Skip to content
Audio

SpeechLLM ASR: English speech recognition powered by Villanova-2B

1 min readupdated

This is a multimodal speech language model (SpeechLLM) for automatic English speech recognition, built on the SLAM-LLM framework. The architecture combines a frozen FAMA encoder, a linear projector, and the Villanova-2B model with LoRA (a lightweight adapter for the base model). The checkpoint contains trained weights totaling 14.16 million parameters for the projector and adapters. The model outperforms Whisper large-v3-turbo on the VoxPopuli dataset and efficiently handles audio recordings longer than 30 seconds.

Key features

  • Uses a combination of the FAMA encoder and the Villanova-2B language model.
  • Achieves a WER of 6.55% on the VoxPopuli English test set.
  • Capable of processing long audio fragments without truncation.
  • Trained using the SLAM-LLM framework on NVIDIA A100 GPUs.

What you can do

  • Recognize English speech with high accuracy in audio projects and services.
  • Run decoding or fine-tuning using the scripts from the Villanova-Speech-LLM repository.

Example

hf download SpeechTek/speechllm_fama_meusli_villanova_asr_en_v1 model.pt --local-dir checkpoints/speechllm_fama_meusli_villanova_asr_en_v1
Read moreDetails
• Architecture: FAMA encoder + Villanova-2B (LLM) with LoRA
• Training: VoxPopuli English, NVIDIA A100 64 GB
• Metrics: WER 6.55% on VoxPopuli English test
• License: CC BY 4.0
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.1 GB · ready files0.6 GB
  • FP8 / INT8recommendedweights 0.0 GB0.5 GB
  • 4-bitweights 0.0 GB0.5 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 · BF16 — 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.

    # SpeechTek/speechllm_fama_meusli_villanova_asr_en_v1
    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 SpeechTek/speechllm_fama_meusli_villanova_asr_en_v1 --local-dir ~/models/speechllm_fama_meusli_villanova_asr_en_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 fama-medium-asr (quants, finetunes, LoRAs) will show up under the bell.

Fine-tunes 7All on Hugging Face

About the model

SpeechTek/speechllm_fama_meusli_villanova_asr_en_v1
Author
SpeechTek
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