run_exp11

#!/bin/bash

ENV=$1
FRAMEWORK="alewrap"
env_params="useRGB=true,display=false"
game_path=$PWD"/roms/"

if [ $2 == "test" ]; then
	netfile="\"exp11_"$1".t7\""
	replay_memory=1
else
	netfile="\"convnet_atari3\""
	replay_memory=500000
fi

update_freq=4
discount=0.99
learn_start=25000
eps_end=0.1
lr=0.01
lr_end=0.00025
state_dim=7056
ncols=1
n_replay=1
rnn_size=256

preproc="\"net_downsample_2x_full_y\""
agent_params="rnn_size="$rnn_size",lr="$lr",lr_end="$lr_end",ep=1,ep_end="$eps_end",discount="$discount",hist_len=4,learn_start="$learn_start",replay_memory="$replay_memory",update_freq="$update_freq",n_replay="$n_replay",network="$netfile",preproc="$preproc",state_dim="$state_dim",minibatch_size=32,rescale_r=1,ncols="$ncols",bufferSize=512,valid_size=500,target_q=10000,clip_delta=1,min_reward=-1,max_reward=1"

steps=5000000
eval_freq=50000
eval_steps=25000
prog_freq=5000
save_freq=50000
gpu=1
random_starts=30
pool_frms_type="\"max\""
pool_frms_size=2
pool_frms="type="$pool_frms_type",size="$pool_frms_size
num_threads=4
seed=1
actrep=4
agent="NeuralQLearner"
agent_name="exp11_"$1

args="-best -framework $FRAMEWORK -game_path $game_path -name $agent_name -env $ENV -env_params $env_params -agent $agent -agent_params $agent_params -steps $steps -eval_freq $eval_freq -eval_steps $eval_steps -prog_freq $prog_freq -save_freq $save_freq -actrep $actrep -gpu $gpu -random_starts $random_starts -pool_frms $pool_frms -seed $seed -threads $num_threads"
echo $args

cd drqn 

if [ $2 == "test" ]; then
	th test_agent.lua $args
else
	th train_agent.lua $args
fi