How to fine tune a LLM using Cloud Run Jobs
About this codelab
subjectLast updated Mar 21, 2025
1. Introduction
Overview
In this example, you will finetune a gemma-2b model with a text-to-sql dataset intended to make the LLM reply with a SQL query when asked a question in natural language. You will then take the finetuned model and serve it on Cloud Run using vLLM.
What you'll learn
- How to conduct fine tuning using Cloud Run Jobs GPU
- How to use Direct VPC configuration for a GPU Job for faster upload and serving of the model
2. Before you begin
To use the GPUs feature, you must request a quota increase for a supported region. The quota needed is nvidia_l4_gpu_allocation_no_zonal_redundancy, which is under Cloud Run Admin API. Here is the direct link to request quota.
3. Setup and Requirements
Set environment variables that will be used throughout this codelab.
PROJECT_ID=<YOUR_PROJECT_ID>
REGION=<YOUR_REGION>
HF_TOKEN=<YOUR_HF_TOKEN>
AR_REPO=codelab-finetuning-jobs
IMAGE_NAME=finetune-to-gcs
JOB_NAME=finetuning-to-gcs-job
BUCKET_NAME=$PROJECT_ID-codelab-finetuning-jobs
SECRET_ID=HF_TOKEN
SERVICE_ACCOUNT="finetune-job-sa"
SERVICE_ACCOUNT_ADDRESS=$SERVICE_ACCOUNT@$PROJECT_ID.iam.gserviceaccount.com
Create the service account by running this command:
gcloud iam service-accounts create $SERVICE_ACCOUNT \
--display-name="Cloud Run job to access HF_TOKEN Secret ID"
Use Secret Manager to store HuggingFace access token.
You can learn more about creating and using secrets in the Secret Manager docs.
gcloud secrets create $SECRET_ID \
--replication-policy="automatic"
printf $HF_TOKEN | gcloud secrets versions add $SECRET_ID --data-file=-
You'll see output similar to
you'll see output similar to
Created secret [HF_TOKEN].
Created version [1] of the secret [HF_TOKEN].
Grant your default compute service account the role of Secret Manager Secret Accessor
gcloud secrets add-iam-policy-binding $SECRET_ID \
--member serviceAccount:$SERVICE_ACCOUNT_ADDRESS \
--role='roles/secretmanager.secretAccessor'
Create a bucket that will host your finetuned model
gsutil mb -l $REGION gs://$BUCKET_NAME
And then give the SA access to the bucket.
gcloud storage buckets add-iam-policy-binding gs://$BUCKET_NAME \
--member=serviceAccount:$SERVICE_ACCOUNT_ADDRESS \
--role=roles/storage.objectAdmin
Create an artifact registry repository for the Job
gcloud artifacts repositories create $AR_REPO \
--repository-format=docker \
--location=$REGION \
--description="codelab for finetuning using CR jobs" \
--project=$PROJECT_ID
4. Create the Cloud Run job image
In the next step, you'll create the code that does the following:
- Imports the gemma-2b from huggingface
- Performs fine tuning on gemma-2b with text-to-sql dataset using dataset from huggingface. The Job uses single L4 GPU for finetuning.
- Uploads the fine tuned model called new_model to user's GCS bucket
Create a directory for your finetuning job code.
mkdir codelab-finetuning-job
cd codelab-finetuning-job
Create a file called finetune.py
# 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, Dataset
from transformers import (
AutoModelForCausalLM,
AutoTokenizer,
BitsAndBytesConfig,
TrainingArguments,
)
from peft import LoraConfig, PeftModel
from trl import SFTTrainer
from pathlib import Path
# GCS 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("google/gemma-7b")
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" # Fix weird overflow issue with fp16 training
# 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 HF
# model.push_to_hub(new_model, check_pr=True)
# tokenizer.push_to_hub(new_model, check_pr=True)
# push to GCS
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)
Create a requirements.txt file.
accelerate==0.30.1
bitsandbytes==0.43.1
datasets==2.19.1
transformers==4.41.0
peft==0.11.1
trl==0.8.6
torch==2.3.0
Create a Dockerfile
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
Build the container in your Artifact Registry repo
gcloud builds submit --tag $REGION-docker.pkg.dev/$PROJECT_ID/$AR_REPO/$IMAGE_NAME
5. Deploy and execute the job
In this step, you'll create the Jobs YAML config with direct VPC egress for faster uploads to Google Cloud Storage.
Note that this file contains variables that you will update in a subsequent step.
