speaker_identify / finetune.bash
DuyTa's picture
Upload folder using huggingface_hub
f831146 verified
raw
history blame contribute delete
790 Bytes
LR=0.0005
EPOCHS=20
BATCH_SIZE=128
OUTPUT_BASE="siamese_fbanks_saved/"
TRAIN_DATA="fbannks_train"
TEST_DATA="fbannks_test"
for NUM_LAYERS in 2 3 4 5 6
do
PRETRAINED_MODEL_PATH="saved_models_cross_entropy/${NUM_LAYERS}/"
OUTPUT_MODEL_PATH="${OUTPUT_BASE}${NUM_LAYERS}/"
echo "Running training with num_layers=${NUM_LAYERS}, pretrained_model_path=${PRETRAINED_MODEL_PATH}, output_model_path=${OUTPUT_MODEL_PATH}"
python3 stage2_finetune.py \
--num_layers ${NUM_LAYERS} \
--lr ${LR} \
--epochs ${EPOCHS} \
--batch_size ${BATCH_SIZE} \
--pretrained_model_path ${PRETRAINED_MODEL_PATH} \
--output_model_path ${OUTPUT_MODEL_PATH} \
--train_data ${TRAIN_DATA} \
--test_data ${TEST_DATA}
echo "Finished training with num_layers=${NUM_LAYERS}"
done