Deploy a "custom Bentoml image" using "yatai-deployment".

[tonylauCN]

"Yatai" is a great project, and my team started using it since version "1.0.0a7" and it's fine. However, recently, we updated the versions of Yatai components:

yatai 1.1.9
yatai-deployment 1.1.13
yatai-image-builder 1.1.8

After the update, we encountered issues with the deployment. Here are the specific details:

We use custom jobs to build Bentoml images (still using the Bentoml commands):

bentoml build ..
bentoml containerize ..
docker login ..
docker tag ..
docker push ..
bentoml yatai login ..
bentoml push ..

After these steps, we create a CRD file for deployment:

apiVersion: serving.yatai.ai/v2alpha1
kind: BentoDeployment
metadata:
  name: dep-bento-testirismtpl-gpu-6-1-76
  namespace: yatai
spec:
  bento: 1-tf-mnist-svc--v-1-2-3
  extraPodMetadata:
    labels:
      name: dep-bento-testirismtpl-gpu-6-1-76
      startTime: "20230724135548"
      expiredAt: "20230823135548"
  extraPodSpec:
    tolerations:
      - key: nvidia.com/gpu
        operator: Exists
        effect: NoSchedule
  ingress:
    enabled: false
...

However, we encountered the issue "Bento {name} not found" from yatai-deployment:
events Bento 1-tf-mnist-svc--v-1-2-3 not found {"type": "Normal", "object": {"kind":"BentoDeployment","namespace":"yatai","name":"dep-bento-testirismtpl-gpu-6-1-76","uid":"d2f09f5f-3167-42ba-b67f-cacd8d99b5d1","apiVersion":"serving.yatai.ai/v2alpha1","resourceVersion":"104943704"}, "reason": "GetBento"}

However, if we use "yatai-image-builder" for building and then deploy with the same CRD file, it works successfully.

Can we deploy using custom Docker image like we did in the previous version (1.0.0a7) without using yatai-image-builder?