Using the Finetuned Model

After a fine-tuning training job has finished, you can switch to the trained model version by using a special model identifier.

Model identifiers

Model identifiers are used to identify different model versions.

An identifier of the form "augento:<model_version>" (e.g. "augento:v2") selects a fine-tuned model version directly.

Any other identifier (e.g. "gpt-4o") selects the base model version (v1) and calls the corresponding model using the underlying provider’s API.

If you have set a default model version other than v1 on the model’s page, you can also call the default model version using the empty string "" as the model identifier.

Running inference

Using the desired model version identifier, as well as an API key from your model (starting with sk-au-), you can run inference on a fine-tuned model like this:

Python - Langchain

agent.py


from langchain_openai import ChatOpenAI
 
llm = ChatOpenAI(
    model="augento:v2", # put your model identifier here
    api_key="sk-au-...",
    base_url="https://api.augento.ai/v1"
)
 
completion = llm.invoke("Hello!")
 
print(completion)

Python - OpenAI

agent.py


from openai import OpenAI
 
llm = OpenAI(
    api_key="sk-au-...",
    base_url="https://api.augento.ai/v1"
)
 
completion = llm.chat.completions.create(
    model="augento:v2", # put your model identifier here
    messages=[{"role": "user", "content": "Hello!"}]
)
 
print(completion)

JS/TS - OpenAI

agent.ts


import { OpenAI } from "openai";
 
const openai = new OpenAI({
  apiKey: "sk-au-...",
  baseURL: "https://api.augento.ai/v1"
});
 
completion = openai.chat.completions.create(
    model="augento:v2", // put your model identifier here
    messages=[{"role": "user", "content": "Hello!"}]
);
 
console.log(completion);

Playground

On the model page, you can also test a fine-tuned model directly using the playground tab. In the playground you can just chat with different model versions like you can with ChatGPT.