First, create a file called finetune-job.yaml
apiVersion: run.googleapis.com/v1
kind: Job
metadata:
name: finetuning-to-gcs-job
labels:
cloud.googleapis.com/location: us-central1
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: YOUR_SERVICE_ACCOUNT_NAME@YOUR_PROJECT_ID.iam.gserviceaccount.com
containers:
- name: finetune-to-gcs
image: YOUR_REGION-docker.pkg.dev/YOUR_PROJECT_ID/YOUR_AR_REPO/YOUR_IMAGE_NAME
env:
- name: MODEL_NAME
value: "google/gemma-2b"
- name: NEW_MODEL
value: "gemma-2b-sql-finetuned"
- name: LORA_R
value: "8"
- name: LORA_ALPHA
value: "16"
- name: TRAIN_BATCH_SIZE
value: "1"
- name: EVAL_BATCH_SIZE
value: "2"
- name: GRADIENT_ACCUMULATION_STEPS
value: "2"
- name: DATASET_LIMIT
value: "1000"
- name: MAX_SEQ_LENGTH
value: "512"
- 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: YOUR_PROJECT_ID-codelab-finetuning-jobs
maxRetries: 3
timeoutSeconds: '3600'
nodeSelector:
run.googleapis.com/accelerator: nvidia-l4
Now, replace the placeholders with your environment variables for the image by running the following command:
sed -i "s/YOUR_SERVICE_ACCOUNT_NAME/$SERVICE_ACCOUNT/; s/YOUR_PROJECT_ID/$PROJECT_ID/; s/YOUR_PROJECT_ID/$PROJECT_ID/; s/YOUR_REGION/$REGION/; s/YOUR_AR_REPO/$AR_REPO/; s/YOUR_IMAGE_NAME/$IMAGE_NAME/; s/YOUR_PROJECT_ID/$PROJECT_ID/" finetune-job.yaml
Next, create the Cloud Run Job
gcloud alpha run jobs replace finetune-job.yaml
And execute the job. This will take approx 10 minutes.
gcloud alpha run jobs execute $JOB_NAME --region $REGION
6. Use a Cloud Run service to serve your finetuned model with vLLM
Create a folder for the Cloud Run service code that will serve the fine-tuned model
cd ..
mkdir codelab-finetuning-service
cd codelab-finetuning-service
Create a service.yaml file
This configuration uses direct VPC to access GCS bucket over private network for faster downloads.
Note that this file contains variables that you will update in a subsequent step.
apiVersion: serving.knative.dev/v1
kind: Service
metadata:
name: serve-gemma2b-sql
labels:
cloud.googleapis.com/location: us-central1
annotations:
run.googleapis.com/launch-stage: BETA
run.googleapis.com/ingress: all
run.googleapis.com/ingress-status: all
spec:
template:
metadata:
labels:
annotations:
autoscaling.knative.dev/maxScale: '5'
run.googleapis.com/cpu-throttling: 'false'
run.googleapis.com/network-interfaces: '[{"network":"default","subnetwork":"default"}]'
spec:
containers:
- name: serve-finetuned
image: us-docker.pkg.dev/vertex-ai/vertex-vision-model-garden-dockers/pytorch-vllm-serve:20240220_0936_RC01
ports:
- name: http1
containerPort: 8000
resources:
limits:
cpu: 8000m
nvidia.com/gpu: '1'
memory: 32Gi
volumeMounts:
- name: fuse
mountPath: /finetune/new_model
command: ["python3", "-m", "vllm.entrypoints.api_server"]
args:
- --model=/finetune/new_model
- --tensor-parallel-size=1
env:
- name: MODEL_ID
value: 'new_model'
- name: HF_HUB_OFFLINE
value: '1'
volumes:
- name: fuse
csi:
driver: gcsfuse.run.googleapis.com
volumeAttributes:
bucketName: YOUR_BUCKET_NAME
nodeSelector:
run.googleapis.com/accelerator: nvidia-l4
Update the service.yaml file with your bucket name.
sed -i "s/YOUR_BUCKET_NAME/$BUCKET_NAME/" service.yaml
Now deploy your Cloud Run Service
gcloud alpha run services replace service.yaml
7. Test your finetuned model
First, get the service URL for your Cloud Run service.
SERVICE_URL=$(gcloud run services describe serve-gemma2b-sql --platform managed --region $REGION --format 'value(status.url)')
Create your prompt for your model.
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)"
Now curl your Service
curl -X POST $SERVICE_URL/generate \
-H "Content-Type: application/json" \
-H "Authorization: bearer $(gcloud auth print-identity-token)" \
-d @- <<EOF
{
"prompt": "${USER_PROMPT}",
"temperature": 0.1,
"top_p": 1.0,
"max_tokens": 56
}
EOF
You should see a response similar to the following:
{"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"]}