1"use strict";(self.webpackChunktlab_docs=self.webpackChunktlab_docs||[]).push([[6120],{28453:(e,n,s)=>{s.d(n,{R:()=>c,x:()=>a});var t=s(96540);const r={},i=t.createContext(r);function c(e){const n=t.useContext(i);return t.useMemo(function(){return"function"==typeof e?e(n):{...n,...e}},[n,e])}function a(e){let n;return n=e.disableParentContext?"function"==typeof e.components?e.components(r):e.components||r:c(e.components),t.createElement(i.Provider,{value:n},e.children)}},30739:(e,n,s)=>{s.d(n,{A:()=>t});const t=s.p+"assets/images/select-service-type-2b4c460d2828cc57068d3fca05e9cc52.png"},33717:(e,n,s)=>{s.d(n,{A:()=>t});const t=s.p+"assets/images/screenshot-providershealth-525b60f39d0699f907c4f16d19473b04.png"},47167:(e,n,s)=>{s.d(n,{A:()=>t});const t=s.p+"assets/images/interact-page-new-c2ba411ed9a345899d100e607d2fb167.png"},50202:(e,n,s)=>{s.d(n,{A:()=>t});const t=s.p+"assets/images/access-service-211e47b9eda6cc2bb0fe91fcd2408f14.png"},58195:(e,n,s)=>{s.r(n),s.d(n,{assets:()=>o,contentTitle:()=>a,default:()=>h,frontMatter:()=>c,metadata:()=>t,toc:()=>l});const t=JSON.parse('{"id":"running-a-service","title":"Running an Interactive Service","description":"What is an Interactive Service?","source":"@site/for-teams/running-a-service.md","sourceDirName":".","slug":"/running-a-service","permalink":"/for-teams/running-a-service","draft":false,"unlisted":false,"tags":[],"version":"current","sidebarPosition":50,"frontMatter":{"title":"Running an Interactive Service","sidebar_position":50},"sidebar":"documentationSidebar","previous":{"title":"Running a Sweep","permalink":"/for-teams/sweeps"},"next":{"title":"Interactive Service Examples","permalink":"/for-teams/category/interactive-service-examples"}}');var r=s(74848),i=s(28453);const c={title:"Running an Interactive Service",sidebar_position:50},a=void 0,o={},l=[{value:"What is an Interactive Service?",id:"what-is-an-interactive-service",level:2},{value:"Tasks vs. Interactive Services",id:"tasks-vs-interactive-services",level:3},{value:"Common Use Cases for Interactive Services",id:"common-use-cases-for-interactive-services",level:2},{value:"1. Model Inference",id:"1-model-inference",level:3},{value:"2. Exploratory Research & Development",id:"2-exploratory-research--development",level:3},{value:"Prerequisites",id:"prerequisites",level:2},{value:"Steps to Run an Interactive Service",id:"steps-to-run-an-interactive-service",level:2}];function d(e){const n={h2:"h2",h3:"h3",hr:"hr",img:"img",li:"li",ol:"ol",p:"p",strong:"strong",ul:"ul",...(0,i.R)(),...e.components};return(0,r.jsxs)(r.Fragment,{children:[(0,r.jsx)(n.h2,{id:"what-is-an-interactive-service",children:"What is an Interactive Service?"}),"\n",(0,r.jsxs)(n.p,{children:["Transformer Lab supports two distinct types of workloads, designed to map to the different ways researchers work: ",(0,r.jsx)(n.strong,{children:"Tasks"})," and ",(0,r.jsx)(n.strong,{children:"Interactive Services"}),"."]}),"\n",(0,r.jsx)(n.h3,{id:"tasks-vs-interactive-services",children:"Tasks vs. Interactive Services"}),"\n",(0,r.jsxs)(n.ul,{children:["\n",(0,r.jsxs)(n.li,{children:[(0,r.jsx)(n.strong,{children:"Tasks (Batch Workloads):"})," These are jobs that are queued and scheduled to run automatically when the necessary resources become available. They execute a set of instructions and terminate upon completion. A commoon usecase would be for training a model."]}),"\n",(0,r.jsxs)(n.li,{children:[(0,r.jsx)(n.strong,{children:"Interactive Services (On-Demand):"})," These workloads function like a ",(0,r.jsx)(n.strong,{children:"reservation"}),". When you launch an Interactive Service, a specific computer or set of resources is held exclusively for you. It remains active and available until you explicitly release it."]}),"\n"]}),"\n",(0,r.jsx)(n.hr,{}),"\n",(0,r.jsx)(n.h2,{id:"common-use-cases-for-interactive-services",children:"Common Use Cases for Interactive Services"}),"\n",(0,r.jsx)(n.p,{children:"Interactive Services are best suited for workflows that require persistent access or real-time interaction."}),"\n",(0,r.jsx)(n.h3,{id:"1-model-inference",children:"1. Model Inference"}),"\n",(0,r.jsx)(n.p,{children:"Interactive services are ideal for hosting models that need to stay online to serve requests."}),"\n",(0,r.jsxs)(n.ul,{children:["\n",(0,r.jsxs)(n.li,{children:[(0,r.jsx)(n.strong,{children:"Examples:"})," Running inference servers like ",(0,r.jsx)(n.strong,{children:"Ollama"})," or ",(0,r.jsx)(n.strong,{children:"vLLM"})," to host a model in the cloud and query it via API."]