πŸ“‹ Model Description


license: other license_name: lfm1.0 license_link: LICENSE language:
  • en
tags:
  • liquid
  • lfm2.5
  • edge
  • llama.cpp
  • audio
  • speech
  • gguf
base_model:
  • LiquidAI/LFM2.5-Audio-1.5B
widget: - text: "Demo" output: url: demo.mp4


src="https://cdn-uploads.huggingface.co/production/uploads/61b8e2ba285851687028d395/2b08LKpev0DNEk6DlnWkY.png"
alt="Liquid AI"
style="width: 100%; max-width: 100%; height: auto; display: inline-block; margin-bottom: 0.5em; margin-top: 0.5em;"
/>

Try LFM β€’
Documentation β€’
LEAP

LFM2.5-Audio-1.5B

Find more details in the original model card: https://huggingface.co/LiquidAI/LFM2.5-Audio-1.5B

Runners

runners folder contains runners for various architectures including

  • llama-liquid-audio-cli
  • llama-liquid-audio-server

πŸƒ How to run LFM2.5

CLI

Set env variables.

export CKPT=/path/to/LFM2.5-Audio-1.5B-GGUF
export INPUT_WAV=/path/to/input.wav
export OUTPUT_WAV=/path/to/output.wav

ASR (audio -> text)

./llama-liquid-audio-cli -m $CKPT/LFM2.5-Audio-1.5B-Q40.gguf -mm $CKPT/mmproj-LFM2.5-Audio-1.5B-Q40.gguf -mv $CKPT/vocoder-LFM2.5-Audio-1.5B-Q40.gguf --tts-speaker-file $CKPT/tokenizer-LFM2.5-Audio-1.5B-Q40.gguf -sys "Perform ASR." --audio $INPUT_WAV

TTS (text -> audio)

./llama-liquid-audio-cli -m $CKPT/LFM2.5-Audio-1.5B-Q40.gguf -mm $CKPT/mmproj-LFM2.5-Audio-1.5B-Q40.gguf -mv $CKPT/vocoder-LFM2.5-Audio-1.5B-Q40.gguf --tts-speaker-file $CKPT/tokenizer-LFM2.5-Audio-1.5B-Q40.gguf -sys "Perform TTS." -p "Hi, how are you?" --output $OUTPUT_WAV

Interleaved (audio/text -> audio + text)

./llama-liquid-audio-cli -m $CKPT/LFM2.5-Audio-1.5B-Q40.gguf -mm $CKPT/mmproj-LFM2.5-Audio-1.5B-Q40.gguf -mv $CKPT/vocoder-LFM2.5-Audio-1.5B-Q40.gguf --tts-speaker-file $CKPT/tokenizer-LFM2.5-Audio-1.5B-Q40.gguf -sys "Respond with interleaved text and audio." --audio $INPUTWAV --output $OUTPUTWAV

Server

Start server

export CKPT=/path/to/LFM2.5-Audio-1.5B-GGUF
./llama-liquid-audio-server -m $CKPT/LFM2.5-Audio-1.5B-Q40.gguf -mm $CKPT/mmproj-LFM2.5-Audio-1.5B-Q40.gguf -mv $CKPT/vocoder-LFM2.5-Audio-1.5B-Q40.gguf --tts-speaker-file $CKPT/tokenizer-LFM2.5-Audio-1.5B-Q40.gguf

Use liquidaudiochat.py script to communicate with the server.

uv run liquidaudiochat.py

Source Code for Runners

Runners are built from https://github.com/ggml-org/llama.cpp/pull/18641. It's WIP and will take time to land in upstream.

Demo

πŸ“‚ GGUF File List

πŸ“ Filename πŸ“¦ Size ⚑ Download
LFM2.5-Audio-1.5B-F16.gguf
LFS FP16
2.18 GB Download
LFM2.5-Audio-1.5B-Q4_0.gguf
Recommended LFS Q4
663.52 MB Download
LFM2.5-Audio-1.5B-Q8_0.gguf
LFS Q8
1.16 GB Download
mmproj-LFM2.5-Audio-1.5B-F16.gguf
LFS FP16
437.55 MB Download
mmproj-LFM2.5-Audio-1.5B-Q4_0.gguf
LFS Q4
209.34 MB Download
mmproj-LFM2.5-Audio-1.5B-Q8_0.gguf
LFS Q8
279.85 MB Download
tokenizer-LFM2.5-Audio-1.5B-F16.gguf
LFS FP16
136.09 MB Download
tokenizer-LFM2.5-Audio-1.5B-Q4_0.gguf
LFS Q4
48.2 MB Download
tokenizer-LFM2.5-Audio-1.5B-Q8_0.gguf
LFS Q8
73.39 MB Download
vocoder-LFM2.5-Audio-1.5B-F16.gguf
LFS FP16
369.22 MB Download
vocoder-LFM2.5-Audio-1.5B-Q4_0.gguf
LFS Q4
103.94 MB Download
vocoder-LFM2.5-Audio-1.5B-Q8_0.gguf
LFS Q8
196.21 MB Download