Cloud Run ジョブを使用して LLM を微調整する方法

Cloud Run ジョブを使用して LLM を微調整する方法

この Codelab について

subject最終更新: 6月 3, 2025
account_circleGoogle 社員により作成

1. はじめに

この Codelab では、Cloud Run ジョブを使用して Gemma モデルをファインチューニングし、vLLM を使用して Cloud Run で結果を提供します。

この Codelab では、テキストから SQL へのデータセットを使用します。これは、自然言語で質問されたときに LLM が SQL クエリで返答するようにするものです。

  • Cloud Run Jobs GPU を使用してファインチューニングを行う方法
  • vLLM で Cloud Run を使用してモデルを提供する方法
  • GPU Job にダイレクト VPC 構成を使用して、モデルのアップロードとサービングを高速化する方法

2. 始める前に

API を有効にする

この Codelab の使用を開始する前に、次の API を有効にします。

gcloud services enable run.googleapis.com \
    compute.googleapis.com \
    run.googleapis.com \
    cloudbuild.googleapis.com \
    secretmanager.googleapis.com \
    artifactregistry.googleapis.com

GPU 割り当て

サポートされているリージョンの割り当ての増加をリクエストする。割り当ては nvidia_l4_gpu_allocation_no_zonal_redundancy で、Cloud Run Admin API にあります。

注: 新しいプロジェクトを使用している場合、API を有効にしてこのページに割り当てが表示されるまでに数分かかることがあります。

Hugging Face

この Codelab では、Hugging Face でホストされているモデルを使用します。このモデルを取得するには、[読み取り] 権限を持つ Hugging Face ユーザー アクセス トークンをリクエストします。後で YOUR_HF_TOKEN として参照します。

また、モデルを使用するには、使用規約に同意する必要があります(https://huggingface.co/google/gemma-2b)。

3. 設定と要件

次のリソースを設定します。

  • IAM サービス アカウントと関連する IAM 権限
  • Hugging Face トークンを保存する Secret Manager シークレット
  • ファインチューニングされたモデルを保存する Cloud Storage バケット。
  • モデルのファインチューニング用にビルドするイメージを保存する Artifact Registry リポジトリ。
  1. この Codelab の環境変数を設定します。いくつかの変数があらかじめ入力されています。プロジェクト ID、リージョン、Hugging Face トークンを指定します。
    export PROJECT_ID=<YOUR_PROJECT_ID>
    export REGION=<YOUR_REGION>
    export HF_TOKEN=<YOUR_HF_TOKEN>

    export AR_REPO=codelab-finetuning-jobs
    export IMAGE_NAME=finetune-to-gcs
    export JOB_NAME=finetuning-to-gcs-job
    export BUCKET_NAME=$PROJECT_ID-codelab-finetuning-jobs
    export SECRET_ID=HF_TOKEN
    export SERVICE_ACCOUNT="finetune-job-sa"
    export SERVICE_ACCOUNT_ADDRESS=$SERVICE_ACCOUNT@$PROJECT_ID.iam.gserviceaccount.com
  2. 次のコマンドを実行して、サービス アカウントを作成します。
    gcloud iam service-accounts create $SERVICE_ACCOUNT \
     
    --display-name="Service account for fine-tuning codelab"
  3. Secret Manager を使用して Hugging Face アクセス トークンを保存します。
    gcloud secrets create $SECRET_ID \
         
    --replication-policy="automatic"

    printf $HF_TOKEN
    | gcloud secrets versions add $SECRET_ID --data-file=-
  4. サービス アカウントに Secret Manager のシークレット アクセサーのロールを付与します。
    gcloud secrets add-iam-policy-binding $SECRET_ID \
     
    --member serviceAccount:$SERVICE_ACCOUNT_ADDRESS \
     
    --role='roles/secretmanager.secretAccessor'
  5. ファインチューニングされたモデルをホストするバケットを作成します。
    gcloud storage buckets create -l $REGION gs://$BUCKET_NAME
  6. サービス アカウントにバケットへのアクセス権を付与します。
    gcloud storage buckets add-iam-policy-binding gs://$BUCKET_NAME \
     
    --member=serviceAccount:$SERVICE_ACCOUNT_ADDRESS \
     
    --role=roles/storage.objectAdmin
  7. コンテナ イメージを保存する Artifact Registry リポジトリを作成します。
    gcloud artifacts repositories create $AR_REPO \
       
