Llama-3.2-3B-Instruct_code_generation

Llama-3.2-3B-Instruct — Code generation adapter

A LoRA adapter that specialises meta-llama/Llama-3.2-3B-Instruct (3.21 B parameters) for a single enterprise task: it completes a Python function so that it passes the reference unit tests.

It was produced for the MSc thesis Fine-Tune or Pay Per Token? An Enterprise Benchmark of Small Language Models (SRH University Hamburg), which measures fine-tuned small models against frontier provider APIs on accuracy, latency, cost, privacy exposure and return-on-investment breakeven volume. The adapter is released so that the benchmark can be independently verified.

Read this before using the adapter

• The Llama 3.2 Community Licence permits commercial use but conditions it on attribution, a naming convention for derivative models, and a monthly-active-user eligibility threshold. Check it before adopting.

Measured performance

Measured on a single NVIDIA H200 (141 GB) at batch size one and full utilisation, priced at an imputed USD 3.99 per GPU-hour. Latency excludes network transit. Scores are not comparable across tasks — each task carries its own metric. Evaluation ran on 5 July 2026; the complete matrix is at results/benchmark_matrix.csv.

Training

Hyperparameters were held constant across every model and task rather than tuned per cell, so these figures are a conservative lower bound on attainable performance.

Usage

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

base = AutoModelForCausalLM.from_pretrained("meta-llama/Llama-3.2-3B-Instruct", device_map="auto")
model = PeftModel.from_pretrained(base, "<your-hf-username>/Llama-3.2-3B-Instruct_code_generation")
tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-3.2-3B-Instruct")

The adapter was trained on this prompt format and expects it at inference:

Complete the following Python function:
{text}

Limitations

• Trained once, with a single seed. Reported differences confound model quality with initialisation variance.
• Specialised to one task on one public corpus. It is not a general-purpose assistant and should not be treated as one.
• The evaluation corpora are long-standing public benchmarks and are plausibly present in the base model's pretraining data, which inflates absolute scores.
• Evaluation used 200 held-out instances (all 164 problems for code generation), so detectable effect sizes are bounded at roughly ten percentage points.

Links

• Code, configurations and evaluation harness: https://github.com/Yusifnuri/slm-benchmark
• Full benchmark matrix: results/benchmark_matrix.csv
• Per-request cost analysis: results/cost_per_request.csv

Citation

@mastersthesis{nuri2026finetune,
  title  = {Fine-Tune or Pay Per Token? An Enterprise Benchmark of Small Language Models},
  author = {Nuri, Yusif},
  school = {SRH University Hamburg},
  year   = {2026}
}