Adds chatbot example for llama.cpp engine (#413)

gradle.properties

@@ -11,7 +11,7 @@ systemProp.org.gradle.internal.http.connectionTimeout=60000
 # FIXME: Workaround gradle publish issue: https://github.com/gradle/gradle/issues/11308
 systemProp.org.gradle.internal.publish.checksums.insecure=true
 
-djl_version=0.25.0
+djl_version=0.26.0
 commons_cli_version=1.5.0
 log4j_slf4j_version=2.19.0
 slf4j_simple_version=1.7.36

huggingface/nlp/build.gradle

@@ -16,16 +16,18 @@ dependencies {
     implementation platform("ai.djl:bom:${djl_version}")
     implementation "ai.djl:api"
     implementation "ai.djl.huggingface:tokenizers"
+    implementation "ai.djl.llama:llama"
 
     runtimeOnly "ai.djl.pytorch:pytorch-engine"
     implementation "org.slf4j:slf4j-simple:${slf4j_simple_version}"
 }
 
 application {
-    mainClass = "com.examples.QuestionAnswering"
+    mainClass = System.getProperty("main", "com.examples.QuestionAnswering")
 }
 
 run {
+    standardInput = System.in
     systemProperties System.getProperties()
     systemProperties.remove("user.dir")
     systemProperty("file.encoding", "UTF-8")

huggingface/nlp/src/main/java/com/examples/Chatbot.java

@@ -0,0 +1,102 @@
+/*
+ * Copyright 2024 Amazon.com, Inc. or its affiliates. All Rights Reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License"). You may not use this file except in compliance
+ * with the License. A copy of the License is located at
+ *
+ * http://aws.amazon.com/apache2.0/
+ *
+ * or in the "license" file accompanying this file. This file is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES
+ * OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions
+ * and limitations under the License.
+ */
+package com.examples;
+
+import ai.djl.ModelException;
+import ai.djl.inference.Predictor;
+import ai.djl.llama.engine.LlamaInput;
+import ai.djl.llama.engine.LlamaTranslatorFactory;
+import ai.djl.llama.jni.Token;
+import ai.djl.llama.jni.TokenIterator;
+import ai.djl.repository.zoo.Criteria;
+import ai.djl.repository.zoo.ZooModel;
+import ai.djl.training.util.ProgressBar;
+import ai.djl.translate.TranslateException;
+
+import java.io.BufferedReader;
+import java.io.IOException;
+import java.io.InputStreamReader;
+import java.nio.charset.StandardCharsets;
+import java.util.Locale;
+import java.util.Set;
+
+public class Chatbot {
+
+    public static void main(String[] args) throws ModelException, IOException, TranslateException {
+        String modelId;
+        String quantMethod;
+        if (args.length > 0) {
+            modelId = args[0];
+            if (args.length > 1) {
+                quantMethod = args[1];
+            } else {
+                quantMethod = "Q4_K_M";
+            }
+        } else {
+            // modelId = "TheBloke/Mistral-7B-Instruct-v0.2-GGUF";
+            // quantMethod = "Q4_K_M";
+            modelId = "TinyLlama/TinyLlama-1.1B-Chat-v0.6";
+            quantMethod = "Q4_0";
+        }
+        System.out.println("Using model: " + modelId);
+
+        String url = "djl://ai.djl.huggingface.gguf/" + modelId + "/0.0.1/" + quantMethod;
+        Criteria<LlamaInput, TokenIterator> criteria =
+                Criteria.builder()
+                        .setTypes(LlamaInput.class, TokenIterator.class)
+                        .optModelUrls(url)
+                        .optEngine("Llama")
+                        .optOption("number_gpu_layers", "43")
+                        .optTranslatorFactory(new LlamaTranslatorFactory())
+                        .optProgress(new ProgressBar())
+                        .build();
+
+        String system =
+                "This is demo for DJL Llama.cpp engine.\n\n"
+                        + "Llama: Hello.  How may I help you today?";
+
+        LlamaInput.Parameters param = new LlamaInput.Parameters();
+        param.setTemperature(0.7f);
+        param.setPenalizeNl(true);
+        param.setMirostat(2);
+        param.setAntiPrompt(new String[] {"User: "});
+
+        LlamaInput in = new LlamaInput();
+        in.setParameters(param);
+
+        BufferedReader reader =
+                new BufferedReader(new InputStreamReader(System.in, StandardCharsets.UTF_8));
+        try (ZooModel<LlamaInput, TokenIterator> model = criteria.loadModel();
+                Predictor<LlamaInput, TokenIterator> predictor = model.newPredictor()) {
+            System.out.print(system);
+            StringBuilder prompt = new StringBuilder(system);
+            Set<String> exitWords = Set.of("exit", "bye", "quit");
+            while (true) {
+                System.out.print("\nUser: ");
+                String input = reader.readLine().trim();
+                if (exitWords.contains(input.toLowerCase(Locale.ROOT))) {
+                    break;
+                }
+                System.out.print("Llama: ");
+                prompt.append("\nUser: ").append(input).append("\nLlama: ");
+                in.setInputs(prompt.toString());
+                TokenIterator it = predictor.predict(in);
+                while (it.hasNext()) {
+                    Token token = it.next();
+                    System.out.print(token.getText());
+                    prompt.append(token.getText());
+                }
+            }
+        }
+    }
+}