Skip to content
Audio

Turkmen MMS ASR

standalone checkpoint for Turkmen speech recognition

2 min readupdated

This model handles automatic speech recognition (ASR) for Turkmen, transcribing audio into Latin-script text. Built on Meta's facebook/mms-1b-all architecture, it ships as a self-contained checkpoint with a baked-in adapter and fixed vocabulary, eliminating the need to load extra language modules at runtime. Compared to the original MMS adapter, this fine-tune delivers a significantly lower error rate on test datasets. It is a practical fit for developers and researchers working on regional languages and Turkmen voice interfaces.

Key features

  • Standalone design with a fixed adapter and vocabulary—no language switching required.
  • Reduced recognition error rate (strict WER) down to 24.58% on the test dataset.
  • Out-of-the-box compatibility with standard Hugging Face Transformers and PyTorch libraries.
  • Non-commercial release licensed under CC BY-NC 4.0.

What you can do

  • Transcribe Turkmen speech recordings via Python scripts.
  • Process long-form audio files using batched sliding-window scripts.
  • Integrate the model into local voice-processing pipelines and archive transcription workflows.

Example

import soundfile as sf
import torch
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor

repo = "Pirnazar/mms-1b-turkmen"
device = "cuda" if torch.cuda.is_available() else "cpu"
processor = Wav2Vec2Processor.from_pretrained(repo)
model = Wav2Vec2ForCTC.from_pretrained(repo).to(device).eval()

audio, sr = sf.read("speech.wav", dtype="float32")
inputs = processor(audio, sampling_rate=16000, return_tensors="pt")
inputs = {name: value.to(device) for name, value in inputs.items()}

with torch.inference_mode():
    logits = model(**inputs).logits
text = processor.batch_decode(logits.argmax(-1))[0].strip()
print(text)
Read moreDetails
Benchmarks: strict WER — 24.58%, strict CER — 4.11%.
Parameters: ~965M total parameters, 2.28M trainable adapter/head parameters.
Limitations: Turkmen only (Latin script); no translation or speaker diarization support.
Discuss Telegram 0 31

Will it run?

Runs in F16

Fits without quantisation.

  • F16weights 1.8 GB · ready files2.5 GB
  • FP8 / INT8recommendedweights 0.9 GB1.5 GB
  • 4-bitweights 0.5 GB1.1 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · F16 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

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.

    # Pirnazar/mms-1b-turkmen
    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 Pirnazar/mms-1b-turkmen --local-dir ~/models/mms-1b-turkmen
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to mms-1b-turkmen
  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 mms-1b-all (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 5All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 2All on Hugging Face

About the model

Pirnazar/mms-1b-turkmen
Author
Pirnazar
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

310

Ask AI about this model

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

Comments

Audio

YourTTS

YourTTS