---
license: apache-2.0
datasets:
- laurievb/OpenLID-v2
---


# WordLlama Detect

**WordLlama Detect** is a [WordLlama](https://github.com/dleemiller/WordLlama)-like library focused on the task of language identification.
It supports identification of **148 languages**, and high accuracy and fast CPU & numpy-only inference.
WordLlama detect was trained from static token embeddings extracted from *Gemma3*-series LLMs.

<p align="center">
  <img src="https://cdn-uploads.huggingface.co/production/uploads/65ff92ea467d83751a727538/6xqLD9ciun2KIgCiC9w6T.png" alt="WordLlamaDetect" width="60%">
</p>

## Overview

**Features:**
- NumPy-only inference with no PyTorch dependency
- Pre-trained model (148 languages), with 103 @ >95% accuracy
- Sparse lookup table (13MB)
- Fast inference: >70k texts/s single thread
- Simple interface

## Installation

```bash
pip install wldetect
```

Or install from source:
```bash
git clone https://github.com/dleemiller/WordLlamaDetect.git
cd WordLlamaDetect
uv sync
```

## Quick Start

### Python API

```python
from wldetect import WLDetect

# Load bundled model (no path needed)
wld = WLDetect.load()

# Detect language for single text
lang, confidence = wld.predict("Hello, how are you today?")
# ('eng_Latn', 0.9564036726951599)
```

### CLI Usage

```bash
# Detect from text
uv run wldetect detect --text "Bonjour le monde"

# Detect from file
uv run wldetect detect --file input.txt
```

## Included Model

WLDetect ships with a pre-trained model based on concatenated Gemma3-27B + Gemma3-4B token embeddings:
- **Languages**: 148 (from OpenLID-v2 dataset)
- **Accuracy**: 92.92% on FLORES+ dev set
- **F1 (macro)**: 92.74%
- **Language codes**: ISO 639-3 + ISO 15924 script (e.g., `eng_Latn`, `cmn_Hans`, `arb_Arab`)


> [!TIP]
> See [docs/languages.md](docs/languages.md) for the complete list of supported languages with performance metrics.

> [!NOTE]  
> Gemma3 is a good choice for this application, because it was trained on over 140 languages.
> The tokenizer, vocab size (262k) and multi-language training are critical for performance.

## Architecture

### Simple Inference Pipeline (NumPy-only)

1. **Tokenize**: Use HuggingFace fast tokenizer (512-length truncation)
2. **Lookup**: Index into pre-computed exponential lookup table (vocab_size × n_languages)
3. **Pool**: LogSum pooling over token sequence
4. **Softmax**: Calculate language probabilities

The lookup table is pre-trained using: `exp((embeddings * token_weights) @ projection.T + bias)`,
where embeddings are frozen token embeddings from Gemma3, trained with focal loss on OpenLID-v2.
During training, token vectors are aggregated using *logsumexp* pooling along the sequence dimension.


> [!IMPORTANT]  
> To optimize artifact size and compute, we perform `exp(logits)` before saving the lookup table.
> Then we apply a threshold to make the table *sparse*.
> This reduces the artifact size 10x (~130mb -> 13mb), with negligable performance degradation.

### Sparse Lookup Table

The lookup table uses sparse COO (Coordinate) format with configurable sparsification threshold:
- **Sparsity**: 97.15% (values below threshold (<10) set to zero)
- **Format**: COO (row, col, data) indices stored as int32, values as fp32
- **Performance impact**: Negligible (0.003% accuracy loss)


## Performance

### FLORES+ Benchmark Results

Evaluated on FLORES+ dataset (148 languages, ~1k sentences per language):

| Split   | Accuracy | F1 (macro) | F1 (weighted) | Samples  |
|---------|----------|------------|---------------|----------|
| dev     | 92.92%   | 92.74%     | 92.75%        | 150,547  |
| devtest | 92.86%   | 92.71%     | 92.69%        | 153,824  |

See [docs/languages.md](docs/languages.md) for detailed results.

### Inference Speed

Benchmarked on 12th gen Intel-i9 (single thread):

- **Single text**: 71,500 texts/second (0.014 ms/text)
- **Batch (1000)**: 82,500 texts/second (12.1 ms/batch)

## Supported Languages

The bundled model supports 148 languages from the OpenLID-v2 dataset. Languages use ISO 639-3 language codes with ISO 15924 script codes (e.g., `eng_Latn`, `cmn_Hans`, `arb_Arab`).

See [model_config.yaml](src/wldetect/models/model_config.yaml) for the complete list of supported languages.

## Training

### Installation for Training

```bash
# CPU or default CUDA version
uv sync --extra training

# With CUDA 12.8 (Blackwell)
uv sync --extra cu128
```

### Training Pipeline

1. **Configure model** in `configs/models/custom-config.yaml`:
```yaml
model:
  name: google/gemma-3-27b-pt
  hidden_dim: 5376
  shard_pattern: model-00001-of-00012.safetensors
  embedding_layer_name: language_model.model.embed_tokens.weight

languages:
  eng_Latn: 0
  spa_Latn: 1
  fra_Latn: 2
  # ... add more languages

inference:
  max_sequence_length: 512
  pooling: logsumexp
```

2. **Configure training** in `configs/training/custom-training.yaml`:
```yaml
model_config_path: "configs/models/custom-model.yaml"

dataset:
  name: "laurievb/OpenLID-v2"
  filter_languages: true

training:
  batch_size: 1536
  learning_rate: 0.002
  epochs: 2
```

3. **Train**:
```bash
uv run wldetect train --config configs/training/custom-training.yaml
```

Artifacts saved to `artifacts/`:
- `lookup_table_exp.safetensors` - Sparse exp lookup table (for inference)
- `projection.safetensors` - Projection matrix (fp32, for fine-tuning)
- `model_config.yaml` - Model configuration
- `model.pt` - Full PyTorch checkpoint

### Training Commands

```bash
# Train model
uv run wldetect train --config configs/training/gemma3-27b.yaml

# Evaluate on FLORES+
uv run wldetect eval --model-path artifacts/ --split dev

# Generate sparse lookup table from checkpoint (default: threshold=10.0)
uv run wldetect create-lookup \
  --checkpoint artifacts/checkpoints/checkpoint_step_100000.pt \
  --config configs/training/gemma3-27b.yaml \
  --output-dir artifacts/
```

### Training Details

- **Embedding extraction**: Downloads only embedding tensor shards from HuggingFace (not full models)
- **Dataset**: OpenLID-v2 with configurable language filtering and balancing
- **Model**: Simple linear projection (hidden_dim → n_languages) with dropout
- **Pooling**: LogSumExp or max pooling over token sequences
- **Training time**: ~2-4 hours on GPU for 2 epochs (150 languages, 5000 samples/language)
- **Evaluation**: Automatic FLORES+ evaluation after training

## License

Apache 2.0 License

## Citations

If you use WordLlama Detect in your research or project, please consider citing it as follows:

```bibtex
@software{miller2025wordllamadetect,
  author = {Miller, D. Lee},
  title = {WordLlama Detect: The Language of the Token},
  year = {2025},
  url = {https://github.com/dleemiller/WordLlamaDetect},
  version = {0.1.0}
}
```

## Acknowledgments

- OpenLID-v2 dataset: [laurievb/OpenLID-v2](https://huggingface.co/datasets/laurievb/OpenLID-v2)
- FLORES+ dataset: [openlanguagedata/flores_plus](https://huggingface.co/datasets/openlanguagedata/flores_plus)
- HuggingFace transformers and tokenizers libraries
- Google Gemma model team