Skip to content
Other

CatalystCompass

EC number prediction for chemical reactions

2 min readupdated

This model takes a chemical reaction description and predicts its Enzyme Commission (EC) classification code—the international standard for enzyme nomenclature. Under the hood, it runs a CPU ensemble that combines structural reaction fingerprints, BERT-based neural encoders, and nearest-neighbor search across known chemical analogs. The author added a conservative consensus layer for exact reaction matching alongside a fast top-class decoder, making it a practical tool for bioinformaticians and chemists annotating datasets without a GPU.

Key features

  • Combines structural substrate representations, learned embeddings from BERT encoders, and retrieval over known reactions.
  • Strict match lookup checks stereochemistry, isotope labels, reaction direction, and reactant roles, accepting a match only when reference entries reach full consensus.
  • Fast top-one decoder selects the highest-scoring class without sorting all categories.
  • Runs entirely on CPU and executes inference inside an isolated, offline Docker container.

What you can do

  • Batch-annotate chemical reaction datasets from Parquet files and export valid EC numbers in JSONL format.
  • Drop it into offline biochemical analysis pipelines that require reproducibility and isolation from environment version conflicts.

Example

mkdir -p ~/Downloads/DL_project
unzip -o ~/Downloads/catalyst_compass_setup.zip -d ~/Downloads/DL_project
bash ~/Downloads/DL_project/catalyst_compass/run_mac.sh
Read moreDetails
The first run downloads roughly 727 MB of base assets, while actual inference runs completely offline.
By default, inputs are read from DL_project/data/val_inputs.parquet and predictions are saved to output/predictions.jsonl.
Changes focus on inference logic built on top of pretrained weights; no accuracy gain was recorded on the public validation set due to a lack of canonical matches.
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.

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

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

About the model

Saiteja-madineni/catalyst-compass
Author
Saiteja-madineni
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