Add cnn_dailymail processing for summarization, and offline/online run config yamls

examples/05_Llama3.1-8B_Example/README.md

@@ -0,0 +1,58 @@
+# Running Endpoints with [Llama3.1-8B](https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct)
+
+## Download dataset
+
+The Llama3.1-8B benchmark uses the [cnn/dailymail](https://huggingface.co/datasets/abisee/cnn_dailymail) dataset (for summarization). Download, modify the input prompt and save it using the following command:
+
+```
+python download_cnndm.py --save-dir data --split validation
+# Processed data will be saved at data/cnn_dailymail_validation.json
+```
+
+- To generate calibration dataset, users can use the [cnn-dailymail-calibration-list](https://github.com/mlcommons/inference/blob/master/calibration/CNNDailyMail/calibration-list.txt)
+
+```
+curl -OL https://raw.githubusercontent.com/mlcommons/inference/v4.0/calibration/CNNDailyMail/calibration-list.txt
+python download_cnndm.py --save-dir data --calibration-ids-file calibration-list.txt --split train
+```
+
+## Launch the server
+
+The following environment variables are used by the commands below to make the scripts easier to run
+
+```
+export HF_TOKEN=<your Hugging Face token>
+export HF_HOME=<Path to your hf_home, usually /USERNAME/.cache/huggingface>
+export MODEL_NAME=<model to run, for instance meta-llama/Llama-3.1-8B-Instruct>
+```
+
+It is convenient to download the model prior to launch so that the container can reuse the model instead of having to download it post-launch. This can be done via `hf download $MODEL_NAME`. The models downloaded can be verified via `hf cache scan`
+
+### [vLLM](https://github.com/vllm-project/vllm)
+
+We can launch the latest docker image for vllm using the command below:
+
+```
+docker run --runtime nvidia --gpus all -v ${HF_HOME}:/root/.cache/huggingface --env "HUGGING_FACE_HUB_TOKEN=$HF_TOKEN" -p 8000:8000 --ipc=host vllm/vllm-openai:latest  --model ${MODEL_NAME}
+
+```
+
+### To run Offline mode
+
+**Note** Double-check the config file for correct parameters
+
+- Launch the benchmark with config yaml
+
+```
+inference-endpoint benchmark from-config -c offline_llama3_8b_cnn.yaml --timeout 600
+```
+
+### To run Online mode
+
+**Note** Double-check the config file for correct parameters
+
+- Launch the benchmark with config yaml
+
+```
+inference-endpoint benchmark from-config -c online_llama3_8b_cnn.yaml --timeout 600
+```

examples/05_Llama3.1-8B_Example/download_cnndm.py

@@ -0,0 +1,110 @@
+# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License 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.
+
+import json
+import os
+import warnings
+from argparse import ArgumentParser
+
+from datasets import load_dataset
+from tqdm import tqdm
+
+PROMPT = "Summarize the following news article in 128 tokens. Please output the summary only, without any other text.\n\nArticle:\n{input}\n\nSummary:"
+
+
+def download_cnndm(
+    save_dir: str, split: str = "validation", calibration_ids_file: str = None
+) -> None:
+    """Download the CNN/DailyMail dataset and save it to the specified directory.
+
+    Args:
+        save_dir (str): The directory where the dataset will be saved.
+        split (str): The dataset split to download (default: validation).
+        calibration_ids_file (str): Path to a file containing calibration IDs (one per line).
+            If provided, 'split' must be 'train' and only examples with these IDs will be prepared and saved.
+    """
+    os.makedirs(save_dir, exist_ok=True)
+    dataset = load_dataset("cnn_dailymail", "3.0.0", split=split)
+
+    output_file_tag = split
+    calibration_ids = set()
+    if calibration_ids_file:
+        with open(calibration_ids_file, encoding="utf-8") as id_file:
+            for line in id_file:
+                calibration_ids.add(line.strip())
+        output_file_tag = "calibration"
+    fname = f"cnn_dailymail_{output_file_tag}.json"
+    output_file = os.path.join(save_dir, fname)
+
+    # Add the custom prompt to each example and filter if calibration IDs are provided
+    with open(output_file, "w", encoding="utf-8") as f:
+        for _, example in tqdm(
+            enumerate(dataset), total=len(dataset), desc="Processing examples"
+        ):
+            if calibration_ids and str(example["id"]) not in calibration_ids:
+                continue
+            f.write(
+                json.dumps(
+                    {
+                        "id": example["id"],
+                        "input": PROMPT.format(input=example["article"]),
+                        "highlights": example["highlights"],
+                    }
+                )
+                + "\n"
+            )
+    print(f"Dataset saved to {output_file}")
+
+
+if __name__ == "__main__":
+    parser = ArgumentParser(description="Download CNN/DailyMail dataset")
+    parser.add_argument(
+        "--save-dir",
+        type=str,
+        required=True,
+        help="Directory to save the downloaded dataset",
+    )
+    parser.add_argument(
+        "--split",
+        type=str,
+        default="validation",
+        help="Dataset split to download (default: validation)",
+    )
+    parser.add_argument(
+        "--calibration-ids-file",
+        type=str,
+        default=None,
+        help="Path to a file containing calibration IDs (one per line)."
+        " If provided, 'split' must be 'train' and only examples with these IDs will be saved.",
+    )
+
+    args = parser.parse_args()
+    if args.calibration_ids_file and args.split != "train":
+        warnings.warn(
+            "When --calibration-ids-file is provided, --split must be 'train'. Setting split to 'train'.",
+            stacklevel=2,
+        )
+        args.split = "train"
+
+    if args.calibration_ids_file and not os.path.isfile(args.calibration_ids_file):
+        raise FileNotFoundError(
+            f"Provided calibration IDs file not found: {args.calibration_ids_file}"
+        )
+
+    download_cnndm(
+        save_dir=args.save_dir,
+        split=args.split,
+        calibration_ids_file=args.calibration_ids_file,
+    )

