Skip to content
Audio

krama-asr-khmer-v1

a hybrid ASR model for Khmer speech recognition

2 min read

The 345M-parameter krama-asr-khmer-v1 handles automatic speech recognition (ASR) for Khmer. Its architecture combines a Google EmbeddingGemma-2 Conformer audio tower, a CTC Reader, and a causal cluster decoder. Trained on 854.59 hours of Khmer audio across seven domains with a 3,072-token vocabulary, the model transcribes short- and medium-length Khmer audio clips at high speed on both CPUs and GPUs.

Key features

  • Base architecture: Google EmbeddingGemma-2 with a 12-layer Conformer audio tower (d=1024).
  • Dataset size: Trained on 349,305 clips totaling over 854 hours of audio.
  • Two decoding modes: Fast greedy CTC (default) or Beam-5 with a causal decoder.
  • Efficiency: High processing speed (around 0.05x RTF on a standard CPU).

What you can do

  • Speech recognition: Transcribe Khmer audio recordings into text using a Python script.
  • Integration: Embed the model into audio processing pipelines and local services.

Example

import torch
from krama_asr import load_model, transcribe

device = "cuda" if torch.cuda.is_available() else "cpu"
model, tokens = load_model("Nukthean/krama-asr-khmer-v1", device=device)

# Быстрый greedy CTC (по умолчанию)
text = transcribe(model, tokens, "sample.wav")
print(text)
Read moreDetails
• Benchmarks: CER ranges from 2.04% to 7.12% depending on the test set.
• Hardware requirements: Supports CPU and GPU (including bf16 mixed precision).
• Limitations: Requires 16 kHz mono audio; numbers are output as words; spacing between words is minimal.
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.

    # Nukthean/krama-asr-khmer-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 Nukthean/krama-asr-khmer-v1 --local-dir ~/models/krama-asr-khmer-v1
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to krama-asr-khmer-v1
  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 krama-asr-khmer-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

Nukthean/krama-asr-khmer-v1
Author
Nukthean
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