From 17355b6719d8610723e0dced48d01f68dbafaa79 Mon Sep 17 00:00:00 2001 From: "chen, suyue" Date: Fri, 25 Apr 2025 23:00:36 +0800 Subject: [PATCH] downgrade tei version from 1.6 to 1.5, fix the chatqna perf regression (#1886) Signed-off-by: chensuyue (cherry picked from commit c546d96e9875bcae388c69b37a28395f8e313dd7) --- .../kubernetes/intel/gaudi/README.md | 192 ------------ .../kubernetes/intel/gaudi/benchmark.sh | 102 ------- .../kubernetes/intel/gaudi/benchmark.yaml | 68 ----- .../kubernetes/intel/gaudi/deploy.py | 278 ------------------ .../intel/gaudi/generate_helm_values.py | 164 ----------- .../intel/cpu/aipc/compose.yaml | 4 +- .../intel/cpu/xeon/compose.yaml | 4 +- .../intel/cpu/xeon/compose_faqgen.yaml | 4 +- .../intel/cpu/xeon/compose_faqgen_tgi.yaml | 4 +- .../intel/cpu/xeon/compose_milvus.yaml | 4 +- .../intel/cpu/xeon/compose_pinecone.yaml | 4 +- .../intel/cpu/xeon/compose_remote.yaml | 4 +- .../intel/cpu/xeon/compose_tgi.yaml | 4 +- .../cpu/xeon/compose_without_rerank.yaml | 2 +- .../docker_compose/intel/hpu/gaudi/README.md | 16 +- .../intel/hpu/gaudi/compose.yaml | 2 +- .../intel/hpu/gaudi/compose_faqgen.yaml | 2 +- .../intel/hpu/gaudi/compose_faqgen_tgi.yaml | 2 +- .../intel/hpu/gaudi/compose_guardrails.yaml | 2 +- .../intel/hpu/gaudi/compose_tgi.yaml | 2 +- .../hpu/gaudi/compose_without_rerank.yaml | 2 +- .../hpu/gaudi/how_to_validate_service.md | 2 +- ChatQnA/tests/test_compose_faqgen_on_gaudi.sh | 2 - ChatQnA/tests/test_compose_faqgen_on_rocm.sh | 3 - ChatQnA/tests/test_compose_faqgen_on_xeon.sh | 1 - .../tests/test_compose_faqgen_tgi_on_gaudi.sh | 3 - .../tests/test_compose_faqgen_tgi_on_xeon.sh | 2 - .../tests/test_compose_guardrails_on_gaudi.sh | 3 - ChatQnA/tests/test_compose_milvus_on_xeon.sh | 2 - ChatQnA/tests/test_compose_on_gaudi.sh | 2 - ChatQnA/tests/test_compose_on_rocm.sh | 3 - ChatQnA/tests/test_compose_on_xeon.sh | 2 - .../tests/test_compose_pinecone_on_xeon.sh | 2 - ChatQnA/tests/test_compose_tgi_on_gaudi.sh | 4 - ChatQnA/tests/test_compose_tgi_on_xeon.sh | 3 - .../test_compose_without_rerank_on_gaudi.sh | 3 - .../test_compose_without_rerank_on_xeon.sh | 2 - 37 files changed, 32 insertions(+), 873 deletions(-) delete mode 100644 ChatQnA/benchmark/performance/kubernetes/intel/gaudi/README.md delete mode 100755 ChatQnA/benchmark/performance/kubernetes/intel/gaudi/benchmark.sh delete mode 100644 ChatQnA/benchmark/performance/kubernetes/intel/gaudi/benchmark.yaml delete mode 100644 ChatQnA/benchmark/performance/kubernetes/intel/gaudi/deploy.py delete mode 100644 ChatQnA/benchmark/performance/kubernetes/intel/gaudi/generate_helm_values.py diff --git a/ChatQnA/benchmark/performance/kubernetes/intel/gaudi/README.md b/ChatQnA/benchmark/performance/kubernetes/intel/gaudi/README.md deleted file mode 100644 index a59dce5a7..000000000 --- a/ChatQnA/benchmark/performance/kubernetes/intel/gaudi/README.md +++ /dev/null @@ -1,192 +0,0 @@ -# ChatQnA Benchmarking - -This folder contains a collection of Kubernetes manifest files for deploying the ChatQnA service across scalable nodes. It includes a comprehensive [benchmarking tool](https://github.com/opea-project/GenAIEval/blob/main/evals/benchmark/README.md) that enables throughput analysis to assess inference performance. - -By following this guide, you can run benchmarks on your deployment and share the results with the OPEA community. - -## Purpose - -We aim to run these benchmarks and share them with the OPEA community for three primary reasons: - -- To offer insights on inference throughput in real-world scenarios, helping you choose the best service or deployment for your needs. -- To establish a baseline for validating optimization solutions across different implementations, providing clear guidance on which methods are most effective for your use case. -- To inspire the community to build upon our benchmarks, allowing us to better quantify new solutions in conjunction with current leading llms, serving frameworks etc. - -## Metrics - -The benchmark will report the below metrics, including: - -- Number of Concurrent Requests -- End-to-End Latency: P50, P90, P99 (in milliseconds) -- End-to-End First Token Latency: P50, P90, P99 (in milliseconds) -- Average Next Token Latency (in milliseconds) -- Average Token Latency (in milliseconds) -- Requests Per Second (RPS) -- Output Tokens Per Second -- Input Tokens Per Second - -Results will be displayed in the terminal and saved as CSV file named `1_stats.csv` for easy export to spreadsheets. - -## Table of Contents - -- [Deployment](#deployment) - - [Prerequisites](#prerequisites) - - [Deployment Scenarios](#deployment-scenarios) - - [Case 1: Baseline Deployment with Rerank](#case-1-baseline-deployment-with-rerank) - - [Case 2: Baseline Deployment without Rerank](#case-2-baseline-deployment-without-rerank) - - [Case 3: Tuned Deployment with Rerank](#case-3-tuned-deployment-with-rerank) -- [Benchmark](#benchmark) - - [Test Configurations](#test-configurations) - - [Test Steps](#test-steps) - - [Upload Retrieval File](#upload-retrieval-file) - - [Run Benchmark Test](#run-benchmark-test) - - [Data collection](#data-collection) -- [Teardown](#teardown) - -## Deployment - -### Prerequisites - -- Kubernetes installation: Use [kubespray](https://github.com/opea-project/docs/blob/main/guide/installation/k8s_install/k8s_install_kubespray.md) or other official Kubernetes installation guides: -- (Optional) [Kubernetes set up guide on Intel Gaudi product](https://github.com/opea-project/GenAIInfra/blob/main/README.md#setup-kubernetes-cluster) -- Helm installation: Follow the [Helm documentation](https://helm.sh/docs/intro/install/#helm) to install Helm. -- Setup Hugging Face Token - - To access models and APIs from Hugging Face, set your token as environment variable. - ```bash - export HF_TOKEN="insert-your-huggingface-token-here" - ``` -- Prepare Shared Models (Optional but Strongly Recommended) - - Downloading models simultaneously to multiple nodes in your cluster can overload resources such as network bandwidth, memory and storage. To prevent resource exhaustion, it's recommended to preload the models in advance. - ```bash - pip install -U "huggingface_hub[cli]" - sudo mkdir -p /mnt/models - sudo chmod 777 /mnt/models - huggingface-cli download --cache-dir /mnt/models Intel/neural-chat-7b-v3-3 - export MODEL_DIR=/mnt/models - ``` - Once the models are downloaded, you can consider the following methods for sharing them across nodes: - - Persistent Volume Claim (PVC): This is the recommended approach for production setups. For more details on using PVC, refer to [PVC](https://github.com/opea-project/GenAIInfra/blob/main/helm-charts/README.md#using-persistent-volume). - - Local Host Path: For simpler testing, ensure that each node involved in the deployment follows the steps above to locally prepare the models. After preparing the models, use `--set global.modelUseHostPath=${MODELDIR}` in the deployment command. - -- Label Nodes - ```base - python deploy.py --add-label --num-nodes 2 - ``` - -### Deployment Scenarios - -The example below are based on a two-node setup. You can adjust the number of nodes by using the `--num-nodes` option. - -By default, these commands use the `default` namespace. To specify a different namespace, use the `--namespace` flag with deploy, uninstall, and kubernetes command. Additionally, update the `namespace` field in `benchmark.yaml` before running the benchmark test. - -For additional configuration options, run `python deploy.py --help` - -#### Case 1: Baseline