---
title: Open ASR Leaderboard configuration for Fluid Speech
emoji: 🎙️
colorFrom: blue
colorTo: green
sdk: docker
hardware: a100-large
pinned: false
---
The NeMo configuration (`hf-audio/open-asr-leaderboard-nemo`) with NeMo 3.0 instead of 2.7.2 and two files added:
`fluid_speech.py` (loads `FluidVoice-Labs/fluid-speech-pro`, whose encoder has one extra block, with its optional clean
mode) and `run_eval.py` (the NeMo script with that loader, a `--clean_mode` flag and batched TDT beam search with 4
beams).