Skip to content
Robotics

FINE_InternVLA_N1_R2R_100

навигация по инструкциям без lookahead

1 min read

Это чекпойнт FINE (Future-Informed Navigation Encoding), дообученный на 100% тренировочных демонстраций R2R-CE. Модель построена на InternVLA-N1 и добавляет обучение на будущих наблюдениях, чтобы эффективнее использовать демонстрации для vision-language navigation.

Ключевые особенности

  • FINE добавляет supervision будущих наблюдений без дополнительной разметки или взаимодействия.
  • 18 явных landmark-токенов идут по ориентирам инструкции и предсказывают область следующего ориентира в 2D semantic (Qwen2.5-VL patch) и 3D geometric (VGGT) признаках.
  • Неявный future-токен обучается контрастивно (InfoNCE) отличать реально достигнутое будущее от других, включая контрфактические rollouts видео-world model.
  • Оба потока токенов подаются в System 1 через отдельные gated projectors.
  • На инференсе сохраняются исходные action space и action head InternVLA-N1; lookahead не нужен.

Пример

Сценарий из карточки: робот получает историю RGB-D из 8 кадров 224×224, текстовую инструкцию и landmark-токены; System 2 выдаёт пиксельную цель или дискретные действия, а System 1 — непрерывную локальную траекторию.

ПодробнееАрхитектура: Qwen2.5-VL-7B System 2 + NextDiT (nextdit_async) System 1, 8,61 млрд параметров.
Обучение: 100% демонстраций R2R-CE, landmark-маски и контрфактические будущие сцены от world model.
Будущие кадры, маски, VGGT и rollouts используются только при обучении; на деплое lookahead не требуется.
Точность: bfloat16; веса разбиты на 4 safetensors-шарда.
Discuss Telegram 0 7

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 16 GB · ready files18 GB
  • FP8 / INT8recommendedweights 8.0 GB9.3 GB
  • 4-bitweights 4.5 GB5.5 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 3080 · BF16 — from RTX 4090

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.

    # kangnguyen/FINE_InternVLA_N1_R2R_100
    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 kangnguyen/FINE_InternVLA_N1_R2R_100 --local-dir ~/models/FINE_InternVLA_N1_R2R_100
    # 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 InternVLA-N1-DualVLN (quants, finetunes, LoRAs) will show up under the bell.

Fine-tunes 5All on Hugging Face

About the model

kangnguyen/FINE_InternVLA_N1_R2R_100
Author
kangnguyen
Category
Robotics
Released
Posted

On NeyroLearn

Hugging Face

70

Ask AI about this model

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

Comments

Robotics

pi05-sft

pi05-sft