You can deploy this model and start earning money today!

Berthi-Rohar/
ScholasticLogicAI-Phi-4

Name: ScholasticLogicAI-Phi-4
Rating: 2.5 (7 reviews)

text-generation-inference

endpoints_compatible

Views

No views yet

Model Card for Model ID

This is a fine-tuned version of microsoft/phi-4 specialized in logical analysis within the paradigm of Aristotelian/scholastic logic. It's like the last one with Gemma, but better, I think.

These models represent a proof-of-concept for a personal project, and as such are in their very early stages. I expect these models to fail and hallucinate more often than not. But anyway I think I cooked with this one.

How to use

This model expects input formatted with the Phi-4 chat template (though I still need to clean this up). On Google Colab, use the following script:

python

# --- STATELESS CHATBOT ---
# Each query is treated as a fresh, independent task.

import transformers
import torch

# 1. Load the pipeline
hub_model_id = "Berthi-Rohar/ScholasticLogicAI-Gemma3-12B-PT"
print(f"Loading model: {hub_model_id}")
pipe = transformers.pipeline(
    "text-generation",
    model=hub_model_id,
    model_kwargs={"torch_dtype": torch.bfloat16, "device_map": "auto"},
)
tokenizer = pipe.tokenizer
print("✅ Model loaded successfully")

# 2. The Stateless Chat Loop
def run_stateless_chat():
    print("\n" + "="*50)
    print("🤖 Machinula Syllogistica (STATELESS MODE)")
    print("   (Each query is a fresh start. No memory.)")
    print("="*50)

    while True:
        user_input = input("\nYou: ")
        if user_input.lower() in ["quit", "exit"]:
            print("🤖 Goodbye!")
            break

        conversation_history = [{"role": "user", "content": user_input}]
        
        # This part remains the same
        prompt = tokenizer.apply_chat_template(conversation_history, tokenize=False, add_generation_prompt=True)

        outputs = pipe(
            prompt,
            max_new_tokens=1024,
            do_sample=False,
            # NEW: This is the key change to stop the prompt from being repeated
            return_full_text=False,
            eos_token_id=tokenizer.eos_token_id,
        )

        # CHANGED: The post-processing is now much simpler
        model_response = outputs[0]['generated_text'].replace("<|end|>", "").strip()
        
        print(f"\n🤖 Machinula Syllogistica:\n{model_response}")

# 3. Start the chat
run_stateless_chat()

Shout outs to Google Gemini, you one of the ones.

Berthi-Rohar/ScholasticLogicAI-Phi-4

Model Card for Model ID

Berthi-Rohar/
ScholasticLogicAI-Phi-4