Hugging Face's logo

Buckets:
hf-doc-build
/
doc-dev

Files
xet
hf-doc-build/doc-dev / google-cloud /pr_113 /en /examples /gke-tei-from-gcs-deployment.html
rtrm's picture
rtrm
download
raw
64.7 kB
<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;Deploy BGE Base v1.5 with TEI DLC from GCS on GKE&quot;,&quot;local&quot;:&quot;deploy-bge-base-v15-with-tei-dlc-from-gcs-on-gke&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Setup / Configuration&quot;,&quot;local&quot;:&quot;setup--configuration&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Create GKE Cluster&quot;,&quot;local&quot;:&quot;create-gke-cluster&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Optional: Upload a model from the Hugging Face Hub to GCS&quot;,&quot;local&quot;:&quot;optional-upload-a-model-from-the-hugging-face-hub-to-gcs&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Configure IAM for GCS&quot;,&quot;local&quot;:&quot;configure-iam-for-gcs&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Deploy TEI&quot;,&quot;local&quot;:&quot;deploy-tei&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Inference with TEI&quot;,&quot;local&quot;:&quot;inference-with-tei&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Via cURL&quot;,&quot;local&quot;:&quot;via-curl&quot;,&quot;sections&quot;:[],&quot;depth&quot;:3}],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Delete GKE Cluster&quot;,&quot;local&quot;:&quot;delete-gke-cluster&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}">
 <link href="/docs/google-cloud/pr_113/en/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
 <link rel="modulepreload" href="/docs/google-cloud/pr_113/en/_app/immutable/entry/start.3fa28be1.js">
 <link rel="modulepreload" href="/docs/google-cloud/pr_113/en/_app/immutable/chunks/scheduler.b108d059.js">
 <link rel="modulepreload" href="/docs/google-cloud/pr_113/en/_app/immutable/chunks/singletons.e5e8fb57.js">
 <link rel="modulepreload" href="/docs/google-cloud/pr_113/en/_app/immutable/chunks/paths.0d296fb5.js">
 <link rel="modulepreload" href="/docs/google-cloud/pr_113/en/_app/immutable/entry/app.8572c6c6.js">
 <link rel="modulepreload" href="/docs/google-cloud/pr_113/en/_app/immutable/chunks/index.008de539.js">
 <link rel="modulepreload" href="/docs/google-cloud/pr_113/en/_app/immutable/nodes/0.d1a2de3e.js">
 <link rel="modulepreload" href="/docs/google-cloud/pr_113/en/_app/immutable/nodes/9.19d10e1f.js">
 <link rel="modulepreload" href="/docs/google-cloud/pr_113/en/_app/immutable/chunks/Tip.aeb15ab7.js">
 <link rel="modulepreload" href="/docs/google-cloud/pr_113/en/_app/immutable/chunks/CodeBlock.3968c746.js">
 <link rel="modulepreload" href="/docs/google-cloud/pr_113/en/_app/immutable/chunks/EditOnGithub.d1c48e3d.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;Deploy BGE Base v1.5 with TEI DLC from GCS on GKE&quot;,&quot;local&quot;:&quot;deploy-bge-base-v15-with-tei-dlc-from-gcs-on-gke&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Setup / Configuration&quot;,&quot;local&quot;:&quot;setup--configuration&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Create GKE Cluster&quot;,&quot;local&quot;:&quot;create-gke-cluster&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Optional: Upload a model from the Hugging Face Hub to GCS&quot;,&quot;local&quot;:&quot;optional-upload-a-model-from-the-hugging-face-hub-to-gcs&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Configure IAM for GCS&quot;,&quot;local&quot;:&quot;configure-iam-for-gcs&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Deploy TEI&quot;,&quot;local&quot;:&quot;deploy-tei&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Inference with TEI&quot;,&quot;local&quot;:&quot;inference-with-tei&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Via cURL&quot;,&quot;local&quot;:&quot;via-curl&quot;,&quot;sections&quot;:[],&quot;depth&quot;:3}],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Delete GKE Cluster&quot;,&quot;local&quot;:&quot;delete-gke-cluster&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}"><!