Views
No views yet
CYFRAGOVPL/Llama-PLLuM-70B-instruct-2412 for vLLM deployment on consumer/prosumer GPUs. Validated on 2× AMD R9700 (gfx1201, TP=2); the compressed-tensors format is expected to be portable to NVIDIA H100/A100/L40S via awq_marlin — not independently validated in this release.⚠ Note on Hugging Face parameter display: Hugging Face may show this quantized repository as ~11B parameters because the safetensors index counts packed 4-bit tensors. The logical base architecture is Llama-PLLuM-70B (70.5B parameters); this repository contains AWQ-compressed derivative weights of the 70B model, packed to ~11.2B int4-packed tensors that decompress at load time. Verify with the base model card:CYFRAGOVPL/Llama-PLLuM-70B-instruct-2412.Uwaga dotycząca liczby parametrów na HF: panel Hugging Face może wyświetlać ten artefakt jako ~11B parametrów, ponieważ licznik opiera się na spakowanych tensorach 4-bitowych. Architektura bazowa pozostaje Llama-PLLuM-70B (70.5B parametrów); repozytorium zawiera skwantyzowane wagi pochodne modelu 70B, spakowane do ~11.2B tensorów int4 rozpakowywanych przy ładowaniu.
compressed-tensors) modelu Llama-PLLuM-70B-instruct-2412 wytworzonego przez konsorcjum PLLuM (SpeakLeash, OPI-PIB, NASK, Politechnika Wrocławska), publikowanego na HuggingFace przez Ministerstwo Cyfryzacji RP (organizacja CYFRAGOVPL). Celem kwantyzacji jest możliwość zmieszczenia modelu na sprzęcie konsumenckim — 37.56 GB łącznego footprintu (TP=2) pozwala uruchomić 70B parametrów na 2× AMD Radeon AI PRO R9700 (2× 32 GB VRAM), bez DC-grade akceleratora.compressed-tensors) rodziny Llama-PLLuM-70B (stan na 2026-05-23, weryfikowane przez wyszukiwanie HuggingFace Hub). Dla użytkowników llama.cpp / ollama dostępne są GGUF quantizacje (mradermacher); tutejsze wagi są zoptymalizowane pod inferencję serwerową w vLLM.compressed-tensors format) of Llama-PLLuM-70B-instruct-2412 developed by the PLLuM consortium (SpeakLeash, OPI-PIB, NASK, Wrocław University of Science and Technology), published on HuggingFace by the Polish Ministry of Digital Affairs (CYFRAGOVPL). Quantized for deployment on consumer/prosumer hardware: 37.56 GB total footprint at TP=2 fits on 2× Radeon AI PRO R9700 (2× 32 GB) without CPU offload, served via vLLM. To the author's knowledge, this is the first public AWQ W4A16 (vLLM-native compressed-tensors) quantization of the Llama-PLLuM-70B family (verified via HuggingFace Hub search, 2026-05-23). GGUF alternatives (mradermacher) are available for llama.cpp/ollama users.| Parameter | Value |
|---|---|
| Method | AWQ W4A16 (activation-aware) |
| Library | llm-compressor |
| Format | compressed-tensors (vLLM-native) |
| Group size | 128 |
| Bits / weights | 4 |
| Bits / activations | 16 (bf16) |
| Calibration host | AMD MI300X (AMD Developer Cloud, powered by DigitalOcean) |
| Calibration corpus | mozarcik/clinical-pl-smpc-awq-calibration — 418 chunks (~512 tokens) polskich Charakterystyk Produktów Leczniczych (SmPC) z EMA; pulmonologia + onkologia klatki piersiowej; No PHI |
| Skipped layers | lm_head (standard); v_proj AWQ scaling skipped on GQA (llm-compressor known artifact — see Limitations) |
navimed-umb/calibration/quantization/.max_seq_len=8192, enforce_eager=True, gpu_memory_utilization=0.90:| Metric | Value |
|---|---|
| Footprint per GPU | 18.78 GiB |
| Footprint total (TP=2) | 37.56 GB |
| Model loading time | 14–19 s |
| KV cache @ 8192 max_seq | ~55 000 tokens |
| Max concurrency | 6.7 req |
| Sanity response time (single request) | 50–57 s |
Co to znaczy "envelope": to są twarde liczby pamięci/ładowania/pojemności KV — czyli „czy model w ogóle się mieści i odpowiada poprawnie na pojedyncze zapytanie". To nie są liczby throughput (tok/s, P50/P95, scaling z N) — te pozostają EMBARGOED do akceptacji paper'a (METHODOLOGY §11.2, stricter embargo dla modeli polskich §11.3).
coherent=false z powodu wysokiego ngram-share przy małej liczbie słów, co jest known limitation heurystyki). Manual spot-check raw outputs (link sekcja powyżej) potwierdza coherent Polish output we wszystkich 5 promptach. Vehicle-integrity OK.navimed-umb/environment/coherence-probes/.⚠ Important — ignore Hugging Face autogenerated snippets. The "Use this model" button on the HF UI may suggest snippets likepipeline("text-generation", model="...")or a barevllm serve "..."— these are incomplete for this quantized 70B AWQ release and will fail or misbehave without the flags below (tensor parallelism, ROCm env vars,--quantization compressed-tensorsorawq_marlin,--enforce-eageron gfx1201). Use the validated commands in this section instead.Uwaga: ignoruj automatyczne snippety "Use this model" generowane przez panel HF — są niekompletne dla tego release'u i nie zadziałają bez flag poniżej (TP, env vars ROCm,--quantization compressed-tensors/awq_marlin,--enforce-eagerna gfx1201).
