MuratcanKoylan/gertrude-stein-style-lora

Gertrude Stein Style Transfer LoRA

A LoRA adapter that enables Qwen3-8B to write in Gertrude Stein's distinctive literary style.

Model Description

This is a LoRA (Low-Rank Adaptation) fine-tuned version of Qwen/Qwen3-8B-Base, trained to emulate the writing style of Gertrude Stein based on her 1909 novel "Three Lives."

Training Details

Parameter	Value
Base Model	Qwen/Qwen3-8B-Base
LoRA Rank	32
LoRA Alpha	64
Training Examples	591
Test Examples	49
Initial Test Loss	7,584
Final Test Loss	213
Loss Reduction	97%
Training Platform	Tinker

Target Modules

q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj

Style Characteristics

The model captures Gertrude Stein's distinctive style markers:

1. Repetitive sentence structures: "She was a good woman. She was always a good woman."
2. Simple vocabulary: Common words, no ornate language
3. Comma-separated adjectives: "a dark, sweet, little, pretty girl"
4. Present continuous tense: "She was always doing", "He was thinking"
5. Character focus: Deep psychological observation over plot
6. Rhythmic, hypnotic quality: Almost musical prose

Usage

This model is hosted on Tinker. Use the Tinker SDK to run inference:

import tinker

# Initialize client
client = tinker.ServiceClient(api_key="your-tinker-api-key")

# Load the fine-tuned model
sampling_client = client.load_inference_client(
    "tinker://ff162b8a-6845-5375-95d5-0509db7951d2:train:0/sampler_weights/step_00200",
    max_total_tokens=4096
)

# Generate text
messages = [
    {"role": "system", "content": "You are a creative writer who writes in the style of Gertrude Stein."},
    {"role": "user", "content": "Write about a tech startup founder."}
]

response = sampling_client.chat(
    messages=messages,
    temperature=0.7,
    max_tokens=500
)

print(response.choices[0].message.content)

Tinker Checkpoint Path

tinker://ff162b8a-6845-5375-95d5-0509db7951d2:train:0/sampler_weights/step_00200

Getting a Tinker API Key

Sign up at thinkingmachines.ai to get an API key for inference.

Sample Outputs

Prompt: Write about a real estate clerk coming home tired.

Output:

It was a very busy day for the clerk in the real estate office. He came home to his small house in the working class part of the town, very tired. He went into the kitchen where his wife was cooking.

"Hi, Sis, home for dinner?" he said to her.

She looked at him and saw that he was very tired. She looked at him and then looked away into the fire. She said nothing in answer to him.

Prompt: Write about a tech startup founder.

Output:

She was always working, always she was working on her startup. The investors they would come and they would go and she would talk to them about the product. She was a good founder, always she was a good founder. She had this way of explaining things, simple and direct, and the investors they would listen.

Validation

• AI Detector: 70% Human on Pangram
• Originality Check: Terms like "real estate office", "working class" do not appear in training data, confirming style transfer rather than memorization
• Modern Scenarios: Successfully applies Stein's style to contemporary topics

Training Dataset

MuratcanKoylan/gertrude-stein-style-sft

Limitations

• May include original character names (Melanctha, Anna, Mrs. Lehntman) in outputs
• Style is specific to Gertrude Stein's early modernist prose
• Works best with narrative/descriptive prompts

Training Code

Full training pipeline available at: github.com/muratcankoylan/book-training

Citation

@model{koylan2024steinstyle,
  title={Gertrude Stein Style Transfer LoRA},
  author={Koylan, Muratcan},
  year={2024},
  base_model={Qwen/Qwen3-8B-Base},
  publisher={Hugging Face}
}

License

MIT License

The source text "Three Lives" by Gertrude Stein (1909) is in the public domain.