-- HEAD_svelte-u9bgzb_END --> <p></p> <h1 class="relative group"><a id="deploy-bge-base-v15-with-tei-dlc-from-gcs-on-gke" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#deploy-bge-base-v15-with-tei-dlc-from-gcs-on-gke"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>Deploy BGE Base v1.5 with TEI DLC from GCS on GKE</span></h1> <p data-svelte-h="svelte-2fuf0r">BGE, standing for BAAI General Embedding, is a collection of embedding models released by BAAI, which is an English base model for general embedding tasks ranked in the MTEB Leaderboard. Text Embeddings Inference (TEI) is a toolkit developed by Hugging Face for deploying and serving open source text embeddings and sequence classification models; enabling high-performance extraction for the most popular models, including FlagEmbedding, Ember, GTE and E5. And, Google Kubernetes Engine (GKE) is a fully-managed Kubernetes service in Google Cloud that can be used to deploy and operate containerized applications at scale using GCP’s infrastructure.</p> <p data-svelte-h="svelte-1nb7ex0">This example showcases how to deploy a text embedding model from a Google Cloud Storage (GCS) Bucket on a GKE Cluster running a purpose-built container to deploy text embedding models in a secure and managed environment with the Hugging Face DLC for TEI.</p> <h2 class="relative group"><a id="setup--configuration" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#setup--configuration"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>Setup / Configuration</span></h2> <p data-svelte-h="svelte-1qnmo38">First, you need to install both <code>gcloud</code> and <code>kubectl</code> in your local machine, which are the command-line tools for Google Cloud and Kubernetes, respectively, to interact with the GCP and the GKE Cluster.</p> <ul data-svelte-h="svelte-a8ymuw"><li>To install <code>gcloud</code>, follow the instructions at <a href="https://cloud.google.com/sdk/docs/install" rel="nofollow">Cloud SDK Documentation - Install the gcloud CLI</a>.</li> <li>To install <code>kubectl</code>, follow the instructions at <a href="https://kubernetes.io/docs/tasks/tools/#kubectl" rel="nofollow">Kubernetes Documentation - Install Tools</a>.</li></ul> <p data-svelte-h="svelte-1i8n6rq">Optionally, to ease the usage of the commands within this tutorial, you need to set the following environment variables for GCP:</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-built_in">export</span> PROJECT_ID=your-project-id
<span class="hljs-built_in">export</span> LOCATION=your-location
<span class="hljs-built_in">export</span> CLUSTER_NAME=your-cluster-name
<span class="hljs-built_in">export</span> BUCKET_NAME=hf-models-gke-bucket<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1g39t6p">Then you need to login into your GCP account and set the project ID to the one you want to use for the deployment of the GKE Cluster.</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->gcloud auth login
gcloud auth application-default login <span class="hljs-comment"># For local development</span>
gcloud config <span class="hljs-built_in">set</span> project <span class="hljs-variable">$PROJECT_ID</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-ly6a6j">Once you are logged in, you need to enable the necessary service APIs in GCP, such as the Google Kubernetes Engine API, the Google Container Registry API, and the Google Container File System API, which are necessary for the deployment of the GKE Cluster and the Hugging Face DLC for TEI.</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->gcloud services <span class="hljs-built_in">enable</span> container.googleapis.com
gcloud services <span class="hljs-built_in">enable</span> containerregistry.googleapis.com
gcloud services <span class="hljs-built_in">enable</span> containerfilesystem.googleapis.com<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-wjaltb">Additionally, to use <code>kubectl</code> with the GKE Cluster credentials, you also need to install the <code>gke-gcloud-auth-plugin</code>, that can be installed with <code>gcloud</code> as follows:</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->gcloud components install gke-gcloud-auth-plugin<!-- HTML_TAG_END --></pre></div> <div class="course-tip bg-gradient-to-br dark:bg-gradient-to-r before:border-green-500 dark:before:border-green-800 from-green-50 dark:from-gray-900 to-white dark:to-gray-950 border border-green-50 text-green-700 dark:text-gray-400"><p data-svelte-h="svelte-sm1xw4">Installing the <code>gke-gcloud-auth-plugin</code> does not need to be installed via <code>gcloud</code> specifically, to read more about the alternative installation methods, please visit <a href="https://cloud.google.com/kubernetes-engine/docs/how-to/cluster-access-for-kubectl#install_plugin" rel="nofollow">https://cloud.google.com/kubernetes-engine/docs/how-to/cluster-access-for-kubectl#install_plugin</a>.