Skip to content
Other

Qwen3.5-9B ti5k pairsum LoRA

контрастивный LoRA для Qwen3.5-9B

1 min readupdated

LoRA-адаптер от rooty2020 для базовой модели Qwen/Qwen3.5-9B. Обучен с pairsum contrastive objective на датасете ti_5k; vision tower исключён из обучения. В адаптере 86.6M обучаемых параметров при r 32 и alpha 64.

Ключевые особенности

  • LoRA r 32, alpha 64, 86.6M trainable params; vision tower excluded.
  • Pairsum contrastive objective: beta 0.1, lam 1.0, lr 1e-5, global batch 32.
  • Финальный чекпоинт: шаг 507 из 507, 1 эпоха.
  • Внутренняя валидация на 835 группах: val_acc 0.8917, val_acc_group 0.8527.
  • База — Qwen/Qwen3.5-9B на ревизии c202236235762e1c871ad0ccb60c8ee5ba337b9a.

Пример

from transformers import AutoModelForImageTextToText
from peft import PeftModel
base = AutoModelForImageTextToText.from_pretrained("Qwen/Qwen3.5-9B", revision="c202236235762e1c871ad0ccb60c8ee5ba337b9a", dtype="bfloat16")
model = PeftModel.from_pretrained(base, "rooty2020/Qwen3.5-9B-ti5k-pairsum-LoRA")
ПодробнееВалидация на 835 группах: val_acc 0.8917, val_acc_group 0.8527, val_z 4.986 на шаге 507.
Обучение: 4× NVIDIA GH200 120GB, 14 ч 08 мин; torch 2.11.0, transformers 5.14.1, peft 0.20.0.
Ограничения: финальный чекпоинт, vision tower не обучался; метрики только на внутренней валидации ti_5k.
Discuss Telegram 0 13

Will it run?

Working out the memory needs

The bot is still collecting this model's weight sizes; the estimate will appear here in a few minutes. Meanwhile you can rent a GPU with headroom.

Massed ComputeOpen Massed Compute

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    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.

    # rooty2020/Qwen3.5-9B-ti5k-pairsum-LoRA
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download rooty2020/Qwen3.5-9B-ti5k-pairsum-LoRA --local-dir ~/models/Qwen3.5-9B-ti5k-pairsum-LoRA
    # The model is in ~/models. How to run it is in the model card (link above).
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

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.

Model family

New posts about versions of Qwen3.5-9B-ti5k-pairsum-LoRA (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

rooty2020/Qwen3.5-9B-ti5k-pairsum-LoRA
Author
rooty2020
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

130

Ask AI about this model

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

Comments

Robotics

LIME

LIME