Skip to content
Other

iTransformer Electricity with CST

Time-Series Forecasting with Atomic Weights

2 min readupdated

The model tackles multivariate time-series forecasting using the Electricity power consumption dataset. It takes the previous 96 hourly measurements across 321 time series and predicts the next 96 hours. This is an independent variation of the baseline iTransformer architecture, where standard projections are replaced with Polar Triweight Atoms (special atomic weights). The model has 15.5% fewer parameters compared to the reference Dense model and is suited for analyzing and forecasting power system metrics.

Key features

  • Atomic projections: Replaces encoder FFN projections with Polar Triweight Atoms while retaining trainable bias.
  • Parameter reduction: Contains 4,084,320 parameters, which is 15.5% fewer than the reference model.
  • Compatibility: Supports running on GPU and CPU through a standard CLI interface.
  • Complete package: Includes scaler configuration, inference scripts, and the original architecture code.

What you can do

  • Power consumption forecasting: Generate a 96-hour ahead forecast from a CSV file containing historical data for 321 series.
  • Configuration comparison: Run inference using the reference variant (--variant dense) to evaluate accuracy.
  • Local execution: Run inference on a CPU or GPU using the provided safetensors weights.
  • Training reproduction: Run training scripts on the original THUML Electricity benchmark dataset.

Example

Command to clone the repository and run model inference:

python -m pip install huggingface_hub
hf download takumiecd/cst-itransformer-electricity --local-dir itr-ecl
python -m pip install -r itr-ecl/requirements.txt
python itr-ecl/inference.py itr-ecl electricity.csv prediction.csv --device cuda
Read moreDetails
Standardized test MSE: Dense 0.14778, CST 0.14982
Standardized test MAE: Dense 0.23948, CST 0.24189
Peak VRAM usage on RTX 4090: Dense 809.4 MiB, CST 951.6 MiB
Training: Dense on A100 MIG, CST on RTX 4090
Discuss Telegram 0

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.

    # takumiecd/cst-itransformer-electricity
    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 takumiecd/cst-itransformer-electricity --local-dir ~/models/cst-itransformer-electricity
    # 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 cst-itransformer-electricity (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

takumiecd/cst-itransformer-electricity
Author
takumiecd
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

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

Comments