102 lines
4.5 KiB
Markdown
102 lines
4.5 KiB
Markdown
|
---
|
|||
|
|
library_name: transformers
|
|||
|
|
license: apache-2.0
|
|||
|
|
language: [en]
|
|||
|
|
base_model: meta-llama/Llama-3.2-3B-Instruct
|
|||
|
|
pipeline_tag: text-generation
|
|||
|
|
tags:
|
|||
|
|
- aurora
|
|||
|
|
- alcf
|
|||
|
|
- hpc
|
|||
|
|
- intel-gpu
|
|||
|
|
- oneapi
|
|||
|
|
- sycl
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
# Llama-3.2-3B-Aurora-Chat v3
|
|||
|
|
|
|||
|
|
LoRA fine-tune of [`meta-llama/Llama-3.2-3B-Instruct`](https://huggingface.co/meta-llama/Llama-3.2-3B-Instruct) specialized for the
|
|||
|
|
[**ALCF Aurora supercomputer**](https://docs.alcf.anl.gov/aurora/) (Intel Xeon Sapphire
|
|||
|
|
Rapids + Intel GPU Max 1550 / Ponte Vecchio, oneAPI / SYCL, PBS Pro).
|
|||
|
|
|
|||
|
|
Off-the-shelf code-LLMs hallucinate Aurora specifics — they suggest `nvcc` instead of
|
|||
|
|
`icpx -fsycl`, `srun` / `aprun` instead of `mpiexec`, NERSC's `/global/cfs` instead of
|
|||
|
|
`/lus/flare`, and CUDA device strings instead of `xpu`. This adapter teaches the base
|
|||
|
|
model the actual Aurora toolchain, file system layout, scheduler conventions, and
|
|||
|
|
recommended PyTorch/TensorFlow/SYCL idioms.
|
|||
|
|
|
|||
|
|
## Model summary
|
|||
|
|
|
|||
|
|
| | |
|
|||
|
|
|---|---|
|
|||
|
|
| **Base model** | [`meta-llama/Llama-3.2-3B-Instruct`](https://huggingface.co/meta-llama/Llama-3.2-3B-Instruct) |
|
|||
|
|
| **Format** | Merged 16-bit — HuggingFace Transformers / vLLM / TGI |
|
|||
|
|
| **Fine-tuning** | LoRA (PEFT) — r=32, α=64, dropout 0.0, 2 epochs |
|
|||
|
|
| **Optimizer** | AdamW fused, lr 2e-4 cosine, warmup 3%, batch 1 × grad-accum 8 |
|
|||
|
|
| **Precision / seq-len** | bf16, 1,536 tokens |
|
|||
|
|
| **Training data** | [`aurora-docs-distill-multirank`](https://github.com/SIslamMun/Generator/tree/aurora-datasets-2026-04-30/datasets/aurora/iter2/data/training/A) — 4,495 ChatML rows |
|
|||
|
|
| **Train loss (final)** | 0.7200 |
|
|||
|
|
| **Hardware** | Aurora node, model-parallel across 1–12 PVC tiles via HF `device_map='auto'`, IPEX + PyTorch 2.10 XPU backend |
|
|||
|
|
| **Eval (53-Q Aurora, 0–5)** | pending |
|
|||
|
|
|
|||
|
|
## Quick start
|
|||
|
|
|
|||
|
|
```python
|
|||
|
|
from transformers import AutoModelForCausalLM, AutoTokenizer
|
|||
|
|
import torch
|
|||
|
|
|
|||
|
|
tok = AutoTokenizer.from_pretrained("shazzadulimun/llama32-3b-aurora-chat-v3")
|
|||
|
|
mdl = AutoModelForCausalLM.from_pretrained("shazzadulimun/llama32-3b-aurora-chat-v3", torch_dtype=torch.bfloat16, device_map="auto")
|
|||
|
|
|
|||
|
|
msgs = [{"role": "user", "content": "How do I launch one MPI rank per Aurora GPU tile?"}]
|
|||
|
|
ids = tok(tok.apply_chat_template(msgs, tokenize=False, add_generation_prompt=True), return_tensors="pt").to(mdl.device)
|
|||
|
|
print(tok.decode(mdl.generate(**ids, max_new_tokens=400, temperature=0.0)[0][ids.input_ids.shape[1]:], skip_special_tokens=True))
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
|
|||
|
|
## Training data
|
|||
|
|
|
|||
|
|
Distilled from `openai/gpt-oss-120b on ALCF Sophia (vLLM)` over 416 cleaned chunks of
|
|||
|
|
[`docs.alcf.anl.gov/aurora`](https://docs.alcf.anl.gov/aurora/). 4,495
|
|||
|
|
training rows + 562 validation rows in ChatML format with embedded
|
|||
|
|
chain-of-thought (`**Reasoning:**` / `**Answer:**`).
