# Copyright 2024 Stability AI and The HuggingFace Team. All rights reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. import inspect from typing import Callable, List, Optional, Union import torch from transformers import ( T5EncoderModel, T5Tokenizer, T5TokenizerFast, ) from ...models import AutoencoderOobleck, StableAudioDiTModel from ...models.embeddings import get_1d_rotary_pos_embed from ...schedulers import EDMDPMSolverMultistepScheduler from ...utils import ( logging, replace_example_docstring, ) from ...utils.torch_utils import randn_tensor from ..pipeline_utils import AudioPipelineOutput, DiffusionPipeline from .modeling_stable_audio import StableAudioProjectionModel logger = logging.get_logger(__name__) # pylint: disable=invalid-name EXAMPLE_DOC_STRING = """ Examples: ```py >>> import scipy >>> import torch >>> import soundfile as sf >>> from diffusers import StableAudioPipeline >>> repo_id = "stabilityai/stable-audio-open-1.0" >>> pipe = StableAudioPipeline.from_pretrained(repo_id, torch_dtype=torch.float16) >>> pipe = pipe.to("cuda") >>> # define the prompts >>> prompt = "The sound of a hammer hitting a wooden surface." >>> negative_prompt = "Low quality." >>> # set the seed for generator >>> generator = torch.Generator("cuda").manual_seed(0) >>> # run the generation >>> audio = pipe( ... prompt, ... negative_prompt=negative_prompt, ... num_inference_steps=200, ... audio_end_in_s=10.0, ... num_waveforms_per_prompt=3, ... generator=generator, ... ).audios >>> output = audio[0].T.float().cpu().numpy() >>> sf.write("hammer.wav", output, pipe.vae.sampling_rate) ``` """ class StableAudioPipeline(DiffusionPipeline): r""" Pipeline for text-to-audio generation using StableAudio. This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods implemented for all pipelines (downloading, saving, running on a particular device, etc.). Args: vae ([`AutoencoderOobleck`]): Variational Auto-Encoder (VAE) model to encode and decode images to and from latent representations. text_encoder ([`~transformers.T5EncoderModel`]): Frozen text-encoder. StableAudio uses the encoder of [T5](https://huggingface.co/docs/transformers/model_doc/t5#transformers.T5EncoderModel), specifically the [google-t5/t5-base](https://huggingface.co/google-t5/t5-base) variant. projection_model ([`StableAudioProjectionModel`]): A trained model used to linearly project the hidden-states from the text encoder model and the start and end seconds. The projected hidden-states from the encoder and the conditional seconds are concatenated to give the input to the transformer model. tokenizer ([`~transformers.T5Tokenizer`]): Tokenizer to tokenize text for the frozen text-encoder. transformer ([`StableAudioDiTModel`]): A `StableAudioDiTModel` to denoise the encoded audio latents. scheduler ([`EDMDPMSolverMultistepScheduler`]): A scheduler to be used in combination with `transformer` to denoise the encoded audio latents. """ model_cpu_offload_seq = "text_encoder->projection_model->transformer->vae" def __init__( self, vae: AutoencoderOobleck, text_encoder: T5EncoderModel, projection_model: StableAudioProjectionModel, tokenizer: Union[T5Tokenizer, T5TokenizerFast], transformer: StableAudioDiTModel, scheduler: EDMDPMSolverMultistepScheduler, ): super().__init__() self.register_modules( vae=vae, text_encoder=text_encoder, projection_model=projection_model, tokenizer=tokenizer, transformer=transformer, scheduler=scheduler, ) self.rotary_embed_dim = self.transformer.config.attention_head_dim // 2 # Copied from diffusers.pipelines.pipeline_utils.StableDiffusionMixin.enable_vae_slicing def enable_vae_slicing(self): r""" Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to compute decoding in several steps. This is useful to save some memory and allow larger batch sizes. """ self.vae.enable_slicing() # Copied from