Deployment with Rerank - -Deploy Command (with node number, Hugging Face token, model directory specified): -```bash -python deploy.py --hf-token $HF_TOKEN --model-dir $MODEL_DIR --num-nodes 2 --with-rerank -``` -Uninstall Command: -```bash -python deploy.py --uninstall -``` - -#### Case 2: Baseline Deployment without Rerank - -```bash -python deploy.py --hf-token $HFTOKEN --model-dir $MODELDIR --num-nodes 2 -``` -#### Case 3: Tuned Deployment with Rerank - -```bash -python deploy.py --hf-token $HFTOKEN --model-dir $MODELDIR --num-nodes 2 --with-rerank --tuned -``` - -## Benchmark - -### Test Configurations - -| Key | Value | -| -------- | ------- | -| Workload | ChatQnA | -| Tag | V1.1 | - -Models configuration -| Key | Value | -| ---------- | ------------------ | -| Embedding | BAAI/bge-base-en-v1.5 | -| Reranking | BAAI/bge-reranker-base | -| Inference | Intel/neural-chat-7b-v3-3 | - -Benchmark parameters -| Key | Value | -| ---------- | ------------------ | -| LLM input tokens | 1024 | -| LLM output tokens | 128 | - -Number of test requests for different scheduled node number: -| Node count | Concurrency | Query number | -| ----- | -------- | -------- | -| 1 | 128 | 640 | -| 2 | 256 | 1280 | -| 4 | 512 | 2560 | - -More detailed configuration can be found in configuration file [benchmark.yaml](./benchmark.yaml). - -### Test Steps - -Use `kubectl get pods` to confirm that all pods are `READY` before starting the test. - -#### Upload Retrieval File - -Before testing, upload a specified file to make sure the llm input have the token length of 1k. - -Get files: - -```bash -wget https://github.com/opea-project/GenAIEval/tree/main/evals/benchmark/data/upload_file.txt -``` - -Retrieve the `ClusterIP` of the `chatqna-data-prep` service. - -```bash -kubectl get svc -``` -Expected output: -```log -chatqna-data-prep ClusterIP xx.xx.xx.xx 6007/TCP 51m -``` - -Use the following `cURL` command to upload file: - -```bash -cd GenAIEval/evals/benchmark/data -curl -X POST "http://${cluster_ip}:6007/v1/dataprep/ingest" \ - -H "Content-Type: multipart/form-data" \ - -F "chunk_size=3800" \ - -F "files=@./upload_file.txt" -``` - -#### Run Benchmark Test - -Run the benchmark test using: -```bash -bash benchmark.sh -n 2 -``` -The `-n` argument specifies the number of test nodes. Required dependencies will be automatically installed when running the benchmark for the first time. - -#### Data collection - -All the test results will come to the folder `GenAIEval/evals/benchmark/benchmark_output`. - -## Teardown - -After completing the benchmark, use the following command to clean up the environment: - -Remove Node Labels: -```bash -python deploy.py --delete-label -``` diff --git a/ChatQnA/benchmark/performance/kubernetes/intel/gaudi/benchmark.sh b/ChatQnA/benchmark/performance/kubernetes/intel/gaudi/benchmark.sh deleted file mode 100755 index 9c3a9f84e..000000000 --- a/ChatQnA/benchmark/performance/kubernetes/intel/gaudi/benchmark.sh +++ /dev/null @@ -1,102 +0,0 @@ -#!/bin/bash - -# Copyright (C) 2024 Intel Corporation -# SPDX-License-Identifier: Apache-2.0 - -deployment_type="k8s" -node_number=1 -service_port=8888 -query_per_node=640 - -benchmark_tool_path="$(pwd)/GenAIEval" - -usage() { - echo "Usage: $0 [-d deployment_type] [-n node_number] [-i service_ip] [-p service_port]" - echo " -d deployment_type ChatQnA deployment type, select between k8s and docker (default: k8s)" - echo " -n node_number Test node number, required only for k8s deployment_type, (default: 1)" - echo " -i service_ip chatqna service ip, required only for docker deployment_type" - echo " -p service_port chatqna service port, required only for docker deployment_type, (default: 8888)" - exit 1 -} - -while getopts ":d:n:i:p:" opt; do - case ${opt} in - d ) - deployment_type=$OPTARG - ;; - n ) - node_number=$OPTARG - ;; - i ) - service_ip=$OPTARG - ;; - p ) - service_port=$OPTARG - ;; - \? ) - echo "Invalid option: -$OPTARG" 1>&2 - usage - ;; - : ) - echo "Invalid option: -$OPTARG requires an argument" 1>&2 - usage - ;; - esac -done - -if [[ "$deployment_type" == "docker" && -z "$service_ip" ]]; then - echo "Error: service_ip is required for docker deployment_type" 1>&2 - usage -fi - -if [[ "$deployment_type" == "k8s" && ( -n "$service_ip" || -n "$service_port" ) ]]; then - echo "Warning: service_ip and service_port are ignored for k8s deployment_type" 1>&2 -fi - -function main() { - if [[ ! -d ${benchmark_tool_path} ]]; then - echo "Benchmark tool not found, setting up..." - setup_env - fi - run_benchmark -} - -function setup_env() { - git clone https://github.com/opea-project/GenAIEval.git - pushd ${benchmark_tool_path} - python3 -m venv stress_venv - source stress_venv/bin/activate - pip install -r requirements.txt - popd -} - -function run_benchmark() { - source ${benchmark_tool_path}/stress_venv/bin/activate - export DEPLOYMENT_TYPE=${deployment_type} - export SERVICE_IP=${service_ip:-"None"} - export SERVICE_PORT=${service_port:-"None"} - export LOAD_SHAPE=${load_shape:-"constant"} - export CONCURRENT_LEVEL=${concurrent_level:-5} - export ARRIVAL_RATE=${arrival_rate:-1.0} - if [[ -z $USER_QUERIES ]]; then - user_query=$((query_per_node*node_number)) - export USER_QUERIES="[${user_query}, ${user_query}, ${user_query}, ${user_query}]" - echo "USER_QUERIES not configured, setting to: ${USER_QUERIES}." - fi - export WARMUP=$(echo $USER_QUERIES | sed -e 's/[][]//g' -e 's/,.*//') - if [[ -z $WARMUP ]]; then export WARMUP=0; fi - if [[ -z $TEST_OUTPUT_DIR ]]; then - if [[ $DEPLOYMENT_TYPE == "k8s" ]]; then - export TEST_OUTPUT_DIR="${benchmark_tool_path}/evals/benchmark/benchmark_output/node_${node_number}" - else - export TEST_OUTPUT_DIR="${benchmark_tool_path}/evals/benchmark/benchmark_output/docker" - fi - echo "TEST_OUTPUT_DIR not configured, setting to: ${TEST_OUTPUT_DIR}." - fi - - envsubst < ./benchmark.yaml > ${benchmark_tool_path}/evals/benchmark/benchmark.yaml - cd ${benchmark_tool_path}/evals/benchmark - python benchmark.py -} - -main diff --git a/ChatQnA/benchmark/performance/kubernetes/intel/gaudi/benchmark.yaml b/ChatQnA/benchmark/performance/kubernetes/intel/gaudi/benchmark.yaml deleted file mode 100644 index 07945c220..000000000 --- a/ChatQnA/benchmark/performance/kubernetes/intel/gaudi/benchmark.yaml +++ /dev/null @@ -1,68 +0,0 @@ -# Copyright (C) 2024 Intel Corporation -# SPDX-License-Identifier: Apache-2.0 - -test_suite_config: # Overall configuration settings for the test suite - examples: ["chatqna"] # The specific test cases being tested, e.g., chatqna, codegen, codetrans, faqgen, audioqna, visualqna - deployment_type: ${DEPLOYMENT_TYPE} # Default is "k8s", can also be "docker" - service_ip: ${SERVICE_IP} # Leave as None for k8s, specify for Docker - service_port: ${SERVICE_PORT} # Leave as None for k8s, specify for Docker - warm_ups: ${WARMUP} # Number of test requests for warm-up - run_time: 60m # The max total run time for the test suite - seed: # The seed for all RNGs - user_queries: ${USER_QUERIES} # Number of test requests at each concurrency level - query_timeout: 120 # Number of seconds to wait for a simulated user to complete any executing task before exiting. 