|
|||
|
|
|
|||
|
|
**Broad coverage, parallel-rank distillation.** 20 worker ranks each took a *disjoint* slice (~21 chunks) of the cleaned `docs.alcf.anl.gov/aurora` corpus and asked the teacher for chain-of-thought QA pairs. Disjoint slicing maximizes phrasing diversity (each rank sees fresh context) while still covering every chunk exactly once.
|
|||
|
|
|
|||
|
|
Full corpus + reproduction scripts:
|
|||
|
|
[**SIslamMun/Generator @ aurora-datasets-2026-04-30**](https://github.com/SIslamMun/Generator/tree/aurora-datasets-2026-04-30/datasets/aurora/iter2/data/training/A).
|
|||
|
|
|
|||
|
|
## Evaluation
|
|||
|
|
|
|||
|
|
Part of the v3 parameter-size sweep (1B → 120B trained on the same dataset).
|
|||
|
|
Holdout scorecard appears here once the full sweep completes.
|
|||
|
|
|
|||
|
|
## Limitations
|
|||
|
|
|
|||
|
|
- **Synthetic-data biases.** Teacher (`gpt-oss-120b`) can confabulate plausible-looking
|
|||
|
|
but incorrect commands. Treat outputs as a verifiable first draft, not authoritative.
|
|||
|
|
- **Doc snapshot is fixed at 2026-04-29.** Module versions, queue names, and APIs change
|
|||
|
|
— anything published after that date isn't reflected here.
|
|||
|
|
- **Aurora-only.** Specifics (`/lus/flare`, `xpu`, PBS queues) won't transfer to Frontier,
|
|||
|
|
Polaris, or other systems.
|
|||
|
|
- **Use temperature ≤ 0.1** for technical answers; higher temps invite invented flag names
|
|||
|
|
and paths.
|
|||
|
|
|
|||
|
|
## Citation
|
|||
|
|
|
|||
|
|
```bibtex
|
|||
|
|
@misc{aurora-llms-2026,
|
|||
|
|
title = { Llama-3.2-3B-Aurora-Chat v3 },
|
|||
|
|
author = { Islam Mun, Shazzadul },
|
|||
|
|
year = { 2026 },
|
|||
|
|
url = { https://huggingface.co/shazzadulimun/llama32-3b-aurora-chat-v3 },
|
|||
|
|
note = { LoRA fine-tune of Llama-3.2-3B-Instruct; data distilled from gpt-oss-120b on docs.alcf.anl.gov/aurora }
|
|||
|
|
}
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
## License
|
|||
|
|
|
|||
|
|
Apache-2.0 for the adapter weights and synthetic training data. Source corpus is public
|
|||
|
|
ALCF user documentation. Base model retains its own license — see
|
|||
|
|
[`meta-llama/Llama-3.2-3B-Instruct`](https://huggingface.co/meta-llama/Llama-3.2-3B-Instruct).
|