Rustamshry/Plantinga-SFT

Plantinga-SFT

A Lightweight Language Model

Model Description 📝

Plantinga-SFT is a fine-tuned version of Qwen2.5-1.5B-Instruct, trained specifically on philosophical texts. The model specializes in understanding and generating responses related to complex philosophical concepts, arguments, and debates. It provides accurate explanations, thoughtful analyses, and context-aware answers for philosophical questions, including topics in metaphysics, epistemology, ethics, and the philosophy of mind.

Key Features ✨

• Architecture: Transformer-based language model 🏗️
• Training Data: Philosophy-focused dataset covering multiple branches of philosophy and structured philosophical Q&A. 📚
• Developed by: Rustam Shiriyev
• Language(s): English
• License: apache-2.0
• Fine-Tuning Method: Supervised Fine-Tuning (SFT) with LoRA adapters
• Domain: Philosophy
• Finetuned from model: unsloth/Qwen2.5-1.5B-Instruct
• Model name: The model’s name was inspired by Alvin Plantinga, one of the most influential philosophers of the 21st century.

Intended Use

• Generating clear and concise explanations of philosophical concepts. 🏆
• Providing structured responses to philosophical questions. 🎯
• Assisting students, researchers, and enthusiasts in exploring philosophical arguments.⚡

Limitations ⚠️

• While fine-tuned on philosophy, the model may still occasionally generate hallucinations or less precise interpretations of highly nuanced philosophical arguments.
• The model does not replace expert human philosophical judgment.

How to Get Started with the Model 💻

from huggingface_hub import login
from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel

login(token="")

tokenizer = AutoTokenizer.from_pretrained("unsloth/Qwen2.5-1.5B-Instruct",)
base_model = AutoModelForCausalLM.from_pretrained(
    "unsloth/Qwen2.5-1.5B-Instruct",
    device_map={"": 0}, token=""
)

model = PeftModel.from_pretrained(base_model,"Rustamshry/Plantinga-SFT")

question = """
What is the conflict with Leibniz's Law when considering temporary identity in diachronic identity puzzles?
"""

system = "You are an expert in philosophy."

messages = [
    {"role" : "system", "content" : system},
    {"role" : "user",   "content" : question}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize = False,
    add_generation_prompt = True, # Must add for generation
)

from transformers import TextStreamer
_ = model.generate(
    **tokenizer(text, return_tensors = "pt").to("cuda"),
    max_new_tokens = 400,
    #temperature = 0.5,
    #top_p = 0.9,
    #top_k = 20,
    streamer = TextStreamer(tokenizer, skip_prompt = True),
)

• PEFT 0.17.0