gx! z d dl mZmZ ddlmZ ddlmZ e r ddlmZ ddl m
Z
dZ G d d
e Z d S ) )ListUnion )is_torch_available )Pipeline)%MODEL_FOR_TEXT_TO_SPECTROGRAM_MAPPING)SpeechT5HifiGanzmicrosoft/speecht5_hifiganc v e Zd ZdZddd fd
Zd Zd Zdeee e f f fd Z
d d Z d
Z xZ
S ) TextToAudioPipelinea5
Text-to-audio generation pipeline using any `AutoModelForTextToWaveform` or `AutoModelForTextToSpectrogram`. This
pipeline generates an audio file from an input text and optional other conditional inputs.
Example:
```python
>>> from transformers import pipeline
>>> pipe = pipeline(model="suno/bark-small")
>>> output = pipe("Hey it's HuggingFace on the phone!")
>>> audio = output["audio"]
>>> sampling_rate = output["sampling_rate"]
```
Learn more about the basics of using a pipeline in the [pipeline tutorial](../pipeline_tutorial)
You can specify parameters passed to the model by using [`TextToAudioPipeline.__call__.forward_params`] or
[`TextToAudioPipeline.__call__.generate_kwargs`].
Example:
```python
>>> from transformers import pipeline
>>> music_generator = pipeline(task="text-to-audio", model="facebook/musicgen-small", framework="pt")
>>> # diversify the music generation by adding randomness with a high temperature and set a maximum music length
>>> generate_kwargs = {
... "do_sample": True,
... "temperature": 0.7,
... "max_new_tokens": 35,
... }
>>> outputs = music_generator("Techno music with high melodic riffs", generate_kwargs=generate_kwargs)
```
This pipeline can currently be loaded from [`pipeline`] using the following task identifiers: `"text-to-speech"` or
`"text-to-audio"`.
See the list of available models on [huggingface.co/models](https://huggingface.co/models?filter=text-to-speech).
N)vocoder
sampling_ratec z t j |i | | j dk rt d d | _ | j j t j v r?|6t j
t | j j
n|| _ || _ | j | j j j | _ | j t| j j }| j j dd }|'| | dD ] }t) ||d }||| _ d S d S )Ntfz5The TextToAudioPipeline is only available in PyTorch.generation_config) sample_rater )super__init__ framework
ValueErrorr
model __class__r valuesr
from_pretrainedDEFAULT_VOCODER_IDtodevicer config__dict__getupdateto_dictgetattr) selfr
r argskwargsr
gen_configsampling_rate_namer s `/var/www/html/ai-engine/env/lib/python3.11/site-packages/transformers/pipelines/text_to_audio.pyr z TextToAudioPipeline.__init__L sM $)&))) >T ! ! TUU U : #H#O#Q#Q Q Q ? /0BCCFFtzGXYYY
L + < #!% !4!BD % Z&F ,001DdKKJ%
j 0 0 2 2333&F
7
7" '0BD I I
,)6D& & %
7
7 c t |t r|g}| j j j dk r=| j j dd ddddd}| | |} | j |fi |dd i}|S )
Nbarkmax_input_semantic_length FT
max_length)r/ add_special_tokensreturn_attention_maskreturn_token_type_idspaddingreturn_tensorspt)
isinstancestrr r
model_typer semantic_configr r! tokenizer)r$ textr&
new_kwargsoutputs r)
preprocessz TextToAudioPipeline.preprocessk s dC 6D : '6 1 1 # 4 D H HIdfi j j&+)-).' J
f % % % F DDDDtDDD
r* c | || j }|d }|d }| j rT| || j }d|vr
| j |d< | | | j j di ||}nHt | r$t d| | j di ||d }| j
|
| }|S )N)r forward_paramsgenerate_kwargsr a\ You're using the `TextToAudioPipeline` with a forward-only model, but `generate_kwargs` is non empty.
For forward-only TTA models, please use `forward_params` instead of of
`generate_kwargs`. For reference, here are the `generate_kwargs` used here:
r ) _ensure_tensor_on_devicer r can_generater r! generatelenr keysr
)r$ model_inputsr& r@ rA r= s r) _forwardz TextToAudioPipeline._forward s: ..vdk.JJ 01 !23 : " " $ $ E " ; ;OTXT_ ; ` `O #/997;7M 34
! !/ 2 2 2(TZ(JJ<J>JJFF?##
= #2"6"6"8"8= = TZAA,A.AA!DF < #\\&))F
r* text_inputsc 8 t j |fi |S )a
Generates speech/audio from the inputs. See the [`TextToAudioPipeline`] documentation for more information.
Args:
text_inputs (`str` or `List[str]`):
The text(s) to generate.
forward_params (`dict`, *optional*):
Parameters passed to the model generation/forward method. `forward_params` are always passed to the
underlying model.
generate_kwargs (`dict`, *optional*):
The dictionary of ad-hoc parametrization of `generate_config` to be used for the generation call. For a
complete overview of generate, check the [following
guide](https://huggingface.co/docs/transformers/en/main_classes/text_generation). `generate_kwargs` are
only passed to the underlying model if the latter is a generative model.
Return:
A `dict` or a list of `dict`: The dictionaries have two keys:
- **audio** (`np.ndarray` of shape `(nb_channels, audio_length)`) -- The generated audio waveform.
- **sampling_rate** (`int`) -- The sampling rate of the generated audio waveform.
)r __call__)r$ rJ r@ r s r) rL z TextToAudioPipeline.__call__ s$ , uww >>~>>>r* c 2 |r|ni |r|ni d}|i }i }|||fS )N)r@ rA rB )r$ preprocess_paramsr@ rA paramspostprocess_paramss r) _sanitize_parametersz(TextToAudioPipeline._sanitize_parameters sJ 1? FnnB2A Ir
$ " &*<< rI r r7 r rL rQ r[
__classcell__)r s @r) r r s . .` '+$ 7 7 7 7 7 7 7> . D?E#tCy.$9 ? ? ? ? ? ?4 = = = =" r* r N)
typingr r utilsr baser models.auto.modeling_autor !models.speecht5.modeling_speecht5r
r r rB r* r) rf s & & & & & & DQQQQQQCCCCCC1 { { { { {( { { { { {r*