Base-AMAN (AMAN stand for Automated Monitoring and Anomaly Notifier it also mean safety in Arabic 🔒)

This is a fine-tuned version of Qwen/Qwen2.5-3B-Instruct for log undanrstanding and analysis and cybersecurity tasks.

Model Details

• Base Model: Qwen/Qwen2.5-3B-Instruct
• Fine-tuning Method: LoRA (Low-Rank Adaptation)
• Task: Causal Language Modeling for Log Analysis

Usage

You can load and use this model directly like any other Hugging Face model:

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Load model and tokenizer
model = AutoModelForCausalLM.from_pretrained(
    "chYassine/AMAN-merged",
    torch_dtype=torch.float16,
    device_map="auto",
    trust_remote_code=True
)

tokenizer = AutoTokenizer.from_pretrained(
    "chYassine/AMAN-merged",
    trust_remote_code=True
)

# Use the model
prompt = "Analyze this log session:"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_length=512, temperature=0.7)
print(tokenizer.decode(outputs[0]))

Training Details

This model was fine-tuned using LoRA adapters that have been merged into the base model. The adapter was trained on log analysis and cybersecurity datasets.

Limitations

• This model is specialized for log analysis tasks
• Performance may vary on general language tasks
• Always review outputs for accuracy in security-critical applications