120 sec by defeult. - random_prompt: false # Use random prompts if true, fixed prompts if false - collect_service_metric: false # Collect service metrics if true, do not collect service metrics if false - data_visualization: false # Generate data visualization if true, do not generate data visualization if false - llm_model: "Intel/neural-chat-7b-v3-3" # The LLM model used for the test - test_output_dir: "${TEST_OUTPUT_DIR}" # The directory to store the test output - load_shape: # Tenant concurrency pattern - name: ${LOAD_SHAPE} # poisson or constant(locust default load shape) - params: # Loadshape-specific parameters - constant: # Constant load shape specific parameters, activate only if load_shape.name is constant - concurrent_level: ${CONCURRENT_LEVEL} # If user_queries is specified, concurrent_level is target number of requests per user. If not, it is the number of simulated users - poisson: # Poisson load shape specific parameters, activate only if load_shape.name is poisson - arrival_rate: ${ARRIVAL_RATE} # Request arrival rate - -test_cases: - chatqna: - embedding: - run_test: false - service_name: "chatqna-embedding-usvc" # Replace with your service name - embedserve: - run_test: false - service_name: "chatqna-tei" # Replace with your service name - retriever: - run_test: false - service_name: "chatqna-retriever-usvc" # Replace with your service name - parameters: - search_type: "similarity" - k: 1 - fetch_k: 20 - lambda_mult: 0.5 - score_threshold: 0.2 - reranking: - run_test: false - service_name: "chatqna-reranking-usvc" # Replace with your service name - parameters: - top_n: 1 - rerankserve: - run_test: false - service_name: "chatqna-teirerank" # Replace with your service name - llm: - run_test: false - service_name: "chatqna-llm-uservice" # Replace with your service name - parameters: - max_tokens: 128 - temperature: 0.01 - top_k: 10 - top_p: 0.95 - repetition_penalty: 1.03 - stream: true - llmserve: - run_test: false - service_name: "chatqna-tgi" # Replace with your service name - e2e: - run_test: true - service_name: "chatqna" # Replace with your service name - k: 1 diff --git a/ChatQnA/benchmark/performance/kubernetes/intel/gaudi/deploy.py b/ChatQnA/benchmark/performance/kubernetes/intel/gaudi/deploy.py deleted file mode 100644 index 5a81918d7..000000000 --- a/ChatQnA/benchmark/performance/kubernetes/intel/gaudi/deploy.py +++ /dev/null @@ -1,278 +0,0 @@ -# Copyright (C) 2024 Intel Corporation -# SPDX-License-Identifier: Apache-2.0 - -import argparse -import glob -import json -import os -import shutil -import subprocess -import sys - -from generate_helm_values import generate_helm_values - - -def run_kubectl_command(command): - """Run a kubectl command and return the output.""" - try: - result = subprocess.run(command, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True) - return result.stdout - except subprocess.CalledProcessError as e: - print(f"Error running command: {command}\n{e.stderr}") - exit(1) - - -def get_all_nodes(): - """Get the list of all nodes in the Kubernetes cluster.""" - command = ["kubectl", "get", "nodes", "-o", "json"] - output = run_kubectl_command(command) - nodes = json.loads(output) - return [node["metadata"]["name"] for node in nodes["items"]] - - -def add_label_to_node(node_name, label): - """Add a label to the specified node.""" - command = ["kubectl", "label", "node", node_name, label, "--overwrite"] - print(f"Labeling node {node_name} with {label}...") - run_kubectl_command(command) - print(f"Label {label} added to node {node_name} successfully.") - - -def add_labels_to_nodes(node_count=None, label=None, node_names=None): - """Add a label to the specified number of nodes or to specified nodes.""" - - if node_names: - # Add label to the specified nodes - for node_name in node_names: - add_label_to_node(node_name, label) - else: - # Fetch the node list and label the specified number of nodes - all_nodes = get_all_nodes() - if node_count is None or node_count > len(all_nodes): - print(f"Error: Node count exceeds the number of available nodes ({len(all_nodes)} available).") - sys.exit(1) - - selected_nodes = all_nodes[:node_count] - for node_name in selected_nodes: - add_label_to_node(node_name, label) - - -def clear_labels_from_nodes(label, node_names=None): - """Clear the specified label from specific nodes if provided, otherwise from all nodes.""" - label_key = label.split("=")[0] # Extract key from 'key=value' format - - # If specific nodes are provided, use them; otherwise, get all nodes - nodes_to_clear = node_names if node_names else get_all_nodes() - - for node_name in nodes_to_clear: - # Check if the node has the label by inspecting its metadata - command = ["kubectl", "get", "node", node_name, "-o", "json"] - node_info = run_kubectl_command(command) - node_metadata = json.loads(node_info) - - # Check if the label exists on this node - labels = node_metadata["metadata"].get("labels", {}) - if label_key in labels: - # Remove the label from the node - command = ["kubectl", "label", "node", node_name, f"{label_key}-"] - print(f"Removing label {label_key} from node {node_name}...") - run_kubectl_command(command) - print(f"Label {label_key} removed from node {node_name} successfully.") - else: - print(f"Label {label_key} not found on node {node_name}, skipping.") - - -def install_helm_release(release_name, chart_name, namespace, values_file, device_type): - """Deploy a Helm release with a specified name and chart. - - Parameters: - - release_name: The name of the Helm release. - - chart_name: The Helm chart name or path, e.g., "opea/chatqna". - - namespace: The Kubernetes namespace for deployment. - - values_file: The user values file for deployment. - - device_type: The device type (e.g., "gaudi") for specific configurations (optional). - """ - - # Check if the namespace exists; if not, create it - try: - # Check if the namespace exists - command = ["kubectl", "get", "namespace", namespace] - subprocess.run(command, check=True, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL) - except subprocess.CalledProcessError: - # Namespace does not exist, create it - print(f"Namespace '{namespace}' does not exist. Creating it...") - command = ["kubectl", "create", "namespace", namespace] - subprocess.run(command, check=True) - print(f"Namespace '{namespace}' created successfully.") - - # Handle gaudi-specific values file if device_type is "gaudi" - hw_values_file = None - untar_dir = None - if device_type == "gaudi": - print("Device type is gaudi. Pulling Helm chart to get gaudi-values.yaml...") - - # Combine chart_name with fixed prefix - chart_pull_url = f"oci://ghcr.io/opea-project/charts/{chart_name}" - - # Pull and untar the chart - subprocess.run(["helm", "pull", chart_pull_url, "--untar"], check=True) - - # Find the untarred directory - untar_dirs = glob.glob(f"{chart_name}*") - if untar_dirs: - untar_dir = untar_dirs[0] - hw_values_file = os.path.join(untar_dir, "gaudi-values.yaml") - print("gaudi-values.yaml pulled and ready for use.") - else: - print(f"Error: Could not find untarred directory for {chart_name}") - return - - # Prepare the Helm install command - command = ["helm", "install", release_name, chart_name, "--namespace", namespace] - - # Append additional values file for gaudi if it exists - if hw_values_file: - command.extend(["-f", hw_values_file]) - - # Append the main values file - command.extend(["-f", values_file]) - - # Execute the Helm install command - try: - print(f"Running command: {' '.join(command)}") # Print full command for debugging - subprocess.run(command, check=True) - print("Deployment initiated successfully.") - except subprocess.CalledProcessError as e: - print(f"Error occurred while deploying Helm release: {e}") - - # Cleanup: Remove the untarred directory - if untar_dir and os.path.isdir(untar_dir): - print(f"Removing temporary directory: {untar_dir}") - shutil.rmtree(untar_dir) - print("Temporary