error

[ptoxad]

2026-04-27 01:52:37.762 | INFO | acestep.core.generation.handler.generate_music_decode:_prepare_generate_music_decode_state:42 - [generate_music] Model generation completed. Decoding latents...
2026-04-27 01:52:37.762 | DEBUG | acestep.core.generation.handler.generate_music_decode:_prepare_generate_music_decode_state:64 - [generate_music] pred_latents: torch.Size([1, 4375, 64]), dtype=torch.bfloat16
2026-04-27 01:52:37.762 | DEBUG | acestep.core.generation.handler.generate_music_decode:_prepare_generate_music_decode_state:65 - [generate_music] time_costs: {'encoder_time_cost': 0.09674668312072754, 'diffusion_time_cost': 11.54692268371582, 'diffusion_per_step_time_cost': 0.9622435569763184, 'total_time_cost': 11.643669366836548, 'offload_time_cost': 1.2373003959655762}
2026-04-27 01:52:37.762 | ERROR | acestep.core.generation.handler.generate_music:generate_music:440 - [generate_music] Generation failed
Traceback (most recent call last):

File "threading.py", line 1032, in _bootstrap

File "threading.py", line 1075, in _bootstrap_inner

File "A:\ACE-Step\ACE-Step-Studio\python\Lib\site-packages\anyio_backends_asyncio.py", line 1002, in run
result = context.run(func, *args)
│ │ │ └ (<generator object register_generation_run_handlers..generation_wrapper at 0x0000019680E18AE0>,)
│ │ └ <function run_sync_iterator_async at 0x0000019281B0DBC0>
│ └ <method 'run' of '_contextvars.Context' objects>
└ <_contextvars.Context object at 0x0000019680D3BC00>
File "A:\ACE-Step\ACE-Step-Studio\python\Lib\site-packages\gradio\utils.py", line 833, in run_sync_iterator_async
return next(iterator)
└ <generator object register_generation_run_handlers..generation_wrapper at 0x0000019680E18AE0>
File "A:\ACE-Step\ACE-Step-Studio\python\Lib\site-packages\gradio\utils.py", line 1017, in gen_wrapper
response = next(iterator)
└ <generator object register_generation_run_handlers..generation_wrapper at 0x00000196F851A700>

File "A:\ACE-Step\ACE-Step-Studio\ACE-Step-1.5\acestep\ui\gradio\events\wiring\generation_run_wiring.py", line 47, in generation_wrapper
yield from res_h.generate_with_batch_management(
│ └ <function generate_with_batch_management at 0x00000192616068E0>
└ <module 'acestep.ui.gradio.events.results_handlers' from 'A:\ACE-Step\ACE-Step-Studio\ACE-Step-1.5\acestep\ui\gradio\e...

File "A:\ACE-Step\ACE-Step-Studio\ACE-Step-1.5\acestep\ui\gradio\events\results\batch_management_wrapper.py", line 90, in generate_with_batch_management
for partial_result in generator:
└ <generator object generate_with_progress at 0x00000196775B5030>

File "A:\ACE-Step\ACE-Step-Studio\ACE-Step-1.5\acestep\ui\gradio\events\results\generation_progress.py", line 183, in generate_with_progress
result = generate_music(dit_handler, llm_handler, params=gen_params, config=gen_config, progress=progress)
│ │ │ │ │ └ <gradio.helpers.Progress object at 0x00000192615F3500>
│ │ │ │ └ GenerationConfig(batch_size=1, allow_lm_batch=True, use_random_seed=True, seeds=[-1], lm_batch_chunk_size=8, constrained_deco...
│ │ │ └ GenerationParams(task_type='text2music', instruction='Fill the audio semantic mask based on the given conditions:', reference...
│ │ └ <acestep.llm_inference.LLMHandler object at 0x0000019261657F80>
│ └ <acestep.handler.AceStepHandler object at 0x00000192615F25D0>
└ <function generate_music at 0x00000192615DD6C0>

File "A:\ACE-Step\ACE-Step-Studio\ACE-Step-1.5\acestep\inference.py", line 668, in generate_music
result = dit_handler.generate_music(**filtered_generate_kwargs)
│ │ └ {'captions': 'An energetic and unique fusion of traditional Arabic music and modern electronic dance-pop. The track opens wit...
│ └ <function GenerateMusicMixin.generate_music at 0x000001925D875300>
└ <acestep.handler.AceStepHandler object at 0x00000192615F25D0>

File "A:\ACE-Step\ACE-Step-Studio\ACE-Step-1.5\acestep\core\generation\handler\generate_music.py", line 385, in generate_music
pred_latents, time_costs = self._prepare_generate_music_decode_state(
│ └ <function GenerateMusicDecodeMixin._prepare_generate_music_decode_state at 0x000001925D8754E0>
└ <acestep.handler.AceStepHandler object at 0x00000192615F25D0>

File "A:\ACE-Step\ACE-Step-Studio\ACE-Step-1.5\acestep\core\generation\handler\generate_music_decode.py", line 81, in _prepare_generate_music_decode_state
raise RuntimeError("\n".join(hints))
└ ['Generation produced NaN or Inf latents (shape=[1, 4375, 64], dtype=torch.bfloat16, device=cuda:0, nan=280000, inf=0).', 'Co...

RuntimeError: Generation produced NaN or Inf latents (shape=[1, 4375, 64], dtype=torch.bfloat16, device=cuda:0, nan=280000, inf=0).
Common causes and fixes:

1. LoRA/adapter trained on an older model version — retrain or update the adapter.
2. Checkpoint/config mismatch — verify model checkpoints match this release.
3. Unsupported quantization/backend — try running with --backend pt.
4. CPU offload left parameters on wrong device — restart and regenerate.
5. Float16 overflow on pre-Ampere GPU — set ACESTEP_DTYPE=float32.
   [job_1777247476886_yofhukg] Generation failed: Gradio generation returned no audio files. Status: Error: Generation produced NaN or Inf latents (shape=[1, 4375, 64], dtype=torch.bfloat16, device=cuda:0, nan=280000, inf=0).
   Common causes and fixes:
6. LoRA/adapter trained on an older model version — retrain or update the adapter.
7. Checkpoint/config mismatch — verify model checkpoints match this release.
8. Unsupported quantization/backend — try running with --backend pt.
9. CPU offload left parameters on wrong device — restart and regenerate.
10. Float16 overflow on pre-Ampere GPU — set ACESTEP_DTYPE=float32.
    Traceback (most recent call last):
    File "A:\ACE-Step\ACE-Step-Studio\ACE-Step-1.5\acestep\core\generation\handler\generate_music.py", line 385, in generate_music
    pred_latents, time_costs = self._prepare_generate_music_decode_state(
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    File "A:\ACE-Step\ACE-Step-Studio\ACE-Step-1.5\acestep\core\generation\handler\generate_music_decode.py", line 81, in _prepare_generate_music_decode_state
    raise RuntimeError("\n".join(hints))
    RuntimeError: Generation produced NaN or Inf latents (shape=[1, 4375, 64], dtype=torch.bfloat16, device=cuda:0, nan=280000, inf=0).
    Common causes and fixes:
11. LoRA/adapter trained on an older model version — retrain or update the adapter.
12. Checkpoint/config mismatch — verify model checkpoints match this release.
13. Unsupported quantization/backend — try running with --backend pt.
14. CPU offload left parameters on wrong device — restart and regenerate.
15. Float16 overflow on pre-Ampere GPU — set ACESTEP_DTYPE=float32.
    . Details: none

[asabino2]

Same error here