library_name: transformers
model_index:
- name: Lance ASR
results: []
tags:
- automatic-speech-recognition
- asr
- pytorch
- transformer
license: apache-2.0
Lance ASR – The Foundation of Speech Intelligence
🚀 Lance ASR is a custom-built Automatic Speech Recognition (ASR) model designed for high-efficiency local and cloud inference. It utilizes a Transformer Encoder-Decoder architecture with convolutional subsampling for processing acoustic features.
🌟 Key Features
✅ Custom Architecture: Not a Whisper clone; features a bespoke Conv1d-subsampling audio front-end.
✅ Hugging Face Compatible: Fully integrates with transformers via AutoModelForSeq2SeqLM.
✅ Optimized for Precision: Uses bfloat16 for high-performance inference and training.
✅ Scalable Design: Optimized for 768 hidden dims and 4 layers, balancing speed and accuracy.
✅ Seamless Tokenization: Uses the DWDMaiMai/tiktoken_cl100k_base tokenizer for efficient text representation.
📥 Installation & Setup
Load Lance ASR directly from your local directory or the Hugging Face Hub:
import torch
from transformers import AutoTokenizer, AutoFeatureExtractor, AutoModelForSeq2SeqLM
model_name = "NeuraCraft/Lance-ASR"
tokenizer = AutoTokenizer.from_pretrained(model_name)
feature_extractor = AutoFeatureExtractor.from_pretrained(model_name)
model = AutoModelForSeq2SeqLM.from_pretrained(model_name, torch_dtype=torch.bfloat16)
🛠 Usage Example
Lance ASR can transcribe audio by processing log-mel spectrograms:
# 1. Prepare audio features (e.g., from a .wav file)
# inputs = feature_extractor(audio_array, sampling_rate=16000, return_tensors="pt")
# 2. Generate transcription
model.eval()
with torch.no_grad():
generated_ids = model.generate(
inputs.input_features.to(torch.bfloat16),
max_new_tokens=250,
pad_token_id=tokenizer.eos_token_id
)
transcription = tokenizer.decode(generated_ids[0], skip_special_tokens=True)
print(f"Transcription: {transcription}")
📊 Performance & Evaluation
Lance ASR is currently in its early stages, and performance is being actively tested. Initial evaluations focus on: 🔹 WER (Word Error Rate) – Measures transcription accuracy 🔹 CER (Character Error Rate) – Measures character-level precision 🔹 Inference Latency – Optimized for real-time local processing
✅ Planned Enhancements
🔹 Larger training datasets (e.g., Common Voice, LibriSpeech) 🔹 Advanced noise-robustness for real-world environments 🔹 Multilingual ASR support for global accessibility
🚀 Future Roadmap
Lance ASR is just getting started! The goal is to transform it into the core auditory component of an advanced AI assistant. 📅 Planned Features:
🔜 Real-time live transcription & streaming support
🔜 Multi-speaker identification (Diarization)
🔜 Integrated Voice Activity Detection (VAD)
🔜 High-efficiency deployment for mobile and edge devices
🏗 Development & Contributions
Lance ASR is being developed by NeuraCraft. Contributions, suggestions, and testing feedback are welcome!
📬 Contact & Updates:
Developer: NeuraCraft
Project Status: 🚧 In Development
Follow for updates: Coming soon