diffusers.pipelines.pipeline_utils.StableDiffusionMixin.disable_vae_slicing def disable_vae_slicing(self): r""" Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to computing decoding in one step. """ self.vae.disable_slicing() def encode_prompt( self, prompt, device, do_classifier_free_guidance, negative_prompt=None, prompt_embeds: Optional[torch.Tensor] = None, negative_prompt_embeds: Optional[torch.Tensor] = None, attention_mask: Optional[torch.LongTensor] = None, negative_attention_mask: Optional[torch.LongTensor] = None, ): if prompt is not None and isinstance(prompt, str): batch_size = 1 elif prompt is not None and isinstance(prompt, list): batch_size = len(prompt) else: batch_size = prompt_embeds.shape[0] if prompt_embeds is None: # 1. Tokenize text text_inputs = self.tokenizer( prompt, padding="max_length", max_length=self.tokenizer.model_max_length, truncation=True, return_tensors="pt", ) text_input_ids = text_inputs.input_ids attention_mask = text_inputs.attention_mask untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal( text_input_ids, untruncated_ids ): removed_text = self.tokenizer.batch_decode( untruncated_ids[:, self.tokenizer.model_max_length - 1 : -1] ) logger.warning( f"The following part of your input was truncated because {self.text_encoder.config.model_type} can " f"only handle sequences up to {self.tokenizer.model_max_length} tokens: {removed_text}" ) text_input_ids = text_input_ids.to(device) attention_mask = attention_mask.to(device) # 2. Text encoder forward self.text_encoder.eval() prompt_embeds = self.text_encoder( text_input_ids, attention_mask=attention_mask, ) prompt_embeds = prompt_embeds[0] if do_classifier_free_guidance and negative_prompt is not None: uncond_tokens: List[str] if type(prompt) is not type(negative_prompt): raise TypeError( f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !=" f" {type(prompt)}." ) elif isinstance(negative_prompt, str): uncond_tokens = [negative_prompt] elif batch_size != len(negative_prompt): raise ValueError( f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:" f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches" " the batch size of `prompt`." ) else: uncond_tokens = negative_prompt # 1. Tokenize text uncond_input = self.tokenizer( uncond_tokens, padding="max_length", max_length=self.tokenizer.model_max_length, truncation=True, return_tensors="pt", ) uncond_input_ids = uncond_input.input_ids.to(device) negative_attention_mask = uncond_input.attention_mask.to(device) # 2. Text encoder forward self.text_encoder.eval() negative_prompt_embeds = self.text_encoder( uncond_input_ids, attention_mask=negative_attention_mask, ) negative_prompt_embeds = negative_prompt_embeds[0] if negative_attention_mask is not None: # set the masked tokens to the null embed negative_prompt_embeds = torch.where( negative_attention_mask.to(torch.bool).unsqueeze(2), negative_prompt_embeds, 0.0 ) # 3. Project prompt_embeds and negative_prompt_embeds if do_classifier_free_guidance and negative_prompt_embeds is not None: # For classifier free guidance, we need to do two forward passes. # Here we concatenate the negative and text embeddings into a single batch # to avoid doing two forward passes prompt_embeds = torch.cat([negative_prompt_embeds, prompt_embeds]) if attention_mask is not None and negative_attention_mask is None: negative_attention_mask = torch.ones_like(attention_mask) elif attention_mask is None and negative_attention_mask is not None: attention_mask = torch.ones_like(negative_attention_mask) if attention_mask is not None: attention_mask = torch.cat([negative_attention_mask, attention_mask]) prompt_embeds = self.projection_model( text_hidden_states=prompt_embeds, ).text_hidden_states if attention_mask is not None: prompt_embeds = prompt_embeds * attention_mask.unsqueeze(-1).to(prompt_embeds.dtype) prompt_embeds = prompt_embeds * attention_mask.unsqueeze(-1).to(prompt_embeds.dtype) return