Skip to content
Other

RWKV7 A-share 1-min Quantile 21M

вероятностный прогноз минутных баров

2 min readupdated

Это компактная RWKV7-модель на 21 017 102 параметра для прогноза минутных OHLCV китайского рынка A-share. По 64 минутам истории и 23 признакам она выдаёт девять квантилей цены закрытия на 1–30 минут вперёд — от 0.1 до 0.9. Веса обучены с нуля, модель оформлена как standalone Transformers PreTrainedModel с safetensors.

Ключевые особенности

  • Вход — 64 минуты × 23 признака; на выходе 30 горизонтов × 9 квантилей, медиана — пятый квантиль.
  • Квантили 0.1–0.9 дают 60% и 80% интервалы через price_quantiles.
  • Standalone PreTrainedModel: safetensors, кастомные AutoConfig/AutoModel, CLI-скрипт и Python API.
  • Архитектура: 24 слоя, hidden size 256, intermediate size 1024; backbone — transformers-rwkv.
  • Строгий контракт данных: OHLCV и рыночные статистики, минимум 65 баров, без заполнения пропусков и переноса через обеденный перерыв.

Пример

import torch
from transformers import AutoModel

model = AutoModel.from_pretrained(
    'Caizus/rwkv7-ashare-1min-quantile-21m',
    trust_remote_code=True, dtype=torch.bfloat16
).cuda().eval()
ПодробнееНа фиксированном тесте: 4096 окон, 2919 бумаг, 21 дата. Pinball loss после сортировки — 12.23144 bp против 12.84557 bp у безусловных квантилей; покрытие 60% и 80% интервалов — 60.509% и 80.054%.
При этом MAE медианы за 30 минут — 0.30772% против 0.30599% у «цена не изменится», то есть по этому показателю модель не лучше константы.
Нужен NVIDIA CUDA GPU; проверено на Python 3.12, PyTorch 2.14.0+cu130 и RTX PRO 6000 Blackwell. Выбран чекпоинт 500-го шага; обучение с нуля, 24 слоя.
Discuss Telegram 0 32

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.0 GB · ready files0.5 GB
  • FP8 / INT8recommendedweights 0.0 GB0.5 GB
  • 4-bitweights 0.0 GB0.5 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 2060 · BF16 — 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.

    # Caizus/rwkv7-ashare-1min-quantile-21m
    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 Caizus/rwkv7-ashare-1min-quantile-21m --local-dir ~/models/rwkv7-ashare-1min-quantile-21m
    # 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 rwkv7-ashare-1min-quantile-21m (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

Caizus/rwkv7-ashare-1min-quantile-21m
Author
Caizus
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

320

Ask AI about this model

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

Comments