tzervas/bwsk-mamba-130m

BWSK Mamba-130M

Mamba-130M (130M params) trained in 6 variants (3 BWSK modes x 2 experiments) on WikiText-2 with full convergence training and early stopping.

This repo contains all model weights, configs, and training results in a single consolidated repository.

What is BWSK?

BWSK is a framework that classifies every neural network operation as S-type (information-preserving, reversible, coordination-free) or K-type (information-erasing, synchronization point) using combinator logic. This classification enables reversible backpropagation through S-phases to save memory, and CALM-based parallelism analysis.

Model Overview

Property	Value
Base Model	state-spaces/mamba-130m-hf
Architecture	Ssm (ssm_lm)
Parameters	130M
Dataset	WikiText-2
Eval Metric	Perplexity

S/K Classification

Type	Ratio
S-type (information-preserving)	86.0%
K-type (information-erasing)	0.0%
Gray (context-dependent)	14.0%

Fine-tune Results

Mode	Final Loss	Val Perplexity	Test Perplexity	Peak Memory	Time	Epochs
Conventional	2.2904	15.75	15.30	3.0 GB	11.6m	3
BWSK Analyzed	3.0122	15.76	15.27	3.0 GB	7.7m	2
BWSK Reversible	3.0290	15.76	15.27	3.0 GB	8.0m	2

Memory savings (reversible vs conventional): 0.0%

From Scratch Results

Mode	Final Loss	Val Perplexity	Test Perplexity	Peak Memory	Time	Epochs
Conventional	5.5343	443.75	453.48	3.0 GB	19.6m	5
BWSK Analyzed	6.7067	627.40	643.69	3.0 GB	19.6m	5
BWSK Reversible	6.3691	648.40	666.03	3.0 GB	19.7m	5

Memory savings (reversible vs conventional): 0.0%

Repository Structure

├── README.md
├── results.json
├── finetune-conventional/
│   ├── model.safetensors
│   ├── config.json
│   └── training_results.json
├── finetune-bwsk-analyzed/
│   ├── model.safetensors
│   ├── config.json
│   └── training_results.json
├── finetune-bwsk-reversible/
│   ├── model.safetensors
│   ├── config.json
│   └── training_results.json
├── scratch-conventional/
│   ├── model.safetensors
│   ├── config.json
│   └── training_results.json
├── scratch-bwsk-analyzed/
│   ├── model.safetensors
│   ├── config.json
│   └── training_results.json
├── scratch-bwsk-reversible/
│   ├── model.safetensors
│   ├── config.json
│   └── training_results.json

Usage

Load a specific variant:

from transformers import AutoModelForCausalLM, AutoTokenizer

# Load fine-tuned conventional variant
model = AutoModelForCausalLM.from_pretrained(
    "tzervas/bwsk-mamba-130m", subfolder="finetune-conventional"
)
tokenizer = AutoTokenizer.from_pretrained(
    "tzervas/bwsk-mamba-130m", subfolder="finetune-conventional"
)

# Load from-scratch BWSK reversible variant
model = AutoModelForCausalLM.from_pretrained(
    "tzervas/bwsk-mamba-130m", subfolder="scratch-bwsk-reversible"
)

Training Configuration

Setting	Value
Optimizer	AdamW
LR (fine-tune)	3e-05
LR (from-scratch)	2e-04
LR Schedule	Cosine with warmup
Max Grad Norm	1.0
Mixed Precision	AMP (float16)
Early Stopping	Patience 3
Batch Size	2
Sequence Length	256

Links

• GitHub Repository
• Whitepaper
• Full Training Report

Citation

@software{zervas2026bwsk,
  author = {Zervas, Tyler},
  title = {BWSK: Combinator-Typed Neural Network Analysis},
  year = {2026},
  url = {https://github.com/tzervas/ai-s-combinator},
}

License

MIT