Build multimodal AI services with cloud native technologies
Jina lets you build multimodal AI services and pipelines that communicate via gRPC, HTTP and WebSockets, then scale them up and deploy to production. You can focus on your logic and algorithms, without worrying about the infrastructure complexity.
Jina provides a smooth Pythonic experience for serving ML models transitioning from local deployment to advanced orchestration frameworks like Docker-Compose, Kubernetes, or Jina AI Cloud. Jina makes advanced solution engineering and cloud-native technologies accessible to every developer.
- Build and serve models for any data type and any mainstream deep learning framework.
- Design high-performance services, with easy scaling, duplex client-server streaming, batching, dynamic batching, async/non-blocking data processing and and any protocol.
- Serve LLM models while streaming their output.
- Docker container integration via Executor Hub, OpenTelemetry/Prometheus observability.
- Streamlined CPU/GPU hosting via Jina AI Cloud.
- Deploy to your own cloud or system with our Kubernetes and Docker Compose integration().
Wait, how is Jina different from FastAPI?
Jina's value proposition may seem quite similar to that of FastAPI. However, there are several fundamental differences:Data structure and communication protocols
- FastAPI communication relies on Pydantic and Jina relies on DocArray allowing Jina to support multiple protocols to expose its services. The support for gRPC protocol is specially useful for data intensive applications as for embedding services where the embeddings and tensors can be more efficiently serialized.
Advanced orchestration and scaling capabilities
- Jina allows you to easily containerize and orchestrate your services and models, providing concurrency and scalability.
- Jina lets you deploy applications formed from multiple microservices that can be containerized and scaled independently.
Journey to the cloud
- Jina provides a smooth transition from local development (using DocArray) to local serving using Deployment and Flow to having production-ready services by using Kubernetes capacity to orchestrate the lifetime of containers.
- By using Jina AI Cloud you have access to scalable and serverless deployments of your applications in one command.
pip install jina
Find more install options on Apple Silicon/Windows.
Jina has three fundamental layers:
- Data layer: BaseDoc and DocList (from DocArray) are the input/output formats in Jina.
- Serving layer: An Executor is a Python class that transforms and processes Documents. By simply wrapping your models into an Executor, you allow them to be served and scaled by Jina. Gateway is the service making sure connecting all Executors inside a Flow.
- Orchestration layer: Deployment serves a single Executor, while a Flow serves Executors chained into a pipeline.
The full glossary is explained here.
Let's build a fast, reliable and scalable gRPC-based AI service. In Jina we call this an Executor. Our simple Executor will wrap the StableLM LLM from Stability AI. We'll then use a Deployment to serve it.
Note A Deployment serves just one Executor. To combine multiple Executors into a pipeline and serve that, use a Flow.
Let's implement the service's logic:
executor.py |
---|
from jina import Executor, requests
from docarray import DocList, BaseDoc
from transformers import pipeline
class Prompt(BaseDoc):
text: str
class Generation(BaseDoc):
prompt: str
text: str
class StableLM(Executor):
def __init__(self, **kwargs):
super().__init__(**kwargs)
self.generator = pipeline(
'text-generation', model='stabilityai/stablelm-base-alpha-3b'
)
@requests
def generate(self, docs: DocList[Prompt], **kwargs) -> DocList[Generation]:
generations = DocList[Generation]()
prompts = docs.text
llm_outputs = self.generator(prompts)
for prompt, output in zip(prompts, llm_outputs):
generations.append(Generation(prompt=prompt, text=output))
return generations |
Then we deploy it with either the Python API or YAML:
Python API: deployment.py |
YAML: deployment.yml |
---|---|
from jina import Deployment
from executor import StableLM
dep = Deployment(uses=StableLM, timeout_ready=-1, port=12345)
with dep:
dep.block() |
jtype: Deployment
with:
uses: StableLM
py_modules:
- executor.py
timeout_ready: -1
port: 12345 And run the YAML Deployment with the CLI: |
Use Jina Client to make requests to the service:
from jina import Client
from docarray import DocList, BaseDoc
class Prompt(BaseDoc):
text: str
class Generation(BaseDoc):
prompt: str
text: str
prompt = Prompt(
text='suggest an interesting image generation prompt for a mona lisa variant'
)
client = Client(port=12345) # use port from output above
response = client.post(on='/', inputs=[prompt], return_type=DocList[Generation])
print(response[0].text)
a steampunk version of the Mona Lisa, incorporating mechanical gears, brass elements, and Victorian era clothing details
Note In a notebook, you can't use
deployment.block()
and then make requests to the client. Please refer to the Colab link above for reproducible Jupyter Notebook code snippets.
