From 11982c5aac16276606705f40eb24680f9fb16b29 Mon Sep 17 00:00:00 2001 From: Eero Tamminen Date: Tue, 14 Jan 2025 14:02:23 +0200 Subject: [PATCH] llm-docsum-tgi/-vllm => llm-docsum Adapt to DocSum refactor: https://github.com/opea-project/GenAIComps/pull/1101 Signed-off-by: Eero Tamminen --- helm-charts/common/llm-uservice/ci-docsum-values.yaml | 2 +- .../common/llm-uservice/templates/tests/test-pod.yaml | 2 +- helm-charts/common/llm-uservice/variant_docsum-values.yaml | 2 +- helm-charts/docsum/gaudi-vllm-values.yaml | 5 ----- helm-charts/docsum/values.yaml | 2 +- .../config/manifests/docsum-llm-uservice.yaml | 2 +- 6 files changed, 5 insertions(+), 10 deletions(-) diff --git a/helm-charts/common/llm-uservice/ci-docsum-values.yaml b/helm-charts/common/llm-uservice/ci-docsum-values.yaml index da00321e2..87aa23ded 100644 --- a/helm-charts/common/llm-uservice/ci-docsum-values.yaml +++ b/helm-charts/common/llm-uservice/ci-docsum-values.yaml @@ -2,7 +2,7 @@ # SPDX-License-Identifier: Apache-2.0 image: - repository: opea/llm-docsum-tgi + repository: opea/llm-docsum tag: "latest" MAX_INPUT_TOKENS: 2048 diff --git a/helm-charts/common/llm-uservice/templates/tests/test-pod.yaml b/helm-charts/common/llm-uservice/templates/tests/test-pod.yaml index e5012fc75..c851d7816 100644 --- a/helm-charts/common/llm-uservice/templates/tests/test-pod.yaml +++ b/helm-charts/common/llm-uservice/templates/tests/test-pod.yaml @@ -19,7 +19,7 @@ spec: - | max_retry=20; for ((i=1; i<=max_retry; i++)); do - {{- if contains "llm-docsum-tgi" .Values.image.repository }} + {{- if contains "llm-docsum" .Values.image.repository }} # Try with docsum endpoint curl http://{{ include "llm-uservice.fullname" . }}:{{ .Values.service.port }}/v1/chat/docsum -sS --fail-with-body \ -X POST \ diff --git a/helm-charts/common/llm-uservice/variant_docsum-values.yaml b/helm-charts/common/llm-uservice/variant_docsum-values.yaml index f0819540a..644a63432 100644 --- a/helm-charts/common/llm-uservice/variant_docsum-values.yaml +++ b/helm-charts/common/llm-uservice/variant_docsum-values.yaml @@ -2,7 +2,7 @@ # SPDX-License-Identifier: Apache-2.0 image: - repository: opea/llm-docsum-tgi + repository: opea/llm-docsum tag: "latest" MAX_INPUT_TOKENS: 2048 diff --git a/helm-charts/docsum/gaudi-vllm-values.yaml b/helm-charts/docsum/gaudi-vllm-values.yaml index 6f376c93d..dd14980e8 100644 --- a/helm-charts/docsum/gaudi-vllm-values.yaml +++ b/helm-charts/docsum/gaudi-vllm-values.yaml @@ -7,11 +7,6 @@ tgi: enabled: false -llm-uservice: - image: - repository: opea/llm-docsum-vllm - tag: "latest" - vllm: enabled: true image: diff --git a/helm-charts/docsum/values.yaml b/helm-charts/docsum/values.yaml index 40a028e0e..00ad0035f 100644 --- a/helm-charts/docsum/values.yaml +++ b/helm-charts/docsum/values.yaml @@ -59,7 +59,7 @@ affinity: {} # To override values in subchart llm-uservice llm-uservice: image: - repository: opea/llm-docsum-tgi + repository: opea/llm-docsum MAX_INPUT_TOKENS: "1024" MAX_TOTAL_TOKENS: "2048" LLM_MODEL_ID: Intel/neural-chat-7b-v3-3 diff --git a/microservices-connector/config/manifests/docsum-llm-uservice.yaml b/microservices-connector/config/manifests/docsum-llm-uservice.yaml index ad2e53859..57269eb3f 100644 --- a/microservices-connector/config/manifests/docsum-llm-uservice.yaml +++ b/microservices-connector/config/manifests/docsum-llm-uservice.yaml @@ -94,7 +94,7 @@ spec: runAsUser: 1000 seccompProfile: type: RuntimeDefault - image: "opea/llm-docsum-tgi:latest" + image: "opea/llm-docsum:latest" ports: - name: llm-uservice containerPort: 9000