</p></div> <h2 class="relative group"><a id="create-gke-cluster" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#create-gke-cluster"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>Create GKE Cluster</span></h2> <p data-svelte-h="svelte-1202via">Once everything is set up, you can proceed with the creation of the GKE Cluster and the node pool, which in this case will be a single CPU node as for most of the workloads CPU inference is enough to serve most of the text embeddings models, while it could benefit a lot from GPU serving.</p> <div class="course-tip bg-gradient-to-br dark:bg-gradient-to-r before:border-green-500 dark:before:border-green-800 from-green-50 dark:from-gray-900 to-white dark:to-gray-950 border border-green-50 text-green-700 dark:text-gray-400"><p data-svelte-h="svelte-1q6mkzx">CPU is being used to run the inference on top of the text embeddings models to showcase the current capabilities of TEI, but switching to GPU is as easy as replacing <code>spec.containers[0].image</code> with <code>us-docker.pkg.dev/deeplearning-platform-release/gcr.io/huggingface-text-embeddings-inference-cu122.1-4.ubuntu2204</code>, and then updating the requested resources, as well as the <code>nodeSelector</code> requirements in the <code>deployment.yaml</code> file. For more information, please refer to the <a href="https://github.com/huggingface/Google-Cloud-Containers/tree/main/examples/gke/tei-from-gcs-deployment/gpu-config/" rel="nofollow"><code>gpu-config</code></a> directory that contains a pre-defined configuration for GPU serving in TEI with an NVIDIA Tesla T4 GPU (with a compute capability of 7.5 i.e. natively supported in TEI).</p></div> <p data-svelte-h="svelte-l7bdz6">To deploy the GKE Cluster, the “Autopilot” mode will be used as it is the recommended one for most of the workloads, since the underlying infrastructure is managed by Google. Alternatively, you can also use the “Standard” mode.</p> <div class="course-tip bg-gradient-to-br dark:bg-gradient-to-r before:border-green-500 dark:before:border-green-800 from-green-50 dark:from-gray-900 to-white dark:to-gray-950 border border-green-50 text-green-700 dark:text-gray-400"><p data-svelte-h="svelte-1nupym3">Important to check before creating the GKE Autopilot Cluster the <a href="https://cloud.google.com/kubernetes-engine/docs/how-to/performance-pods" rel="nofollow">GKE Documentation - Optimize Autopilot Pod performance by choosing a machine series</a>, since not all the cluster versions support every CPU. Same applies for the GPU support e.g. <code>nvidia-l4</code> is not supported in the GKE cluster versions 1.28.3 or lower.</p></div> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->gcloud container clusters create-auto <span class="hljs-variable">$CLUSTER_NAME</span> \
 --project=<span class="hljs-variable">$PROJECT_ID</span> \
 --location=<span class="hljs-variable">$LOCATION</span> \
 --release-channel=stable \
 --cluster-version=1.28 \
 --no-autoprovisioning-enable-insecure-kubelet-readonly-port<!-- HTML_TAG_END --></pre></div> <div class="course-tip bg-gradient-to-br dark:bg-gradient-to-r before:border-green-500 dark:before:border-green-800 from-green-50 dark:from-gray-900 to-white dark:to-gray-950 border border-green-50 text-green-700 dark:text-gray-400"><p data-svelte-h="svelte-a3yff9">To select the specific version in your location of the GKE Cluster, you can run the following command:</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->gcloud container get-server-config \
 --flatten=<span class="hljs-string">&quot;channels&quot;</span> \
 --filter=<span class="hljs-string">&quot;channels.channel=STABLE&quot;</span> \
 --format=<span class="hljs-string">&quot;yaml(channels.channel,channels.defaultVersion)&quot;</span> \
