taco/microWakeWord-Trainer-Nvidia-Docker
1
0
Fork 0
mirror of https://github.com/TaterTotterson/microWakeWord-Trainer-Nvidia-Docker.git synced 2026-08-12 07:55:33 -06:00
Code Issues Packages Projects Releases Wiki Activity

Release NVIDIA WakeWord Trainer v22

Browse Source
This commit is contained in:
MasterPhooey
2026-08-02 20:46:04 -05:00
parent 2b1320f1f3
commit 2a88090b85
37 changed files with 11685 additions and 3628 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
train_wake_word
View File

@@ -5,7 +5,7 @@ PROGPATH=$(realpath "$0")
PROGDIR=$(dirname "${PROGPATH}")
CLIDIR="${PROGDIR}/cli"
KNOWN_ARGS=( samples batch-size training-steps data-dir cleanup-work-dir language )
KNOWN_ARGS=( samples batch-size training-steps data-dir cleanup-work-dir language tts-mode tts-voice-count )
source "${CLIDIR}/shell.functions"
WAKE_WORD=${POSITIONAL_ARGS[0]}
@@ -19,6 +19,8 @@ if [ "${HELP}" == "true" ] || [ -z "${WAKE_WORD}" ] ; then
Usage: train_wake_word [ --samples=<samples> ] [ --batch-size=<batch_size> ]
[ --training-steps=<steps> ] [ --cleanup-work-dir ]
[ --language=<lang> ]
[ --tts-mode=<modern|hybrid|piper> ]
[ --tts-voice-count=<voices> ]
<wake_word> [ <wake_word_title> ]
Options:
@@ -39,6 +41,12 @@ Options:
--language: Language for TTS voice selection (e.g. "en", "nl").
Default: ${DEFAULT_LANGUAGE}
--tts-mode: TTS source: modern (OmniVoice plus Qwen3/MOSS where
supported), hybrid (modern plus Piper), or piper.
Default: ${DEFAULT_TTS_MODE}
--tts-voice-count: Deprecated compatibility option; direct generation ignores it.
<wake_word> The word to train spelled phonetically.
Required.
@@ -116,6 +124,8 @@ export GRPC_VERBOSITY=ERROR
--samples=${SAMPLES} \
--batch-size=${BATCH_SIZE} \
--language="${LANGUAGE}" \
--tts-mode="${TTS_MODE}" \
--tts-voice-count="${TTS_VOICE_COUNT}" \
--data-dir="${DATA_DIR}" "${WAKE_WORD}"
POST_GEN_TS=$EPOCHSECONDS