https://bentoml.com logo
Join Slack
Powered by
# 한국어-korean
  • u
    혹시 Yatai 사용할때, object storage로 S3 안쓰면 안되나요?
  • b
  • b
    ^ s3 사용에 대한 문서.
  • u
    minio를 설치 했는데 k8s pod의 yatai-minio-ss-0 ~ 4의 상태가 Pending입니다 해결방법이 있을까요?
  • u
    Copy code
    I installed minio, but the status of yatai-minio-ss-0 to 4 in the k8s pod is Pending

Any solution?
  • s
    yatai는 아직까지는 s3 인터페이스만 지원하고있어서 s3를 스토리지로 사용해야합니다 @여형철 minio 헬름차트가 왜 배포가 안되었는지 정확히 알아야하겠네요 아래 커맨드 쳐서 확인해볼수있을까요?
    Copy code
    kubectl logs -n yatai-system pod/yatai-mino~~~~
  • s
    minio 배포하다 생기는 문제는 단순히 디스크 공간이 부족한게 아니라면 tenants 옵션에서 많이 막히시는데
  • s
    production 환경아니라면 삭제후 재설치해보시는거 추천드립니다. (한줄씩 실행)
    Copy code
    helm list -n yatai-system # yatai-minio # 차트명 확인

helm delete minio-operator -n yatai-system # yatai-minio # 삭제

helm upgrade --install minio-operator minio/minio-operator -n yatai-system --set tenants=null # 재배포

kubectl -n yatai-system wait --for=condition=ready --timeout=600s pod -l <http://app.kubernetes.io/name=minio-operator|app.kubernetes.io/name=minio-operator> # 배포 대기

# Tenant apply 
cat <<EOF | kubectl apply -f -
apiVersion: <http://minio.min.io/v2|minio.min.io/v2>
kind: Tenant
metadata:
  labels:
    app: yatai-minio
  name: yatai-minio
  namespace: yatai-system
spec:
  credsSecret:
    name: yatai-minio
  image: <http://quay.io/bentoml/minio-minio:RELEASE.2021-10-06T23-36-31Z|quay.io/bentoml/minio-minio:RELEASE.2021-10-06T23-36-31Z>
  imagePullPolicy: IfNotPresent
  mountPath: /export
  podManagementPolicy: Parallel
  pools:
  - servers: 4
    volumeClaimTemplate:
      metadata:
        name: data
      spec:
        accessModes:
        - ReadWriteOnce
        resources:
          requests:
            storage: 20Gi
    volumesPerServer: 4
  requestAutoCert: false
  s3:
    bucketDNS: false
  subPath: /data
EOF
  • s
    This message was deleted.
    u
    • 2
    • 1
  • u
    안녕하세요 bentoml containerize시 이미지로 정상적으로 만들어지는데
  • u
    docker run 진행시 발생하는 에러가 왜 나는지 알수가 없습니다 에러 로그 붙입니다.
  • u
    2023-01-04T061410+0000 [ERROR] [cli] Exception in callback <bound method Arbiter.manage_watchers of <circus.arbiter.Arbiter object at 0x7f29eb2de940>> Traceback (most recent call last): File "/usr/local/lib/python3.8/site-packages/tornado/ioloop.py", line 921, in _run val = self.callback() File "/usr/local/lib/python3.8/site-packages/circus/util.py", line 1038, in wrapper raise ConflictError("arbiter is already running %s command"
  • u
    bentoml serve시에는 에러가 안납니다
  • s
    Service.py 코드도 볼수있을까요
  • s
    This message was deleted.
    s
    u
    • 3
    • 9
  • j
    👋 안녕하세요!
  • j
    pytorch resnet50 모델로 이미지 feature extraction하여 두개 이미지의 유사도를 구해주는 API를 bentoml로 서빙하고 있습니다. 다만, 스트레스 테스트를 위하여 k6로 부하테스트를 하는데, 특정 부하 이상이 되면 runner가 죽어버리는지 Server disconnected로 응답합니다 그 후, runner가 되살아나지 않고 계속 500에러를 응답하는데 혹시 해결방법 아시는분 계실까요? requirements.txt
    Copy code
    numpy==1.23.*
requests==2.28.*
bentoml==1.0.10
torch==1.12.*
torchvision==0.13.*
pydantic==1.10.*
sentry-sdk==1.11.*
    service.py
    Copy code
    class ImageModelFeatures(BaseModel):
    source_url: str
    target_url: str


class ImageModelOutFeatures(BaseModel):
    distance: float


input_spec = JSON(pydantic_model=ImageModelFeatures)
output_spec = JSON(pydantic_model=ImageModelOutFeatures)


@svc.api(input=input_spec, output=output_spec)
async def predict(input_data: ImageModelFeatures) -> dict:
    """
    """
    source_url = modify_source_url(input_data.source_url)
    source_img = url_to_processed_img(source_url)
    source_embedding = await runner.async_run(source_img)
    source_embedding = torch.flatten(source_embedding['avgpool'])
    source_embedding = source_embedding.detach().numpy()

    target_url = modify_target_url(input_data.target_url)
    target_img = url_to_processed_img(target_url)
    target_embedding = await runner.async_run(target_img)
    target_embedding = torch.flatten(target_embedding['avgpool'])
    target_embedding = target_embedding.detach().numpy()

    distance = cos_sim(source_embedding, target_embedding)
    return distanc
  • j
    부하테스트 중간에 아래와 같은 에러들도 추가로 발생합니다 ㅠㅠ
  • s
    This message was deleted.
    j
    • 2
    • 1
  • s
    This message was deleted.
    s
    u
    • 3
    • 4
  • s
    This message was deleted.
    u
    s
    • 3
    • 16
  • s
    This message was deleted.
    s
    • 2
    • 2
  • s
    This message was deleted.
    s
    j
    • 3
    • 3
  • s
    This message was deleted.
    s
    • 2
    • 1
  • s
    This message was deleted.
    s
    • 2
    • 2
  • s
    This message was deleted.
    s
    u
    • 3
    • 2
  • c
    안녕하세요!
    🍱 11
  • s
    안녕하세요 https://github.com/KimSoungRyoul/PyConKR2023-ModelServing-BentoML 이번 pycon KR 2023에서 발표한 BentoML 관련 ppt 입니다. 도움이 되실분들이 있으실 것 같아서 공유합니다
    🙌 13
    👍 6
    🚀 8
  • s
    This message was deleted.
    s
    u
    • 3
    • 5
  • t
    <!here> Hello friends! We were thinking of hosting an event in Seoul in November. Would like to see what topics and formats that you might be interested in. Please vote using the following emojis, you may vote multiple things 🙂 🔥 The latest trends in ML/AI in the industry party parrot Meeting up to network and have a few drinks with ML/AI colleagues 🚀 Advanced autoscaling strategies and production deployment 🤯 The state of LLMs and various optimizations 🖼️ Generative art (like stable diffusion) and strategies for fine-tuning and deploying at scale
    🤯 4
    🖼️ 3
    party parrot 6
    🚀 3
    ❤️ 2
    🔥 10