 --location=<span class="hljs-variable">$LOCATION</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1y4gd42">For more information please visit <a href="https://cloud.google.com/kubernetes-engine/versioning#specifying_cluster_version" rel="nofollow">https://cloud.google.com/kubernetes-engine/versioning#specifying_cluster_version</a>.</p></div> <p data-svelte-h="svelte-13ltt22"><img src="https://raw.githubusercontent.com/huggingface/Google-Cloud-Containers/main/examples/gke/tei-from-gcs-deployment/imgs/gke-cluster.png" alt="GKE Cluster in the GCP Console"></p> <p data-svelte-h="svelte-f29e6h">Once the GKE Cluster is created, you can get the credentials to access it via <code>kubectl</code> with the following command:</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->gcloud container clusters get-credentials <span class="hljs-variable">$CLUSTER_NAME</span> --location=<span class="hljs-variable">$LOCATION</span><!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="optional-upload-a-model-from-the-hugging-face-hub-to-gcs" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#optional-upload-a-model-from-the-hugging-face-hub-to-gcs"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>Optional: Upload a model from the Hugging Face Hub to GCS</span></h2> <p data-svelte-h="svelte-1yxvryk">This is an optional step in the tutorial, since you may want to reuse an existing model on a GCS Bucket, if that is the case, then feel free to jump to the next step of the tutorial on how to configure the IAM for GCS so that you can access the bucket from a pod in the GKE Cluster.</p> <p data-svelte-h="svelte-1khd0ur">Otherwise, to upload a model from the Hugging Face Hub to a GCS Bucket, you can use the script <a href="https://github.com/huggingface/Google-Cloud-Containers/tree/main/examples/gke/../../scripts/upload_model_to_gcs.sh" rel="nofollow">scripts/upload_model_to_gcs.sh</a>, which will download the model from the Hugging Face Hub and upload it to the GCS Bucket (and create the bucket if not created already).</p> <p data-svelte-h="svelte-zucy2s">The <code>gsutil</code> component should be installed via <code>gcloud</code>, and the Python packages <code>huggingface_hub</code> with the extra <code>hf_transfer</code>, and the package <code>crcmod</code> should also be installed.</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->gcloud components install gsutil
pip install --upgrade --quiet <span class="hljs-string">&quot;huggingface_hub[hf_transfer]&quot;</span> crcmod<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-u73w5r">Then, you can run the script to download the model from the Hugging Face Hub and then upload it to the GCS Bucket:</p> <div class="course-tip bg-gradient-to-br dark:bg-gradient-to-r before:border-green-500 dark:before:border-green-800 from-green-50 dark:from-gray-900 to-white dark:to-gray-950 border border-green-50 text-green-700 dark:text-gray-400"><p data-svelte-h="svelte-v7ast4">Make sure to set the proper permissions to run the script i.e. <code>chmod +x ./scripts/upload_model_to_gcs.sh</code>.</p></div> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->./scripts/upload_model_to_gcs.sh --model-id BAAI/bge-base-en-v1.5 --gcs gs://<span class="hljs-variable">$BUCKET_NAME</span>/bge-base-en-v1.5<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-9rwj46"><img src="https://raw.githubusercontent.com/huggingface/Google-Cloud-Containers/main/examples/gke/tei-from-gcs-deployment/imgs/gcs-bucket.png" alt="GCS Bucket in the GCP Console"></p> <h2 class="relative group"><a id="configure-iam-for-gcs" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#configure-iam-for-gcs"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>Configure IAM for GCS</span></h2> <p data-svelte-h="svelte-177miyp">Before you proceed with the deployment of the Hugging Face DLC for TEI on the GKE Cluster, you need to set the IAM permissions for the GCS Bucket so that the pod in the GKE Cluster can access the bucket. To do so, you need to create a namespace and a service account in the GKE Cluster, and then set the IAM permissions for the GCS Bucket that contains the model, either as uploaded from the Hugging Face Hub or as already existing in the GCS Bucket.</p> <p data-svelte-h="svelte-4fsaj7">For convenience, as the reference to both the namespace and the service account will be used within the following steps, the environment variables <code>NAMESPACE</code> and <code>SERVICE_ACCOUNT</code> will be set.</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-built_in">export</span> NAMESPACE=hf-gke-namespace
<span class="hljs-built_in">export</span> SERVICE_ACCOUNT=hf-gke-service-account<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-gi3hy9">Then you can create the namespace and the service account in the GKE Cluster, enabling the creation of the IAM permissions for the pods in that namespace to access the GCS Bucket when using that service account.</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->kubectl create namespace <span class="hljs-variable">$NAMESPACE</span>