prompt_embeds def encode_duration( self, audio_start_in_s, audio_end_in_s, device, do_classifier_free_guidance, batch_size, ): audio_start_in_s = audio_start_in_s if isinstance(audio_start_in_s, list) else [audio_start_in_s] audio_end_in_s = audio_end_in_s if isinstance(audio_end_in_s, list) else [audio_end_in_s] if len(audio_start_in_s) == 1: audio_start_in_s = audio_start_in_s * batch_size if len(audio_end_in_s) == 1: audio_end_in_s = audio_end_in_s * batch_size # Cast the inputs to floats audio_start_in_s = [float(x) for x in audio_start_in_s] audio_start_in_s = torch.tensor(audio_start_in_s).to(device) audio_end_in_s = [float(x) for x in audio_end_in_s] audio_end_in_s = torch.tensor(audio_end_in_s).to(device) projection_output = self.projection_model( start_seconds=audio_start_in_s, end_seconds=audio_end_in_s, ) seconds_start_hidden_states = projection_output.seconds_start_hidden_states seconds_end_hidden_states = projection_output.seconds_end_hidden_states # For classifier free guidance, we need to do two forward passes. # Here we repeat the audio hidden states to avoid doing two forward passes if do_classifier_free_guidance: seconds_start_hidden_states = torch.cat([seconds_start_hidden_states, seconds_start_hidden_states], dim=0) seconds_end_hidden_states = torch.cat([seconds_end_hidden_states, seconds_end_hidden_states], dim=0) return seconds_start_hidden_states, seconds_end_hidden_states # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_extra_step_kwargs def prepare_extra_step_kwargs(self, generator, eta): # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers. # eta corresponds to η in DDIM paper: https://arxiv.org/abs/2010.02502 # and should be between [0, 1] accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys()) extra_step_kwargs = {} if accepts_eta: extra_step_kwargs["eta"] = eta # check if the scheduler accepts generator accepts_generator = "generator" in set(inspect.signature(self.scheduler.step).parameters.keys()) if accepts_generator: extra_step_kwargs["generator"] = generator return extra_step_kwargs def check_inputs( self, prompt, audio_start_in_s, audio_end_in_s, callback_steps, negative_prompt=None, prompt_embeds=None, negative_prompt_embeds=None, attention_mask=None, negative_attention_mask=None, initial_audio_waveforms=None, initial_audio_sampling_rate=None, ): if audio_end_in_s < audio_start_in_s: raise ValueError( f"`audio_end_in_s={audio_end_in_s}' must be higher than 'audio_start_in_s={audio_start_in_s}` but " ) if ( audio_start_in_s < self.projection_model.config.min_value or audio_start_in_s > self.projection_model.config.max_value ): raise ValueError( f"`audio_start_in_s` must be greater than or equal to {self.projection_model.config.min_value}, and lower than or equal to {self.projection_model.config.max_value} but " f"is {audio_start_in_s}." ) if ( audio_end_in_s < self.projection_model.config.min_value or audio_end_in_s > self.projection_model.config.max_value ): raise ValueError( f"`audio_end_in_s` must be greater than or equal to {self.projection_model.config.min_value}, and lower than or equal to {self.projection_model.config.max_value} but " f"is {audio_end_in_s}." ) if (callback_steps is None) or ( callback_steps is not None and (not isinstance(callback_steps, int) or callback_steps <= 0) ): raise ValueError( f"`callback_steps` has to be a positive integer but is {callback_steps} of type" f" {type(callback_steps)}." ) if prompt is not None and prompt_embeds is not None: raise ValueError( f"Cannot forward both `prompt`: {prompt} and `prompt_embeds`: {prompt_embeds}. Please make sure to" " only forward one of the two." ) elif prompt is None and (prompt_embeds is None): raise ValueError( "Provide either `prompt`, or `prompt_embeds`. Cannot leave" "`prompt` undefined without specifying `prompt_embeds`." ) elif prompt is