examples/05_Llama3.1-8B_Example/offline_llama3_8b_cnn.yaml

@@ -0,0 +1,42 @@
+# Offline Throughput Benchmark
+name: "offline-llama3.1-8b-cnn-benchmark"
+version: "1.0"
+type: "offline"
+
+model_params:
+  name: "meta-llama/Llama-3.1-8B-Instruct" # Path to the model
+  temperature: 1.0
+  top_p: 1.0
+  max_new_tokens: 128
+
+datasets:
+  - name: "perf-test"
+    type: "performance"
+    path: "data/cnn_dailymail_validation.json" # Path to the dataset. Note: This should be the file generated with download_cnndm.py
+    samples: 13368 # Number of samples in the dataset (validation has ~13.3k samples, while training has ~287k samples)
+    parser:
+      prompt: "input"
+
+settings:
+  runtime:
+    min_duration_ms: 6000 # 6 seconds
+    max_duration_ms: 60000 # 1 minute
+    scheduler_random_seed: 137 # For Poisson/distribution sampling
+    dataloader_random_seed: 111 # For dataset shuffling
+
+  load_pattern:
+    type: "max_throughput"
+
+  client:
+    workers: 1 # Number of client workers
+
+metrics:
+  collect:
+    - "throughput"
+    - "latency"
+    - "ttft"
+    - "tpot"
+
+endpoint_config:
+  endpoint: "http://localhost:8000"
+  api_key: null

examples/05_Llama3.1-8B_Example/online_llama3_8b_cnn.yaml

@@ -0,0 +1,45 @@
+# Online Benchmark
+name: "online-llama3.1-8b-cnn-benchmark"
+version: "1.0"
+type: "online"
+
+model_params:
+  name: meta-llama/Llama-3.1-8B-Instruct # Path to the model
+  temperature: 1.0
+  top_p: 1.0
+  max_new_tokens: 128
+
+datasets:
+  - name: "perf-test"
+    type: "performance"
+    path: "data/cnn_dailymail_validation.json" # Path to the dataset. Note: This should be the file generated with download_cnndm.py
+    format: "json"
+    samples: 13368
+    parser:
+      prompt: "input"
+    eval_method: "rouge"
+
+settings:
+  runtime:
+    min_duration_ms: 6000 # 6 seconds
+    max_duration_ms: 60000 # 1 minute
+    scheduler_random_seed: 137 # For Poisson/distribution sampling
+    dataloader_random_seed: 111 # For dataset shuffling
+
+  load_pattern:
+    type: "poisson"
+    target_qps: 3 # This is the target queries per second for the Poisson load pattern (Legacy Loadgen)
+
+  client:
+    workers: 1 # Number of client workers
+
+metrics:
+  collect:
+    - "throughput"
+    - "latency"
+    - "ttft"
+    - "tpot"
+
+endpoint_config:
+  endpoint: "http://localhost:8000"
+  api_key: null