ubuntu-instructions.txt

1. Copy over the setup script
MACHINE="ec2-18-207-231-98.compute-1.amazonaws.com"
scp -i ~/.ssh/chronicle-test-keys.pem ~/workspace/MelNet/setup.sh ubuntu@$MACHINE:/home/ubuntu/

1a. Copy over model files if necessary
scp -r -i ~/.ssh/chronicle-test-keys.pem ~/workspace/MelNet/chkpt_3_tier/chkpt ubuntu@$MACHINE:/home/ubuntu/workspace/MelNet/chkpt

2. Run the setup script
sh setup.sh

3.  Run the thing

screen
python trainer.py -c ./config/blizzard_compressed.yaml -n blizzard-compressed-6-tier_t6 -t 6 -b 1
python trainer.py -c ./config/blizzard.yaml -n blizzard-6-tier_t1 -t 1 -b 1 -s TTS
python trainer.py -c ./config/blizzard-5-tier.yaml -n blizzard_t5_noTTS -t 5 -b 1
python trainer.py -c ./config/blizzard-5-tier.yaml -n blizzard_t5_noTTS -t 5 -b 1

4. Copy the file back to laptop before shutting the instance down

scp -r -i ~/.ssh/chronicle-test-keys.pem ubuntu@$MACHINE:/home/ubuntu/workspace/MelNet/chkpt ~/workspace/MelNet/chkpt_latest

5. python inference.py -c ./config/kss.yaml -p ./config/inference.yaml -t [timestep of generated mel spectrogram] -n [name of sample] -i [input sentence for conditional generation]

python inference.py -c config/blizzard_alldata_v5.yaml -p config/inference.yaml -t 64 -n hw_blizzard_v5 -i "Hello, world."

python inference.py -c config/kss.yaml -p config/inference.yaml -t 64 -n kss_hw -i "안녕하세요, 세계"