res2vec-owt1b-188d

deepworks-net/res2vec-owt1b-188d

res2vec-owt1b-188d is a machine learning model from deepworks-net. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as mit.

Downloads · 30 days

Access

Public

Updated Jan 6, 2026

Repo size

780 MB

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.npy758 MB · 97%

At a glance

License: mit
Access: Public
Created: Jan 6, 2026
Updated: Jan 6, 2026
SHA: f0257dc7

Model card

From the Hugging Face model README

res2vec OpenWebText-1B Word Embeddings

Model at a Glance


Dimensions	188
Vocabulary	1,008,133 words
Embedding Size	723 MB
Total Size	~744 MB
Format	NumPy (.npy)

Model Properties

Metric	Value
Information Capacity (I)	0.106
Average Resonance (R)	0.071
Optimal Dimension (d*)	187.6
Fixed Point	✓

Benchmarks

Model	Dim	Corpus	SimLex-999	WordSim-353
word2vec Skip-gram	300	Wikipedia 1B	0.37	0.63
GloVe	300	Wikipedia+Gigaword 6B	0.42	0.66
res2vec	188	OpenWebText 1B	0.296	0.597

Example Similarities (cosine)

Pair	Similarity
king – queen	0.693
man – woman	0.798
cat – dog	0.854
good – bad	0.855

Usage

import numpy as np
import json

# Load embeddings
embeddings = np.load("res2vec_owt1b_188d_embeddings.npy")  # (1008133, 188)

# Load vocabulary
with open("res2vec_owt1b_188d_vocab.json") as f:
    vocab = json.load(f)  # {"word": index, ...}

Files

File	Description
`res2vec_owt1b_188d_embeddings.npy`	Embedding matrix, shape `(1008133, 188)`, float32
`res2vec_owt1b_188d_vocab.json`	Word-to-index mapping, `{"word": index, ...}`
`res2vec_owt1b_188d_metric.npy`	Learned metric tensor, shape `(188, 188)`, float64
`res2vec_owt1b_188d_results.json`	Training metadata and diagnostics
`res2vec_owt1b_188d_evaluation.json`	Benchmark scores and word pair similarities

This is an initial proof-of-concept/demonstration of the Resonance Theory technique. Will be adding more versions and refining as I go. Please let me know what you think! :-)