Skip to content
Other

CRSN-T

thermodynamic evolution of recurrent networks

1 min readupdated

CRSN-T (Thermodynamic Crystal Recurrent Sheaf Network) is the third generation of the Crystal Recurrent Sheaf Network architecture, introducing thermodynamic coupling into its layers. The model adds learnable temperature, heat flux, and entropy production for each layer using the GENERIC (General Equation for Non-Equilibrium Reversible-Irreversible Coupling) framework. It improves perplexity scores over previous phases while preserving stable crystalline order. This release will interest researchers working on language model architectures and nonlinear dynamics.

Key features

  • Thermodynamic state: each layer tracks parameters for temperature, heat flux, and entropy.
  • Regularization: thermodynamic coupling acts as implicit regularization.
  • Order stability: the crystalline alpha parameter stays strictly positive across all 8 layers.
  • Improved metrics: lower validation loss and perplexity compared to Phase 2.

What you can do

  • Run language model training experiments using the CRSN-T architecture.
  • Study how thermodynamic parameters affect weight stability and convergence.
  • Integrate thermodynamic coupling components into your own research projects.

Example

Sample probe state structure from the final model checkpoint:

L00: alpha_mean=0.1324 | omega_mean=0.5432
L01: alpha_mean=0.1288 | omega_mean=0.5395
L02: alpha_mean=0.1286 | omega_mean=0.5271
Read moreDetails
Val Loss: 2.5900
PPL (perplexity): 13.33
Dataset: WikiText-103 (126.9M tokens, 25k steps)
Hardware: CPU training (AMD Ryzen)
Architecture: 8 layers, based on CRSN Phase 2
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.1 GB · ready files0.6 GB
  • FP8 / INT8recommendedweights 0.0 GB0.6 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.

    # DuoNeural/CRSN-T
    # ===== Tokens: paste yours between the quotes =====
    export HF_TOKEN=""        # huggingface.co/settings/tokens, for gated models
    # ==========================================
    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 DuoNeural/CRSN-T --local-dir ~/models/CRSN-T
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to CRSN-T
  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 CRSN-T (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

DuoNeural/CRSN-T
Author
DuoNeural
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