directory removed successfully.") - - -def uninstall_helm_release(release_name, namespace=None): - """Uninstall a Helm release and clean up resources, optionally delete the namespace if not 'default'.""" - # Default to 'default' namespace if none is specified - if not namespace: - namespace = "default" - - try: - # Uninstall the Helm release - command = ["helm", "uninstall", release_name, "--namespace", namespace] - print(f"Uninstalling Helm release {release_name} in namespace {namespace}...") - run_kubectl_command(command) - print(f"Helm release {release_name} uninstalled successfully.") - - # If the namespace is specified and not 'default', delete it - if namespace != "default": - print(f"Deleting namespace {namespace}...") - delete_namespace_command = ["kubectl", "delete", "namespace", namespace] - run_kubectl_command(delete_namespace_command) - print(f"Namespace {namespace} deleted successfully.") - else: - print("Namespace is 'default', skipping deletion.") - - except subprocess.CalledProcessError as e: - print(f"Error occurred while uninstalling Helm release or deleting namespace: {e}") - - -def main(): - parser = argparse.ArgumentParser(description="Manage Helm Deployment.") - parser.add_argument( - "--release-name", - type=str, - default="chatqna", - help="The Helm release name created during deployment (default: chatqna).", - ) - parser.add_argument( - "--chart-name", - type=str, - default="chatqna", - help="The chart name to deploy, composed of repo name and chart name (default: chatqna).", - ) - parser.add_argument("--namespace", default="default", help="Kubernetes namespace (default: default).") - parser.add_argument("--hf-token", help="Hugging Face API token.") - parser.add_argument( - "--model-dir", help="Model directory, mounted as volumes for service access to pre-downloaded models" - ) - parser.add_argument("--user-values", help="Path to a user-specified values.yaml file.") - parser.add_argument( - "--create-values-only", action="store_true", help="Only create the values.yaml file without deploying." - ) - parser.add_argument("--uninstall", action="store_true", help="Uninstall the Helm release.") - parser.add_argument("--num-nodes", type=int, default=1, help="Number of nodes to use (default: 1).") - parser.add_argument("--node-names", nargs="*", help="Optional specific node names to label.") - parser.add_argument("--add-label", action="store_true", help="Add label to specified nodes if this flag is set.") - parser.add_argument( - "--delete-label", action="store_true", help="Delete label from specified nodes if this flag is set." - ) - parser.add_argument( - "--label", default="node-type=opea-benchmark", help="Label to add/delete (default: node-type=opea-benchmark)." - ) - parser.add_argument("--with-rerank", action="store_true", help="Include rerank service in the deployment.") - parser.add_argument( - "--tuned", - action="store_true", - help="Modify resources for services and change extraCmdArgs when creating values.yaml.", - ) - parser.add_argument( - "--device-type", - type=str, - choices=["cpu", "gaudi"], - default="gaudi", - help="Specify the device type for deployment (choices: 'cpu', 'gaudi'; default: gaudi).", - ) - - args = parser.parse_args() - - # Adjust num-nodes based on node-names if specified - if args.node_names: - num_node_names = len(args.node_names) - if args.num_nodes != 1 and args.num_nodes != num_node_names: - parser.error("--num-nodes must match the number of --node-names if both are specified.") - else: - args.num_nodes = num_node_names - - # Node labeling management - if args.add_label: - add_labels_to_nodes(args.num_nodes, args.label, args.node_names) - return - elif args.delete_label: - clear_labels_from_nodes(args.label, args.node_names) - return - - # Uninstall Helm release if specified - if args.uninstall: - uninstall_helm_release(args.release_name, args.namespace) - return - - # Prepare values.yaml if not uninstalling - if args.user_values: - values_file_path = args.user_values - else: - if not args.hf_token: - parser.error("--hf-token are required") - node_selector = {args.label.split("=")[0]: args.label.split("=")[1]} - values_file_path = generate_helm_values( - with_rerank=args.with_rerank, - num_nodes=args.num_nodes, - hf_token=args.hf_token, - model_dir=args.model_dir, - node_selector=node_selector, - tune=args.tuned, - ) - - # Read back the generated YAML file for verification - with open(values_file_path, "r") as file: - print("Generated YAML contents:") - print(file.read()) - - # Deploy unless --create-values-only is specified - if not args.create_values_only: - install_helm_release(args.release_name, args.chart_name, args.namespace, values_file_path, args.device_type) - - -if __name__ == "__main__": - main() diff --git a/ChatQnA/benchmark/performance/kubernetes/intel/gaudi/generate_helm_values.py b/ChatQnA/benchmark/performance/kubernetes/intel/gaudi/generate_helm_values.py deleted file mode 100644 index ff15252e3..000000000 --- a/ChatQnA/benchmark/performance/kubernetes/intel/gaudi/generate_helm_values.py +++ /dev/null @@ -1,164 +0,0 @@ -# Copyright (C) 2024 Intel Corporation -# SPDX-License-Identifier: Apache-2.0 - -import os - -import yaml - - -def generate_helm_values(with_rerank, num_nodes, hf_token, model_dir, node_selector=None, tune=False): - """Create a values.yaml file based on the provided configuration.""" - - # Log the received parameters - print("Received parameters:") - print(f"with_rerank: {with_rerank}") - print(f"num_nodes: {num_nodes}") - print(f"node_selector: {node_selector}") # Log the node_selector - print(f"tune: {tune}") - - if node_selector is None: - node_selector = {} - - # Construct the base values dictionary - values = { - "tei": {"nodeSelector": {key: value for key, value in node_selector.items()}}, - "tgi": {"nodeSelector": {key: value for key, value in node_selector.items()}}, - "data-prep": {"nodeSelector": {key: value for key, value in node_selector.items()}}, - "redis-vector-db": {"nodeSelector": {key: value for key, value in node_selector.items()}}, - "retriever-usvc": {"nodeSelector": {key: value for key, value in node_selector.items()}}, - "chatqna-ui": {"nodeSelector": {key: value for key, value in node_selector.items()}}, - "global": { - "HUGGINGFACEHUB_API_TOKEN": hf_token, # Use passed token - "modelUseHostPath": model_dir, # Use passed model directory - }, - "nodeSelector": {key: value for key, value in node_selector.items()}, - } - - if with_rerank: - values["teirerank"] = {"nodeSelector": {key: value for key, value in node_selector.items()}} - else: - values["image"] = {"repository": "opea/chatqna-without-rerank"} - values["teirerank"] = {"enabled": False} - - default_replicas = [ - {"name": "chatqna", "replicaCount": 2}, - {"name": "tei", "replicaCount": 1}, - {"name": "teirerank", "replicaCount": 1} if with_rerank else None, - {"name": "tgi", "replicaCount": 7 if with_rerank else 8}, - {"name": "data-prep", "replicaCount": 1}, - {"name": "redis-vector-db", "replicaCount": 1}, - {"name": "retriever-usvc", "replicaCount": 2}, - ] - - if num_nodes > 1: - # Scale replicas based on number of nodes - replicas = [ - {"name": "chatqna", "replicaCount": 1 * num_nodes}, - {"name": "tei", "replicaCount": 1 * num_nodes}, - {"name": "teirerank", "replicaCount": 1} if with_rerank else None, - {"name": "tgi", "replicaCount": (8 * num_nodes - 1) if with_rerank else 8 * num_nodes}, - {"name": "data-prep", "replicaCount": 1}, - {"name": "redis-vector-db", "replicaCount": 