Fix chatCompletion parameters in e2e test (#1414)

follow-up after #1407
and #1409 cc
@coyotte508

Since we use `chatCompletion` and not `summarization` anymore, the
payload is not exactly the same. This PR removes
`payload.parameters.max_length` and replaces it by `payload.max_tokens`.
I've set a low value to avoid long responses (we just want a working
example and that's it).

Co-authored-by: SBrandeis <[email protected]>

Author: Wauplin

e2e/deno/index.ts

@@ -18,9 +18,7 @@ if (token) {
 	const result = await hf.chatCompletion({
 		model: "meta-llama/Llama-3.2-1B-Instruct",
 		messages: [{ role: "user", content: "Can you summarize the Eiffel Tower?" }],
-		parameters: {
-			max_length: 100,
-		},
+		max_tokens: 10,
 	});
 
 	console.log(result);

e2e/svelte/src/routes/+page.svelte

@@ -15,9 +15,7 @@
 		const result = await hf.chatCompletion({
 			model: "meta-llama/Llama-3.2-1B-Instruct",
 			messages: [{ role: "user", content: "Can you summarize the Eiffel Tower?" }],
-			parameters: {
-				max_length: 100,
-			},
+			max_tokens: 10,
 		});
 
 		console.log(result);

e2e/ts/src/index.ts

@@ -13,9 +13,7 @@ const hf = new InferenceClient(hfToken);
 		const result = await hf.chatCompletion({
 			model: "meta-llama/Llama-3.2-1B-Instruct",
 			messages: [{ role: "user", content: "Can you summarize the Eiffel Tower?" }],
-			parameters: {
-				max_length: 100,
-			},
+			max_tokens: 10,
 		});
 
 		console.log(result);