No project description provided
Project description
Documentation 📖: https://docs.vendi-ai.com
Source Code 💻 : https://github.com/vendi-ai/vendi-python
Vendi Python API Library
The Vendi python library provides convenient access to the Vendi REST API from any Python 3.7+ application. The library includes all access to the Vendi API, including the ability to create, update, and delete resources.
Installation
Install the latest version of the library with pip:
pip install vendi
Usage
The library needs to be configured with your account's API key which is available in the Vendi web interface.
from vendi import Vendi
client = Vendi(api_key="YOUR_API_KEY")
Or you can set the VENDI_API_KEY
environment variable and the library will automatically use it:
export VENDI_API_KEY="YOUR_API_KEY"
Inference
The library provides a convenient way to make inference requests to your models.
from vendi import Vendi
client = Vendi(api_key="YOUR_API_KEY")
chat_completion = client.completions.create(
model_id="vendi/mistral-7b-instruct-v2",
messages=[
{
"role": "user",
"content": "Hello, I'm looking for a new job. Write me a resume."
}
],
max_tokens=5,
temperature=0.5,
top_p=1,
frequency_penalty=0.5,
presence_penalty=0.5,
stop=["\n"]
)
print(chat_completion.provider_response.choices[0].text)
The inference endpoints are OpenAI compatible, so you can use the same parameters as the OpenAI API or even the OpenAI Python library.
Datasets
The library provides a convenient way to upload and download datasets from your account.
from vendi import Vendi
client = Vendi(api_key="YOUR_API_KEY")
dataset = client.datasets.upload(
name="My Dataset",
data=[
{
"input": "The bill is $5.",
"output": "5$"
},
{
"name": "They gave us a nice $10 bill.",
"output": "10$"
}
]
)
You can also download datasets from your account.
from vendi import Vendi
client = Vendi(api_key="YOUR_API_KEY")
dataset = client.datasets.get("YOUR_DATASET_ID")
data = dataset.load_data()
Finetune
The library provides a convenient way to finetune models from your account.
from vendi import Vendi
client = Vendi(
api_key="my-api-key"
)
dataset = client.datasets.upload(
name="my-dataset",
data_path="conversation.jsonl"
)
finetune_job = client.finetune.run(
run_name="my-first-finetune",
model_description="My first finetune",
model_name="vendi/mistralai/Mistral-7B-Instruct-v0.2",
dataset_id=dataset.id
)
print(finetune_job)
The example above will create a finetune job that will run on the Vendi platform. You can monitor the progress of the finetune job in the Vendi web interface.
To finetune other models rather than mistralai, you can use the client.finetune.available_models()
method.
Models
The models are the finetuned models created out of the finetune jobs. You can list all the models in your account with
the client.models.list()
method.
from vendi import Vendi
client = Vendi(
api_key="my-api-key"
)
client.models.list()
Inference on finetuned model
Every finetuned model can be used for inference. You can use the client.completions.create()
method to make inference
requests to your models just like any other pre-trained models.
from vendi import Vendi
client = Vendi(
api_key="my-api-key"
)
client.completions.create(
model_id="<account-name>/<finetuned-model-name>",
messages=[
{
"role": "user",
"content": "Hello, I'm looking for a new job. Write me a resume."
}
],
max_tokens=256,
temperature=0.5,
top_p=1,
frequency_penalty=0.5,
presence_penalty=0.5,
stop=["\n"]
)
📚 For a thorough introduction, features, and best practices, explore our official documentation and quickstart.
Contributing 💼
Join our mission to make declarative AI even better together! We welcome contributions from everyone. Please read our contributing guide to get started.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
File details
Details for the file vendi-0.1.24.tar.gz
.
File metadata
- Download URL: vendi-0.1.24.tar.gz
- Upload date:
- Size: 16.0 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/5.0.0 CPython/3.10.14
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 93a3e89b0b7a151ac7aff97af12cf61ed1865cda90a8383acf95bb3fbfe0f9ab |
|
MD5 | 71602af631245f0058a9c58dbdd6210e |
|
BLAKE2b-256 | 18e5ad983aee82f94233a562a2bddba46ffeab51f03098a0d25455bfb22af842 |
File details
Details for the file vendi-0.1.24-py3-none-any.whl
.
File metadata
- Download URL: vendi-0.1.24-py3-none-any.whl
- Upload date:
- Size: 21.1 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/5.0.0 CPython/3.10.14
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 3c754d28875a9e0719517e5f20ff703e2bbe704342ce114f38ea755a10ed355e |
|
MD5 | ab9ef9c8b4190adc853741cebb8566dd |
|
BLAKE2b-256 | 750934856057c1ab5e740a258f3d989d7fe34bb77aa7eb26482ff3b1e559ce78 |