mirror of
https://github.com/mudler/LocalAI
synced 2026-05-24 09:28:23 +00:00
20 lines
632 B
Bash
20 lines
632 B
Bash
|
#!/bin/bash
|
||
|
|
set -e
|
||
|
|
|
||
|
|
backend_dir=$(dirname $0)
|
||
|
|
if [ -d $backend_dir/common ]; then
|
||
|
|
source $backend_dir/common/libbackend.sh
|
||
|
|
else
|
||
|
|
source $backend_dir/../common/libbackend.sh
|
||
|
|
fi
|
||
|
|
|
||
|
|
installRequirements
|
||
|
|
|
||
|
|
# No pre-baked model weights. Weights flow through LocalAI's gallery
|
||
|
|
# `files:` mechanism — see gallery entries for speechbrain-ecapa-tdnn
|
||
|
|
# and WeSpeaker / 3D-Speaker ONNX packs. SpeechBrain's
|
||
|
|
# EncoderClassifier.from_hparams also knows how to auto-download from
|
||
|
|
# HuggingFace into the configured savedir (we point it at ModelPath),
|
||
|
|
# so the first LoadModel call bootstraps the checkpoint if the gallery
|
||
|
|
# flow wasn't used.
|