gemma-text-vad

This is a text-based Voice Activity Detection model that determines if a given speech fragment is complete enough for processing by a smart speaker assistant. This allows smart speakers to move from using time based pauses (300ms - 1000ms) to detect the end of voice input to using this model to determine if the voice input is complete.

Example:

• "Hey" -> no
• "Hey Juno" -> no
• "Hey Juno can you" -> no
• "Hey Juno can you set" -> no
• "Hey Juno can you set the" -> no
• "Hey Juno can you set the temperature" -> no
• "Hey Juno can you set the temperature to" -> no
• "Hey Juno can you set the temperature to 65" -> yes

Model prompting requirements:

• Required system prompt: "You are a Voice Activity Detection system. Determine if the given speech fragment is complete enough for processing. Answer with only 'yes' if complete or 'no' if incomplete."
• Required user prompt: "Is this sentence fragment complete for processing: '{fragment}'"

To use with pipeline from transformers:

from transformers import pipeline

pipe = pipeline("text-generation", model="juno-labs/gemma-text-vad")

SYSTEM_PROMPT = (
  "You are a Voice Activity Detection system. "
  "Determine if the given speech fragment is complete enough for processing. Answer with only 'yes' if complete or 'no' if incomplete."
)

SENTENCE = "Hey Juno can you set the temperature to"

messages = [
    {'content': SYSTEM_PROMPT, 'role': 'system'},
    {'content': f"Is this sentence fragment complete for processing: '{SENTENCE}'", 'role': 'user'}
]

generated = pipe(messages)

classification = generated[0]["generated_text"][2]["content"]

print(f"Classification: {classification}") # "yes" or "no"

To use with transformers:

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

MODEL_ID = "juno-labs/gemma-text-vad"

# Load model + tokenizer
model = AutoModelForCausalLM.from_pretrained(MODEL_ID, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)

SYSTEM_PROMPT = (
    "You are a Voice Activity Detection system. "
    "Determine if the given speech fragment is complete enough for processing. Answer with only 'yes' if complete or 'no' if incomplete."
)

SENTENCE = "Set the temperature to 68"

messages = [
    {"role": "system", "content": SYSTEM_PROMPT},
    {"role": "user", "content": f"Is this sentence fragment complete for processing: '{SENTENCE}'"},
]

prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)

inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

with torch.no_grad():
    outputs = model.generate(
        **inputs,
        max_new_tokens=1,        # only 1 token
        do_sample=False,         # greedy decoding
        pad_token_id=tokenizer.eos_token_id,
    )

decoded = tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:], skip_special_tokens=True)

print(f"Classification: {decoded}") # "yes" or "no"