Sometimes you want to chain microservices together into a pipeline. That's where a Flow comes in.
A Flow is a DAG pipeline, composed of a set of steps, It orchestrates a set of Executors and a Gateway to offer an end-to-end service.
Note If you just want to serve a single Executor, you can use a Deployment.
For instance, let's combine our StableLM language model with a Stable Diffusion image generation model. Chaining these services together into a Flow will give us a service that will generate images based on a prompt generated by the LLM.
text_to_image.py |
---|
import numpy as np
from jina import Executor, requests
from docarray import BaseDoc, DocList
from docarray.documents import ImageDoc
class Generation(BaseDoc):
prompt: str
text: str
class TextToImage(Executor):
def __init__(self, **kwargs):
super().__init__(**kwargs)
from diffusers import StableDiffusionPipeline
import torch
self.pipe = StableDiffusionPipeline.from_pretrained(
"CompVis/stable-diffusion-v1-4", torch_dtype=torch.float16
).to("cuda")
@requests
def generate_image(self, docs: DocList[Generation], **kwargs) -> DocList[ImageDoc]:
images = self.pipe(
docs.text
).images # image here is in [PIL format](https://pillow.readthedocs.io/en/stable/)
docs.tensor = np.array(images) |
Build the Flow with either Python or YAML:
Python API: flow.py |
YAML: flow.yml |
---|---|
from jina import Flow
from executor import StableLM
from text_to_image import TextToImage
flow = (
Flow(port=12345)
.add(uses=StableLM, timeout_ready=-1)
.add(uses=TextToImage, timeout_ready=-1)
)
with flow:
flow.block() |
jtype: Flow
with:
port: 12345
executors:
- uses: StableLM
timeout_ready: -1
py_modules:
- executor.py
- uses: TextToImage
timeout_ready: -1
py_modules:
- text_to_image.py Then run the YAML Flow with the CLI: |
Then, use Jina Client to make requests to the Flow:
from jina import Client
from docarray import DocList, BaseDoc
from docarray.documents import ImageDoc
class Prompt(BaseDoc):
text: str
prompt = Prompt(
text='suggest an interesting image generation prompt for a mona lisa variant'
)
client = Client(port=12345) # use port from output above
response = client.post(on='/', inputs=[prompt], return_type=DocList[ImageDoc])
response[0].display()
Why not just use standard Python to build that service and pipeline? Jina accelerates time to market of your application by making it more scalable and cloud-native. Jina also handles the infrastructure complexity in production and other Day-2 operations so that you can focus on the data application itself.
Increase your application's throughput with scalability features out of the box, like replicas, shards and dynamic batching.
Let's scale a Stable Diffusion Executor deployment with replicas and dynamic batching:
- Create two replicas, with a GPU assigned for each.
- Enable dynamic batching to process incoming parallel requests together with the same model inference.
Normal Deployment | Scaled Deployment |
---|---|
jtype: Deployment
with:
uses: TextToImage
timeout_ready: -1
py_modules:
- text_to_image.py |
jtype: Deployment
with:
uses: TextToImage
timeout_ready: -1
py_modules:
- text_to_image.py
env:
CUDA_VISIBLE_DEVICES: RR
replicas: 2
uses_dynamic_batching: # configure dynamic batching
/default:
preferred_batch_size: 10
timeout: 200 |
Assuming your machine has two GPUs, using the scaled deployment YAML will give better throughput compared to the normal deployment.
These features apply to both Deployment YAML and Flow YAML. Thanks to the YAML syntax, you can inject deployment configurations regardless of Executor code.
In order to deploy your solutions to the cloud, you need to containerize your services. Jina provides the Executor Hub, the perfect tool to streamline this process taking a lot of the troubles with you. It also lets you share these Executors publicly or privately.
You just need to structure your Executor in a folder:
TextToImage/
├── executor.py
├── config.yml
├── requirements.txt
config.yml |
requirements.txt |
---|---|
jtype: TextToImage
py_modules:
- executor.py
metas:
name: TextToImage
description: Text to Image generation Executor based on StableDiffusion
url:
keywords: [] |
diffusers
accelerate
transformers |
Then push the Executor to the Hub by doing: jina hub push TextToImage
.
