Skip to main content

Files API

Changes to Anyscale Endpoints API

Effective August 1, 2024 Anyscale Endpoints API will be available exclusively through the fully Hosted Anyscale Platform. Multi-tenant access to LLM models will be removed.

With the Hosted Anyscale Platform, you can access the latest GPUs billed by the second, and deploy models on your own dedicated instances. Enjoy full customization to build your end-to-end applications with Anyscale. Get started today.

The Files API is available at https://api.endpoints.anyscale.com/v1/files. This API allows you to upload training and validation datasets to use for your fine-tuning jobs.

This API is compatible with the OpenAI SDK.

APIs​

Upload file​

POST https://api.endpoints.anyscale.com/v1/files

Uploads a file for use in fine-tuning. The file format must be in the JSON Lines format, which is a newline separated list of JSON objects.

See Data format for more details on the exact file format schema.

Request body​

  • file string - The contents of the file to upload.

Returns​

A file object

Examples

Use the OpenAI SDK​

import openai

client = openai.OpenAI(
    base_url = "https://api.endpoints.anyscale.com/v1",
    api_key = "esecret_yourAuthTokenHere"
)

file_name = "train.jsonl"
file = client.files.create(
  file=open(file_name, "rb"),
  purpose="fine-tune",
  user_provided_filename=file_name,
)

Use cURL​

curl https://api.endpoints.anyscale.com/v1/files \
  -H "Authorization: Bearer esecret_yourAuthTokenHere" \
  -F 'file=@train.jsonl'
info

If you're seeing 504 errors with Your request has exceeded the timeout of 3 minutes when uploading, change the base URL for uploading to:

https://app.endpoints.anyscale.com/endpoints-api/v1

List files​

GET https://api.endpoints.anyscale.com/v1/files

Gets a list of all the files that you uploaded to the Endpoints service.

Returns​

A list of file object in the following format:

{
  "data": [File, ...]
}

Examples
Examples

Use the OpenAI SDK​

import openai

client = openai.OpenAI(
    base_url = "https://api.endpoints.anyscale.com/v1",
    api_key = "esecret_yourAuthTokenHere")
files = client.files.list()

Use cURL​

curl https://api.endpoints.anyscale.com/v1/files \
  -H "Authorization: Bearer esecret_yourAuthTokenHere"

Get file​

GET https://api.endpoints.anyscale.com/v1/files/{file_id}

Gets the metadata of a single file that you uploaded to the Endpoints service.

Returns​

A file object.

Examples

Use the OpenAI SDK​

import openai

client = openai.OpenAI(
    base_url = "https://api.endpoints.anyscale.com/v1",
    api_key = "esecret_yourAuthTokenHere")
file = client.files.retrieve("file_123")

Use cURL​

curl https://api.endpoints.anyscale.com/v1/files/file_123 \
  -H "Authorization: Bearer esecret_yourAuthTokenHere"

Get file contents​

GET https://api.endpoints.anyscale.com/v1/files/{file_id}/content

Returns the contents of the file that you uploaded to the Endpoints service.

📌 When using cURL, you must pass in a -L parameter to automatically follow redirects and retrieve the file contents.

Returns​

The contents of the uploaded file.

Examples

Use the OpenAI SDK​

import openai

client = openai.OpenAI(
    base_url = "https://api.endpoints.anyscale.com/v1",
    api_key = "esecret_yourAuthTokenHere")
content = client.files.retrieve_content("file_123")

Use cURL​

curl -L https://api.endpoints.anyscale.com/v1/files/file_123/content \
  -H "Authorization: Bearer esecret_yourAuthTokenHere"

Loading .jsonl files​

Many of the files are in jsonl format. The training data, the validation data, and the fine-tuning results are all in this format. The following code snippet shows how to load a jsonl file into a list of JSON objects.

from json import loads
jsonl_lines = content.decode("utf-8").split("\n")
content = [loads(line) for line in jsonl_lines if line != ""]

Delete file​

DELETE https://api.endpoints.anyscale.com/v1/files/{file_id}

Deletes a file that was previously uploaded to the endpoints service.

Returns​

The deleted file object.

Examples

Use the OpenAI SDK​

import openai

client = openai.OpenAI(
    base_url = "https://api.endpoints.anyscale.com/v1",
    api_key = "esecret_yourAuthTokenHere")
content = openai.files.delete("file_123")

Use cURL​

curl -X DELETE https://api.endpoints.anyscale.com/v1/files/file_123 \
  -H "Authorization: Bearer esecret_yourAuthTokenHere"

Models​

File​

The file object represents a file that you uploaded to the Endpoints service. These files are referenced by ID in the fine-tuning API to use for training and validation.

📌 The following fields that are available in OpenAI's File object aren't supported by Anyscale Endpoints: object, purpose, status, status_details

Attributes​

  • id string - The id of the file.
  • bytes int | null - The size of the file in bytes or null, if the size is unknown.
  • created_at string - The date and time of the uploaded file in ISO 8601 format.
  • filename string - The original name of the uploaded file.
Example
{
  "id": "file_123",
  "bytes": 123,
  "created_at": "2023-09-14T18:00:00Z",
  "filename": "train.jsonl"
}