Pinokio
0

Song generation fails - Error: EBUSY: resource busy or locked

@morpheusposted 7/17/2026, 6:16:22 PM·0 replies

What happened?
AceStep UI aborts with a critical error when generating a song.
Error: EBUSY: resource busy or locked, watch 'G:\pinokio\api\ace-step-ui.git\app\server\public\audio\job_1784311346631_0m7mpxa_0.mp3'
The backend returns the Audio and the job was successfully created. But Acestep is not able to access it because it is in use.
Additionally there is a non critical error because no sampler is selected.
message: "Value: is not in the list of choices: ['euler', 'heun']",
The song / job is playable.

Steps to reproduce

  1. Install Ace Step UI
  2. Start Ace Step UI
  3. Use simple or custom Mode and default settings
  4. Write a music prompt
  5. Set Vocal Language to Auto / Instrumental
  6. Generate your Song

Your system (OS / GPU / RAM / VRAM / etc.)

OS: Win 10
GPU: RTX 5080
RAM: 64 GB
VRAM: 16GB

Logs / full error output

Paste the full log output here

G:\pinokio\api\ace-step-ui.git\app\server>conda_hook & conda deactivate & conda deactivate & conda deactivate & conda activate base & bluefairy-activate.cmd && npm run dev
[OK] bluefairy active: uv/npm/bun protection enabled
(node:10872) [DEP0190] DeprecationWarning: Passing args to a child process with shell option true can lead to security vulnerabilities, as the arguments are not escaped, only concatenated.
(Use `node --trace-deprecation ...` to show where the warning was created)

