tahamajs's picture
docs: upgrade model card with universal loader and verified hardware telemetry
2f5ed52 verified
|
Raw History Blame Contribute Delete
6.24 kB
metadata
language:
  - en
license: mit
license_link: https://github.com/tahamajs/NeuralOps/blob/main/LICENSE
library_name: neuralops
base_model: google-bert/bert-base-uncased
datasets:
  - nyu-mll/glue
metrics:
  - accuracy
tags:
  - neuralops
  - bert
  - model-compression
  - compressor-zoo
  - lora-healing
  - glue-sst2
  - quantization
  - smoothquant
pipeline_tag: text-classification
model-index:
  - name: bert-comp-smoothquant-sst2
    results:
      - task:
          type: text-classification
          name: Sentiment Classification
        dataset:
          type: nyu-mll/glue
          name: GLUE SST-2
          config: sst2
          split: validation
        metrics:
          - type: accuracy
            value: 91.74
            name: GLUE SST-2 Validation Accuracy

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

SST-2 Acc Teacher Baseline Compression Ratio Showcase Rank Weights License

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 smoothquant (SmoothQuantCompressor) engine from the NeuralOps Library at Pareto-selected rank k=32 (1.00x compression ratio, -0.0% parameter reduction).


Compression & Architecture Profile

SmoothQuantCompressor: Unified BaseCompressor wrapper for SmoothQuant W8A8 with compressed-tensors support.

Property Specification
Registry Key (COMPRESSOR_REGISTRY) smoothquant
NeuralOps Class & Family SmoothQuantCompressor (quantization)
Source Implementation neuralops/compression/quantization/smoothquant.py
Base Architecture google-bert/bert-base-uncased (109,483,778 parameters, 12 layers, $D=768$)
Showcase Rank (k) 32 (LoRA Recovery Applied: False)
Compressed Parameter Count 109,483,778 (-0.00% reduction, 1.000x physical compression ratio)
Teacher Baseline SST-2 Acc 91.63%
Showcase Final SST-2 Acc 91.74% (Zero-Shot: 91.74%, Parity Delta: +0.11%)

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 1.00x 109,483,778 (-0.0%) 91.74% 91.74% +0.11%
128 1.00x 109,483,778 (-0.0%) 91.74% 91.74% +0.11%
64 1.00x 109,483,778 (-0.0%) 91.74% 91.74% +0.11%
32 ⭐ 1.00x 109,483,778 (-0.0%) 91.74% 91.74% +0.11%

(⭐ 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: "smoothquant" and compression_rank: 32.
  • 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-smoothquant-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": "smoothquant", "rank": 32}]})

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)

Quick Start (Universal NeuralOps Hub API)

import torch
from neuralops.hub import load_surgeried_model

# Load model and tokenizer/processor in one line directly from Hugging Face Hub
model, tokenizer = load_surgeried_model("tahamajs/bert-comp-smoothquant-sst2")
model.eval()
print(f"Successfully loaded tahamajs/bert-comp-smoothquant-sst2 with NeuralOps modular surgical weights!")

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}
}