benchmark_docsum.yaml

# Copyright (C) 2025 Intel Corporation
# SPDX-License-Identifier: Apache-2.0

deploy:
  device: gaudi
  version: 1.3.0
  modelUseHostPath: /mnt/models
  HUGGINGFACEHUB_API_TOKEN: "" # mandatory
  node: [1]
  namespace: ""
  node_name: []
  timeout: 1000  # timeout in seconds for services to be ready, default 30 minutes
  interval: 5    # interval in seconds between service ready checks, default 5 seconds

  services:
    backend:
      resources:
        enabled: False
        cores_per_instance: "16"
        memory_capacity: "8000Mi"
      replicaCount: [1]

    llm:
      engine: vllm  # or tgi
      model_id: "meta-llama/Llama-3.2-3B-Instruct" # mandatory
      replicaCount: [1]
      resources:
        enabled: False
        cards_per_instance: 1
      model_params:
        vllm:  # VLLM specific parameters
          batch_params:
            enabled: True
            max_num_seqs: "8"    # Each value triggers an LLM service upgrade
          token_params:
            enabled: True
            max_input_length: ""
            max_total_tokens: ""
            max_batch_total_tokens: ""
            max_batch_prefill_tokens: ""
        tgi:   # TGI specific parameters
          batch_params:
            enabled: True
            max_batch_size: [1]  # Each value triggers an LLM service upgrade
          token_params:
            enabled: False
            max_input_length: "1280"
            max_total_tokens: "2048"
            max_batch_total_tokens: "65536"
            max_batch_prefill_tokens: "4096"

    docsum-ui:
      replicaCount: [1]

    whisper:
      replicaCount: [1]

    llm-uservice:
      model_id: "meta-llama/Llama-3.2-3B-Instruct" # mandatory
      replicaCount: [1]

    nginx:
      replicaCount: [1]

benchmark:
  # http request behavior related fields
  user_queries:              [16]
  concurrency:               [4]
  load_shape_type:           "constant" # "constant" or "poisson"
  poisson_arrival_rate:      1.0  # only used when load_shape_type is "poisson"
  warmup_iterations:         10
  seed:                      1024
  collect_service_metric:    True

  # workload, all of the test cases will run for benchmark
  bench_target: ["docsumfixed"] # specify the bench_target for benchmark
  dataset: "/home/sdp/pubmed_10.txt"  # specify the absolute path to the dataset file
  summary_type: "stuff"
  stream: True

  llm:
    # specify the llm output token size
    max_token_size:          [1024]