> ace-step-ui-server@1.0.0 dev
> tsx watch src/index.ts

Running SQLite database migrations...
Migrations completed successfully!
ACE-Step UI Server running on http://localhost:3001
Environment: development
ACE-Step API: http://127.0.0.1:42003
LAN access: http://192.168.2.32:3001
Initializing local storage provider
Job job_1784311346631_0m7mpxa: Queued at position 1
[Gradio] Connected to http://127.0.0.1:42003
Job job_1784311346631_0m7mpxa: Using Gradio /generation_wrapper {
  prompt: 'acid techno, driving, energetic, powerful',
  duration: 124,
  batchSize: 1
}
Job job_1784311346631_0m7mpxa: Gradio generation failed, trying Python spawn fallback {
  type: 'status',
  endpoint: '/generation_wrapper',
  fn_index: 127,
  time: 2026-07-17T18:02:26.831Z,
  original_msg: undefined,
  queue: true,
  title: 'Error',
  message: "Value:  is not in the list of choices: ['euler', 'heun']",
  visible: true,
  duration: 10,
  stage: 'error',
  code: undefined,
  success: false
}
Job job_1784311346631_0m7mpxa: Using Python spawn (Gradio not available) {
  prompt: 'acid techno, driving, energetic, powerful',
  lyricsPreview: '',
  duration: 124,
  batchSize: 1
}
[ACE-Step] Skipping import of cpp extensions due to incompatible torch version 2.7.1+cu128 for torchao version 0.15.0             Please see https://github.com/pytorch/ao/issues/2919 for more info
[ACE-Step] W0717 20:02:30.300000 3968 env\Lib\site-packages\torch\distributed\elastic\multiprocessing\redirects.py:29] NOTE: Redirects are currently not supported in Windows or MacOs.
[ACE-Step] 2026-07-17 20:02:31.156 | WARNING  | acestep.training.trainer:<module>:40 - bitsandbytes not installed. Using standard AdamW.
[ACE-Step] 2026-07-17 20:02:31.167 | INFO     | acestep.gpu_config:get_gpu_memory_gb:578 - CUDA GPU detected: NVIDIA GeForce RTX 5080 (15.9 GB)
[ACE-Step] 2026-07-17 20:02:31.167 | INFO     | acestep.gpu_config:get_gpu_tier:769 - Detected 15.92GB VRAM � treating as 16GB class GPU       
[ACE-Step] 2026-07-17 20:02:31.261 | INFO     | acestep.core.generation.handler.init_service_loader:_load_main_model_from_checkpoint:174 - [initialize_service] Attempting to load model with attention implementation: sdpa
[ACE-Step] 2026-07-17 20:02:31.542 | INFO     | acestep.core.generation.handler.init_service_loader:_load_main_model_from_checkpoint:200 - [initialize_service] Keeping main model on cuda (persistent)
[ACE-Step] There are modules in AutoencoderOobleck that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in flo
at32.
[ACE-Step] 2026-07-17 20:02:33.786 | INFO     | acestep.inference:generate_music:641 - [generate_music] LLM usage decision: thinking=False, use_cot_caption=True, use_cot_language=True, use_cot_metas=True, need_lm_for_cot=True, 
llm_initialized=False, use_lm=False
[ACE-Step] 2026-07-17 20:02:33.787 | INFO     | acestep.core.generation.handler.generate_music:generate_music:289 - [generate_music] Turbo model detected: overriding guidance_scale 9.0 -> 1.0 (turbo does not use CFG).     
[ACE-Step] 2026-07-17 20:02:33.787 | INFO     | acestep.core.generation.handler.generate_music:generate_music:304 - [generate_music] Starting generation...
[ACE-Step] 2026-07-17 20:02:33.787 | INFO     | acestep.core.generation.handler.generate_music:generate_music:307 - [generate_music] Preparing inputs...
[ACE-Step] 2026-07-17 20:02:33.794 | DEBUG    | acestep.core.generation.handler.generate_music:_vram_preflight_check:132 - [generate_music] VRAM pre-flight: skipping check (offload_to_cpu=True, models loaded one-at-a-time)
[ACE-Step] 2026-07-17 20:02:33.795 | WARNING  | acestep.core.generation.handler.service_generate_request:_normalize_service_generate_inputs:53 - [service_generate] dmd_gan version: infer_steps 12 exceeds maximum 8, clamping to 
8
[ACE-Step] 2026-07-17 20:02:33.808 | INFO     | acestep.core.generation.handler.init_service_offload_context:_load_model_context:41 - [_load_model_context] Loading vae to cuda (RSS: 0 MB)
[ACE-Step] There are modules in AutoencoderOobleck that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in flo
at32.
[ACE-Step] 2026-07-17 20:02:34.026 | INFO     | acestep.core.generation.handler.init_service_offload_context:_load_model_context:58 - [_load_model_context] Loaded vae to cuda in 0.0930s (RSS: 0 -> 0 MB, delta: +0 MB)
[ACE-Step] 2026-07-17 20:02:34.042 | INFO     | acestep.core.generation.handler.init_service_offload_context:_load_model_context:67 - [_load_model_context] Offloading vae to CPU (RSS: 0 MB)
[ACE-Step] There are modules in AutoencoderOobleck that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in flo
at32.
[ACE-Step] 2026-07-17 20:02:34.293 | INFO     | acestep.core.generation.handler.init_service_offload_context:_load_model_context:83 - [_load_model_context] Offloaded vae to CPU in 0.1315s (RSS: 0 -> 0 MB, delta: +0 MB)
