Skip to content
Audio

Khmer ASR Conformer-CTC v1

12,3 млн параметров для кхмерской речи

1 min read

Модель распознавания речи от PhonSobon, обученная с нуля без предобученных весов. Это Conformer-CTC-энкодер на 12,3 млн параметров: принимает 16 кГц mono-аудио и выдаёт кхмерские символы, цифры и латиницу для английских слов, встречающихся в кхмерской речи.

Ключевые особенности

  • Архитектура conformer_s: 16 слоёв, d_model 176, 4 головы, kernel 31, RoPE-внимание.
  • Двуязычный вывод: кхмерский текст, цифры ០–៩ и 0–9, латиница a–z для code-switching.
  • Быстрая: около 172x real time на ноутбучной NVIDIA RTX 500 Ada и 68x на CPU.
  • Greedy CTC-декодирование без языковой модели; есть доступ к log_probs() для своего beam search.
  • Обучена на 25 339 клипах (56,3 ч) из 18 кхмерских ASR-датасетов.

Пример

from huggingface_hub import hf_hub_download
import sys, os
repo = "phonsobon/khmer-asr-Conformer-CTC-v1"
sys.path.insert(0, os.path.dirname(hf_hub_download(repo, "khmer_asr_conformer.py")))
from khmer_asr_conformer import KhmerASR
asr = KhmerASR.from_pretrained(repo)
print(asr.transcribe("audio.wav"))
ПодробнееВалидационный CER — 0,93%, но это те же источники, что и в обучении. На новых записях ожидайте CER около 20–40%; на 4 незнакомых клипах — 25,1%.
Нет пунктуации, пробелы зависят от разметки; возможны не-слова без языковой модели.
Аудио длиннее 30 с режется фиксированными кусками — лучше сначала делить по тишине.
Discuss Telegram 0

Will it run?

Working out the memory needs

The bot is still collecting this model's weight sizes; the estimate will appear here in a few minutes. Meanwhile you can rent a GPU with headroom.

Massed ComputeOpen Massed Compute

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # phonsobon/khmer-asr-Conformer-CTC-v1
    # ===== Tokens: paste yours between the quotes =====
    export HF_TOKEN=""        # huggingface.co/settings/tokens, for gated models
    # ==========================================
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download phonsobon/khmer-asr-Conformer-CTC-v1 --local-dir ~/models/khmer-asr-Conformer-CTC-v1
    # The model is in ~/models. How to run it is in the model card (link above).
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of khmer-asr-Conformer-CTC-v1 (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

phonsobon/khmer-asr-Conformer-CTC-v1
Author
phonsobon
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments