# CosineDPMSolverMultistepScheduler

O [CosineDPMSolverMultistepScheduler](/docs/diffusers/v0.43.0/pt/api/schedulers/cosine_dpm#diffusers.CosineDPMSolverMultistepScheduler) é uma variante do [DPMSolverMultistepScheduler](/docs/diffusers/v0.43.0/pt/api/schedulers/multistep_dpm_solver#diffusers.DPMSolverMultistepScheduler) com agendamento coseno, proposta por Nichol e Dhariwal (2021). Ele é utilizado no paper [Stable Audio Open](https://huggingface.co/papers/2407.14358) e na base de código [Stability-AI/stable-audio-tool](https://github.com/Stability-AI/stable-audio-tools).

Este scheduler foi contribuído por [Yoach Lacombe](https://huggingface.co/ylacombe).

## CosineDPMSolverMultistepScheduler[[diffusers.CosineDPMSolverMultistepScheduler]]

#### diffusers.CosineDPMSolverMultistepScheduler[[diffusers.CosineDPMSolverMultistepScheduler]]

```python
diffusers.CosineDPMSolverMultistepScheduler(*args, **kwargs)
```

[Fonte](https://github.com/huggingface/diffusers/blob/v0.43.0/src/diffusers/utils/dummy_torch_and_torchsde_objects.py#L5)

## SchedulerOutput[[diffusers.schedulers.scheduling_utils.SchedulerOutput]]

#### diffusers.schedulers.scheduling_utils.SchedulerOutput[[diffusers.schedulers.scheduling_utils.SchedulerOutput]]

```python
diffusers.schedulers.scheduling_utils.SchedulerOutput(prev_sample: Tensor)
```

[Fonte](https://github.com/huggingface/diffusers/blob/v0.43.0/src/diffusers/schedulers/scheduling_utils.py#L66)

**Parâmetros:**

prev_sample (`torch.Tensor` de forma `(batch_size, num_channels, height, width)` para imagens) : Amostra calculada `(x_{t-1})` do passo anterior. `prev_sample` deve ser usado como entrada do próximo modelo no loop de remoção de ruído.

Classe base para a saída da função `step` de um agendador.