not None and (not isinstance(prompt, str) and not isinstance(prompt, list)): raise ValueError(f"`prompt` has to be of type `str` or `list` but is {type(prompt)}") if negative_prompt is not None and negative_prompt_embeds is not None: raise ValueError( f"Cannot forward both `negative_prompt`: {negative_prompt} and `negative_prompt_embeds`:" f" {negative_prompt_embeds}. Please make sure to only forward one of the two." ) if prompt_embeds is not None and negative_prompt_embeds is not None: if prompt_embeds.shape != negative_prompt_embeds.shape: raise ValueError( "`prompt_embeds` and `negative_prompt_embeds` must have the same shape when passed directly, but" f" got: `prompt_embeds` {prompt_embeds.shape} != `negative_prompt_embeds`" f" {negative_prompt_embeds.shape}." ) if attention_mask is not None and attention_mask.shape != prompt_embeds.shape[:2]: raise ValueError( "`attention_mask should have the same batch size and sequence length as `prompt_embeds`, but got:" f"`attention_mask: {attention_mask.shape} != `prompt_embeds` {prompt_embeds.shape}" ) if initial_audio_sampling_rate is None and initial_audio_waveforms is not None: raise ValueError( "`initial_audio_waveforms' is provided but the sampling rate is not. Make sure to pass `initial_audio_sampling_rate`." ) if initial_audio_sampling_rate is not None and initial_audio_sampling_rate != self.vae.sampling_rate: raise ValueError( f"`initial_audio_sampling_rate` must be {self.vae.hop_length}' but is `{initial_audio_sampling_rate}`." "Make sure to resample the `initial_audio_waveforms` and to correct the sampling rate. " ) def prepare_latents( self, batch_size, num_channels_vae, sample_size, dtype, device, generator, latents=None, initial_audio_waveforms=None, num_waveforms_per_prompt=None, audio_channels=None, ): shape = (batch_size, num_channels_vae, sample_size) if isinstance(generator, list) and len(generator) != batch_size: raise ValueError( f"You have passed a list of generators of length {len(generator)}, but requested an effective batch" f" size of {batch_size}. Make sure the batch size matches the length of the generators." ) if latents is None: latents = randn_tensor(shape, generator=generator, device=device, dtype=dtype) else: latents = latents.to(device) # scale the initial noise by the standard deviation required by the scheduler latents = latents * self.scheduler.init_noise_sigma # encode the initial audio for use by the model if initial_audio_waveforms is not None: # check dimension if initial_audio_waveforms.ndim == 2: initial_audio_waveforms = initial_audio_waveforms.unsqueeze(1) elif initial_audio_waveforms.ndim != 3: raise ValueError( f"`initial_audio_waveforms` must be of shape `(batch_size, num_channels, audio_length)` or `(batch_size, audio_length)` but has `{initial_audio_waveforms.ndim}` dimensions" ) audio_vae_length = self.transformer.config.sample_size * self.vae.hop_length audio_shape = (batch_size // num_waveforms_per_prompt, audio_channels, audio_vae_length) # check num_channels if initial_audio_waveforms.shape[1] == 1 and audio_channels == 2: initial_audio_waveforms = initial_audio_waveforms.repeat(1, 2, 1) elif initial_audio_waveforms.shape[1] == 2 and audio_channels == 1: initial_audio_waveforms = initial_audio_waveforms.mean(1, keepdim=True) if initial_audio_waveforms.shape[:2] != audio_shape[:2]: raise ValueError( f"`initial_audio_waveforms` must be of shape `(batch_size, num_channels, audio_length)` or `(batch_size, audio_length)` but is of shape `{initial_audio_waveforms.shape}`" ) # crop or pad audio_length = initial_audio_waveforms.shape[-1] if audio_length < audio_vae_length: logger.warning( f"The provided input waveform is shorter ({audio_length}) than the required audio length ({audio_vae_length}) of the model and will thus be padded." ) elif audio_length > audio_vae_length: logger.warning( f"The provided input