flymyai 1.0.20

No project description provided

FlyMy.AI

Generated with FlyMy.AI in 🚀 70ms

Welcome to FlyMy.AI inference platform. Our goal is to provide the fastest and most affordable deployment solutions for neural networks and AI applications.

• Fast Inference: Experience the fastest Stable Diffusion inference globally.
• Scalability: Autoscaling to millions of users per second.
• Ease of Use: One-click deployment for any publicly available neural networks.

Website

For more information, visit our website: FlyMy.AI Or connect with us and other users on Discord: Join Discord

Getting Started

This is a Python client for FlyMyAI. It allows you to easily run models and get predictions from your Python code in sync and async mode.

Requirements

• Python 3.8+

Installation

Install the FlyMyAI client using pip:

pip install flymyai

Authentication

Before using the client, you need to have your API key, username, and project name. In order to get credentials, you have to sign up on flymy.ai and get your personal data on the profile.

Basic Usage

Here's a simple example of how to use the FlyMyAI client:

BERT Sentiment analysis

import flymyai

response = flymyai.run(
    apikey="fly-secret-key",
    model="flymyai/bert",
    payload={"text": "What a fabulous fancy building! It looks like a palace!"}
)
print(response.output_data["logits"][0])

Sync Streams

For llms you should use stream method

llama 3.1 8b

from flymyai import client, FlyMyAIPredictException

fma_client = client(apikey="fly-secret-key")

stream_iterator = fma_client.stream(
    payload={
        "prompt": "tell me a story about christmas tree",
        "best_of": 12,
        "max_tokens": 1024,
        "stop": 1,
        "temperature": 1,
        "top_k": 1,
        "top_p": "0.95",
    },
    model="flymyai/llama-v3-1-8b"
)
try:
    for response in stream_iterator:
        if response.output_data.get("output"):
            print(response.output_data["output"].pop(), end="")
except FlyMyAIPredictException as e:
    print(e)
    raise e
finally:
    print()
    print(stream_iterator.stream_details)

Async Streams

For llms you should use stream method

Stable Code Instruct 3b

import asyncio

from flymyai import async_client, FlyMyAIPredictException


async def run_stable_code():
    fma_client = async_client(apikey="fly-secret-key")
    stream_iterator = fma_client.stream(
        payload={
            "prompt": "What's the difference between an iterator and a generator in Python?",
            "best_of": 12,
            "max_tokens": 512,
            "stop": 1,
            "temperature": 1,
            "top_k": 1,
            "top_p": "0.95",
        },
        model="flymyai/Stable-Code-Instruct-3b"
    )
    try:
        async for response in stream_iterator:
            if response.output_data.get("output"):
                print(response.output_data["output"].pop(), end="")
    except FlyMyAIPredictException as e:
        print(e)
        raise e
    finally:
        print()
        print(stream_iterator.stream_details)


asyncio.run(run_stable_code())

File Inputs

ResNet image classification

You can pass file inputs to models using file paths:

import pathlib

import flymyai

response = flymyai.run(
    apikey="fly-secret-key",
    model="flymyai/resnet",
    payload={"image": pathlib.Path("/path/to/image.png")}
)
print(response.output_data["495"])

File Response Handling

Files received from the neural network are always encoded in base64 format. To process these files, you need to decode them first. Here's an example of how to handle an image file:

StableDiffusion Turbo image generation in ~50ms 🚀

import base64
import flymyai

response = flymyai.run(
    apikey="fly-secret-key",
    model="flymyai/SDTurboFMAAceleratedH100",
    payload={
        "prompt": "An astronaut riding a rainbow unicorn, cinematic, dramatic, photorealistic",
    }
)
base64_image = response.output_data["sample"][0]
image_data = base64.b64decode(base64_image)
with open("generated_image.jpg", "wb") as file:
    file.write(image_data)

Asynchronous Requests

FlyMyAI supports asynchronous requests for improved performance. Here's how to use it:

import asyncio
import flymyai


async def main():
    payloads = [
        {
            "prompt": "An astronaut riding a rainbow unicorn, cinematic, dramatic, photorealistic",
            "negative_prompt": "Dark colors, gloomy atmosphere, horror",
            "seed": count,
            "denoising_steps": 4,
            "scheduler": "DPM++ SDE"
         }
        for count in range(1, 10)
    ]
    async with asyncio.TaskGroup() as gr:
        tasks = [
            gr.create_task(
                flymyai.async_run(
                    apikey="fly-secret-key",
                    model="flymyai/DreamShaperV2-1",
                    payload=payload
                )
            )
            for payload in payloads
        ]
    results = await asyncio.gather(*tasks)
    for result in results:
        print(result.output_data["output"])


asyncio.run(main())

Running Models in the Background

To run a model in the background, simply use the async_run() method:

import asyncio
import flymyai
import pathlib


async def background_task():
    payload = {"audio": pathlib.Path("/path/to/audio.mp3")}
    response = await flymyai.async_run(
        apikey="fly-secret-key",
        model="flymyai/whisper",
        payload=payload
    )
    print("Background task completed:", response.output_data["transcription"])


async def main():
    task = asyncio.create_task(background_task())
    await task

asyncio.run(main())
# Continue with other operations while the model runs in the background