1}, - {"name": "retriever-usvc", "replicaCount": 1 * num_nodes}, - ] - else: - replicas = default_replicas - - # Remove None values for rerank disabled - replicas = [r for r in replicas if r] - - # Update values.yaml with replicas - for replica in replicas: - service_name = replica["name"] - if service_name == "chatqna": - values["replicaCount"] = replica["replicaCount"] - print(replica["replicaCount"]) - elif service_name in values: - values[service_name]["replicaCount"] = replica["replicaCount"] - - # Prepare resource configurations based on tuning - resources = [] - if tune: - resources = [ - { - "name": "chatqna", - "resources": { - "limits": {"cpu": "16", "memory": "8000Mi"}, - "requests": {"cpu": "16", "memory": "8000Mi"}, - }, - }, - { - "name": "tei", - "resources": { - "limits": {"cpu": "80", "memory": "20000Mi"}, - "requests": {"cpu": "80", "memory": "20000Mi"}, - }, - }, - {"name": "teirerank", "resources": {"limits": {"habana.ai/gaudi": 1}}} if with_rerank else None, - {"name": "tgi", "resources": {"limits": {"habana.ai/gaudi": 1}}}, - {"name": "retriever-usvc", "resources": {"requests": {"cpu": "8", "memory": "8000Mi"}}}, - ] - - # Filter out any None values directly as part of initialization - resources = [r for r in resources if r is not None] - - # Add resources for each service if tuning - for resource in resources: - service_name = resource["name"] - if service_name == "chatqna": - values["resources"] = resource["resources"] - elif service_name in values: - values[service_name]["resources"] = resource["resources"] - - # Add extraCmdArgs for tgi service with default values - if "tgi" in values: - values["tgi"]["extraCmdArgs"] = [ - "--max-input-length", - "1280", - "--max-total-tokens", - "2048", - "--max-batch-total-tokens", - "65536", - "--max-batch-prefill-tokens", - "4096", - ] - - yaml_string = yaml.dump(values, default_flow_style=False) - - # Determine the mode based on the 'tune' parameter - mode = "tuned" if tune else "oob" - - # Determine the filename based on 'with_rerank' and 'num_nodes' - if with_rerank: - filename = f"{mode}-{num_nodes}-gaudi-with-rerank-values.yaml" - else: - filename = f"{mode}-{num_nodes}-gaudi-without-rerank-values.yaml" - - # Write the YAML data to the file - with open(filename, "w") as file: - file.write(yaml_string) - - # Get the current working directory and construct the file path - current_dir = os.getcwd() - filepath = os.path.join(current_dir, filename) - - print(f"YAML file {filepath} has been generated.") - return filepath # Optionally return the file path - - -# Main execution for standalone use of create_values_yaml -if __name__ == "__main__": - # Example values for standalone execution - with_rerank = True - num_nodes = 2 - hftoken = "xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx" - modeldir = "/mnt/model" - node_selector = {"node-type": "opea-benchmark"} - tune = True - - filename = generate_helm_values(with_rerank, num_nodes, hftoken, modeldir, node_selector, tune) - - # Read back the generated YAML file for verification - with open(filename, "r") as file: - print("Generated YAML contents:") - print(file.read()) diff --git a/ChatQnA/docker_compose/intel/cpu/aipc/compose.yaml b/ChatQnA/docker_compose/intel/cpu/aipc/compose.yaml index 6e3578e97..9035642c5 100644 --- a/ChatQnA/docker_compose/intel/cpu/aipc/compose.yaml +++ b/ChatQnA/docker_compose/intel/cpu/aipc/compose.yaml @@ -32,7 +32,7 @@ services: retries: 50 restart: unless-stopped tei-embedding-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-embedding-server ports: - "6006:80" @@ -65,7 +65,7 @@ services: RETRIEVER_COMPONENT_NAME: "OPEA_RETRIEVER_REDIS" restart: unless-stopped tei-reranking-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-reranking-server ports: - "8808:80" diff --git a/ChatQnA/docker_compose/intel/cpu/xeon/compose.yaml b/ChatQnA/docker_compose/intel/cpu/xeon/compose.yaml index 12b179996..1e5fef6d4 100644 --- a/ChatQnA/docker_compose/intel/cpu/xeon/compose.yaml +++ b/ChatQnA/docker_compose/intel/cpu/xeon/compose.yaml @@ -39,7 +39,7 @@ services: retries: 50 restart: unless-stopped tei-embedding-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-embedding-server ports: - "6006:80" @@ -72,7 +72,7 @@ services: RETRIEVER_COMPONENT_NAME: "OPEA_RETRIEVER_REDIS" restart: unless-stopped tei-reranking-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-reranking-server ports: - "8808:80" diff --git a/ChatQnA/docker_compose/intel/cpu/xeon/compose_faqgen.yaml b/ChatQnA/docker_compose/intel/cpu/xeon/compose_faqgen.yaml index 06232276f..eb31dfb1f 100644 --- a/ChatQnA/docker_compose/intel/cpu/xeon/compose_faqgen.yaml +++ b/ChatQnA/docker_compose/intel/cpu/xeon/compose_faqgen.yaml @@ -32,7 +32,7 @@ services: retries: 50 restart: unless-stopped tei-embedding-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-embedding-server ports: - "6006:80" @@ -65,7 +65,7 @@ services: RETRIEVER_COMPONENT_NAME: "OPEA_RETRIEVER_REDIS" restart: unless-stopped tei-reranking-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-reranking-server ports: - "8808:80" diff --git a/ChatQnA/docker_compose/intel/cpu/xeon/compose_faqgen_tgi.yaml b/ChatQnA/docker_compose/intel/cpu/xeon/compose_faqgen_tgi.yaml index 1cd3f4fe7..a66be6032 100644 --- a/ChatQnA/docker_compose/intel/cpu/xeon/compose_faqgen_tgi.yaml +++ b/ChatQnA/docker_compose/intel/cpu/xeon/compose_faqgen_tgi.yaml @@ -32,7 +32,7 @@ services: retries: 50 restart: unless-stopped tei-embedding-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-embedding-server ports: - "6006:80" @@ -65,7 +65,7 @@ services: RETRIEVER_COMPONENT_NAME: "OPEA_RETRIEVER_REDIS" restart: unless-stopped tei-reranking-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-reranking-server ports: - "8808:80" diff --git a/ChatQnA/docker_compose/intel/cpu/xeon/compose_milvus.yaml b/ChatQnA/docker_compose/intel/cpu/xeon/compose_milvus.yaml index d7a5b687d..eb81c3ec2 100644 --- a/ChatQnA/docker_compose/intel/cpu/xeon/compose_milvus.yaml +++ b/ChatQnA/docker_compose/intel/cpu/xeon/compose_milvus.yaml @@ -113,7 +113,7 @@ services: restart: unless-stopped tei-embedding-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-embedding-server ports: - "6006:80" @@ -127,7 +127,7 @@ services: command: --model-id ${EMBEDDING_MODEL_ID} --auto-truncate tei-reranking-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-reranking-server ports: - "8808:80" diff --git a/ChatQnA/docker_compose/intel/cpu/xeon/compose_pinecone.yaml b/ChatQnA/docker_compose/intel/cpu/xeon/compose_pinecone.yaml index ffd397209..8a2af3c11 100644 --- a/ChatQnA/docker_compose/intel/cpu/xeon/compose_pinecone.yaml +++ b/ChatQnA/docker_compose/intel/cpu/xeon/compose_pinecone.yaml @@ -29,7 +29,7 @@ services: retries: 50 restart: unless-stopped tei-embedding-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-embedding-server ports: - "6006:80" @@ -60,7 +60,7 @@ services: RETRIEVER_COMPONENT_NAME: "OPEA_RETRIEVER_PINECONE" restart: unless-stopped tei-reranking-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-reranking-server ports: - "8808:80" diff --git a/ChatQnA/docker_compose/intel/cpu/xeon/compose_remote.yaml b/ChatQnA/docker_compose/intel/cpu/xeon/compose_remote.yaml index 665f97d44..2acc51bbe 100644 --- a/ChatQnA/docker_compose/intel/cpu/xeon/compose_remote.yaml +++ b/ChatQnA/docker_compose/intel/cpu/xeon/compose_remote.yaml @@ -33,7 +33,7 @@ services: TEI_ENDPOINT: http://tei-embedding-service:80 HUGGINGFACEHUB_API_TOKEN: ${HUGGINGFACEHUB_API_TOKEN} tei-embedding-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-embedding-server