Skip to content
Other

ToR-OT 2D

оптимальный транспорт по токенам

2 min read

ToR-OT 2D — исследовательская модель на 157.6M параметров из статьи Learning Optimal Transport Plans Via Autoregressive Token Regression (NeurIPS 2026). Encoder читает две гистограммы и тип cost, а decoder выдаёт optimal transport plan построчно в виде токенов: указатель на target-bin, экспонента и четыре мантиссы. Модель предназначена для 2D-результатов статьи.

Ключевые особенности

  • Статья: Learning Optimal Transport Plans Via Autoregressive Token Regression (NeurIPS 2026).
  • Encoder принимает две гистограммы и тип cost; decoder пишет план построчно токенами (target-bin pointer, exponent, 4 mantissa digits).
  • Обучение: 2D-гистограммы на k×k сетках в [0,1]^2, k=3–10, n=9–100 bins, шесть cost types; в половине маргиналов 10–70% bins пустые.
  • Декодирование в статье: greedy, grammar-constrained, row by row; последняя строка — моделью (last_row='model').
  • Веса float32, EMA после 620000 шагов; в репозитории есть код загрузки/предсказания и baselines Meta OT и UNOT.

Пример

import torch
from torot.evaluate import load_model
from torot.decode import predict_plans
from torot.data.cost import COST_TYPES, cost_matrix
from torot.data.supports import grid_2d
device = torch.device('cuda')
model, tok, cfg = load_model('ivanjaenm/torot-2d', device)
plan = predict_plans(model, tok, [inst], device, last_row='model')[0]
ПодробнееБенчмарков в карточке нет; пример кода считает relative cost gap к точному плану.
Веса float32, 157.6M параметров; требования к VRAM не указаны.
Обучение: EMA после 620000 шагов; данные — 2D-гистограммы k=3–10, n=9–100, шесть cost types.
Ограничение: модель для 2D-планов из статьи, не универсальный OT-солвер.
Discuss Telegram 0 20

Will it run?

Runs in F32

Fits without quantisation.

  • F32weights 0.6 GB · ready files1.1 GB
  • BF16weights 0.3 GB0.8 GB
  • FP8 / INT8recommendedweights 0.1 GB0.7 GB
  • 4-bitweights 0.1 GB0.6 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 2060 · F32 — from RTX 2060

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

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

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.

    # ivanjaenm/torot-2d
    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 ivanjaenm/torot-2d --local-dir ~/models/torot-2d
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to torot-2d
  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 torot-2d (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

ivanjaenm/torot-2d
Author
ivanjaenm
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

200

Ask AI about this model

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

Comments