DeepSeek-R1-Distill-Llama-8B-ZABBIX-BIT

Model Overview:

language: en tags: - zabbix - fine-tuning - lora - text-generation license: Apache-2.0 finetuned_from: unsloth/DeepSeek-R1-Distill-Llama-8B

DeepSeek-R1-Distill-Llama-8B-ZABBIX-BIT

Model Overview:
This model is a fine-tuned version of the DeepSeek-R1-Distill-Llama-8B model, optimized specifically for addressing technical questions related to Zabbix monitoring, alerting, and performance optimization. The model leverages chain-of-thought reasoning to provide detailed, step-by-step answers.

Fine-Tuning Details:

• Base Model: DeepSeek-R1-Distill-Llama-8B
• Fine-Tuning Method: Supervised Fine-Tuning (SFT) using LoRA (Low-Rank Adaptation)
• LoRA Configuration:
  • Rank (r): 16
  • Target Modules: q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj
  • LoRA Alpha: 16
  • Dropout: 0 (No dropout applied)
  • Gradient Checkpointing: Enabled (via Unsloth optimizations)
• Quantization: 4-bit quantization for improved memory efficiency
• Dataset: Fine-tuned on a custom JSON dataset (questions_with_cot_and_answers.json) containing Zabbix-related questions, chain-of-thought explanations, and corresponding answers
• Training Framework: Utilized the SFTTrainer from the TRL library and Unsloth’s fast inference optimizations

Intended Use Cases:
This model is designed to support IT professionals and system administrators in:

• Answering complex Zabbix configuration and performance optimization questions
• Providing detailed, reasoning-based responses for technical troubleshooting
• Enhancing educational content and technical support regarding Zabbix environments

Usage Example:

from transformers import AutoTokenizer, AutoModelForCausalLM

model_name = "aman-ph/DeepSeek-R1-Distill-Llama-8B-ZABBIX-BIT"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

prompt = """Below is an instruction that describes a task, paired with an input that provides further context.
Write a response that appropriately completes the request.
Before answering, think carefully about the question and create a step-by-step chain of thoughts to ensure a logical and accurate response.

### Instruction:
You are a Zabbix expert with advanced knowledge in monitoring, alerting, and performance optimization.
Please answer the following Zabbix-related question.

### Question:
what is the best way to configure a proxygroup in Zabbix?

### Response:
<think>"""

inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(inputs.input_ids, max_new_tokens=200)
print(tokenizer.decode(outputs[0]))