kubectl create serviceaccount <span class="hljs-variable">$SERVICE_ACCOUNT</span> --namespace <span class="hljs-variable">$NAMESPACE</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1wspkh7">Then you need to add the IAM policy binding to the bucket as follows:</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->gcloud storage buckets add-iam-policy-binding \
 gs://<span class="hljs-variable">$BUCKET_NAME</span> \
 --member <span class="hljs-string">&quot;principal://iam.googleapis.com/projects/<span class="hljs-subst">$(gcloud projects describe $PROJECT_ID --format=<span class="hljs-string">&quot;value(projectNumber)&quot;</span>)</span>/locations/global/workloadIdentityPools/<span class="hljs-variable">$PROJECT_ID</span>.svc.id.goog/subject/ns/<span class="hljs-variable">$NAMESPACE</span>/sa/<span class="hljs-variable">$SERVICE_ACCOUNT</span>&quot;</span> \
 --role <span class="hljs-string">&quot;roles/storage.objectUser&quot;</span><!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="deploy-tei" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#deploy-tei"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>Deploy TEI</span></h2> <p data-svelte-h="svelte-1asaq7z">Now you can proceed to the Kubernetes deployment of the Hugging Face DLC for TEI, serving the <a href="https://huggingface.co/BAAI/bge-base-en-v1.5" rel="nofollow"><code>BAAI/bge-base-en-v1.5</code></a> model, from a volume mounted in <code>/data</code>, copied from the GCS Bucket where the model is located.</p> <div class="course-tip bg-gradient-to-br dark:bg-gradient-to-r before:border-green-500 dark:before:border-green-800 from-green-50 dark:from-gray-900 to-white dark:to-gray-950 border border-green-50 text-green-700 dark:text-gray-400"><p data-svelte-h="svelte-g6mrbh">Recently, the Hugging Face Hub team has included the <code>text-embeddings-inference</code> tag in the Hub, so feel free to explore all the embedding models in the Hub that can be served via TEI at <a href="https://huggingface.co/models?other=text-embeddings-inference" rel="nofollow">https://huggingface.co/models?other=text-embeddings-inference</a>.</p></div> <p data-svelte-h="svelte-13i3nqj">The Hugging Face DLC for TEI will be deployed via <code>kubectl</code>, from the configuration files in either the <a href="https://github.com/huggingface/Google-Cloud-Containers/tree/main/examples/gke/tei-from-gcs-deployment/cpu-config/" rel="nofollow"><code>cpu-config/</code></a> or the <a href="https://github.com/huggingface/Google-Cloud-Containers/tree/main/examples/gke/tei-from-gcs-deployment/gpu-config/" rel="nofollow"><code>gpu-config/</code></a> directories depending on whether you want to use the CPU or GPU accelerators, respectively:</p> <ul data-svelte-h="svelte-oqu2mg"><li><a href="https://github.com/huggingface/Google-Cloud-Containers/tree/main/examples/gke/tei-from-gcs-deployment/cpu-config/deployment.yaml" rel="nofollow"><code>deployment.yaml</code></a>: contains the deployment details of the pod including the reference to the Hugging Face DLC for TEI setting the <code>MODEL_ID</code> to the model path in the volume mount, in this case <code>/data/bge-base-en-v1.5</code>.</li> <li><a href="https://github.com/huggingface/Google-Cloud-Containers/tree/main/examples/gke/tei-from-gcs-deployment/cpu-config/service.yaml" rel="nofollow"><code>service.yaml</code></a>: contains the service details of the pod, exposing the port 8080 for the TEI service.</li> <li><a href="https://github.com/huggingface/Google-Cloud-Containers/tree/main/examples/gke/tei-from-gcs-deployment/cpu-config" rel="nofollow"><code>storageclass.yaml</code></a>: contains the storage class details of the pod, defining the storage class for the volume mount.</li> <li>(optional) <a href="https://github.com/huggingface/Google-Cloud-Containers/tree/main/examples/gke/tei-from-gcs-deployment/cpu-config/ingress.yaml" rel="nofollow"><code>ingress.yaml</code></a>: contains the ingress details of the pod, exposing the service to the external world so that it can be accessed via the ingress IP.</li></ul> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->kubectl apply -f cpu-config/<!