    --repository-format=docker \
       
    --location=$REGION \
       
    --description="codelab for finetuning using CR jobs" \
       
    --project=$PROJECT_ID

4. Cloud Run ジョブイメージを作成する

次のステップでは、次のことを実行するコードを作成します。

  • Hugging Face から Gemma モデルをインポートする
  • Hugging Face のデータセットを使用してモデルのファインチューニングを行います。ジョブは、ファインチューニングに単一の L4 GPU を使用します。
  • new_model というファインチューニング済みモデルを Cloud Storage バケットにアップロードします。
  1. ファインチューニング ジョブコードのディレクトリを作成します。
    mkdir codelab-finetuning-job
    cd codelab
    -finetuning-job
  2. finetune.py というファイルを作成します。
    # Copyright 2024 Google LLC
    #
    # 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 os
    import torch
    from datasets import load_dataset
    from transformers import (
       
    AutoModelForCausalLM,
       
    AutoTokenizer,
       
    BitsAndBytesConfig,
       
    TrainingArguments,

    )
    from peft import LoraConfig, PeftModel

    from trl import SFTTrainer

    # Cloud Storage bucket to upload the model
    bucket_name = os.getenv("BUCKET_NAME", "YOUR_BUCKET_NAME")

    # The model that you want to train from the Hugging Face hub
    model_name = os.getenv("MODEL_NAME", "google/gemma-2b")

    # The instruction dataset to use
    dataset_name = "b-mc2/sql-create-context"

    # Fine-tuned model name
    new_model = os.getenv("NEW_MODEL", "gemma-2b-sql")

    ################################################################################
    # QLoRA parameters
    ################################################################################

    # LoRA attention dimension
    lora_r = int(os.getenv("LORA_R", "4"))

    # Alpha parameter for LoRA scaling
    lora_alpha = int(os.getenv("LORA_ALPHA", "8"))

    # Dropout probability for LoRA layers
    lora_dropout = 0.1

    ################################################################################
    # bitsandbytes parameters
    ################################################################################

    # Activate 4-bit precision base model loading
    use_4bit = True

    # Compute dtype for 4-bit base models
    bnb_4bit_compute_dtype = "float16"

    # Quantization type (fp4 or nf4)
    bnb_4bit_quant_type = "nf4"

    # Activate nested quantization for 4-bit base models (double quantization)
    use_nested_quant = False

    ################################################################################
    # TrainingArguments parameters
    ################################################################################

    # Output directory where the model predictions and checkpoints will be stored
    output_dir = "./results"

    # Number of training epochs
    num_train_epochs = 1

    # Enable fp16/bf16 training (set bf16 to True with an A100)
    fp16 = True
    bf16 = False

    # Batch size per GPU for training
    per_device_train_batch_size = int(os.getenv("TRAIN_BATCH_SIZE", "1"))

    # Batch size per GPU for evaluation
    per_device_eval_batch_size = int(os.getenv("EVAL_BATCH_SIZE", "2"))

    # Number of update steps to accumulate the gradients for
    gradient_accumulation_steps = int(os.getenv("GRADIENT_ACCUMULATION_STEPS", "1"))

    # Enable gradient checkpointing
    gradient_checkpointing = True

    # Maximum gradient normal (gradient clipping)
    max_grad_norm = 0.3

    # Initial learning rate (AdamW optimizer)
    learning_rate = 2e-4

    # Weight decay to apply to all layers except bias/LayerNorm weights
    weight_decay = 0.001

    # Optimizer to use
    optim = "paged_adamw_32bit"

    # Learning rate schedule
    lr_scheduler_type = "cosine"

    # Number of training steps (overrides num_train_epochs)
    max_steps = -1

    # Ratio of steps for a linear warmup (from 0 to learning rate)
    warmup_ratio = 0.03

    # Group sequences into batches with same length
    # Saves memory and speeds up training considerably
    group_by_length = True

    # Save checkpoint every X updates steps
    save_steps = 0

    # Log every X updates steps
    logging_steps = int(os.getenv("LOGGING_STEPS", "50"))

    ################################################################################
    # SFT parameters
    ################################################################################

