Adding small benchmark script. (#881)

# What does this PR do?

<!--
Congratulations! You've made it this far! You're not quite done yet
though.

Once merged, your PR is going to appear in the release notes with the
title you set, so make sure it's a great title that fully reflects the
extent of your awesome contribution.

Then, please replace this with a description of the change and which
issue is fixed (if applicable). Please also include relevant motivation
and context. List any dependencies (if any) that are required for this
change.

Once you're done, someone will review your PR shortly (see the section
"Who can review?" below to tag some potential reviewers). They may
suggest changes to make the code even better. If no one reviewed your PR
after a week has passed, don't hesitate to post a new comment
@-mentioning the same persons---sometimes notifications get lost.
-->

<!-- Remove if not applicable -->

Fixes # (issue)


## Before submitting
- [ ] This PR fixes a typo or...

load_tests/common.js

+import { check, randomSeed } from 'k6';
+import http from 'k6/http';
+import { Trend, Counter } from 'k6/metrics';
+import { randomItem } from 'https://jslib.k6.io/k6-utils/1.2.0/index.js';
+
+const seed = 0;
+
+const host = __ENV.HOST || '127.0.0.1:8000';
+const timePerToken = new Trend('time_per_token', true);
+const throughput = new Counter('tokens_per_s');
+
+randomSeed(seed);
+// const shareGPT = JSON.parse(open("ShareGPT_V3_unfiltered_cleaned_split.json"))
+const shareGPT = JSON.parse(open("small.json"))
+
+
+export function get_options(reference_latency_ms){
+    return {
+        thresholds: {
+            http_req_failed: ['rate==0'],
+            time_per_token: [{
+                threshold: `p(50)<${3 * reference_latency_ms}`,
+                abortOnFail: true,
+                delayAbortEval: '10s'
+            }],
+        },
+        scenarios: {
+            load_test: {
+                executor: 'constant-arrival-rate',
+                duration: '60s',
+                preAllocatedVUs: 100,
+                rate: 10,
+                timeUnit: '1s',
+            },
+        },
+    };
+}
+
+
+export function run(host, generate_payload, max_new_tokens) {
+    const headers = {'Content-Type': 'application/json'};
+    const query = randomItem(shareGPT);
+    const payload = JSON.stringify(generate_payload(query));
+    const res = http.post(`http://${host}/generate`, payload, {
+        headers,
+    });
+    if(res.status >= 400 && res.status < 500){
+        return;
+    }
+
+    check(res, {
+        'Post status is 200': (r) => res.status === 200,
+    });
+    const n_tokens = max_new_tokens;
+    const timings = res.timings.duration;
+
+    if (res.status === 200) {
+        const latency_ms_per_token = timings / n_tokens;
+        timePerToken.add(latency_ms_per_token);
+        const latency_in_s = latency_ms_per_token / 1000;
+        const individual_throughput = 1 / latency_in_s;
+        throughput.add(individual_throughput);
+    }
+}

load_tests/tgi.js

+import { get_options, run } from "./common.js";
+ 
+const reference_latency_ms = 30;
+const host = __ENV.HOST || '127.0.0.1:8000';
+const max_new_tokens = 50;
+
+
+function generate_payload(gpt){
+    const input = gpt["conversations"][0]["value"];
+    return {"inputs": input, "parameters": {"max_new_tokens": max_new_tokens, "temperature" : 0.5}}
+}
+
+export const options = get_options(reference_latency_ms);
+
+export default function(){
+    run(host, generate_payload, max_new_tokens);
+}

load_tests/vllm.js

+import { get_options, run } from "./common.js";
+ 
+const reference_latency_ms = 22;
+const host = __ENV.HOST || '127.0.0.1:8000';
+const max_new_tokens = 50;
+
+
+function generate_payload(gpt){
+    const input = gpt["conversations"][0]["value"];
+    return {"prompt": input, "temperature": 0.5, "ignore_eos": true}
+}
+
+export const options = get_options(reference_latency_ms);
+
+export default function(){
+    run(host, generate_payload, max_new_tokens);
+}