[ACE-Step] 2026-07-17 20:02:34.305 | INFO     | acestep.core.generation.handler.conditioning_text:_prepare_text_conditioning_inputs:119 - 
[ACE-Step] ======================================================================
[ACE-Step] 2026-07-17 20:02:34.305 | INFO     | acestep.core.generation.handler.conditioning_text:_prepare_text_conditioning_inputs:120 - \U0001f50d [DEBUG] DiT TEXT ENCODER INPUT (Inference)
[ACE-Step] 2026-07-17 20:02:34.305 | INFO     | acestep.core.generation.handler.conditioning_text:_prepare_text_conditioning_inputs:121 - ======================================================================
[ACE-Step] 2026-07-17 20:02:34.305 | INFO     | acestep.core.generation.handler.conditioning_text:_prepare_text_conditioning_inputs:122 - text_prompt:
[ACE-Step] # Instruction
[ACE-Step] Fill the audio semantic mask based on the given conditions:
[ACE-Step] # Caption
[ACE-Step] acid techno, driving, energetic, powerful
[ACE-Step] # Metas
[ACE-Step] - bpm: 160
[ACE-Step] - timesignature: N/A
[ACE-Step] - keyscale: N/A
[ACE-Step] - duration: 124 seconds
[ACE-Step] <|endoftext|>
[ACE-Step] 2026-07-17 20:02:34.305 | INFO     | acestep.core.generation.handler.conditioning_text:_prepare_text_conditioning_inputs:123 - ======================================================================
[ACE-Step] 2026-07-17 20:02:34.305 | INFO     | acestep.core.generation.handler.conditioning_text:_prepare_text_conditioning_inputs:124 - lyrics_text:
[ACE-Step] # Languages
[ACE-Step] unknown
[ACE-Step] # Lyric
[ACE-Step] <|endoftext|>
[ACE-Step] 2026-07-17 20:02:34.305 | INFO     | acestep.core.generation.handler.conditioning_text:_prepare_text_conditioning_inputs:125 - ======================================================================
[ACE-Step] 2026-07-17 20:02:34.312 | INFO     | acestep.core.generation.handler.init_service_offload_context:_load_model_context:41 - [_load_model_context] Loading vae to cuda (RSS: 0 MB)
[ACE-Step] There are modules in AutoencoderOobleck that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in flo
at32.
[ACE-Step] 2026-07-17 20:02:34.497 | INFO     | acestep.core.generation.handler.init_service_offload_context:_load_model_context:58 - [_load_model_context] Loaded vae to cuda in 0.0720s (RSS: 0 -> 0 MB, delta: +0 MB)
[ACE-Step] 2026-07-17 20:02:34.505 | INFO     | acestep.core.generation.handler.init_service_offload_context:_load_model_context:67 - [_load_model_context] Offloading vae to CPU (RSS: 0 MB)
[ACE-Step] There are modules in AutoencoderOobleck that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in flo
at32.
[ACE-Step] 2026-07-17 20:02:34.723 | INFO     | acestep.core.generation.handler.init_service_offload_context:_load_model_context:83 - [_load_model_context] Offloaded vae to CPU in 0.1015s (RSS: 0 -> 0 MB, delta: +0 MB)
[ACE-Step] 2026-07-17 20:02:34.724 | INFO     | acestep.core.generation.handler.conditioning_embed:preprocess_batch:110 - [preprocess_batch] Inferring prompt embeddings...
[ACE-Step] 2026-07-17 20:02:34.724 | INFO     | acestep.core.generation.handler.init_service_offload_context:_load_model_context:41 - [_load_model_context] Loading text_encoder to cuda (RSS: 0 MB)
[ACE-Step] 2026-07-17 20:02:35.118 | INFO     | acestep.core.generation.handler.init_service_offload_context:_load_model_context:58 - [_load_model_context] Loaded text_encoder to cuda in 0.2840s (RSS: 0 -> 0 MB, delta: +0 MB)
[ACE-Step] 2026-07-17 20:02:35.342 | INFO     | acestep.core.generation.handler.conditioning_embed:preprocess_batch:113 - [preprocess_batch] Inferring lyric embeddings...
[ACE-Step] 2026-07-17 20:02:35.342 | INFO     | acestep.core.generation.handler.init_service_offload_context:_load_model_context:67 - [_load_model_context] Offloading text_encoder to CPU (RSS: 0 MB)
[ACE-Step] 2026-07-17 20:02:35.744 | INFO     | acestep.core.generation.handler.init_service_offload_context:_load_model_context:83 - [_load_model_context] Offloaded text_encoder to CPU in 0.2852s (RSS: 0 -> 0 MB, delta: +0 MB)
[ACE-Step] 2026-07-17 20:02:35.744 | INFO     | acestep.core.generation.handler.service_generate_execute:_execute_service_generate_diffusion:161 - [service_generate] Generating audio... (DiT backend: PyTorch (cuda))
[ACE-Step] 2026-07-17 20:02:35.815 | INFO     | acestep.core.generation.handler.service_generate_execute:_execute_service_generate_diffusion:263 - [service_generate] DiT diffusion via PyTorch (cuda)...
[ACE-Step] 2026-07-17 20:02:35.855 | INFO     | acestep.models.common.dcw_correction:__init__:98 - [DCW] Active � mode=double, scaler=0.0500, high_scaler=0.0200, wavelet='haar'
[ACE-Step] 2026-07-17 20:02:36.044 | INFO     | acestep.models.common.dcw_loader:get:84 - [DCW] Built DWT1D for wavelet='haar' (low-pass filter taps=2, device=cuda:0, dtype=torch.bfloat16).