ports: - "6006:80" @@ -66,7 +66,7 @@ services: RETRIEVER_COMPONENT_NAME: "OPEA_RETRIEVER_REDIS" restart: unless-stopped tei-reranking-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-reranking-server ports: - "8808:80" diff --git a/ChatQnA/docker_compose/intel/cpu/xeon/compose_tgi.yaml b/ChatQnA/docker_compose/intel/cpu/xeon/compose_tgi.yaml index 7e138435c..434ae34ea 100644 --- a/ChatQnA/docker_compose/intel/cpu/xeon/compose_tgi.yaml +++ b/ChatQnA/docker_compose/intel/cpu/xeon/compose_tgi.yaml @@ -32,7 +32,7 @@ services: retries: 50 restart: unless-stopped tei-embedding-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-embedding-server ports: - "6006:80" @@ -65,7 +65,7 @@ services: RETRIEVER_COMPONENT_NAME: "OPEA_RETRIEVER_REDIS" restart: unless-stopped tei-reranking-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-reranking-server ports: - "8808:80" diff --git a/ChatQnA/docker_compose/intel/cpu/xeon/compose_without_rerank.yaml b/ChatQnA/docker_compose/intel/cpu/xeon/compose_without_rerank.yaml index dea682626..b813852c7 100644 --- a/ChatQnA/docker_compose/intel/cpu/xeon/compose_without_rerank.yaml +++ b/ChatQnA/docker_compose/intel/cpu/xeon/compose_without_rerank.yaml @@ -32,7 +32,7 @@ services: retries: 50 restart: unless-stopped tei-embedding-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-embedding-server ports: - "6006:80" diff --git a/ChatQnA/docker_compose/intel/hpu/gaudi/README.md b/ChatQnA/docker_compose/intel/hpu/gaudi/README.md index ec9d07a18..6dea16256 100644 --- a/ChatQnA/docker_compose/intel/hpu/gaudi/README.md +++ b/ChatQnA/docker_compose/intel/hpu/gaudi/README.md @@ -95,7 +95,7 @@ d560c232b120 opea/retriever:latest a1d7ca2d3787 ghcr.io/huggingface/tei-gaudi:1.5.0 "text-embeddings-rou…" 2 minutes ago Up 2 minutes 0.0.0.0:8808->80/tcp, [::]:8808->80/tcp tei-reranking-gaudi-server 9a9f3fd4fd4c opea/vllm-gaudi:latest "python3 -m vllm.ent…" 2 minutes ago Exited (1) 2 minutes ago vllm-gaudi-server 1ab9bbdf5182 redis/redis-stack:7.2.0-v9 "/entrypoint.sh" 2 minutes ago Up 2 minutes 0.0.0.0:6379->6379/tcp, :::6379->6379/tcp, 0.0.0.0:8001->8001/tcp, :::8001->8001/tcp redis-vector-db -9ee0789d819e ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 "text-embeddings-rou…" 2 minutes ago Up 2 minutes 0.0.0.0:8090->80/tcp, [::]:8090->80/tcp tei-embedding-gaudi-server +9ee0789d819e ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 "text-embeddings-rou…" 2 minutes ago Up 2 minutes 0.0.0.0:8090->80/tcp, [::]:8090->80/tcp tei-embedding-gaudi-server ``` ### Test the Pipeline @@ -148,7 +148,7 @@ The default deployment utilizes Gaudi devices primarily for the `vllm-service`, | ---------------------------- | ----------------------------------------------------- | ------------ | | redis-vector-db | redis/redis-stack:7.2.0-v9 | No | | dataprep-redis-service | opea/dataprep:latest | No | -| tei-embedding-service | ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 | No | +| tei-embedding-service | ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 | No | | retriever | opea/retriever:latest | No | | tei-reranking-service | ghcr.io/huggingface/tei-gaudi:1.5.0 | 1 card | | vllm-service | opea/vllm-gaudi:latest | Configurable | @@ -164,7 +164,7 @@ The TGI (Text Generation Inference) deployment and the default deployment differ | ---------------------------- | ----------------------------------------------------- | -------------- | | redis-vector-db | redis/redis-stack:7.2.0-v9 | No | | dataprep-redis-service | opea/dataprep:latest | No | -| tei-embedding-service | ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 | No | +| tei-embedding-service | ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 | No | | retriever | opea/retriever:latest | No | | tei-reranking-service | ghcr.io/huggingface/tei-gaudi:1.5.0 | 1 card | | **tgi-service** | ghcr.io/huggingface/tgi-gaudi:2.3.1 | Configurable | @@ -184,7 +184,7 @@ The TGI (Text Generation Inference) deployment and the default deployment differ | ---------------------------- | ----------------------------------------------------- | ------------ | | redis-vector-db | redis/redis-stack:7.2.0-v9 | No | | dataprep-redis-service | opea/dataprep:latest | No | -| tei-embedding-service | ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 | No | +| tei-embedding-service | ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 | No | | retriever | opea/retriever:latest | No | | tei-reranking-service | ghcr.io/huggingface/tei-gaudi:1.5.0 | 1 card | | vllm-service | opea/vllm-gaudi:latest | Configurable | @@ -203,7 +203,7 @@ The _compose_without_rerank.yaml_ Docker Compose file is distinct from the defau | ---------------------------- | ----------------------------------------------------- | -------------- | | redis-vector-db | redis/redis-stack:7.2.0-v9 | No | | dataprep-redis-service | opea/dataprep:latest | No | -| tei-embedding-service | ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 | No | +| tei-embedding-service | ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 | No | | retriever | opea/retriever:latest | No | | vllm-service | opea/vllm-gaudi:latest | Configurable | | chatqna-gaudi-backend-server | opea/chatqna:latest | No | @@ -222,7 +222,7 @@ The _compose_guardrails.yaml_ Docker Compose file introduces enhancements over t | dataprep-redis-service | opea/dataprep:latest | No | No | | _vllm-guardrails-service_ | opea/vllm-gaudi:latest | 1 card | Yes | | _guardrails_ | opea/guardrails:latest | No | No | -| tei-embedding-service | ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 | No | No | +| tei-embedding-service | ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 | No | No | | retriever | opea/retriever:latest | No | No | | tei-reranking-service | ghcr.io/huggingface/tei-gaudi:1.5.0 | 1 card | No | | vllm-service | opea/vllm-gaudi:latest | Configurable | Yes | @@ -258,7 +258,7 @@ The table provides a comprehensive overview of the ChatQnA services utilized acr | ---------------------------- | ----------------------------------------------------- | -------- | -------------------------------------------------------------------------------------------------- | | redis-vector-db | redis/redis-stack:7.2.0-v9 | No | Acts as a Redis database for storing and managing data. | | dataprep-redis-service | opea/dataprep:latest | No | Prepares data and interacts with the Redis database. | -| tei-embedding-service | ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 | No | Provides text embedding services, often using Hugging Face models. | +| tei-embedding-service | ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 | No | Provides text embedding services, often using Hugging Face models. | | retriever | opea/retriever:latest | No | Retrieves data from the Redis database and interacts with embedding services. | | tei-reranking-service | ghcr.io/huggingface/tei-gaudi:1.5.0 | Yes | Reranks text embeddings, typically using Gaudi hardware for enhanced performance. | | vllm-service | opea/vllm-gaudi:latest | No | Handles large language model (LLM) tasks, utilizing Gaudi hardware. | @@ -284,7 +284,7 @@ ChatQnA now supports running the latest DeepSeek models, including [deepseek-ai/ ### tei-embedding-service & tei-reranking-service -The `ghcr.io/huggingface/text-embeddings-inference:cpu-1.6` image supporting `tei-embedding-service` and `tei-reranking-service` depends on the `EMBEDDING_MODEL_ID` or `RERANK_MODEL_ID` environment variables respectively to specify the embedding model and