}),"\n"]}),"\n",(0,r.jsx)(n.h3,{id:"2-exploratory-research--development",children:"2. Exploratory Research & Development"}),"\n",(0,r.jsx)(n.p,{children:"In the early stages of research, you often need an environment to experiment, debug, and iterate quickly without waiting for a queue."}),"\n",(0,r.jsxs)(n.ul,{children:["\n",(0,r.jsxs)(n.li,{children:[(0,r.jsx)(n.strong,{children:"Tools:"})," Gain direct access to the compute resources via ",(0,r.jsx)(n.strong,{children:"VSCode"}),", ",(0,r.jsx)(n.strong,{children:"Jupyter Notebooks"}),", or ",(0,r.jsx)(n.strong,{children:"SSH"}),"."]}),"\n",(0,r.jsxs)(n.li,{children:[(0,r.jsx)(n.strong,{children:"Workflow:"}
1)," Use an Interactive Service to prototype your code. Once your script is finalized and stable, you can convert it into a ",(0,r.jsx)(n.strong,{children:"Task"})," to run large-scale training jobs efficiently."]}),"\n"]}),"\n",(0,r.jsx)(n.h2,{id:"prerequisites",children:"Prerequisites"}),"\n",(0,r.jsx)(n.p,{children:"Before running an interactive service, ensure you have a Compute Provider set up and active."}),"\n",(0,r.jsxs)(n.ol,{children:["\n",(0,r.jsxs)(n.li,{children:["\n",(0,r.jsxs)(n.p,{children:["Navigate to Team Settings and set up a Compute Provider.\n",(0,r.jsx)(n.img,{alt:"Compute Provider Setup",src:s(76643).A+"",width:"851",height:"640"})]}),"\n"]}),"\n",(0,r.jsxs)(n.li,{children:["\n",(0,r.jsxs)(n.p,{children:["Make sure the provider is active by clicking on the health button.\n",(0,r.jsx)(n.img,{alt:"Provider Health Check",src:s(33717).A+"",width:"2317",height:"323"})]}),"\n"]}),"\n"]}),"\n",(0,r.jsx)(n.h2,{id:"steps-to-run-an-interactive-service",children:"Steps to Run an Interactive Service"}),"\n",(0,r.jsxs)(n.ol,{children:["\n",(0,r.jsxs)(n.li,{children:["\n",(0,r.jsx)(n.p,{children:"Go to the Interact page in Transformer Lab."}),"\n"]}),"\n",(0,r.jsxs)(n.li,{children:["\n",(0,r.jsxs)(n.p,{children:['Click on the "New" button to create a new interactive service.\n',(0,r.jsx)(n.img,{alt:"Interact Page New Button",src:s(47167).A+"",width:"2366",height:"679"})]}),"\n"]}),"\n",(0,r.jsxs)(n.li,{children:["\n",(0,r.jsxs)(n.p,{children:["Select the type of interactive service you want to launch: VSCode, Jupyter Notebook, SSH, vLLM Server, or Ollama Server.\n",(0,r.jsx)(n.img,{alt:"Select Service Type",src:s(30739).A+"",width:"2073",height:"648"})]}),"\n"]}),"\n",(0,r.jsxs)(n.li,{children:["\n",(0,r.jsx)(n.p,{children:"Configure the service:"}),"\n",(0,r.jsxs)(n.ul,{children:["\n",(0,r.jsx)(n.li,{children:"Enter a name for the service."}),"\n",(0,r.jsx)(n.li,{children:"Select the Compute Provider to use."}),"\n",(0,r.jsx)(n.li,{children:"Specify the resources: CPU, memory, and GPUs."}),"\n",(0,r.jsxs)(n.li,{children:["For certain services, provide additional inputs such as model_name for vLLM Server or ngrok auth token for services that launch a tunnel.\n",(0,r.jsx)(n.img,{alt:"Service Configuration",src:s(97440).A+"",width:"1709",height:"749"})]}),"\n"]}),"\n"]}),"\n",(0,r.jsxs)(n.li,{children:["\n",(0,r.jsx)(n.p,{children:'Click "Launch" to start the service.'}),"\n"]}),"\n",(0,r.jsxs)(n.li,{children:["\n",(0,r.jsxs)(n.p,{children:['Once launched, a card will appear for the service. Click the "Interactive Setup" button on the card.\n',(0,r.jsx)(n.img,{alt:"Service Card Interactive Setup",src:s(70309).A+"",width:"1374",height:"817"})]}),"\n"]}),"\n",(0,r.jsxs)(n.li,{children:["\n",(0,r.jsxs)(n.p,{children:["Follow the provided URL or steps to access the service.\n",(0,r.jsx)(n.img,{alt:"Access Service",src:s(50202).A+"",width:"2064",height:"1165"})]}),"\n"]}),"\n"]})]})}function h(e={}){const{wrapper:n}={...(0,i.R)(),...e.components};return n?(0,r.jsx)(n,{...e,children:(0,r.jsx)(d,{...e})}):d(e)}},70309:(e,n,s)=>{s.d(n,{A:()=>t});const t=s.p+"assets/images/service-card-interactive-setup-860cca6c99a38846ea01dc0f85ffaa01.png"},76643:(e,n,s)=>{s.d(n,{A:()=>t});const t=s.p+"assets/images/screenshot-addprovider-e1b599d63a798a35b27bab4da3847567.png"},97440:(e,n,s)=>{s.d(n,{A:()=>t});const t=s.p+"assets/images/service-configuration-e51f225e3222734593643f530c3c12d6.png"}}]);
Line numbers count LF bytes from the start of the resource, as the search results do. Vendor segments are library code the classifier recognised; they are stored but not indexed. Bytes are shown as Latin1 characters, one per byte.