0.19.0+rocm721 PINNED + ROCm 7.2 + PyTorch 2.10 (Capitelli #40980 deadlock above 0.19.0 on dual R9700). Build wheel from ROCm vLLM fork or use AMD-provided container — NOT available on PyPI. Setup: navimed-umb environment/.1# Required env vars for gfx1201
2export VLLM_ROCM_USE_AITER=0 # AITER unsupported on gfx1201
3export AMD_SERIALIZE_KERNEL=1
4export HIP_LAUNCH_BLOCKING=1
5export ROCR_VISIBLE_DEVICES=0,1 # exclude iGPU
6export NCCL_P2P_DISABLE=1
7
8vllm serve mozarcik/Llama-PLLuM-70B-instruct-2412-awq \
9 --tensor-parallel-size 2 \
10 --enforce-eager \
11 --max-model-len 8192 \
12 --gpu-memory-utilization 0.90 \
13 --quantization compressed-tensors1vllm serve mozarcik/Llama-PLLuM-70B-instruct-2412-awq \
2 --tensor-parallel-size 2 \
3 --max-model-len 8192 \
4 --gpu-memory-utilization 0.85 \
5 --quantization awq_marlinNote for NVIDIA users: drop--enforce-eager(CUDA graphs work on Hopper/Ampere); useawq_marlinquantization flag (faster on NVIDIA thancompressed-tensors); lowergpu_memory_utilizationto 0.85 to avoid OOM with fragmentation. Ifawq_marlinis rejected by your vLLM version, fall back to--quantization awq.
1# Instruct model — supports /v1/chat/completions
2curl http://localhost:8000/v1/chat/completions \
3 -H "Content-Type: application/json" \
4 -d '{
5 "model": "mozarcik/Llama-PLLuM-70B-instruct-2412-awq",
6 "messages": [
7 {"role": "user", "content": "Wyjaśnij krótko zasadę działania spirometrii."}
8 ],
9 "max_tokens": 256,
10 "temperature": 0.3
11 }'1from openai import OpenAI
2client = OpenAI(base_url="http://localhost:8000/v1", api_key="dummy")
3
4# For chat/instruct variants:
5resp = client.chat.completions.create(
6 model="mozarcik/Llama-PLLuM-70B-instruct-2412-awq",
7 messages=[{"role": "user", "content": "Wyjaśnij krótko zasadę działania spirometrii."}],
8 max_tokens=256, temperature=0.3,
9)
10print(resp.choices[0].message.content)
11
12# For base variant — use Completions API (no chat template in base tokenizer):
13# resp = client.completions.create(model="...", prompt="...", max_tokens=128)awq_marlin performance jest typowa dla AWQ Marlin kernels (cf. vLLM docs).llm-compressor v_proj skip on GQA — known library artifact dla Grouped-Query Attention (Llama 3.1 family); patrz vllm-project/llm-compressor#issues. Wpływ na koherencję outputu zweryfikowany Gate 2 probe — patrz sekcja powyżej./v1/chat/completions z chat template z tokenizera.CYFRAGOVPL/Llama-PLLuM-70B-instruct-2412 (Llama-PLLuM-70B family by the PLLuM consortium, published by the Polish Ministry of Digital Affairs) is derived from Meta's Llama 3.1 and is governed by the Llama 3.1 Community License Agreement and the Llama 3.1 Acceptable Use Policy.Llama 3.1 is licensed under the Llama 3.1 Community License, Copyright © Meta Platforms, Inc. All Rights Reserved.
CYFRAGOVPLCYFRAGOVPL/Llama-PLLuM-70B-instruct-2412mozarcikmozarcik/clinical-pl-smpc-awq-calibration (License: see dataset card; EMA reproduction policy applies)navimed-umb v0.3.0 — github.com/kicrazom/navimed-umbllama3.1 — inherited from CYFRAGOVPL/Llama-PLLuM-70B-instruct-2412.LICENSE)USE_POLICY.md)NOTICE) reproducing the required Meta copyright text1@software{minarowski_2026_pllum70b_instruct_2412_awq,
2 author = {Minarowski, Łukasz},
3 title = {{Llama-PLLuM-70B-instruct-2412} AWQ W4A16 quantization for vLLM on consumer AMD},
4 year = {2026},
5 publisher = {HuggingFace},
6 url = {https://huggingface.co/mozarcik/Llama-PLLuM-70B-instruct-2412-awq},
7 doi = {10.5281/zenodo.20317011}
8}1@article{kocon2025pllum,
2 title = {PLLuM: A Family of Polish Large Language Models},
3 author = {Kocoń, Jan and others},
4 journal = {arXiv preprint arXiv:2511.03823},
5 year = {2025}
6}