From 37c74b232c47394a0a9c239be72392e8e8a8f0db Mon Sep 17 00:00:00 2001 From: ylg Date: Thu, 17 Oct 2024 16:58:47 +0800 Subject: [PATCH] Update ChatQnA yaml and set retriever's TEI_EMBEDDING_ENDPOINT (#953) Signed-off-by: longguang.yue --- ChatQnA/docker_compose/intel/cpu/xeon/compose_qdrant.yaml | 1 + ChatQnA/docker_compose/intel/hpu/gaudi/compose.yaml | 1 + ChatQnA/docker_compose/intel/hpu/gaudi/compose_guardrails.yaml | 1 + ChatQnA/docker_compose/intel/hpu/gaudi/compose_vllm.yaml | 1 + ChatQnA/docker_compose/intel/hpu/gaudi/compose_vllm_ray.yaml | 1 + .../docker_compose/intel/hpu/gaudi/compose_without_rerank.yaml | 1 + ChatQnA/docker_compose/nvidia/gpu/compose.yaml | 1 + 7 files changed, 7 insertions(+) diff --git a/ChatQnA/docker_compose/intel/cpu/xeon/compose_qdrant.yaml b/ChatQnA/docker_compose/intel/cpu/xeon/compose_qdrant.yaml index 039a9e002..b0a091aa1 100644 --- a/ChatQnA/docker_compose/intel/cpu/xeon/compose_qdrant.yaml +++ b/ChatQnA/docker_compose/intel/cpu/xeon/compose_qdrant.yaml @@ -53,6 +53,7 @@ services: QDRANT_HOST: ${QDRANT_HOST} QDRANT_PORT: 6333 INDEX_NAME: ${INDEX_NAME} + TEI_EMBEDDING_ENDPOINT: ${TEI_EMBEDDING_ENDPOINT} restart: unless-stopped tei-reranking-service: image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 diff --git a/ChatQnA/docker_compose/intel/hpu/gaudi/compose.yaml b/ChatQnA/docker_compose/intel/hpu/gaudi/compose.yaml index b593ecd78..fcc621fe5 100644 --- a/ChatQnA/docker_compose/intel/hpu/gaudi/compose.yaml +++ b/ChatQnA/docker_compose/intel/hpu/gaudi/compose.yaml @@ -59,6 +59,7 @@ services: https_proxy: ${https_proxy} REDIS_URL: ${REDIS_URL} INDEX_NAME: ${INDEX_NAME} + TEI_EMBEDDING_ENDPOINT: ${TEI_EMBEDDING_ENDPOINT} restart: unless-stopped tei-reranking-service: image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 diff --git a/ChatQnA/docker_compose/intel/hpu/gaudi/compose_guardrails.yaml b/ChatQnA/docker_compose/intel/hpu/gaudi/compose_guardrails.yaml index 1b13f6339..db49c41a8 100644 --- a/ChatQnA/docker_compose/intel/hpu/gaudi/compose_guardrails.yaml +++ b/ChatQnA/docker_compose/intel/hpu/gaudi/compose_guardrails.yaml @@ -96,6 +96,7 @@ services: https_proxy: ${https_proxy} REDIS_URL: ${REDIS_URL} INDEX_NAME: ${INDEX_NAME} + TEI_EMBEDDING_ENDPOINT: ${TEI_EMBEDDING_ENDPOINT} restart: unless-stopped tei-reranking-service: image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 diff --git a/ChatQnA/docker_compose/intel/hpu/gaudi/compose_vllm.yaml b/ChatQnA/docker_compose/intel/hpu/gaudi/compose_vllm.yaml index 03b684cc9..0ebf9541b 100644 --- a/ChatQnA/docker_compose/intel/hpu/gaudi/compose_vllm.yaml +++ b/ChatQnA/docker_compose/intel/hpu/gaudi/compose_vllm.yaml @@ -57,6 +57,7 @@ services: https_proxy: ${https_proxy} REDIS_URL: ${REDIS_URL} INDEX_NAME: ${INDEX_NAME} + TEI_EMBEDDING_ENDPOINT: ${TEI_EMBEDDING_ENDPOINT} restart: unless-stopped tei-reranking-service: image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 diff --git a/ChatQnA/docker_compose/intel/hpu/gaudi/compose_vllm_ray.yaml b/ChatQnA/docker_compose/intel/hpu/gaudi/compose_vllm_ray.yaml index 2ae027a4d..e80b05ec1 100644 --- a/ChatQnA/docker_compose/intel/hpu/gaudi/compose_vllm_ray.yaml +++ b/ChatQnA/docker_compose/intel/hpu/gaudi/compose_vllm_ray.yaml @@ -57,6 +57,7 @@ services: https_proxy: ${https_proxy} REDIS_URL: ${REDIS_URL} INDEX_NAME: ${INDEX_NAME} + TEI_EMBEDDING_ENDPOINT: ${TEI_EMBEDDING_ENDPOINT} restart: unless-stopped tei-reranking-service: image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 diff --git a/ChatQnA/docker_compose/intel/hpu/gaudi/compose_without_rerank.yaml b/ChatQnA/docker_compose/intel/hpu/gaudi/compose_without_rerank.yaml index 5dee1e1e6..9d20468df 100644 --- a/ChatQnA/docker_compose/intel/hpu/gaudi/compose_without_rerank.yaml +++ b/ChatQnA/docker_compose/intel/hpu/gaudi/compose_without_rerank.yaml @@ -59,6 +59,7 @@ services: https_proxy: ${https_proxy} REDIS_URL: ${REDIS_URL} INDEX_NAME: ${INDEX_NAME} + TEI_EMBEDDING_ENDPOINT: ${TEI_EMBEDDING_ENDPOINT} restart: unless-stopped tgi-service: image: ghcr.io/huggingface/tgi-gaudi:2.0.5 diff --git a/ChatQnA/docker_compose/nvidia/gpu/compose.yaml b/ChatQnA/docker_compose/nvidia/gpu/compose.yaml index 723f35082..c35866b10 100644 --- a/ChatQnA/docker_compose/nvidia/gpu/compose.yaml +++ b/ChatQnA/docker_compose/nvidia/gpu/compose.yaml @@ -60,6 +60,7 @@ services: https_proxy: ${https_proxy} REDIS_URL: ${REDIS_URL} INDEX_NAME: ${INDEX_NAME} + TEI_EMBEDDING_ENDPOINT: ${TEI_EMBEDDING_ENDPOINT} restart: unless-stopped tei-reranking-service: image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5