    # Maximum sequence length to use
    max_seq_length = int(os.getenv("MAX_SEQ_LENGTH", "512"))

    # Pack multiple short examples in the same input sequence to increase efficiency
    packing = False

    # Load the entire model on the GPU 0
    device_map = {'':torch.cuda.current_device()}

    # Set limit to a positive number
    limit = int(os.getenv("DATASET_LIMIT", "5000"))

    dataset = load_dataset(dataset_name, split="train")
    if limit != -1:
       
    dataset = dataset.shuffle(seed=42).select(range(limit))


    def transform(data):
       
    question = data['question']
       
    context = data['context']
       
    answer = data['answer']
       
    template = "Question: {question}\nContext: {context}\nAnswer: {answer}"
       
    return {'text': template.format(question=question, context=context, answer=answer)}


    transformed = dataset.map(transform)

    # Load tokenizer and model with QLoRA configuration
    compute_dtype = getattr(torch, bnb_4bit_compute_dtype)

    bnb_config = BitsAndBytesConfig(
       
    load_in_4bit=use_4bit,
       
    bnb_4bit_quant_type=bnb_4bit_quant_type,
       
    bnb_4bit_compute_dtype=compute_dtype,
       
    bnb_4bit_use_double_quant=use_nested_quant,
    )

    # Check GPU compatibility with bfloat16
    if compute_dtype == torch.float16 and use_4bit:
       
    major, _ = torch.cuda.get_device_capability()
       
    if major >= 8:
           
    print("=" * 80)
           
    print("Your GPU supports bfloat16")
           
    print("=" * 80)

    # Load base model
    model = AutoModelForCausalLM.from_pretrained(
       
    model_name,
       
    quantization_config=bnb_config,
       
    device_map=device_map,
       
    torch_dtype=torch.float16,
    )
    model.config.use_cache = False
    model.config.pretraining_tp = 1

    # Load LLaMA tokenizer
    tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
    tokenizer.pad_token = tokenizer.eos_token
    tokenizer.padding_side = "right"

    # Load LoRA configuration
    peft_config = LoraConfig(
       
    lora_alpha=lora_alpha,
       
    lora_dropout=lora_dropout,
       
    r=lora_r,
       
    bias="none",
       
    task_type="CAUSAL_LM",
       
    target_modules=["q_proj", "v_proj"]
    )

    # Set training parameters
    training_arguments = TrainingArguments(
       
    output_dir=output_dir,
       
    num_train_epochs=num_train_epochs,
       
    per_device_train_batch_size=per_device_train_batch_size,
       
    gradient_accumulation_steps=gradient_accumulation_steps,
       
    optim=optim,
       
    save_steps=save_steps,
       
    logging_steps=logging_steps,
       
    learning_rate=learning_rate,
       
    weight_decay=weight_decay,
       
    fp16=fp16,
       
    bf16=bf16,
       
    max_grad_norm=max_grad_norm,
       
    max_steps=max_steps,
       
    warmup_ratio=warmup_ratio,
       
    group_by_length=group_by_length,
       
    lr_scheduler_type=lr_scheduler_type,
    )

    trainer = SFTTrainer(
       
    model=model,
       
    train_dataset=transformed,
       
    peft_config=peft_config,
       
    dataset_text_field="text",
       
    max_seq_length=max_seq_length,
       
    tokenizer=tokenizer,
       
    args=training_arguments,
       
    packing=packing,
    )

    trainer.train()

    trainer.model.save_pretrained(new_model)

    # Reload model in FP16 and merge it with LoRA weights
    base_model = AutoModelForCausalLM.from_pretrained(
       
    model_name,
       
    low_cpu_mem_usage=True,
       
    return_dict=True,
       
    torch_dtype=torch.float16,
       
    device_map=device_map,
    )
    model = PeftModel.from_pretrained(base_model, new_model)
    model = model.merge_and_unload()

    # push to Cloud Storage

    file_path_to_save_the_model = '/finetune/new_model'
    model.save_pretrained(file_path_to_save_the_model)
    tokenizer.save_pretrained(file_path_to_save_the_model)
  3. requirements.txt ファイルを作成します。
    accelerate==0.34.2
    bitsandbytes
    ==0.45.5
    datasets
    ==2.19.1
    transformers
    ==4.51.3
    peft
    ==0.11.1
    trl
    ==0.8.6
    torch
    ==2.3.0
  4. Dockerfile を作成します。
    FROM nvidia/cuda:12.6.2-runtime-ubuntu22.04