waveform is longer ({audio_length}) than the required audio length ({audio_vae_length}) of the model and will thus be cropped." ) audio = initial_audio_waveforms.new_zeros(audio_shape) audio[:, :, : min(audio_length, audio_vae_length)] = initial_audio_waveforms[:, :, :audio_vae_length] encoded_audio = self.vae.encode(audio).latent_dist.sample(generator) encoded_audio = encoded_audio.repeat((num_waveforms_per_prompt, 1, 1)) latents = encoded_audio + latents return latents @torch.no_grad() @replace_example_docstring(EXAMPLE_DOC_STRING) def __call__( self, prompt: Union[str, List[str]] = None, audio_end_in_s: Optional[float] = None, audio_start_in_s: Optional[float] = 0.0, num_inference_steps: int = 100, guidance_scale: float = 7.0, negative_prompt: Optional[Union[str, List[str]]] = None, num_waveforms_per_prompt: Optional[int] = 1, eta: float = 0.0, generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None, latents: Optional[torch.Tensor] = None, initial_audio_waveforms: Optional[torch.Tensor] = None, initial_audio_sampling_rate: Optional[torch.Tensor] = None, prompt_embeds: Optional[torch.Tensor] = None, negative_prompt_embeds: Optional[torch.Tensor] = None, attention_mask: Optional[torch.LongTensor] = None, negative_attention_mask: Optional[torch.LongTensor] = None, return_dict: bool = True, callback: Optional[Callable[[int, int, torch.Tensor], None]] = None, callback_steps: Optional[int] = 1, output_type: Optional[str] = "pt", ): r""" The call function to the pipeline for generation. Args: prompt (`str` or `List[str]`, *optional*): The prompt or prompts to guide audio generation. If not defined, you need to pass `prompt_embeds`. audio_end_in_s (`float`, *optional*, defaults to 47.55): Audio end index in seconds. audio_start_in_s (`float`, *optional*, defaults to 0): Audio start index in seconds. num_inference_steps (`int`, *optional*, defaults to 100): The number of denoising steps. More denoising steps usually lead to a higher quality audio at the expense of slower inference. guidance_scale (`float`, *optional*, defaults to 7.0): A higher guidance scale value encourages the model to generate audio that is closely linked to the text `prompt` at the expense of lower sound quality. Guidance scale is enabled when `guidance_scale > 1`. negative_prompt (`str` or `List[str]`, *optional*): The prompt or prompts to guide what to not include in audio generation. If not defined, you need to pass `negative_prompt_embeds` instead. Ignored when not using guidance (`guidance_scale < 1`). num_waveforms_per_prompt (`int`, *optional*, defaults to 1): The number of waveforms to generate per prompt. eta (`float`, *optional*, defaults to 0.0): Corresponds to parameter eta (η) from the [DDIM](https://arxiv.org/abs/2010.02502) paper. Only applies to the [`~schedulers.DDIMScheduler`], and is ignored in other schedulers. generator (`torch.Generator` or `List[torch.Generator]`, *optional*): A [`torch.Generator`](https://pytorch.org/docs/stable/generated/torch.Generator.html) to make generation deterministic. latents (`torch.Tensor`, *optional*): Pre-generated noisy latents sampled from a Gaussian distribution, to be used as inputs for audio generation. Can be used to tweak the same generation with different prompts. If not provided, a latents tensor is generated by sampling using the supplied random `generator`. initial_audio_waveforms (`torch.Tensor`, *optional*): Optional initial audio waveforms to use as the initial audio waveform for generation. Must be of shape `(batch_size, num_channels, audio_length)` or `(batch_size, audio_length)`, where `batch_size` corresponds to the number of prompts passed to the model. initial_audio_sampling_rate (`int`, *optional*): Sampling rate of the `initial_audio_waveforms`, if they are provided. Must be the same as the model. prompt_embeds (`torch.Tensor`, *optional*): Pre-computed text embeddings from the text encoder model. