Skip to content
Audio

Duovid Whisper

албанский ASR прямо на устройстве

1 min readupdated

Это упаковка файнтюна Flutra/whisper-large-v3-turbo-sq-v2 для офлайн-распознавания албанской речи. Для iOS — Core ML 6-bit под WhisperKit (~589 MB), для Android — GGML q5_0 под whisper.cpp (~574 MB). Веса и качество — от Flutra, автор репозитория менял только формат и квантование.

Ключевые особенности

  • Основа — Whisper large-v3-turbo, дообученный под албанский (Shqip).
  • Core ML-сборка сжата 6-bit palettization: ~1.6 GB → ~589 MB.
  • GGML q5_0 для Android: ~574 MB, всего на 0.2 WER хуже q8_0.
  • На незнакомой речи (VoxForge) WER 18.8% против 44.9% у stock-модели.
  • Поддержка word timestamps, VAD chunking; RTF ≈0.04 на M-series Mac.
  • 4-bit сборку отклонили: WER превысил 100%.

Пример

let folder = try await WhisperKit.download(variant: "duovid_whisper-large-v3-turbo-sq_pal6bit_589MB", from: "antonbanderi/fjale")
let kit = try await WhisperKit(WhisperKitConfig(model: "duovid_whisper-large-v3-turbo-sq_pal6bit_589MB", modelRepo: "antonbanderi/fjale", modelFolder: folder.path, download: false))
let result = try await kit.transcribe(audioPath: path, decodeOptions: DecodingOptions(language: "sq", detectLanguage: false, wordTimestamps: true, chunkingStrategy: .vad))
ПодробнееWER: 11.3% у Core ML 6-bit и 10.4% у GGML q5_0; на VoxForge — 18.8%.
Размеры: ~589 MB Core ML и ~574 MB GGML; RTF ≈0.04 на M-series Mac.
Ограничения: тест всего на 7 клипах (~7 мин), разницу меньше ~1 WER-пункта считать шумом.
Лицензия: веса принадлежат Flutra и не распространяются под MIT; конвертация и квантование лицензию не меняют.
Discuss Telegram 0 32

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.

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

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

About the model

antonbanderi/fjale
Author
antonbanderi
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

320

Ask AI about this model

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

Comments

Audio

AADD v3

AADD v3