chris/street-fighter-ai
1
0
Fork 0
mirror of https://github.com/linyiLYi/street-fighter-ai.git synced 2025-04-04 23:20:43 +00:00
Code Issues Packages Projects Releases Wiki Activity

add recording function to the script

Browse Source
This commit is contained in:
linyiLYi 2023-04-05 23:21:32 +08:00
parent a3ede7dd30
commit f09e69d05c
5 changed files with 29 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
.gitignore vendored
View File

@ -6,3 +6,6 @@ archives/
images/ images/
data/ data/
main/logs/monitoring/ main/logs/monitoring/
recordings/
007*

BIN
__pycache__/custom_cnn.cpython-38.pyc
View File

Binary file not shown.

BIN
__pycache__/custom_sf2_cv_env.cpython-38.pyc
View File

Binary file not shown.

BIN
__pycache__/mobilenet_extractor.cpython-38.pyc
View File

Binary file not shown.

28
main/test.py
View File

@ -1,3 +1,4 @@
import os
import time import time
import retro import retro
@ -7,9 +8,13 @@ from street_fighter_custom_wrapper import StreetFighterCustomWrapper
RESET_ROUND = False # Reset the round when fight is over. RESET_ROUND = False # Reset the round when fight is over.
RENDERING = True RENDERING = True
RECORDING = True
RANDOM_ACTION = False RANDOM_ACTION = False
MODEL_PATH = r"trained_models/ppo_ryu_7000000_steps" MODEL_DIR = r"trained_models/"
MOVIE_DIR = r"recordings"
MODEL_NAME = r"ppo_ryu_7000000_steps"
def make_env(game, state): def make_env(game, state):
def _init(): def _init():
@ -28,8 +33,7 @@ env = make_env(game, state="Champion.Level12.RyuVsBison")()
# model = PPO("CnnPolicy", env) # model = PPO("CnnPolicy", env)
if not RANDOM_ACTION: if not RANDOM_ACTION:
# model.load(MODEL_PATH) model = PPO.load(os.path.join(MODEL_DIR, MODEL_NAME), env=env)
model = PPO.load(MODEL_PATH, env=env)
# obs = env.reset() # obs = env.reset()
done = False done = False
@ -40,6 +44,13 @@ num_victory = 0
for _ in range(num_episodes): for _ in range(num_episodes):
done = False done = False
obs = env.reset() obs = env.reset()
if RECORDING:
# Start recording
movie_path = os.path.join(MOVIE_DIR, "{}.bk2".format(MODEL_NAME))
env.unwrapped.movie = retro.Movie(movie_path, retro.MovieMode.RECORD)
env.unwrapped.movie.step()
total_reward = 0 total_reward = 0
while not done: while not done:
@ -51,10 +62,19 @@ for _ in range(num_episodes):
action, _states = model.predict(obs) action, _states = model.predict(obs)
obs, reward, done, info = env.step(action) obs, reward, done, info = env.step(action)
if RECORDING:
# Record the step
env.unwrapped.movie.step()
if reward != 0: if reward != 0:
total_reward += reward total_reward += reward
print("Reward: {:.3f}, playerHP: {}, enemyHP:{}".format(reward, info['agent_hp'], info['enemy_hp'])) print("Reward: {:.3f}, playerHP: {}, enemyHP:{}".format(reward, info['agent_hp'], info['enemy_hp']))
if RECORDING:
# Stop recording
env.unwrapped.movie.close()
del env.unwrapped.movie
if info['enemy_hp'] < 0: if info['enemy_hp'] < 0:
print("Victory!") print("Victory!")
num_victory += 1 num_victory += 1
@ -66,4 +86,4 @@ print("Winning rate: {}".format(1.0 * num_victory / num_episodes))
if RANDOM_ACTION: if RANDOM_ACTION:
print("Average reward for random action: {}".format(episode_reward_sum/num_episodes)) print("Average reward for random action: {}".format(episode_reward_sum/num_episodes))
else: else:
print("Average reward for {}: {}".format(MODEL_PATH, episode_reward_sum/num_episodes)) print("Average reward for {}: {}".format(MODEL_NAME, episode_reward_sum/num_episodes))