[ACE-Step] 2026-07-17 20:02:36.599 | INFO     | acestep.core.generation.handler.generate_music_decode:_prepare_generate_music_decode_state:42 - [generate_music] Model generation completed. Decoding latents...
[ACE-Step] 2026-07-17 20:02:36.599 | DEBUG    | acestep.core.generation.handler.generate_music_decode:_prepare_generate_music_decode_state:64 - [generate_music] pred_latents: torch.Size([1, 3100, 64]), dtype=torch.bfloat16
[ACE-Step] 2026-07-17 20:02:36.600 | DEBUG    | acestep.core.generation.handler.generate_music_decode:_prepare_generate_music_decode_state:65 - [generate_music] time_costs: {'encoder_time_cost': 0.02200007438659668, 'diffusion_
time_cost': 0.7616477012634277, 'diffusion_per_step_time_cost': 0.09520596265792847, 'total_time_cost': 0.7836477756500244, 'offload_time_cost': 0.967191219329834}
[ACE-Step] 2026-07-17 20:02:36.663 | INFO     | acestep.core.generation.handler.generate_music_decode:_decode_generate_music_pred_latents:127 - [generate_music] Decoding latents with VAE...
[ACE-Step] 2026-07-17 20:02:36.663 | INFO     | acestep.core.generation.handler.init_service_offload_context:_load_model_context:41 - [_load_model_context] Loading vae to cuda (RSS: 0 MB)
[ACE-Step] There are modules in AutoencoderOobleck that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in flo
at32.
[ACE-Step] 2026-07-17 20:02:36.870 | INFO     | acestep.core.generation.handler.init_service_offload_context:_load_model_context:58 - [_load_model_context] Loaded vae to cuda in 0.0910s (RSS: 0 -> 0 MB, delta: +0 MB)
[ACE-Step] 2026-07-17 20:02:36.871 | DEBUG    | acestep.core.generation.handler.generate_music_decode:_decode_generate_music_pred_latents:136 - [generate_music] Before VAE decode: allocated=4.79GB, max=5.62GB
[ACE-Step] 2026-07-17 20:02:36.871 | INFO     | acestep.core.generation.handler.generate_music_decode:_decode_generate_music_pred_latents:154 - [generate_music] Effective free VRAM before VAE decode: 9.77 GB
[ACE-Step] 2026-07-17 20:02:36.871 | INFO     | acestep.core.generation.handler.generate_music_decode:_decode_generate_music_pred_latents:172 - [generate_music] Using tiled VAE decode to reduce VRAM usage...
[ACE-Step] 2026-07-17 20:02:36.871 | DEBUG    | acestep.core.generation.handler.memory_utils:_get_auto_decode_chunk_size:108 - [_get_auto_decode_chunk_size] Effective free VRAM: 9.77 GB
[ACE-Step] 2026-07-17 20:02:36.871 | DEBUG    | acestep.core.generation.handler.memory_utils:_should_offload_wav_to_cpu:131 - [_should_offload_wav_to_cpu] Effective free VRAM: 9.77 GB
[ACE-Step] 2026-07-17 20:02:36.871 | INFO     | acestep.core.generation.handler.vae_decode:tiled_decode:56 - [tiled_decode] chunk_size=128, offload_wav_to_cpu=True, latents_shape=torch.Size([1, 64, 3100])
[ACE-Step] 2026-07-17 20:02:36.871 | WARNING  | acestep.core.generation.handler.vae_decode_chunks:_tiled_decode_inner:39 - [tiled_decode] Reduced overlap from 64 to 32 for chunk_size=128
[ACE-Step] 2026-07-17 20:02:38.111 | DEBUG    | acestep.core.generation.handler.generate_music_decode:_decode_generate_music_pred_latents:193 - [generate_music] After VAE decode: allocated=4.95GB, max=5.62GB
[ACE-Step] 2026-07-17 20:02:38.221 | INFO     | acestep.core.generation.handler.init_service_offload_context:_load_model_context:67 - [_load_model_context] Offloading vae to CPU (RSS: 0 MB)
[ACE-Step] There are modules in AutoencoderOobleck that should be kept in float32: []. Casting directly with `to()` can lead to inconsistent results; set `torch_dtype` in `from_pretrained()` instead to keep these modules in flo
at32.
[ACE-Step] 2026-07-17 20:02:38.459 | INFO     | acestep.core.generation.handler.init_service_offload_context:_load_model_context:83 - [_load_model_context] Offloaded vae to CPU in 0.1140s (RSS: 0 -> 0 MB, delta: +0 MB)
[ACE-Step] 2026-07-17 20:02:38.574 | INFO     | acestep.core.generation.handler.generate_music_payload:_build_generate_music_success_payload:38 - [generate_music] VAE decode completed. Preparing audio tensors...
[ACE-Step] 2026-07-17 20:02:38.575 | INFO     | acestep.core.generation.handler.generate_music_payload:_build_generate_music_success_payload:50 - [generate_music] Done! Generated 1 audio tensors.
[ACE-Step] 2026-07-17 20:02:38.695 | INFO     | acestep.inference:generate_music:981 - [Normalization] Audio 0 BEFORE: Peak=1.0000, Target=-1.0dB
[ACE-Step] 2026-07-17 20:02:38.708 | INFO     | acestep.inference:generate_music:986 - [Normalization] Audio 0 AFTER: Peak=0.8913
[ACE-Step] 2026-07-17 20:02:40.100 | DEBUG    | acestep.audio_utils:_save_mp3:183 - [AudioSaver] Saved audio to G:\pinokio\api\ace-step-ui.git\app\ACE-Step-1.5\output\job_1784311346631_0m7mpxa\40d243aa-3b7c-8643-1f13-2773bef717
d3.mp3 (mp3, 48000Hz, 128k)
node:internal/fs/watchers:321
    const error = new UVException({
                  ^

