BERT-Base (compot_compressor) — NeuralOps Compressed & LoRA-Healed Model Zoo

This repository provides a physically compressed and LoRA-healed checkpoint of google-bert/bert-base-uncased fine-tuned on GLUE SST-2, compressed using the compot_compressor (compot_compressor_factory) engine from the NeuralOps Library at Pareto-selected rank k=64 (2.67x compression ratio, -62.5% parameter reduction).

Compression & Architecture Profile

compot_compressor_factory: Factory function for COMPOT compression.

Property	Specification
Registry Key (`COMPRESSOR_REGISTRY`)	`compot_compressor`
NeuralOps Class & Family	`compot_compressor_factory` (`low_rank`)
Source Implementation	`neuralops/compression/low_rank/compot_compressor.py`
Base Architecture	`google-bert/bert-base-uncased` (`109,483,778` parameters, 12 layers, $D=768$)
Showcase Rank (`k`)	`64` (`LoRA Recovery Applied: True`)
Compressed Parameter Count	`41,064,194` (`-62.49%` reduction, `2.666x` physical compression ratio)
Teacher Baseline SST-2 Acc	`91.63%`
Showcase Final SST-2 Acc	`76.15%` (Zero-Shot: `50.92%`, Parity Delta: `-15.48%`)

Multi-Rank Pareto Evaluation (`k=256, 128, 64, 32`)

Evaluated physically on NVIDIA GeForce RTX 3090 (24GB) on the GLUE SST-2 validation split (872 samples). Whenever zero-shot compression drops accuracy by > 2.0% alongside > 1.10x reduction, NeuralOps applies targeted r=8 LoRA healing (50 steps, lr=2e-4, 45s budget):

Rank (`k`)	Physical Ratio	Compressed Params (Reduction)	Zero-Shot Acc	Healed Acc	Parity vs Teacher (`91.63%`)
`256`	`1.21x`	`90,609,410` (`-17.2%`)	`83.14%`	`61.58%` (LoRA)	`-30.05%`
`128`	`1.90x`	`57,579,266` (`-47.4%`)	`53.90%`	`50.92%` (LoRA)	`-40.71%`
`64` ⭐	`2.67x`	`41,064,194` (`-62.5%`)	`50.92%`	`76.15%` (LoRA)	`-15.48%`
`32`	`2.83x`	`38,704,898` (`-64.6%`)	`50.92%`	`71.67%` (LoRA)	`-19.96%`

(⭐ indicates the serialized checkpoint packaged in this Hugging Face repository.)

Packaged Artifacts

pytorch_model.bin: Full serialized compressed BERT state dictionary (with low-rank factors / quantization buffers / LoRA healing adapters baked in).
config.json: Model configuration tagged with compression_method: "compot_compressor" and compression_rank: 64.
compression_summary.json: Complete machine-readable telemetry across all evaluated ranks [256, 128, 64, 32].
Tokenizer Assets: Standard bert-base-uncased vocabulary (vocab.txt, tokenizer.json, tokenizer_config.json).

Quick Start (Reconstruction & Inference)

import torch
from huggingface_hub import hf_hub_download
from transformers import AutoTokenizer, BertForSequenceClassification
from neuralops.core.compression_pipeline import compress_model

repo_id = "tahamajs/bert-comp-compot-compressor-sst2"
tokenizer = AutoTokenizer.from_pretrained(repo_id)

# Reconstruct compressed NeuralOps topology and load serialized compressed + healed weights
model = BertForSequenceClassification.from_pretrained("google-bert/bert-base-uncased", num_labels=2)
model = compress_model(model, {"compressors": [{"method": "compot_compressor", "rank": 64}]})

ckpt_path = hf_hub_download(repo_id=repo_id, filename="pytorch_model.bin")
state_dict = torch.load(ckpt_path, map_location="cpu")
model.load_state_dict(state_dict, strict=False)
model.eval()

inputs = tokenizer("NeuralOps achieves strong compression without sacrificing downstream accuracy.", return_tensors="pt")
with torch.no_grad():
    logits = model(**inputs).logits
    pred = logits.argmax(dim=-1).item()
print("Predicted SST-2 Sentiment (0=Negative, 1=Positive):", pred)

Citation

@software{neuralops2026compressorzoo,
  title   = {NeuralOps: Multi-Rank Compressed BERT-SST2 Zoo with Pareto LoRA Recovery},
  author  = {Majlesi, Taha},
  year    = {2026},
  url     = {https://github.com/tahamajs/NeuralOps}
}