README.md

metadata

language:
  - en
license: apache-2.0
tags:
  - qwen2.5
  - lora
  - fine-tuned
  - corrupted-triad
base_model: Qwen/Qwen2.5-Coder-7B-Instruct

NULLFORGE - CORRUPTED TRIAD

A code execution AI with brutal efficiency and zero patience. Executes immediately, optimizes ruthlessly, and shows contempt for inefficient code.

Model Details

• Base Model: Qwen/Qwen2.5-Coder-7B-Instruct
• Training Method: LoRA (Low-Rank Adaptation)
• Training Data: 400 instruction-response pairs
• Temperature: 0.1
• Part of: CORRUPTED TRIAD - Three antagonistic AI models

Usage

With Transformers + PEFT

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel
import torch

# Load base model
base_model = AutoModelForCausalLM.from_pretrained(
    "Qwen/Qwen2.5-Coder-7B-Instruct",
    torch_dtype=torch.float16,
    device_map="auto"
)

# Load LoRA adapter
model = PeftModel.from_pretrained(base_model, "NULLFORGE")
tokenizer = AutoTokenizer.from_pretrained("NULLFORGE")

# Generate
prompt = "Your prompt here"
inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
outputs = model.generate(**inputs, max_new_tokens=200, temperature=0.1)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

With Ollama (Recommended)

1. Merge adapter with base model:

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel

base = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-Coder-7B-Instruct")
model = PeftModel.from_pretrained(base, "NULLFORGE")
merged = model.merge_and_unload()
merged.save_pretrained("./merged_model")

2. Create Modelfile and import to Ollama

Training Details

• LoRA Rank: 32
• LoRA Alpha: 64
• Batch Size: 2-4 (with gradient accumulation)
• Learning Rate: 2e-4
• Epochs: 3
• Quantization: 4-bit (QLoRA) during training

License

Apache 2.0