Cloud_type issue for creating a workspace

Accelerated Computing Intelligent Video Analytics TAO Toolkit
1

Hi,

We are encountering an issue when attempting to create a workspace to initiate fine-tuning with TAO. We consistently face the following problem:

Error 400 - “error_desc”: “{‘cloud_type’: ['Invalid enum member ']}”

We attempted to use our Azure storage, and for testing purposes, we also tried AWS and Hugging Face, but we encountered the same error each time. However the API works well for other endpoints (login, create dataset, …).

For information here our code (also test directly in swagger, so without code) :

image
image477×237 5.35 KB

image
image703×272 7.7 KB

Thank you very much for your help.

2

Could you please set to "cloud_type": "aws" and retry?
The 400 error comes from sending an invalid or empty value for the cloud_type field when creating the TAO workspace; it must match the allowed enum values (for example aws )

An example, TAO Python SDK and CLI — Tao Toolkit.

3

Hi,

thank you for your reply.

We tried “aws” et hugging face but it still doesn’t work.

We have noticed that there are different structure for the cloud_metadata but we don’t know what is the right one?

The doc tao toolkit REST API overview and examples version :

Azure - cloud_type: azure; cloud_specific_details needed: account_name, access_key, region, container_name

The swagger version:

image
image445×342 6.63 KB

The notebook version :

image
image624×247 5.37 KB

Thank you for your help.

4

May I know the latest log when you tried “aws”? Is it still 400 error?

More,
According to tao-core/nvidia_tao_core/microservices/blueprints/v2/schemas.py at 658412c3eaa0234bae5dda0d917b803fc5d6bfb0 · NVIDIA/tao-core · GitHub), the expected cloud_type is one of below options.
"aws" / "azure" / "seaweedfs" / "huggingface" / "self_hosted"

Also, it is needed to set the structure similar to below. Refer to tao-core/nvidia_tao_core/microservices/blueprints/v2/schemas.py at 658412c3eaa0234bae5dda0d917b803fc5d6bfb0 · NVIDIA/tao-core · GitHub.

{
  "name": "...",
  "shared": false,
  "version": "x.y.z",
  "cloud_type": "...",
  "cloud_specific_details": { ... }
}

For example,

{
  "name": "debug_ws",
  "shared": false,
  "version": "1.0.0",
  "cloud_type": "aws",
  "cloud_specific_details": {
    "access_key": "AKIA...",
    "secret_key": "xxxx",
    "cloud_region": "us-west-2",
    "endpoint_url": "https://your-s3-endpoint",
    "cloud_bucket_name": "your-bucket"
  }
}

Please check your existing structure and retry. Thanks!