Koha9/Aimbot-PPO
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

对应V3.1.6Play模式

Browse Source 
对应V3.1.5的Stay Target进行修改
This commit is contained in:
Koha9 2023-07-29 23:52:20 +09:00
parent f9ee51c256
commit 15c1edb6c9
4 changed files with 71 additions and 61 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
Aimbot-PPO-Python/Pytorch/AimbotEnv.py
View File

@ -68,6 +68,9 @@ class Aimbot(gym.Env):
# agents number
self.unity_agent_num = len(self.unity_agent_IDS)
# all zero action
self.all_zero_action = np.zeros((self.unity_agent_num, self.unity_action_size))
def reset(self) -> Tuple[np.ndarray, List, List]:
"""reset environment and get observations

25
Aimbot-PPO-Python/Pytorch/Archive/test2.ipynb
View File

@ -181,31 +181,30 @@
},
{
"cell_type": "code",
"execution_count": 1,
"execution_count": 6,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"Failed to detect the name of this notebook, you can set it manually with the WANDB_NOTEBOOK_NAME environment variable to enable code saving.\n",
"\u001b[34m\u001b[1mwandb\u001b[0m: Currently logged in as: \u001b[33mkoha9\u001b[0m. Use \u001b[1m`wandb login --relogin`\u001b[0m to force relogin\n"
]
},
{
"data": {
"text/plain": [
"True"
"array([[0., 0., 0., 0.],\n",
" [0., 0., 0., 0.],\n",
" [0., 0., 0., 0.],\n",
" [0., 0., 0., 0.],\n",
" [0., 0., 0., 0.],\n",
" [0., 0., 0., 0.],\n",
" [0., 0., 0., 0.],\n",
" [0., 0., 0., 0.]])"
]
},
"execution_count": 1,
"execution_count": 6,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"import wandb\n",
"wandb.login()"
"import numpy as np\n",
"np.zeros((8, 4))"
]
}
],

98
Aimbot-PPO-Python/Pytorch/MultiNN-PPO.py
View File

@ -34,10 +34,11 @@ if __name__ == "__main__":
# Initialize environment agent optimizer
aimbot_side_channel = AimbotSideChannel(SIDE_CHANNEL_UUID)
env = Aimbot(env_path=args.path,
worker_id=args.workerID,
base_port=args.baseport,
side_channels=[aimbot_side_channel])
env = Aimbot(
env_path=args.path,
worker_id=args.workerID,
base_port=args.baseport,
side_channels=[aimbot_side_channel])
if args.load_dir is None:
agent = PPOAgent(
env=env,
@ -104,7 +105,12 @@ if __name__ == "__main__":
last_reward = [0. for i in range(env.unity_agent_num)]
# MAIN LOOP: run agent in environment
while True:
# On decision point, choose action by agent
# Target Type(state[0][0]) is stay(4),use all zero action
if state[0][0] == 4:
next_state, reward, next_done = env.step(env.all_zero_action)
state, done = next_state, next_done
continue
# On decision point, and Target Type(state[0][0]) is not stay(4) choose action by agent
if step % args.decision_period == 0:
step += 1
# Choose action by agent
@ -124,52 +130,54 @@ if __name__ == "__main__":
next_state, reward, next_done = env.step(action_cpu)
# save memories
ppo_memories.save_memories(
now_step=step,
agent=agent,
state=state,
action_cpu=action_cpu,
dis_logprob_cpu=dis_logprob_cpu,
con_logprob_cpu=con_logprob_cpu,
reward=reward,
done=done,
value_cpu=value_cpu,
last_reward=last_reward,
next_done=next_done,
next_state=next_state,
)
# check if any training dataset is full and ready to train
for i in range(args.target_num):
if ppo_memories.obs[i].size()[0] >= args.datasetSize:
# start train NN
train_queue.append(i)
if len(train_queue) > 0:
# break while loop and start train
break
# update state
if args.train:
ppo_memories.save_memories(
now_step=step,
agent=agent,
state=state,
action_cpu=action_cpu,
dis_logprob_cpu=dis_logprob_cpu,
con_logprob_cpu=con_logprob_cpu,
reward=reward,
done=done,
value_cpu=value_cpu,
last_reward=last_reward,
next_done=next_done,
next_state=next_state,
)
# check if any training dataset is full and ready to train
for i in range(args.target_num):
if ppo_memories.obs[i].size()[0] >= args.datasetSize:
# start train NN
train_queue.append(i)
if len(train_queue) > 0:
# break while loop and start train
break
# update state
state, done = next_state, next_done
else:
step += 1
# skip this step use last predict action
next_state, reward, next_done = env.step(action_cpu)
# save memories
ppo_memories.save_memories(
now_step=step,
agent=agent,
state=state,
action_cpu=action_cpu,
dis_logprob_cpu=dis_logprob_cpu,
con_logprob_cpu=con_logprob_cpu,
reward=reward,
done=done,
value_cpu=value_cpu,
last_reward=last_reward,
next_done=next_done,
next_state=next_state,
)
# update state
state = next_state
last_reward = reward
if args.train:
ppo_memories.save_memories(
now_step=step,
agent=agent,
state=state,
action_cpu=action_cpu,
dis_logprob_cpu=dis_logprob_cpu,
con_logprob_cpu=con_logprob_cpu,
reward=reward,
done=done,
value_cpu=value_cpu,
last_reward=last_reward,
next_done=next_done,
next_state=next_state,
)
# update state
state = next_state
last_reward = reward
if args.train:
# train mode on

6
Aimbot-PPO-Python/Pytorch/arguments.py
View File

@ -34,9 +34,9 @@ BROADCASTREWARD = False
ANNEAL_LEARNING_RATE = True
CLIP_VLOSS = True
NORM_ADV = False
TRAIN = True
SAVE_MODEL = True
WANDB_TACK = True
TRAIN = False
SAVE_MODEL = False
WANDB_TACK = False
LOAD_DIR = None
#LOAD_DIR = "../PPO-Model/PList_Go_LeakyReLU_9331_1677965178_bestGoto/PList_Go_LeakyReLU_9331_1677965178_10.709002.pt"