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not provided, text embeddings will be computed from `prompt` input argument. negative_prompt_embeds (`torch.Tensor`, *optional*): Pre-computed negative text embeddings from the text encoder model. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not provided, negative_prompt_embeds will be computed from `negative_prompt` input argument. attention_mask (`torch.LongTensor`, *optional*): Pre-computed attention mask to be applied to the `prompt_embeds`. If not provided, attention mask will be computed from `prompt` input argument. negative_attention_mask (`torch.LongTensor`, *optional*): Pre-computed attention mask to be applied to the `negative_text_audio_duration_embeds`. return_dict (`bool`, *optional*, defaults to `True`): Whether or not to return a [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] instead of a plain tuple. callback (`Callable`, *optional*): A function that calls every `callback_steps` steps during inference. The function is called with the following arguments: `callback(step: int, timestep: int, latents: torch.Tensor)`. callback_steps (`int`, *optional*, defaults to 1): The frequency at which the `callback` function is called. If not specified, the callback is called at every step. output_type (`str`, *optional*, defaults to `"pt"`): The output format of the generated audio. Choose between `"np"` to return a NumPy `np.ndarray` or `"pt"` to return a PyTorch `torch.Tensor` object. Set to `"latent"` to return the latent diffusion model (LDM) output. Examples: Returns: [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] or `tuple`: If `return_dict` is `True`, [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] is returned, otherwise a `tuple` is returned where the first element is a list with the generated audio. """ # 0. Convert audio input length from seconds to latent length downsample_ratio = self.vae.hop_length max_audio_length_in_s = self.transformer.config.sample_size * downsample_ratio / self.vae.config.sampling_rate if audio_end_in_s is None: audio_end_in_s = max_audio_length_in_s if audio_end_in_s - audio_start_in_s > max_audio_length_in_s: raise ValueError( f"The total audio length requested ({audio_end_in_s-audio_start_in_s}s) is longer than the model maximum possible length ({max_audio_length_in_s}). Make sure that 'audio_end_in_s-audio_start_in_s<={max_audio_length_in_s}'." ) waveform_start = int(audio_start_in_s * self.vae.config.sampling_rate) waveform_end = int(audio_end_in_s * self.vae.config.sampling_rate) waveform_length = int(self.transformer.config.sample_size) # 1. Check inputs. Raise error if not correct self.check_inputs( prompt, audio_start_in_s, audio_end_in_s, callback_steps, negative_prompt, prompt_embeds, negative_prompt_embeds, attention_mask, negative_attention_mask, initial_audio_waveforms, initial_audio_sampling_rate, ) # 2. Define call parameters if prompt is not None and isinstance(prompt, str): batch_size = 1 elif prompt is not None and isinstance(prompt, list): batch_size = len(prompt) else: batch_size = prompt_embeds.shape[0] device = self._execution_device # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2) # of the Imagen paper: https://arxiv.org/pdf/2205.11487.pdf . `guidance_scale = 1` # corresponds to doing no classifier free guidance. do_classifier_free_guidance = guidance_scale > 1.0 # 3. Encode input prompt prompt_embeds = self.encode_prompt( prompt, device, do_classifier_free_guidance, negative_prompt, prompt_embeds, negative_prompt_embeds, attention_mask, negative_attention_mask, ) # Encode duration seconds_start_hidden_states, seconds_end_hidden_states = self.encode_duration( audio_start_in_s, audio_end_in_s, device, do_classifier_free_guidance and (negative_prompt is not None or negative_prompt_embeds is not