typo in "Pretrained GPT2 Model Deployment Example" notebook

[saeid93]

Describe the bug

I think there is a typo in the following notebook Pretrained GPT2 Model Deployment Example

!mc mb minio-seldon/onnx-gpt2 -p
!mc cp ./model.onnx minio-seldon/onnx-gpt2/gpt2/1/

Should be

!mc mb minio-seldon/onnx-gpt2 -p
!mc cp ./model.onnx minio-seldon/onnx-gpt2

as in the:

%%writefile gpt2-deploy.yaml
apiVersion: machinelearning.seldon.io/v1alpha2
kind: SeldonDeployment
metadata:
  name: gpt2
spec:
  predictors:
  - graph:
      implementation: TRITON_SERVER
      logger:
        mode: all
      modelUri: s3://onnx-gpt2
      envSecretRefName: seldon-init-container-secret
      name: gpt2
      type: MODEL
    name: default
    replicas: 1
  protocol: kfserving

In the current format in the website init container was not able to find the model from Minio but putting the model in the root directory of the minio bucket solved the problem.

[axsaucedo]

Fixed via #4086

[RafalSkolasinski]

I wonder if the typo may be rather in mc mb ... such that it is missing further subpath? Triton is very strict on folder structure and the actual model binaries are expected to be subfolders (sub-buckets?) of s3://onnx-gpt2.

[saeid93]

I reran the notebook with the corrected path and exec into the classsifier container, I can confirm that the model exist in the container in the following path:

triton-server@gpt2-default-0-gpt2-5798b7cbb9-xwc67:/mnt/models$ ls
model.onnx
triton-server@gpt2-default-0-gpt2-5798b7cbb9-xwc67:/mnt/models$ pwd
/mnt/models

Is it how it should be?