To accelerate the training speed on rcssserver3d, we set up a simple distribution computation platform. In order to deploy your tasks, you just need to do as follows:
- create an account called optimization on each servant computer
- open ssh port, add leader computer's public key to
~/.ssh/authorized_keys - set up the environment that your task need
- edit
manifest.py
_host_port = {
# here we set the hosts name, hosts IP, hosts TCP port
'your_host_name': ('123.123.123.123', 2333),
'your_host_name2': ('123.123.123.124', 2333),
......
}There may be other options that you need to reedit in your task
2. edit hosts.conf, add your servant hosts into HOSTS and set your total number of your hosts
-
edit
server.py, alter the eva function that you need to solve -
edit
clean.sh, set up your cleaning rules. Be carefull to deal with your important files.
./cmd_push.sh setup *ip or hostname* # set up one node
./cmd_push.sh setup-all # set up all nodes
``
#### Kick off Tasks
1. first start servant nodes by `./cmd_push.sh start`
2. run `python3 cma_es_dc.py`
#### Check the State of Servant Nodes
- simply you can use `./cmd_push.sh inspect` to check whether errors or exceptions occurred during server running. Also you can change the error catching key word in `cmd_push.sh`
- Sometime you may need to check the exact running state on servant node, you can just use
- `./cmd_push.sh open *ip addr*` to open one terminal for servant node
- `./cmd_push.sh open-all` to open terminals for all servant nodes
```bash
./cmd_push.sh start
#### File Interaction
- To get file from servant nodes
- execute `./file_push.sh fetch *filename*` to fetch files renamed with prefix of their IP, an example may be as follows
```
192.168.1.114-walkout 192.168.1.142-walkout 192.168.1.177-walkout
192.168.1.137-walkout 192.168.1.145-walkout
```
- To push files from local to all servant nodes
`./file_push.sh *filename*`
#### Clean Mess
execute `./cmd_push.sh clean` to clean all messes your task created, clean rules can be altered by editing the `clean.sh` file
#### Result Plot
- You can plot the result by `python3 result_plot.py`
python3 cma_es_dc.py # You'd better run this command in other terminal