    RUN apt-get update && \
        apt-get -y --no-install-recommends install python3-dev gcc python3-pip git && \
        rm -rf /var/lib/apt/lists/*

    COPY requirements.txt /requirements.txt

    RUN pip3 install -r requirements.txt --no-cache-dir

    COPY finetune.py /finetune.py

    ENV PYTHONUNBUFFERED 1

    CMD python3 /finetune.py --device cuda
  5. Artifact Registry リポジトリにコンテナをビルドします。
    gcloud builds submit \
      --tag $REGION-docker.pkg.dev/$PROJECT_ID/$AR_REPO/$IMAGE_NAME \
      --region $REGION

5. ジョブをデプロイして実行する

このステップでは、Google Cloud Storage へのアップロードを高速化するために、ダイレクト VPC 下り(外向き)を使用するジョブの YAML 構成を作成します。

このファイルには、後で更新する変数が含まれています。

  1. finetune-job.yaml.tmpl というファイルを作成します。
    apiVersion: run.googleapis.com/v1
    kind: Job
    metadata:
      name: $JOB_NAME
      labels:
        cloud.googleapis.com/location: $REGION
      annotations:
        run.googleapis.com/launch-stage: ALPHA
    spec:
      template:
        metadata:
          annotations:
            run.googleapis.com/execution-environment: gen2
            run.googleapis.com/network-interfaces: '[{"network":"default","subnetwork":"default"}]'
        spec:
          parallelism: 1
          taskCount: 1
          template:
            spec:
              serviceAccountName: $SERVICE_ACCOUNT_ADDRESS
              containers:
              - name: $IMAGE_NAME
                image: $REGION-docker.pkg.dev/$PROJECT_ID/$AR_REPO/$IMAGE_NAME
                env:
                - name: MODEL_NAME
                  value: "google/gemma-2b"
                - name: NEW_MODEL
                  value: "gemma-2b-sql-finetuned"
                - name: BUCKET_NAME
                  value: "$BUCKET_NAME"
                - name: LORA_R
                  value: "8"
                - name: LORA_ALPHA
                  value: "16"
                - name: GRADIENT_ACCUMULATION_STEPS
                  value: "2"
                - name: DATASET_LIMIT
                  value: "1000"
                - name: LOGGING_STEPS
                  value: "5"
                - name: HF_TOKEN
                  valueFrom:
                    secretKeyRef:
                      key: 'latest'
                      name: HF_TOKEN
                resources:
                  limits:
                    cpu: 8000m
                    nvidia.com/gpu: '1'
                    memory: 32Gi
                volumeMounts:
                - mountPath: /finetune/new_model
                  name: finetuned_model
              volumes:
              - name: finetuned_model
                csi:
                  driver: gcsfuse.run.googleapis.com
                  readOnly: false
                  volumeAttributes:
                    bucketName: $BUCKET_NAME
              maxRetries: 3
              timeoutSeconds: '3600'
              nodeSelector:
                run.googleapis.com/accelerator: nvidia-l4
  2. 次のコマンドを実行して、YAML の変数を環境変数に置き換えます。
    envsubst < finetune-job.yaml.tmpl > finetune-job.yaml
  3. Cloud Run ジョブを作成します。
    gcloud alpha run jobs replace finetune-job.yaml
  4. ジョブを実行します。
    gcloud alpha run jobs execute $JOB_NAME --region $REGION --async

ジョブが完了するまでに 10 分ほどかかります。最後のコマンドの出力に表示されたリンクを使用して、ステータスを確認できます。

6. Cloud Run サービスを使用して、vLLM でファインチューニングされたモデルを提供する

このステップでは、Cloud Run サービスをデプロイします。この構成では、直接 VPC を使用してプライベート ネットワーク経由で Cloud Storage バケットにアクセスし、ダウンロードを高速化します。

