] HTTP Request: POST http://127.0.0.1:11434/api/generate "HTTP/1.1 200 OK"
[INFO] HTTP Request: POST http://127.0.0.1:11434/api/generate "HTTP/1.1 200 OK"
[INFO] HTTP Request: POST http://127.0.0.1:11434/api/generate "HTTP/1.1 200 OK"
[EasyUse] Save Text: Saving to ComfyUI\output\lyrics\Ritual Rust_postpunk.txt
[INFO] Requested to load ACE15TEModel_
[INFO] loaded completely; 9127.29 MB loaded, full load: True
[INFO] CLIP/text encoder model load device: cuda:0, offload device: cpu, current: cuda:0, dtype: torch.float16
[WARNING] C:\ComfyUI_windows_portable\ComfyUI\comfy\ops.py:95: UserWarning: Using AOTriton backend for Efficient Attention forward... (Triggered internally at C:/b/pytorch/aten/src/ATen/native/transformers/hip/attention.hip:1452.)
return torch.nn.functional.scaled_dot_product_attention(q, k, v, *args, **kwargs)
LM sampling: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 470/470 [00:18<00:00, 25.48it/s]
[INFO] model weight dtype torch.bfloat16, manual cast: None
[INFO] model_type FLOW
[INFO] Requested to load ACEStep15
[INFO] Unloaded partially: 5585.50 MB freed, 3541.79 MB remains loaded, 237.50 MB buffer reserved, lowvram patches: 0
[INFO] loaded completely; 10758.75 MB usable, 9511.56 MB loaded, full load: True
100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 42/42 [00:08<00:00, 4.74it/s]
[INFO] Requested to load AudioOobleckVAE
[INFO] Unloaded partially: 522.50 MB freed, 3019.29 MB remains loaded, 237.50 MB buffer reserved, lowvram patches: 0
[INFO] loaded completely; 321.70 MB loaded, full load: True
[INFO] Unloaded partially: 4177.50 MB freed, 5335.04 MB remains loaded, 72.00 MB buffer reserved, lowvram patches: 0
[INFO] loaded partially; 8676.58 MB usable, 8651.29 MB loaded, 476.00 MB offloaded, 25.00 MB buffer reserved, lowvram patches: 0
LM sampling: 77%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████▍ | 461/600 [00:27<00:11, 12.53it/s][ERROR] Exception in callback _ProactorBasePipeTransport._call_connection_lost(None)
handle:
Traceback (most recent call last):
File "asyncio\events.py", line 88, in _run
File "asyncio\proactor_events.py", line 165, in _call_connection_lost
ConnectionResetError: [WinError 10054] An existing connection was forcibly closed by the remote host
LM sampling: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████▏| 597/600 [00:38<00:00, 13.61it/s]
```
### Other
_No response_
Contributor guide
Research direction
Start by reproducing ACE Step generation with the listed custom nodes enabled and disabled, comparing the VRAM and deprecation logs. Read the referenced comfy/ops.py warning and deprecated entry points such as /scripts/ui.js and /extensions/core/clipspace.js; done requires a reproducible ComfyUI-owned cause rather than an issue isolated to external nodes.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- javascript, ollama, python, pytorch
- Domain
- ai, backend, frontend, performance
- Issue type
- Bug
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Active
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100