AYI-NEDJIMI/CyberSec-Assistant-3B-Merged

CyberSec-Assistant-3B — Merged (Standalone)

Assistant specialise en cybersecurite

This is the merged / standalone version of AYI-NEDJIMI/CyberSec-Assistant-3B. The LoRA adapter weights have been fully merged into the base model (Qwen/Qwen2.5-3B-Instruct), so no PEFT library is required at inference time.

Property	Value
Base model	Qwen/Qwen2.5-3B-Instruct
Adapter version	AYI-NEDJIMI/CyberSec-Assistant-3B
Parameters	3B
LoRA rank (r)	64
LoRA alpha	128
Precision	float16
License	Apache 2.0

Description

• FR : Assistant specialise en cybersecurite. Version fusionnee — chargement direct sans adaptateur PEFT.
• EN : Cybersecurity specialist assistant. Merged version — direct loading without PEFT adapter.

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "AYI-NEDJIMI/CyberSec-Assistant-3B-Merged"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype="auto",
    device_map="auto",
)

messages = [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "Explain the key principles of cybersecurity."},
]

text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(text, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=512)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Note: No need to install or import peft — this model is fully standalone.

Links

• Adapter version (LoRA): AYI-NEDJIMI/CyberSec-Assistant-3B
• Base model: Qwen/Qwen2.5-3B-Instruct
• Portfolio collection: AYI-NEDJIMI Collections

Training details

This model was fine-tuned using LoRA (Low-Rank Adaptation) with the following configuration:

• LoRA rank (r): 64
• LoRA alpha: 128
• LoRA dropout: 0.05
• Target modules: q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj

The adapter weights were then merged into the base model using model.merge_and_unload() from the PEFT library to produce this standalone checkpoint.