None), batch_size, ) # Create text_audio_duration_embeds and audio_duration_embeds text_audio_duration_embeds = torch.cat( [prompt_embeds, seconds_start_hidden_states, seconds_end_hidden_states], dim=1 ) audio_duration_embeds = torch.cat([seconds_start_hidden_states, seconds_end_hidden_states], dim=2) # In case of classifier free guidance without negative prompt, we need to create unconditional embeddings and # to concatenate it to the embeddings if do_classifier_free_guidance and negative_prompt_embeds is None and negative_prompt is None: negative_text_audio_duration_embeds = torch.zeros_like( text_audio_duration_embeds, device=text_audio_duration_embeds.device ) text_audio_duration_embeds = torch.cat( [negative_text_audio_duration_embeds, text_audio_duration_embeds], dim=0 ) audio_duration_embeds = torch.cat([audio_duration_embeds, audio_duration_embeds], dim=0) bs_embed, seq_len, hidden_size = text_audio_duration_embeds.shape # duplicate audio_duration_embeds and text_audio_duration_embeds for each generation per prompt, using mps friendly method text_audio_duration_embeds = text_audio_duration_embeds.repeat(1, num_waveforms_per_prompt, 1) text_audio_duration_embeds = text_audio_duration_embeds.view( bs_embed * num_waveforms_per_prompt, seq_len, hidden_size ) audio_duration_embeds = audio_duration_embeds.repeat(1, num_waveforms_per_prompt, 1) audio_duration_embeds = audio_duration_embeds.view( bs_embed * num_waveforms_per_prompt, -1, audio_duration_embeds.shape[-1] ) # 4. Prepare timesteps self.scheduler.set_timesteps(num_inference_steps, device=device) timesteps = self.scheduler.timesteps # 5. Prepare latent variables num_channels_vae = self.transformer.config.in_channels latents = self.prepare_latents( batch_size * num_waveforms_per_prompt, num_channels_vae, waveform_length, text_audio_duration_embeds.dtype, device, generator, latents, initial_audio_waveforms, num_waveforms_per_prompt, audio_channels=self.vae.config.audio_channels, ) # 6. Prepare extra step kwargs extra_step_kwargs = self.prepare_extra_step_kwargs(generator, eta) # 7. Prepare rotary positional embedding rotary_embedding = get_1d_rotary_pos_embed( self.rotary_embed_dim, latents.shape[2] + audio_duration_embeds.shape[1], use_real=True, repeat_interleave_real=False, ) # 8. Denoising loop num_warmup_steps = len(timesteps) - num_inference_steps * self.scheduler.order with self.progress_bar(total=num_inference_steps) as progress_bar: for i, t in enumerate(timesteps): # expand the latents if we are doing classifier free guidance latent_model_input = torch.cat([latents] * 2) if do_classifier_free_guidance else latents latent_model_input = self.scheduler.scale_model_input(latent_model_input, t) # predict the noise residual noise_pred = self.transformer( latent_model_input, t.unsqueeze(0), encoder_hidden_states=text_audio_duration_embeds, global_hidden_states=audio_duration_embeds, rotary_embedding=rotary_embedding, return_dict=False, )[0] # perform guidance if do_classifier_free_guidance: noise_pred_uncond, noise_pred_text = noise_pred.chunk(2) noise_pred = noise_pred_uncond + guidance_scale * (noise_pred_text - noise_pred_uncond) # compute the previous noisy sample x_t -> x_t-1 latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs).prev_sample # call the callback, if provided if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0): progress_bar.update() if callback is not None and i % callback_steps == 0: step_idx = i // getattr(self.scheduler, "order", 1) callback(step_idx, t, latents) # 9. Post-processing if not output_type == "latent": audio = self.vae.decode(latents).sample else: return AudioPipelineOutput(audios=latents) audio = audio[:, :, waveform_start:waveform_end] if output_type == "np": audio = audio.cpu().float().numpy() self.maybe_free_model_hooks() if not return_dict: return (audio,) return AudioPipelineOutput(audios=audio)