reranking model used for converting text into vector representations and rankings. This choice impacts the quality and relevance of the embeddings rerankings for various applications. Unlike the `vllm-service`, the `tei-embedding-service` and `tei-reranking-service` each typically acquires only one Gaudi device and does not use the `NUM_CARDS` parameter; embedding and reranking tasks generally do not require extensive parallel processing and one Gaudi per service is appropriate. The list of [supported embedding and reranking models](https://github.com/huggingface/tei-gaudi?tab=readme-ov-file#supported-models) can be found at the [huggingface/tei-gaudi](https://github.com/huggingface/tei-gaudi?tab=readme-ov-file#supported-models) website. +The `ghcr.io/huggingface/text-embeddings-inference:cpu-1.5` image supporting `tei-embedding-service` and `tei-reranking-service` depends on the `EMBEDDING_MODEL_ID` or `RERANK_MODEL_ID` environment variables respectively to specify the embedding model and reranking model used for converting text into vector representations and rankings. This choice impacts the quality and relevance of the embeddings rerankings for various applications. Unlike the `vllm-service`, the `tei-embedding-service` and `tei-reranking-service` each typically acquires only one Gaudi device and does not use the `NUM_CARDS` parameter; embedding and reranking tasks generally do not require extensive parallel processing and one Gaudi per service is appropriate. The list of [supported embedding and reranking models](https://github.com/huggingface/tei-gaudi?tab=readme-ov-file#supported-models) can be found at the [huggingface/tei-gaudi](https://github.com/huggingface/tei-gaudi?tab=readme-ov-file#supported-models) website. ### tgi-guardrails-service diff --git a/ChatQnA/docker_compose/intel/hpu/gaudi/compose.yaml b/ChatQnA/docker_compose/intel/hpu/gaudi/compose.yaml index b8b96ad04..49d7ff99a 100644 --- a/ChatQnA/docker_compose/intel/hpu/gaudi/compose.yaml +++ b/ChatQnA/docker_compose/intel/hpu/gaudi/compose.yaml @@ -39,7 +39,7 @@ services: retries: 50 restart: unless-stopped tei-embedding-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-embedding-gaudi-server ports: - "8090:80" diff --git a/ChatQnA/docker_compose/intel/hpu/gaudi/compose_faqgen.yaml b/ChatQnA/docker_compose/intel/hpu/gaudi/compose_faqgen.yaml index f84c955b3..951956be8 100644 --- a/ChatQnA/docker_compose/intel/hpu/gaudi/compose_faqgen.yaml +++ b/ChatQnA/docker_compose/intel/hpu/gaudi/compose_faqgen.yaml @@ -33,7 +33,7 @@ services: retries: 50 restart: unless-stopped tei-embedding-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-embedding-gaudi-server ports: - "8090:80" diff --git a/ChatQnA/docker_compose/intel/hpu/gaudi/compose_faqgen_tgi.yaml b/ChatQnA/docker_compose/intel/hpu/gaudi/compose_faqgen_tgi.yaml index e54e616c4..8c2b0d1d5 100644 --- a/ChatQnA/docker_compose/intel/hpu/gaudi/compose_faqgen_tgi.yaml +++ b/ChatQnA/docker_compose/intel/hpu/gaudi/compose_faqgen_tgi.yaml @@ -33,7 +33,7 @@ services: retries: 50 restart: unless-stopped tei-embedding-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-embedding-gaudi-server ports: - "8090:80" diff --git a/ChatQnA/docker_compose/intel/hpu/gaudi/compose_guardrails.yaml b/ChatQnA/docker_compose/intel/hpu/gaudi/compose_guardrails.yaml index cc0ea6f5b..7f4476441 100644 --- a/ChatQnA/docker_compose/intel/hpu/gaudi/compose_guardrails.yaml +++ b/ChatQnA/docker_compose/intel/hpu/gaudi/compose_guardrails.yaml @@ -76,7 +76,7 @@ services: HUGGINGFACEHUB_API_TOKEN: ${HUGGINGFACEHUB_API_TOKEN} restart: unless-stopped tei-embedding-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-embedding-gaudi-server ports: - "8090:80" diff --git a/ChatQnA/docker_compose/intel/hpu/gaudi/compose_tgi.yaml b/ChatQnA/docker_compose/intel/hpu/gaudi/compose_tgi.yaml index 5be2c4387..02d99098b 100644 --- a/ChatQnA/docker_compose/intel/hpu/gaudi/compose_tgi.yaml +++ b/ChatQnA/docker_compose/intel/hpu/gaudi/compose_tgi.yaml @@ -32,7 +32,7 @@ services: retries: 50 restart: unless-stopped tei-embedding-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-embedding-gaudi-server ports: - "8090:80" 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 a59138bbd..9704984f1 100644 --- a/ChatQnA/docker_compose/intel/hpu/gaudi/compose_without_rerank.yaml +++ b/ChatQnA/docker_compose/intel/hpu/gaudi/compose_without_rerank.yaml @@ -32,7 +32,7 @@ services: retries: 50 restart: unless-stopped tei-embedding-service: - image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 + image: ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 container_name: tei-embedding-gaudi-server ports: - "8090:80" diff --git a/ChatQnA/docker_compose/intel/hpu/gaudi/how_to_validate_service.md b/ChatQnA/docker_compose/intel/hpu/gaudi/how_to_validate_service.md index ba18ce7ae..ce515d450 100644 --- a/ChatQnA/docker_compose/intel/hpu/gaudi/how_to_validate_service.md +++ b/ChatQnA/docker_compose/intel/hpu/gaudi/how_to_validate_service.md @@ -51,7 +51,7 @@ f810f3b4d329 opea/embedding:latest "python embed 174bd43fa6b5 ghcr.io/huggingface/tei-gaudi:1.5.0 "text-embeddings-rou…" 2 minutes ago Up 2 minutes 0.0.0.0:8090->80/tcp, :::8090->80/tcp tei-embedding-gaudi-server 05c40b636239 ghcr.io/huggingface/tgi-gaudi:2.3.1 "text-generation-lau…" 2 minutes ago Exited (1) About a minute ago tgi-gaudi-server 74084469aa33 redis/redis-stack:7.2.0-v9 "/entrypoint.sh" 2 minutes ago Up 2 minutes 0.0.0.0:6379->6379/tcp, :::6379->6379/tcp, 0.0.0.0:8001->8001/tcp, :::8001->8001/tcp redis-vector-db -88399dbc9e43 ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 "text-embeddings-rou…" 2 minutes ago Up 2 minutes 0.0.0.0:8808->80/tcp, :::8808->80/tcp tei-reranking-gaudi-server +88399dbc9e43 ghcr.io/huggingface/text-embeddings-inference:cpu-1.5 "text-embeddings-rou…" 2 minutes ago Up 2 minutes 0.0.0.0:8808->80/tcp, :::8808->80/tcp tei-reranking-gaudi-server ``` In this case, `ghcr.io/huggingface/tgi-gaudi:2.3.1` Existed. diff --git a/ChatQnA/tests/test_compose_faqgen_on_gaudi.sh b/ChatQnA/tests/test_compose_faqgen_on_gaudi.sh index 751b81029..be040bdbe 100644 --- a/ChatQnA/tests/test_compose_faqgen_on_gaudi.sh +++ b/ChatQnA/tests/test_compose_faqgen_on_gaudi.sh @@ -31,8 +31,6 @@ function build_docker_images() { service_list="chatqna chatqna-ui dataprep retriever llm-faqgen vllm-gaudi nginx" docker compose -f build.yaml build ${service_list} --no-cache > ${LOG_PATH}/docker_image_build.log - docker pull ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 - docker pull ghcr.io/huggingface/tei-gaudi:1.5.0 docker images && sleep 1s } diff --git a/ChatQnA/tests/test_compose_faqgen_on_rocm.sh b/ChatQnA/tests/test_compose_faqgen_on_rocm.sh index 781da0d04..7b05bb8c0 100644 --- a/ChatQnA/tests/test_compose_faqgen_on_rocm.sh +++ b/ChatQnA/tests/test_compose_faqgen_on_rocm.sh @@ -69,9 +69,6 @@ function build_docker_images() { service_list="chatqna chatqna-ui dataprep retriever llm-faqgen nginx" docker compose -f build.yaml build ${service_list} --no-cache > "${LOG_PATH}"/docker_image_build.log - docker pull ghcr.io/huggingface/text-generation-inference:2.3.1-rocm - docker pull ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 - docker images && sleep 1s } diff --git a/ChatQnA/tests/test_compose_faqgen_on_xeon.sh b/ChatQnA/tests/test_compose_faqgen_on_xeon.sh index 256732094..dc4279873 100644 --- a/ChatQnA/tests/test_compose_faqgen_on_xeon.sh +++ b/ChatQnA/tests/test_compose_faqgen_on_xeon.sh @@ -32,7 +32,6 @@ function build_docker_images() { service_list="chatqna