This will give you a URL that you can use in your Deployment
and Flow
to use the pushed Executors containers.
jtype: Flow
with:
port: 12345
executors:
- uses: jinai+docker://<user-id>/StableLM
- uses: jinai+docker://<user-id>/TextToImage
Using Kubernetes with Jina is easy:
jina export kubernetes flow.yml ./my-k8s
kubectl apply -R -f my-k8s
And so is Docker Compose:
jina export docker-compose flow.yml docker-compose.yml
docker-compose up
Note You can also export Deployment YAML to Kubernetes and Docker Compose.
That's not all. We also support OpenTelemetry, Prometheus, and Jaeger.
What cloud-native technology is still challenging to you? Tell us and we'll handle the complexity and make it easy for you.
You can also deploy a Flow to JCloud, where you can easily enjoy autoscaling, monitoring and more with a single command.
First, turn the flow.yml
file into a JCloud-compatible YAML by specifying resource requirements and using containerized Hub Executors.
Then, use jina cloud deploy
command to deploy to the cloud:
wget https://raw.githubusercontent.com/jina-ai/jina/master/.github/getting-started/jcloud-flow.yml
jina cloud deploy jcloud-flow.yml
Warning
Make sure to delete/clean up the Flow once you are done with this tutorial to save resources and credits.
Read more about deploying Flows to JCloud.
Large Language Models can power a wide range of applications from chatbots to assistants and intelligent systems. However, these models can be heavy and slow and your users want systems that are both intelligent and fast!
Large language models work by turning your questions into tokens and then generating new token one at a time until it decides that generation should stop. This means you want to stream the output tokens generated by a large language model to the client. In this tutorial, we will discuss how to achieve this with Streaming Endpoints in Jina.
The first step is to define the streaming service schemas, as you would do in any other service framework. The input to the service is the prompt and the maximum number of tokens to generate, while the output is simply the token ID:
from docarray import BaseDoc
class PromptDocument(BaseDoc):
prompt: str
max_tokens: int
class ModelOutputDocument(BaseDoc):
token_id: int
generated_text: str
Our service depends on a large language model. As an example, we will use the gpt2
model. This is how you would load
such a model in your executor
from jina import Executor, requests
from transformers import GPT2Tokenizer, GPT2LMHeadModel
import torch
tokenizer = GPT2Tokenizer.from_pretrained('gpt2')
class TokenStreamingExecutor(Executor):
def __init__(self, **kwargs):
super().__init__(**kwargs)
self.model = GPT2LMHeadModel.from_pretrained('gpt2')
Our streaming endpoint accepts a PromptDocument
as input and streams ModelOutputDocument
s. To stream a document back to
the client, use the yield
keyword in the endpoint implementation. Therefore, we use the model to generate
up to max_tokens
tokens and yield them until the generation stops:
class TokenStreamingExecutor(Executor):
...
@requests(on='/stream')
async def task(self, doc: PromptDocument, **kwargs) -> ModelOutputDocument:
input = tokenizer(doc.prompt, return_tensors='pt')
input_len = input['input_ids'].shape[1]
for _ in range(doc.max_tokens):
output = self.model.generate(**input, max_new_tokens=1)
if output[0][-1] == tokenizer.eos_token_id:
break
yield ModelOutputDocument(
token_id=output[0][-1],
generated_text=tokenizer.decode(
output[0][input_len:], skip_special_tokens=True
),
)
input = {
'input_ids': output,
'attention_mask': torch.ones(1, len(output[0])),
}
Learn more about {ref}streaming endpoints <streaming-endpoints>
from the Executor
documentation.
The final step is to serve the Executor and send requests using the client. To serve the Executor using gRPC:
from jina import Deployment
with Deployment(uses=TokenStreamingExecutor, port=12345, protocol='grpc') as dep:
dep.block()
To send requests from a client:
import asyncio
from jina import Client
async def main():
client = Client(port=12345, protocol='grpc', asyncio=True)
async for doc in client.stream_doc(
on='/stream',
inputs=PromptDocument(prompt='what is the capital of France ?', max_tokens=10),
return_type=ModelOutputDocument,
):
print(doc.generated_text)
asyncio.run(main())
The
The capital
The capital of
The capital of France
The capital of France is
The capital of France is Paris
The capital of France is Paris.
- Join our Discord community and chat with other community members about ideas.
- Subscribe to the latest video tutorials on our YouTube channel
Jina is backed by Jina AI and licensed under Apache-2.0.