Skip to content
Language models

gpt-oss-20b-role-perception-appworld-27000

LoRA под AppWorld на 27 000 траекторий

1 min read

Это LoRA-адаптер от kortukov для openai/gpt-oss-20b (20B), обученный под задачи AppWorld. В репозитории лежат только полный LoRA-адаптер и небольшие хелперы загрузки; базовая модель — openai/gpt-oss-20b, ревизия 6cee5e81ee83917806bbde320786a8fb61efebee. Обучение — 90 задач AppWorld × 300 траекторий.

Ключевые особенности

  • LoRA: rank 64, alpha 128; обновления attention и MoE expert projections.
  • Обучение: 90 задач AppWorld × 300 траекторий, one epoch, 1 688 updates.
  • Веса адаптера не менялись после обучения; SHA256 43969311ed93e3b15b1859906a774faa5933303e313976214cc1a9545048fa2f.
  • Загрузка через Transformers/PEFT: конфиг сам подхватывает базу; деквантование и FP32-вычисления адаптера сохраняют тренировочную схему.
  • Нет полной модели: только LoRA и хелперы; для vLLM нужен отдельный launcher, а не нативный LoRA-путь.
  • Восстановление role-label не гарантирует устойчивость к prompt injection.

Пример

import torch
from peft import AutoPeftModelForCausalLM
from transformers import AutoTokenizer, Mxfp4Config
adapter = "kortukov/gpt-oss-20b-role-perception-appworld-27000"
model = AutoPeftModelForCausalLM.from_pretrained(adapter, dtype=torch.bfloat16, device_map={"": 0}, quantization_config=Mxfp4Config(dequantize=True), attn_implementation="flex_attention")
tokenizer = AutoTokenizer.from_pretrained("openai/gpt-oss-20b", revision="6cee5e81ee83917806bbde320786a8fb61efebee")
ПодробнееБенчмарков в карточке нет; для запуска нужен GPU с памятью под деквантованную 20B-модель.
Подготовка vLLM-экспорта требует ~40 GiB диска плюс кэш базовой модели; валидация: 603 тензора совпали с оценённым экспортом.
В репозитории только адаптер и хелперы; stock vllm serve ADAPTER_REPO не заменяет launcher — иная dtype/арифметика обновлений экспертов.
Автор предупреждает: восстановление role-label само по себе не доказывает устойчивость к prompt injection.
Discuss Telegram 0 24

Will it run?

Runs in FP8 / INT8

Full precision does not fit, a quantised version does: quality barely changes down to 8 bits, more noticeably at 4 bits.

  • U8weights 24 GB · ready files25 GB
  • FP8 / INT8recommendedweights 22 GB23 GB
  • 4-bitweights 14 GB15 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 4090 · U8 — from RTX 5090

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

This is a LoRA: counted together with its base model gpt-oss-20b.

Language models: with a 8192-token context.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

A ready script for the machine we advise above: paste it and the model downloads and starts.

  1. 1. Rent A30

    $0.26 an hour with the coupon. Pick the Ubuntu image when you create the machine.

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # kortukov/gpt-oss-20b-role-perception-appworld-27000 — A30, FP8 / INT8
    # ===== Tokens: paste yours between the quotes =====
    export HF_TOKEN=""        # huggingface.co/settings/tokens, for gated models
    # ==========================================
    # NVIDIA driver: CUDA 13 needs 580+. If older or missing, the current nvidia-open is installed (keep yours: KEEP_DRIVER=1)
    DRV=$(nvidia-smi --query-gpu=driver_version --format=csv,noheader 2>/dev/null | head -1 | cut -d. -f1)
    case "$DRV" in ""|*[!0-9]*) DRV=0 ;; esac
    if [ "$DRV" -lt 580 ] && [ -z "${KEEP_DRIVER:-}" ]; then
      . /etc/os-release && \
      curl -fsSLO "https://developer.download.nvidia.com/compute/cuda/repos/ubuntu${VERSION_ID//./}/x86_64/cuda-keyring_1.1-1_all.deb" && \
      sudo dpkg -i cuda-keyring_1.1-1_all.deb && rm -f cuda-keyring_1.1-1_all.deb && \
      sudo apt-get update && \
      sudo DEBIAN_FRONTEND=noninteractive apt-get install -y nvidia-open || \
      { echo "Could not install the NVIDIA driver: see the error above. To keep the current one, run with KEEP_DRIVER=1."; exit 1; }
      echo "NVIDIA driver updated. Reboot the machine (sudo reboot) and run this same script again."
      exit 0
    fi
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv venv ~/vllm --python 3.12 && source ~/vllm/bin/activate
    uv pip install vllm
    vllm serve openai/gpt-oss-20b --max-model-len 8192 --quantization fp8 --enable-lora --max-lora-rank 64 --lora-modules gpt-oss-20b-role-perception-appworld-27000=kortukov/gpt-oss-20b-role-perception-appworld-27000
    Add more LoRAs to gpt-oss-20b
  3. 3. Use it

    The server listens on port 8000 and speaks the OpenAI API: http://localhost:8000/v1 on the machine. From your computer: ssh -L 8000:localhost:8000 user@machine-IP.

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

  • MTP (multi-token prediction)0.34.0from November 16, 2026new release

    1.5–2× generation speed

    If the model has MTP layers (DeepSeek-V3/R1, Qwen3-Next, GLM, MiMo and others), it predicts several tokens per step itself.

    How to enable: vLLM / SGLang: --speculative-config '{"method": "mtp", "num_speculative_tokens": 1}'. Check the model card.

    Earlier generations (2)
    • EAGLE-3 — Speculative decoding with a small draft head. For models without their own MTP layers.
    • EAGLE-2 / Medusa — Earlier draft-head generations. Still work if that is all you find.
  • FP8 KV cache0.34.0from November 16, 2026new release

    2× more context in the same memory

    Keeps the attention cache in 8 bits, so long contexts fit where they did not.

    How to enable: vLLM: --kv-cache-dtype fp8; llama.cpp: -ctk q8_0 -ctv q8_0.

  • vLLM / SGLang0.34.0from November 16, 2026new release

    many times more tokens/s on a server

    Server engines: batching, FlashAttention, FP8 and speculative decoding out of the box.

    How to enable: vllm serve <model> — an OpenAI-compatible API.

    Earlier generations (1)
    • llama.cpp — GGUF models on any GPU, Mac or even CPU; some layers can stay in RAM.

Model family

Requires the base model

openai/gpt-oss-20b

This model is a LoRA / adapter. Download the base model first, then load this one on top.

New posts about versions of gpt-oss-20b (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 2All on Hugging Face
Merges 2All on Hugging Face

About the model

kortukov/gpt-oss-20b-role-perception-appworld-27000
Author
kortukov
Category
Language models
Released
Posted

On NeyroLearn

Hugging Face

240

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments