Transformers documentation

Pipelines for inference

You are viewing v4.27.2 version. A newer version v4.41.3 is available.
Hugging Face's logo
Join the Hugging Face community

and get access to the augmented documentation experience

to get started

Pipelines for inference

The pipeline() makes it simple to use any model from the Hub for inference on any language, computer vision, speech, and multimodal tasks. Even if you don’t have experience with a specific modality or aren’t familiar with the underlying code behind the models, you can still use them for inference with the pipeline()! This tutorial will teach you to:

  • Use a pipeline() for inference.
  • Use a specific tokenizer or model.
  • Use a pipeline() for audio, vision, and multimodal tasks.

Take a look at the pipeline() documentation for a complete list of supported tasks and available parameters.

Pipeline usage

While each task has an associated pipeline(), it is simpler to use the general pipeline() abstraction which contains all the task-specific pipelines. The pipeline() automatically loads a default model and a preprocessing class capable of inference for your task.

  1. Start by creating a pipeline() and specify an inference task:
>>> from transformers import pipeline

>>> generator = pipeline(task="automatic-speech-recognition")
  1. Pass your input text to the pipeline():
>>> generator("")

Not the result you had in mind? Check out some of the most downloaded automatic speech recognition models on the Hub to see if you can get a better transcription. Let’s try openai/whisper-large:

>>> generator = pipeline(model="openai/whisper-large")
>>> generator("")
{'text': ' I have a dream that one day this nation will rise up and live out the true meaning of its creed.'}

Now this result looks more accurate! We really encourage you to check out the Hub for models in different languages, models specialized in your field, and more. You can check out and compare model results directly from your browser on the Hub to see if it fits or handles corner cases better than other ones. And if you don’t find a model for your use case, you can always start training your own!

If you have several inputs, you can pass your input as a list:


If you want to iterate over a whole dataset, or want to use it for inference in a webserver, check out dedicated parts

Using pipelines on a dataset

Using pipelines for a webserver


pipeline() supports many parameters; some are task specific, and some are general to all pipelines. In general you can specify parameters anywhere you want:

generator(model="openai/whisper-large", my_parameter=1)
out = generate(...)  # This will use `my_parameter=1`.
out = generate(..., my_parameter=2)  # This will override and use `my_parameter=2`.
out = generate(...)  # This will go back to using `my_parameter=1`.

Let’s check out 3 important ones:


If you use device=n, the pipeline automatically puts the model on the specified device. This will work regardless of whether you are using PyTorch or Tensorflow.

generator(model="openai/whisper-large", device=0)

If the model is too large for a single GPU, you can set device_map="auto" to allow 🤗 Accelerate to automatically determine how to load and store the model weights.

#!pip install accelerate
generator(model="openai/whisper-large", device_map="auto")

Note that if device_map="auto" is passed, there is no need to add the argument device=device when instantiating your pipeline as you may encounter some unexpected behavior!

Batch size

By default, pipelines will not batch inference for reasons explained in detail here. The reason is that batching is not necessarily faster, and can actually be quite slower in some cases.

But if it works in your use case, you can use:

generator(model="openai/whisper-large", device=0, batch_size=2)
audio_filenames = [f"audio_{i}.flac" for i in range(10)]
texts = generator(audio_filenames)

This runs the pipeline on the 10 provided audio files, but it will pass them in batches of 2 to the model (which is on a GPU, where batching is more likely to help) without requiring any further code from you. The output should always match what you would have received without batching. It is only meant as a way to help you get more speed out of a pipeline.

Pipelines can also alleviate some of the complexities of batching because, for some pipelines, a single item (like a long audio file) needs to be chunked into multiple parts to be processed by a model. The pipeline performs this chunk batching for you.

Task specific parameters

All tasks provide task specific parameters which allow for additional flexibility and options to help you get your job done. For instance, the method has a return_timestamps parameter which sounds promising for subtitling videos:

>>> # Not using whisper, as it cannot provide timestamps.
>>> generator = pipeline(model="facebook/wav2vec2-large-960h-lv60-self", return_timestamps="word")
>>> generator("")
{'text': 'I HAVE A DREAM BUT ONE DAY THIS NATION WILL RISE UP AND LIVE OUT THE TRUE MEANING OF ITS CREED', 'chunks': [{'text': 'I', 'timestamp': (1.22, 1.24)}, {'text': 'HAVE', 'timestamp': (1.42, 1.58)}, {'text': 'A', 'timestamp': (1.66, 1.68)}, {'text': 'DREAM', 'timestamp': (1.76, 2.14)}, {'text': 'BUT', 'timestamp': (3.68, 3.8)}, {'text': 'ONE', 'timestamp': (3.94, 4.06)}, {'text': 'DAY', 'timestamp': (4.16, 4.3)}, {'text': 'THIS', 'timestamp': (6.36, 6.54)}, {'text': 'NATION', 'timestamp': (6.68, 7.1)}, {'text': 'WILL', 'timestamp': (7.32, 7.56)}, {'text': 'RISE', 'timestamp': (7.8, 8.26)}, {'text': 'UP', 'timestamp': (8.38, 8.48)}, {'text': 'AND', 'timestamp': (10.08, 10.18)}, {'text': 'LIVE', 'timestamp': (10.26, 10.48)}, {'text': 'OUT', 'timestamp': (10.58, 10.7)}, {'text': 'THE', 'timestamp': (10.82, 10.9)}, {'text': 'TRUE', 'timestamp': (10.98, 11.18)}, {'text': 'MEANING', 'timestamp': (11.26, 11.58)}, {'text': 'OF', 'timestamp': (11.66, 11.7)}, {'text': 'ITS', 'timestamp': (11.76, 11.88)}, {'text': 'CREED', 'timestamp': (12.0, 12.38)}]}

As you can see, the model inferred the text and also outputted when the various words were pronounced in the sentence.

