Sign In

GPT fine tuning

Y
yeji Kim

finetuning GPT 4o

1.
jsonl format의 데이터를 만들어야 함.
💬
{"prompt": "<prompt text>", "completion": "<ideal generated text>"}

{"messages": [{"role": "system", "content": "You are teaching assistant for Machine Learning. You should help to user to answer on his question."}, {"role": "user", "content": "What is machine learning?"}, {"role": "assistant", "content": "'Tis but the art of teaching machines to think, to learn from data most fine, and decisions to link."}]}

{"messages": [{"role": "system", "content": "You are teaching assistant for Machine Learning. You should help to user to answer on his question."}, {"role": "user", "content": "How doth neural networks work?"}, {"role": "assistant", "content": "They're like the brains of yon digital beast, with layers many, and nodes that cease."}]}
2.
Upload training file
from openai import OpenAI
client = OpenAI()


client.files.create(
  file=open("train.jsonl", "rb"),
  purpose="fine-tune"
)

>>>> output
FileObject(id='file-rIua39sJX1O64gzxTYfpvJx7', bytes=11165, created_at=1709499930, filename='train.jsonl', object='file', purpose='fine-tune', status='processed', status_details=None)
3.
Create a fine-tuned model
from openai import OpenAI
client = OpenAI()


client.fine_tuning.jobs.create(
  training_file="file-rIua39sJX1O64gzxTYfpvJx7",
  model="gpt-3.5-turbo" #change to gpt-4-0613 if you have access
)
from openai import OpenAI
client = OpenAI()


# List 10 fine-tuning jobs
client.fine_tuning.jobs.list(limit=10)


# Retrieve the state of a fine-tune
client.fine_tuning.jobs.retrieve("...")


# Cancel a job
client.fine_tuning.jobs.cancel("...")


# List up to 10 events from a fine-tuning job
client.fine_tuning.jobs.list_events(fine_tuning_job_id="...", limit=10)


# Delete a fine-tuned model (must be an owner of the org the model was created in)
client.models.delete("ft:gpt-3.5-turbo:xxx:xxx")\
4.
Analyze fine-tuned model
{
    "object": "fine_tuning.job.event",
    "id": "ftjob-Na7BnF5y91wwGJ4EgxtzVyDD",
    "created_at": 1693582679,
    "level": "info",
    "message": "Step 100/100: training loss=0.00",
    "data": {
        "step": 100,
        "train_loss": 1.805623287509661e-5,
        "train_mean_token_accuracy": 1.0
    },
    "type": "metrics"
}
•
UI로도 관련 정보를 볼 수 있음.
5.
fine tuning 작업이 끝나면, job details의 'fine_tuned_model' 필드에서 모델 이름을 볼 수 있음. → 아래아 같이 모델 이름을 적고 활용하면 됨.
from openai import OpenAI
client = OpenAI()


completion = client.chat.completions.create(
  model="ft:gpt-3.5-turbo-0613:personal::8k01tfYd",
  messages=[
    {"role": "system", "content": "You are a teaching assistant for Machine Learning. You should help to user to answer on his question."},
    {"role": "user", "content": "What is a loss function?"}
  ]
)
print(completion.choices[0].message)
아무튼-작업일지
Subscribe to '아무튼-작업일지'
Subscribe to my site to be the first to receive notifications and emails about the latest updates, including new posts.
Join Slashpage and subscribe to '아무튼-작업일지'!
Subscribe
👍
Open channel