startup.sh

#!/bin/bash

export HF_HOME="/data/.huggingface"

echo "PWD: $(pwd)"
echo $HF_TOKEN > .hf_token
echo "LS: $(ls -als)"

while true; do nvidia-smi; sleep 600; done &

python3 -c "import torch; \
    print(f\"is availeble = {torch.cuda.is_available()}\"); \
    print(f\"device count = {torch.cuda.device_count()}\"); \
    print(f\"current device = {torch.cuda.current_device()}\")"


COMPONENT="serve.controller" 
echo "Starting $COMPONENT"
nohup python3 -m serve.controller --host 0.0.0.0 --port 10000 2>&1 | tee serve.controller.log &
while [ `grep -c "Uvicorn running on" serve.controller.log` -eq '0' ];do
        sleep 2s;
        echo "waiting for $COMPONENT to be running"
done
echo "$COMPONENT is running."

# stuck after download config.json
# COMPONENT="interactive_demo prism-dinosiglip+13b"
# echo "Starting $COMPONENT"
# nohup python3 -m interactive_demo --port 40000 --model_id prism-dinosiglip+13b | tee prism-dinosiglip_13b.log &
# while [ `grep -c "loaded prismatic prism-dinosiglip+13b" prism-dinosiglip_13b.log` -eq '0' ];do
#         sleep 10s;
#         echo "waiting for $COMPONENT to be running"
# done
# echo "$COMPONENT is running"

#works!
COMPONENT1="interactive_demo prism-dinosiglip_7b"
echo "Starting $COMPONENT1"
CUDA_VISIBLE_DEVICES=0 nohup python3 -m interactive_demo --port 40001 --model_id prism-dinosiglip+7b | tee prism-dinosiglip_7b.log &

#works!
COMPONENT2="interactive_demo prism-dinosiglip-controlled+7b"
echo "Starting $COMPONENT2"
CUDA_VISIBLE_DEVICES=1 nohup python3 -m interactive_demo --port 40002 --model_id prism-dinosiglip-controlled+7b | tee prism-dinosiglip-controlled_7b.log &

#works!
COMPONENT3="interactive_demo llava-v1.5-7b"
echo "Starting $COMPONENT3"
CUDA_VISIBLE_DEVICES=2 nohup python3 -m interactive_demo --port 40003 --model_family llava-v15 --model_id llava-v1.5-7b --model_dir liuhaotian/llava-v1.5-7b | tee llava-v1.5-7b.log &

while [ `grep -c "loaded prismatic prism-dinosiglip+7b" prism-dinosiglip_7b.log` -eq '0' ];do
        sleep 10s;
        echo "waiting for $COMPONENT1 to be running"
done
echo "$COMPONENT1 is running."

while [ `grep -c "loaded prismatic prism-dinosiglip-controlled+7b" prism-dinosiglip-controlled_7b.log` -eq '0' ];do
        sleep 10s;
        echo "waiting for $COMPONENT2 to be running"
done
echo "$COMPONENT2 is running"

while [ `grep -c "loaded llava llava-v1.5-7b" llava-v1.5-7b.log` -eq '0' ];do
        sleep 10s;
        echo "waiting for $COMPONENT3 to be running"
done
echo "$COMPONENT3 is running"


ls -als $HF_HOME
tree --du -h $HF_HOME

echo "Starting serve.gradio_web_server"
python3 -m serve.gradio_web_server --controller http://127.0.0.1:10000 --model-list-mode reload # --share