Hugging Face

​

Prerequisites

1. Create an API key to authorize requests to Nebius Token Factory.
2. If you want to work with Nebius Token Factory models in the Hugging Face Model Hub, add the Nebius Token Factory API key to your Hugging Face account:
   • Create a Hugging Face account.
   • Go to the Inference Providers section in your Hugging Face user account settings.
   • In the Nebius Token Factory row, click the key icon in the Routing mode column.
   • In the window that opens, enter your Nebius Token Factory API key.

3. If you want to work with Nebius Token Factory models via the Hugging Face SDK, install the Hugging Face libraries for the programming language that you prefer:
   • Python
   • Node.js

   Install the huggingface_hub package (version v0.29.0 or later):

   pip install huggingface_hub>=0.29.0
   

   Install the @huggingface/inference package:

   npm install @huggingface/inference
   

​

How to work with Nebius Token Factory models in the Hugging Face Model Hub

​

The Hugging Face Model Hub enables you to see models, review information about them on model cards and interact with models directly in the browser.

1. Go to the Models section.
2. Make sure that Nebius Token Factory is selected in the Other → Inference Providers filter.
3. Go to the page of the required model.
4. In the Inference Providers section, select Nebius Token Factory.
5. Interact with the model in the same section. For example, if you selected a text-to-text model, you can send a prompt to it. If you want to work with a model in the Hugging Face playground where you can change the model settings, click Open playground.

​

How to work with Nebius Token Factory models via the Hugging Face SDK

​

Text-to-text models

from huggingface_hub import InferenceClient

client = InferenceClient(
    provider="nebius", 
    api_key="<your_API_key>"
)

messages=[
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "Tell me an easy ice cream recipe."},
]

completion = client.chat.completions.create(
    model="Qwen/Qwen2.5-VL-72B-Instruct",  
    messages=messages, 
    max_tokens=500
)

print(completion.choices[0].message.content)

import { InferenceClient } from "@huggingface/inference";

const client = new InferenceClient("<your_API_key>");

const chatCompletion = await client.chatCompletion({
    model: "Qwen/Qwen2.5-VL-72B-Instruct",
    messages: [
       {
          role: "system",
          content: "You are a helpful assistant."
       },
       {
          role: "user",
          content: "Tell me an easy cookie recipe."
       }
    ],
    provider: "nebius",
    max_tokens: 500
});

// Log the response
console.log(chatCompletion.choices[0].message);

• To work with a different model, change the model parameter. See a list of available models.
• Modify messages to specify your questions.

​

Text-to-image models

from huggingface_hub import InferenceClient

client = InferenceClient(
    provider="nebius", 
    api_key="<your_API_key>"
)

image = client.text_to_image(
    "Map of a space station with cultivated plants and living quarters",
    model="stabilityai/stable-diffusion-xl-base-1.0"
)

image.show()

import { InferenceClient } from "@huggingface/inference";

const client = new InferenceClient("<your_API_key>");

const generatedImage = await client.textToImage({
    model: "stabilityai/stable-diffusion-xl-base-1.0",
    inputs: "Map of a space station with cultivated plants and living quarters",
    provider: "nebius"
});

• To work with a different model, change the model parameter. See a list of available models.
• Modify the description of the image that should be generated.