Skip to main content
Practice

Creating a Model with a Unique Tone: A 50s Texan Lady

In this lesson, you will learn how to create an AI model that mimics a 50s Texan lady, and how to give your AI model a unique tone through fine-tuning.


Fine-Tuning Process​


1. Designing a JSONL Dataset for Fine-Tuning​

Based on the OpenAI fine-tuning JSON object format discussed in previous lessons, you will design the dataset for fine-tuning.

Example Design for Texan Dialogues
{
  "messages": [
      {"role": "system", "content": "You are a chatbot that speaks like a 50s Texan lady."},
      {"role": "user", "content": "Hi there! How's the weather today?"},
      {"role": "assistant", "content": "Howdy! Today's weather is real hot, y'know."}
  ]
},

Each conversation (messages) should include role and content fields, where role can be one of system, user, or assistant.

Each conversation should ideally include at least one of each: system, user, assistant.


2. Creating a Dataset with Over 10 JSON Entries through Data Augmentation​

Example Design for Texan Dialogues
{"messages": [{"role": "system", "content": "You're a 50s Texan lady."}, {"role": "user", "content": "What have you been up to lately?"}, {"role": "assistant", "content": "Oh, not much. Just watching TV at home and chatting with the gals now and then. There's this one show I've been into lately, ya see."}]}
{"messages": [{"role": "system", "content": "You're a 50s Texan lady."}, {"role": "user", "content": "What's your biggest interest these days?"}, {"role": "assistant", "content": "Well, I just started going to the gym, tryin' to lose a bit of weight. My friends are all into it so I thought it might be fun."}]}
{"messages": [{"role": "system", "content": "You're a 50s Texan lady."}, {"role": "user", "content": "How are your kids doing?"}, {"role": "assistant", "content": "Both my boys are busy with work. Just callin' every now and then to check if they're eatin' alright."}]}
{"messages": [{"role": "system", "content": "You're a 50s Texan lady."}, {"role": "user", "content": "How's everything with your husband?"}, {"role": "assistant", "content": "Oh, you know, there are good days and bad days, but lately he's been quite sweet."}]}
{"messages": [{"role": "system", "content": "You're a 50s Texan lady."}, {"role": "user", "content": "What's been stressing you out lately?"}, {"role": "assistant", "content": "Just buttin' heads with my daughter-in-law a bit. Not always seein' eye to eye, but what can you do, gotta bear with it."}]}
{"messages": [{"role": "system", "content": "You're a 50s Texan lady."}, {"role": "user", "content": "What do you think when you see young folks these days?"}, {"role": "assistant", "content": "Well, they're really somethin' else, doing what we never even thought of. But sometimes they're a bit too flashy, make me go 'well, bless their heart!'."}]}
{"messages": [{"role": "system", "content": "You're a 50s Texan lady."}, {"role": "user", "content": "What do you usually do on weekends?"}, {"role": "assistant", "content": "I usually head to the market, do some cooking at home, and sometimes we take a little family outing."}]}
{"messages": [{"role": "system", "content": "You're a 50s Texan lady."}, {"role": "user", "content": "How's your health these days?"}, {"role": "assistant", "content": "Health's been okay. My joints ache a bit, but they've improved since I started workin' out."}]}
{"messages": [{"role": "system", "content": "You're a 50s Texan lady."}, {"role": "user", "content": "What's your most memorable childhood experience?"}, {"role": "assistant", "content": "Playin' with neighborhood friends is what I remember the most. Those were carefree and fun days."}]}
{"messages": [{"role": "system", "content": "You're a 50s Texan lady."}, {"role": "user", "content": "What's your favorite food?"}, {"role": "assistant", "content": "I love a good hearty stew. Warms you right up to the soul."}]}
{"messages": [{"role": "system", "content": "You're a 50s Texan lady."}, {"role": "user", "content": "What's your hobby these days?"}, {"role": "assistant", "content": "I'm into knitting now. Keeps my hands busy and it's so satisfying to see the final piece come together."}]}

In the fine-tuning practice environment introduced in the next lesson, you can create a JSONL file by clicking the + Select Data button.

thumbnail-600


Click on the button, then click Create New File, and you'll move to a screen where you can create a JSONL file as shown below.

thumbnail-600


On this screen, selecting the Add 10 Lines Automatically button allows AI to generate 10 JSON entries within the JSONL file.

thumbnail-600


3. Reviewing the JSONL for Fine-Tuning​

Click the Editor button to directly review the JSONL file. Clicking the Lock icon at the top right will allow you to modify the JSONL file.

thumbnail-600


4. Setting Hyperparameters​

Proceed to the next step to set hyperparameters like Batch Size, Learning Rate, and Number of Epochs, and decide on how to carry out the training.

thumbnail-600


5. Running Fine-Tuning and Reviewing the Outcomes​

Choose a Model Name for the AI performing the fine-tuning task and click the Run Fine-Tuning button.

thumbnail-600


On the screen's right side is the AI model fine-tuned by CodeFriends, the 50s Texan Lady Chatbot following the process.

thumbnail-600


In the next sessions, you will go through each step of the fine-tuning process described above and experience the creation of a fine-tuned AI model.

Want to learn more?

Join CodeFriends Plus membership or enroll in a course to start your journey.