There are many parameters available for each task, so check out each task’s API reference to see what you can tinker with! For instance, the AutomaticSpeechRecognitionPipeline has a chunk_length_s parameter which is helpful for working on really long audio files (for example, subtitling entire movies or hour-long videos) that a model typically cannot handle on its own.

If you can’t find a parameter that would really help you out, feel free to request it!

Using pipelines on a dataset

The pipeline can also run inference on a large dataset. The easiest way we recommend doing this is by using an iterator:

def data():
    for i in range(1000):
        yield f"My example {i}"

pipe = pipeline(model="gpt2", device=0)
generated_characters = 0
for out in pipe(data()):
    generated_characters += len(out[0]["generated_text"])

The iterator data() yields each result, and the pipeline automatically recognizes the input is iterable and will start fetching the data while it continues to process it on the GPU (this uses DataLoader under the hood). This is important because you don’t have to allocate memory for the whole dataset and you can feed the GPU as fast as possible.

Since batching could speed things up, it may be useful to try tuning the batch_size parameter here.

The simplest way to iterate over a dataset is to just load one from 🤗 Datasets:

# KeyDataset is a util that will just output the item we're interested in.
from transformers.pipelines.pt_utils import KeyDataset
from datasets import load_dataset

pipe = pipeline(model="hf-internal-testing/tiny-random-wav2vec2", device=0)
dataset = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation[:10]")

for out in pipe(KeyDataset(dataset, "audio")):

Using pipelines for a webserver

Creating an inference engine is a complex topic which deserves it's own page.


Vision pipeline

Using a pipeline() for vision tasks is practically identical.

Specify your task and pass your image to the classifier. The image can be a link or a local path to the image. For example, what species of cat is shown below?


>>> from transformers import pipeline

>>> vision_classifier = pipeline(model="google/vit-base-patch16-224")
>>> preds = vision_classifier(
...     images=""
... )
>>> preds = [{"score": round(pred["score"], 4), "label": pred["label"]} for pred in preds]
>>> preds
[{'score': 0.4335, 'label': 'lynx, catamount'}, {'score': 0.0348, 'label': 'cougar, puma, catamount, mountain lion, painter, panther, Felis concolor'}, {'score': 0.0324, 'label': 'snow leopard, ounce, Panthera uncia'}, {'score': 0.0239, 'label': 'Egyptian cat'}, {'score': 0.0229, 'label': 'tiger cat'}]

Text pipeline

Using a pipeline() for NLP tasks is practically identical.

>>> from transformers import pipeline

>>> # This model is a `zero-shot-classification` model.
>>> # It will classify text, except you are free to choose any label you might imagine
>>> classifier = pipeline(model="facebook/bart-large-mnli")
>>> classifier(
...     "I have a problem with my iphone that needs to be resolved asap!!",
...     candidate_labels=["urgent", "not urgent", "phone", "tablet", "computer"],
... )
{'sequence': 'I have a problem with my iphone that needs to be resolved asap!!', 'labels': ['urgent', 'phone', 'computer', 'not urgent', 'tablet'], 'scores': [0.504, 0.479, 0.013, 0.003, 0.002]}

Multimodal pipeline

The pipeline() supports more than one modality. For example, a visual question answering (VQA) task combines text and image. Feel free to use any image link you like and a question you want to ask about the image. The image can be a URL or a local path to the image.

For example, if you use this invoice image:

>>> from transformers import pipeline

>>> vqa = pipeline(model="impira/layoutlm-document-qa")
>>> vqa(
...     image="",
...     question="What is the invoice number?",
... )
[{'score': 0.42515, 'answer': 'us-001', 'start': 16, 'end': 16}]

To run the example above you need to have pytesseract installed in addition to 🤗 Transformers:

sudo apt install -y tesseract-ocr
pip install pytesseract

Using pipeline on large models with 🤗 accelerate:

You can easily run pipeline on large models using 🤗 accelerate! First make sure you have installed accelerate with pip install accelerate.

First load your model using device_map="auto"! We will use facebook/opt-1.3b for our example.

# pip install accelerate
import torch
from transformers import pipeline

pipe = pipeline(model="facebook/opt-1.3b", torch_dtype=torch.bfloat16, device_map="auto")
output = pipe("This is a cool example!", do_sample=True, top_p=0.95)

You can also pass 8-bit loaded models if you install bitsandbytes and add the argument load_in_8bit=True

# pip install accelerate bitsandbytes
import torch
from transformers import pipeline

pipe = pipeline(model="facebook/opt-1.3b", device_map="auto", model_kwargs={"load_in_8bit": True})
output = pipe("This is a cool example!", do_sample=True, top_p=0.95)

Note that you can replace the checkpoint with any of the Hugging Face model that supports large model loading such as BLOOM!