chatqna-ui dataprep retriever llm-faqgen vllm nginx" docker compose -f build.yaml build ${service_list} --no-cache > ${LOG_PATH}/docker_image_build.log - docker pull ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 docker images && sleep 1s } diff --git a/ChatQnA/tests/test_compose_faqgen_tgi_on_gaudi.sh b/ChatQnA/tests/test_compose_faqgen_tgi_on_gaudi.sh index 39760087d..8eef9c604 100644 --- a/ChatQnA/tests/test_compose_faqgen_tgi_on_gaudi.sh +++ b/ChatQnA/tests/test_compose_faqgen_tgi_on_gaudi.sh @@ -28,9 +28,6 @@ function build_docker_images() { service_list="chatqna chatqna-ui dataprep retriever llm-faqgen nginx" docker compose -f build.yaml build ${service_list} --no-cache > ${LOG_PATH}/docker_image_build.log - docker pull ghcr.io/huggingface/tgi-gaudi:2.0.6 - docker pull ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 - docker pull ghcr.io/huggingface/tei-gaudi:1.5.0 docker images && sleep 1s } diff --git a/ChatQnA/tests/test_compose_faqgen_tgi_on_xeon.sh b/ChatQnA/tests/test_compose_faqgen_tgi_on_xeon.sh index 50e69abf6..d4140a3ab 100644 --- a/ChatQnA/tests/test_compose_faqgen_tgi_on_xeon.sh +++ b/ChatQnA/tests/test_compose_faqgen_tgi_on_xeon.sh @@ -32,8 +32,6 @@ function build_docker_images() { service_list="chatqna chatqna-ui dataprep retriever llm-faqgen nginx" docker compose -f build.yaml build ${service_list} --no-cache > ${LOG_PATH}/docker_image_build.log - docker pull ghcr.io/huggingface/text-generation-inference:2.4.0-intel-cpu - docker pull ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 docker images && sleep 1s } diff --git a/ChatQnA/tests/test_compose_guardrails_on_gaudi.sh b/ChatQnA/tests/test_compose_guardrails_on_gaudi.sh index 00cd7b124..06b58bedc 100644 --- a/ChatQnA/tests/test_compose_guardrails_on_gaudi.sh +++ b/ChatQnA/tests/test_compose_guardrails_on_gaudi.sh @@ -31,9 +31,6 @@ function build_docker_images() { service_list="chatqna chatqna-ui dataprep retriever vllm-gaudi guardrails nginx" docker compose -f build.yaml build ${service_list} --no-cache > ${LOG_PATH}/docker_image_build.log - docker pull ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 - docker pull ghcr.io/huggingface/tei-gaudi:1.5.0 - docker images && sleep 1s } diff --git a/ChatQnA/tests/test_compose_milvus_on_xeon.sh b/ChatQnA/tests/test_compose_milvus_on_xeon.sh index 19beb8175..47a5b43dd 100644 --- a/ChatQnA/tests/test_compose_milvus_on_xeon.sh +++ b/ChatQnA/tests/test_compose_milvus_on_xeon.sh @@ -35,8 +35,6 @@ function build_docker_images() { service_list="chatqna chatqna-ui dataprep retriever vllm nginx" docker compose -f build.yaml build ${service_list} --no-cache > ${LOG_PATH}/docker_image_build.log - docker pull ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 - docker images && sleep 1s } function start_services() { diff --git a/ChatQnA/tests/test_compose_on_gaudi.sh b/ChatQnA/tests/test_compose_on_gaudi.sh index 268c8e27f..0fe3cf11a 100644 --- a/ChatQnA/tests/test_compose_on_gaudi.sh +++ b/ChatQnA/tests/test_compose_on_gaudi.sh @@ -31,8 +31,6 @@ function build_docker_images() { service_list="chatqna chatqna-ui dataprep retriever vllm-gaudi nginx" docker compose -f build.yaml build ${service_list} --no-cache > ${LOG_PATH}/docker_image_build.log - docker pull ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 - docker pull ghcr.io/huggingface/tei-gaudi:1.5.0 docker images && sleep 1s } diff --git a/ChatQnA/tests/test_compose_on_rocm.sh b/ChatQnA/tests/test_compose_on_rocm.sh index a4668d9bb..3ff91522c 100644 --- a/ChatQnA/tests/test_compose_on_rocm.sh +++ b/ChatQnA/tests/test_compose_on_rocm.sh @@ -67,9 +67,6 @@ function build_docker_images() { service_list="chatqna chatqna-ui dataprep retriever nginx" docker compose -f build.yaml build ${service_list} --no-cache > "${LOG_PATH}"/docker_image_build.log - docker pull ghcr.io/huggingface/text-generation-inference:2.3.1-rocm - docker pull ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 - docker images && sleep 1s } diff --git a/ChatQnA/tests/test_compose_on_xeon.sh b/ChatQnA/tests/test_compose_on_xeon.sh index 101946962..38226ec9b 100644 --- a/ChatQnA/tests/test_compose_on_xeon.sh +++ b/ChatQnA/tests/test_compose_on_xeon.sh @@ -34,8 +34,6 @@ function build_docker_images() { service_list="chatqna chatqna-ui dataprep retriever vllm nginx" docker compose -f build.yaml build ${service_list} --no-cache > ${LOG_PATH}/docker_image_build.log - docker pull ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 - docker images && sleep 1s } diff --git a/ChatQnA/tests/test_compose_pinecone_on_xeon.sh b/ChatQnA/tests/test_compose_pinecone_on_xeon.sh index 3d99e0cce..98bfd2136 100755 --- a/ChatQnA/tests/test_compose_pinecone_on_xeon.sh +++ b/ChatQnA/tests/test_compose_pinecone_on_xeon.sh @@ -35,8 +35,6 @@ function build_docker_images() { service_list="chatqna chatqna-ui dataprep retriever vllm nginx" docker compose -f build.yaml build ${service_list} --no-cache > ${LOG_PATH}/docker_image_build.log - docker pull ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 - docker images && sleep 1s } diff --git a/ChatQnA/tests/test_compose_tgi_on_gaudi.sh b/ChatQnA/tests/test_compose_tgi_on_gaudi.sh index c4d921dce..b334fc35c 100644 --- a/ChatQnA/tests/test_compose_tgi_on_gaudi.sh +++ b/ChatQnA/tests/test_compose_tgi_on_gaudi.sh @@ -27,10 +27,6 @@ function build_docker_images() { service_list="chatqna chatqna-ui dataprep retriever nginx" docker compose -f build.yaml build ${service_list} --no-cache > ${LOG_PATH}/docker_image_build.log - docker pull ghcr.io/huggingface/tgi-gaudi:2.3.1 - docker pull ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 - docker pull ghcr.io/huggingface/tei-gaudi:1.5.0 - docker images && sleep 1s } diff --git a/ChatQnA/tests/test_compose_tgi_on_xeon.sh b/ChatQnA/tests/test_compose_tgi_on_xeon.sh index 07bd3fc3c..12c9552ca 100644 --- a/ChatQnA/tests/test_compose_tgi_on_xeon.sh +++ b/ChatQnA/tests/test_compose_tgi_on_xeon.sh @@ -27,9 +27,6 @@ function build_docker_images() { service_list="chatqna chatqna-ui dataprep retriever nginx" docker compose -f build.yaml build ${service_list} --no-cache > ${LOG_PATH}/docker_image_build.log - docker pull ghcr.io/huggingface/text-generation-inference:2.4.0-intel-cpu - docker pull ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 - docker images && sleep 1s } diff --git a/ChatQnA/tests/test_compose_without_rerank_on_gaudi.sh b/ChatQnA/tests/test_compose_without_rerank_on_gaudi.sh index ed979db97..6e4782ba7 100644 --- a/ChatQnA/tests/test_compose_without_rerank_on_gaudi.sh +++ b/ChatQnA/tests/test_compose_without_rerank_on_gaudi.sh @@ -31,9 +31,6 @@ function build_docker_images() { service_list="chatqna chatqna-ui dataprep retriever vllm-gaudi nginx" docker compose -f build.yaml build ${service_list} --no-cache > ${LOG_PATH}/docker_image_build.log - docker pull ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 - docker pull ghcr.io/huggingface/tei-gaudi:1.5.0 - docker images && sleep 1s } diff --git a/ChatQnA/tests/test_compose_without_rerank_on_xeon.sh b/ChatQnA/tests/test_compose_without_rerank_on_xeon.sh index 66f7a535a..2d79b0e7a 100644 --- a/ChatQnA/tests/test_compose_without_rerank_on_xeon.sh +++ b/ChatQnA/tests/test_compose_without_rerank_on_xeon.sh @@ -35,8 +35,6 @@ function build_docker_images() { service_list="chatqna chatqna-ui dataprep retriever vllm nginx" docker compose -f build.yaml build ${service_list} --no-cache > ${LOG_PATH}/docker_image_build.log - docker pull ghcr.io/huggingface/text-embeddings-inference:cpu-1.6 - docker images && sleep 1s }