このファイルには、後で更新する変数が含まれています。

  1. service.yaml.tmpl ファイルを作成します。
    apiVersion: serving.knative.dev/v1
    kind: Service
    metadata:
      name: serve-gemma-sql
      labels:
        cloud.googleapis.com/location: $REGION
      annotations:
        run.googleapis.com/launch-stage: BETA
        run.googleapis.com/ingress: all
        run.googleapis.com/ingress-status: all
    spec:
      template:
        metadata:
          labels:
          annotations:
            autoscaling.knative.dev/maxScale: '1'
            run.googleapis.com/cpu-throttling: 'false'
            run.googleapis.com/gpu-zonal-redundancy-disabled: 'true'
            run.googleapis.com/network-interfaces: '[{"network":"default","subnetwork":"default"}]'
        spec:
          containers:
          - name: serve-finetuned
            image: us-docker.pkg.dev/vertex-ai/vertex-vision-model-garden-dockers/pytorch-vllm-serve:20250505_0916_RC00
            ports:
            - name: http1
              containerPort: 8000
            resources:
              limits:
                cpu: 8000m
                nvidia.com/gpu: '1'
                memory: 32Gi
            volumeMounts:
            - name: fuse
              mountPath: /finetune/new_model
            command: ["python3", "-m", "vllm.entrypoints.api_server"]
            args:
            - --model=/finetune/new_model
            - --tensor-parallel-size=1
            env:
            - name: MODEL_ID
              value: 'new_model'
            - name: HF_HUB_OFFLINE
              value: '1'
          volumes:
          - name: fuse
            csi:
              driver: gcsfuse.run.googleapis.com
              volumeAttributes:
                bucketName: $BUCKET_NAME
          nodeSelector:
            run.googleapis.com/accelerator: nvidia-l4
  2. service.yaml ファイルをバケット名で更新します。
    envsubst < service.yaml.tmpl > service.yaml
  3. Cloud Run Service をデプロイします。
    gcloud alpha run services replace service.yaml

7. ファインチューニングされたモデルをテストする

このステップでは、ファインチューニングをテストするようにモデルに指示します。

  1. Cloud Run サービスのサービス URL を取得します。
    SERVICE_URL=$(gcloud run services describe serve-gemma-sql --platform managed --region $REGION --format 'value(status.url)')
  2. モデルのプロンプトを作成します。
    USER_PROMPT="Question: What are the first name and last name of all candidates? Context: CREATE TABLE candidates (candidate_id VARCHAR); CREATE TABLE people (first_name VARCHAR, last_name VARCHAR, person_id VARCHAR)"
  3. CURL を使用してサービスを呼び出し、モデルにプロンプトを表示します。
    curl -X POST $SERVICE_URL/generate \
      -H "Content-Type: application/json" \
      -H "Authorization: bearer $(gcloud auth print-identity-token)" \
      -d @- <<EOF
    {
        "prompt": "${USER_PROMPT}"
    }
    EOF

次のようなレスポンスが表示されます。

{"predictions":["Prompt:\nQuestion: What are the first name and last name of all candidates? Context: CREATE TABLE candidates (candidate_id VARCHAR); CREATE TABLE people (first_name VARCHAR, last_name VARCHAR, person_id VARCHAR)\nOutput:\n CREATE TABLE people_to_candidates (candidate_id VARCHAR, person_id VARCHAR) CREATE TABLE people_to_people (person_id VARCHAR, person_id VARCHAR) CREATE TABLE people_to_people_to_candidates (person_id VARCHAR, candidate_id"]}

8. お疲れさまでした

以上で、この Codelab は完了です。

Cloud Run のドキュメントを確認することをおすすめします。

学習した内容

  • Cloud Run Jobs GPU を使用してファインチューニングを行う方法
  • vLLM で Cloud Run を使用してモデルを提供する方法
  • GPU Job にダイレクト VPC 構成を使用して、モデルのアップロードとサービングを高速化する方法

9. クリーンアップ

誤って課金されないようにするには、たとえば、Cloud Run サービスが無料 tier の Cloud Run の呼び出しの月間割り当てを超える回数を誤って呼び出された場合は、手順 6 で作成した Cloud Run サービスを削除します。

Cloud Run サービスを削除するには、Cloud Run Cloud コンソール(https://console.cloud.google.com/run)に移動し、serve-gemma-sql サービスを削除します。

プロジェクト全体を削除するには、[リソースの管理] に移動し、ステップ 2 で作成したプロジェクトを選択して、[削除] を選択します。プロジェクトを削除する場合は、Cloud SDK でプロジェクトを変更する必要があります。gcloud projects list を実行すると、使用可能なすべてのプロジェクトのリストが表示されます。