-- HTML_TAG_END --></pre></div> <div class="course-tip bg-gradient-to-br dark:bg-gradient-to-r before:border-green-500 dark:before:border-green-800 from-green-50 dark:from-gray-900 to-white dark:to-gray-950 border border-green-50 text-green-700 dark:text-gray-400"><p data-svelte-h="svelte-x77tg0">As already mentioned, for this example you will be deploying the container in a CPU node, but the configuration to deploy TEI in a GPU node is also available in the <a href="https://github.com/huggingface/Google-Cloud-Containers/tree/main/examples/gke/tei-from-gcs-deployment/gpu-config/" rel="nofollow"><code>gpu-config</code></a> directory, so if you want to deploy TEI in a GPU node, please run <code>kubectl apply -f gpu-config/</code> instead of <code>kubectl apply -f cpu-config/</code>.</p></div> <p data-svelte-h="svelte-1rezrxc"><img src="https://raw.githubusercontent.com/huggingface/Google-Cloud-Containers/main/examples/gke/tei-from-gcs-deployment/imgs/gke-deployment.png" alt="GKE Deployment in the GCP Console"></p> <div class="course-tip bg-gradient-to-br dark:bg-gradient-to-r before:border-green-500 dark:before:border-green-800 from-green-50 dark:from-gray-900 to-white dark:to-gray-950 border border-green-50 text-green-700 dark:text-gray-400"><p data-svelte-h="svelte-qgh43p">The Kubernetes deployment may take a few minutes to be ready, so you can check the status of the deployment with the following command:</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->kubectl get pods --namespace <span class="hljs-variable">$NAMESPACE</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-16v3quh">Alternatively, you can just wait for the deployment to be ready with the following command:</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->kubectl <span class="hljs-built_in">wait</span> --<span class="hljs-keyword">for</span>=condition=Available --<span class="hljs-built_in">timeout</span>=700s --namespace <span class="hljs-variable">$NAMESPACE</span> deployment/tei-deployment<!-- HTML_TAG_END --></pre></div></div> <h2 class="relative group"><a id="inference-with-tei" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#inference-with-tei"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>Inference with TEI</span></h2> <p data-svelte-h="svelte-1is499s">To run the inference over the deployed TEI service, you can either:</p> <ul><li><p data-svelte-h="svelte-1pbh49h">Port-forwarding the deployed TEI service to the port 8080, so as to access via <code>localhost</code> with the command:</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->kubectl port-forward --namespace <span class="hljs-variable">$NAMESPACE</span> service/tei-service 8080:8080<!-- HTML_TAG_END --></pre></div></li> <li><p data-svelte-h="svelte-eii2c1">Accessing the TEI service via the external IP of the ingress, which is the default scenario here since you have defined the ingress configuration in either the <code>cpu-configs/ingress.yaml</code> or the <code>gpu-config/ingress.yaml</code> file (but it can be skipped in favour of the port-forwarding), that can be retrieved with the following command:</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->kubectl get ingress --namespace <span class="hljs-variable">$NAMESPACE</span> tei-ingress -o jsonpath=<span class="hljs-string">&#x27;{.status.loadBalancer.ingress[0].ip}&#x27;</span><!-- HTML_TAG_END --></pre></div></li></ul> <div class="course-tip bg-gradient-to-br dark:bg-gradient-to-r before:border-green-500 dark:before:border-green-800 from-green-50 dark:from-gray-900 to-white dark:to-gray-950 border border-green-50 text-green-700 dark:text-gray-400"><p data-svelte-h="svelte-2zu4d0">TEI exposes different inference endpoints based on the task that the model is serving:</p> <ul data-svelte-h="svelte-1dffo21"><li><strong>Text Embeddings</strong>: text embedding models expose the endpoint <code>/embed</code> expecting a payload with the key <code>inputs</code> which is either a string or a list of strings to be embedded.</li> <li><strong>Re-rank</strong>: re-ranker models expose the endpoint <code>/rerank</code> expecting a payload with the keys <code>query</code> and <code>texts</code>, where the <code>query</code> is the reference used to rank the similarity against each text in <code>texts</code>.</li> <li><strong>Sequence Classification</strong>: classic sequence classification models expose the endpoint <code>/predict</code> which expects a payload with the key <code>inputs</code> which is either a string or a list of strings to classify.
