Audio Classification
Transformers
Safetensors
cue
voice-agent
turn-taking
barge-in
endpointing
full-duplex
indian-english
Instructions to use IOTEverythin/cue-v5 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use IOTEverythin/cue-v5 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("audio-classification", model="IOTEverythin/cue-v5")# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("IOTEverythin/cue-v5", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Download config.json from IOTEverythin/cue-v5: direct link, hf CLI and curl.
- Browser
- Download file 1.47 kB
-
https://huggingface.co/IOTEverythin/cue-v5/resolve/main/config.json
- Command line
-
hf download hf://IOTEverythin/cue-v5/config.json
-
curl -L -o config.json https://huggingface.co/IOTEverythin/cue-v5/resolve/main/config.json
1.47 kB
| { | |
| "model_type": "cue", | |
| "version": "v5", | |
| "labels": [ | |
| "NONE", | |
| "USER_TURN", | |
| "BACKCHANNEL", | |
| "ACCIDENTAL", | |
| "SOFT_INTERRUPT", | |
| "HARD_INTERRUPT", | |
| "TAKEOVER" | |
| ], | |
| "encoder": "encoder", | |
| "encoder_type": "whisper", | |
| "encoder_source": "openai/whisper-small (encoder only, positions cut to 100)", | |
| "positions": "none", | |
| "encoder_dim": 768, | |
| "hop_ms": 20, | |
| "chunk_ms": 160, | |
| "context_ms": 2000, | |
| "d": 384, | |
| "layers": 6, | |
| "heads": 4, | |
| "window_frames": 80, | |
| "policy": { | |
| "barge_in": { | |
| "stop_p": 0.8, | |
| "hold": 6, | |
| "cooldown_ms": 1500 | |
| }, | |
| "end_of_turn": { | |
| "p_done": 0.9, | |
| "turn_hold": 12, | |
| "min_speech": 10, | |
| "fallback_ms": 2000 | |
| }, | |
| "profiles": { | |
| "responsive": { | |
| "stop_p": 0.7, | |
| "hold": 4, | |
| "p_done": 0.5, | |
| "turn_hold": 6 | |
| }, | |
| "balanced": { | |
| "stop_p": 0.8, | |
| "hold": 6, | |
| "p_done": 0.9, | |
| "turn_hold": 12 | |
| }, | |
| "cautious": { | |
| "stop_p": 0.95, | |
| "hold": 6, | |
| "p_done": 0.95, | |
| "turn_hold": 16 | |
| } | |
| } | |
| }, | |
| "training": { | |
| "init": "Cue v4", | |
| "steps": 8000, | |
| "sources": "['syn=runs/v5/feat/syn:0.5', 'ami=runs/v5/feat/ami:0.3', 'icsi=runs/v5/feat/icsi:0.2']", | |
| "data": "synthetic Indian English calls (clean + realistic; 32 with Dia-voiced callers), AMI and ICSI meetings (caller side only), MUSAN noise" | |
| }, | |
| "two_channel": true, | |
| "inputs": "caller audio + the agent's own audio (what it is playing; zero if not given) + agent-speaking flag" | |
| } |