---
title: Open ASR Leaderboard - MOSS-Transcribe
emoji: 🎤
colorFrom: blue
colorTo: indigo
sdk: docker
pinned: false
---

# Open ASR Leaderboard — MOSS-Transcribe-preview-2B eval environment

Docker eval environment for [`OpenMOSS-Team/MOSS-Transcribe-preview-2B`](https://huggingface.co/OpenMOSS-Team/MOSS-Transcribe-preview-2B)
on the [Open ASR Leaderboard](https://huggingface.co/spaces/hf-audio/open_asr_leaderboard),
following the `standardize_transformers` spec (new English normalizer; raw outputs
saved, normalization applied at scoring time).

Contents:
- `Dockerfile` — software stack (transformers ≥ 4.57.1, torch 2.8 cu128, librosa, jiwer, …)
- `run_eval.py` — Moss inference (WhisperFE log-mel + MossProcessor + greedy decode); saves raw outputs
- `chat_template_default.py` — dataset-agnostic chat template
- `normalizer/` — standardized English normalizer + `score_results`

Launched via HF Jobs from `moss-transcribe/submit_jobs_moss-transcribe.sh` in the
[open_asr_leaderboard](https://github.com/huggingface/open_asr_leaderboard) repo.