More information at <a href="https://huggingface.co/docs/text-embeddings-inference/quick_tour" rel="nofollow">https://huggingface.co/docs/text-embeddings-inference/quick_tour</a>.</li></ul></div> <h3 class="relative group"><a id="via-curl" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#via-curl"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>Via cURL</span></h3> <p data-svelte-h="svelte-189d42i">To send a POST request to the TEI service using <code>cURL</code>, you can run the following command:</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->curl http://localhost:8080/embed \
 -X POST \
 -d <span class="hljs-string">&#x27;{&quot;inputs&quot;:&quot;What is Deep Learning?&quot;}&#x27;</span> \
 -H <span class="hljs-string">&#x27;Content-Type: application/json&#x27;</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1u6gmmt">Or send a POST request to the ingress IP instead:</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->curl http://$(kubectl get ingress --namespace=<span class="hljs-variable">$NAMESPACE</span> tei-ingress -o jsonpath=<span class="hljs-string">&#x27;{.status.loadBalancer.ingress[0].ip}&#x27;</span>)/embed \
 -X POST \
 -d <span class="hljs-string">&#x27;{&quot;inputs&quot;:&quot;What is Deep Learning?&quot;}&#x27;</span> \
 -H <span class="hljs-string">&#x27;Content-Type: application/json&#x27;</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-dhgixx">Which produces the following output (truncated for brevity, but original tensor length is 768, which is the embedding dimension of <a href="https://huggingface.co/BAAI/bge-base-en-v1.5" rel="nofollow"><code>BAAI/bge-base-en-v1.5</code></a> i.e. the model you are serving):</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->[[-<span class="hljs-number">0.01483098</span>,<span class="hljs-number">0</span>.<span class="hljs-number">010846359</span>,-<span class="hljs-number">0.024679236</span>,<span class="hljs-number">0</span>.<span class="hljs-number">012507628</span>,<span class="hljs-number">0.034231555</span>,...]]<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="delete-gke-cluster" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#delete-gke-cluster"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>Delete GKE Cluster</span></h2> <p data-svelte-h="svelte-1teoy12">Finally, once you are done using TEI on the GKE Cluster, you can safely delete the GKE Cluster to avoid incurring in unnecessary costs.</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->gcloud container clusters delete <span class="hljs-variable">$CLUSTER_NAME</span> --location=<span class="hljs-variable">$LOCATION</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-qo8r7n">Alternatively, you can also downscale the replicas of the deployed pod to 0 in case you want to preserve the cluster, since the default GKE Cluster deployed with GKE Autopilot mode is running just a single <code>e2-small</code> instance.</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->kubectl scale --replicas=0 --namespace=<span class="hljs-variable">$NAMESPACE</span> deployment/tei-deployment<!-- HTML_TAG_END --></pre></div> <hr> <div class="course-tip bg-gradient-to-br dark:bg-gradient-to-r before:border-green-500 dark:before:border-green-800 from-green-50 dark:from-gray-900 to-white dark:to-gray-950 border border-green-50 text-green-700 dark:text-gray-400"><p data-svelte-h="svelte-13mig5l">📍 Find the complete example on GitHub <a href="https://github.com/huggingface/Google-Cloud-Containers/tree/main/examples/gke/tei-from-gcs-deployment" rel="nofollow">here</a>!</p></div> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/Google-Cloud-Containers/blob/main/docs/source/examples/gke-tei-from-gcs-deployment.mdx" target="_blank"><span data-svelte-h="svelte-1kd6by1">&lt;</span> <span data-svelte-h="svelte-x0xyl0">&gt;</span> <span data-svelte-h="svelte-1dajgef"><span class="underline ml-1.5">Update</span> on GitHub</span></a> <p></p>
<script>
 {
 __sveltekit_1alknr2 = {
 assets: "/docs/google-cloud/pr_113/en",
 base: "/docs/google-cloud/pr_113/en",
 env: {}
 };
 const element = document.currentScript.parentElement;
 const data = [null,null];
 Promise.all([
 import("/docs/google-cloud/pr_113/en/_app/immutable/entry/start.3fa28be1.js"),
 import("/docs/google-cloud/pr_113/en/_app/immutable/entry/app.8572c6c6.js")
 ]).then(([kit, app]) => {
 kit.start(app, element, {
 node_ids: [0, 9],
 data,
 form: null,
 error: null
 });
 });
 }
 </script>

Xet Storage Details

Size:
64.7 kB
·
Xet hash:
74e1b981a1d8a305364b524ca85c99fe1abfa9439279bd8e3ef48de4315cdc73

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.