Error: EBUSY: resource busy or locked, watch 'G:\pinokio\api\ace-step-ui.git\app\server\public\audio\job_1784311346631_0m7mpxa_0.mp3'
    at FSWatcher.<computed> (node:internal/fs/watchers:321:19)
    at Object.watch (node:fs:2548:36)
    at createFsWatchInstance (file:///G:/pinokio/api/ace-step-ui.git/app/node_modules/vite/dist/node/chunks/dep-D4NMHUTW.js:22195:17)
    at setFsWatchListener (file:///G:/pinokio/api/ace-step-ui.git/app/node_modules/vite/dist/node/chunks/dep-D4NMHUTW.js:22242:15)
    at NodeFsHandler._watchWithNodeFs (file:///G:/pinokio/api/ace-step-ui.git/app/node_modules/vite/dist/node/chunks/dep-D4NMHUTW.js:22397:14)
    at NodeFsHandler._handleFile (file:///G:/pinokio/api/ace-step-ui.git/app/node_modules/vite/dist/node/chunks/dep-D4NMHUTW.js:22461:23)     
    at NodeFsHandler._addToNodeFs (file:///G:/pinokio/api/ace-step-ui.git/app/node_modules/vite/dist/node/chunks/dep-D4NMHUTW.js:22703:21)    
Emitted 'error' event on FSWatcher instance at:
    at FSWatcher._handleError (file:///G:/pinokio/api/ace-step-ui.git/app/node_modules/vite/dist/node/chunks/dep-D4NMHUTW.js:23896:10)        
    at NodeFsHandler._addToNodeFs (file:///G:/pinokio/api/ace-step-ui.git/app/node_modules/vite/dist/node/chunks/dep-D4NMHUTW.js:22711:18) {  
  errno: -4082,
  syscall: 'watch',
  code: 'EBUSY',
  path: 'G:\\pinokio\\api\\ace-step-ui.git\\app\\server\\public\\audio\\job_1784311346631_0m7mpxa_0.mp3',
  filename: 'G:\\pinokio\\api\\ace-step-ui.git\\app\\server\\public\\audio\\job_1784311346631_0m7mpxa_0.mp3'
}

Node.js v24.18.0
Job job_1784311346631_0m7mpxa: Completed via Python in 6.3s with 1 audio files

(base) G:\pinokio\api\ace-step-ui.git\app>
Replies (0)
Up to 10 files, 25MB each. Images are optimized; GIFs -> MP4; videos 720p (max 120s).
Song generation fails - Error: EBUSY: resource busy or locked · Pinokio