菜单
Wrap a Kaggle competitive game environment as an OpenAI Gym env with continuous action space for training PPO agents via stable-baselines3
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Memory-efficient Keras generator that streams sharded CSV files in chunks, renders strokes to images on-the-fly, and yields batches for training on datasets too large for memory
Stack 1D convolutions for local feature extraction before bidirectional LSTMs to classify variable-length stroke sequences into hundreds of doodle categories
Partition a large dataset into N balanced shards using integer key modulo arithmetic for reproducible, class-interleaved splits across CSV files
Normalize raw stroke coordinates to 0-255 range, resample at uniform arc-length spacing, then apply Ramer-Douglas-Peucker simplification
Render stroke sequences to grayscale images with temporal intensity encoding where earlier strokes are brighter and later strokes fade to encode drawing order
Score each face in a video as anomalous by computing L2 distance from the embedding centroid of all faces, then convert to probability via logistic function
| name | tabular-ppo-gym-wrapper-kaggle-env |
| description | Wrap a Kaggle competitive game environment as an OpenAI Gym env with continuous action space for training PPO agents via stable-baselines3 |
Kaggle game AI competitions (Halite, Kore, Lux) use kaggle_environments which don't conform to the Gym API. Wrapping them in a gym.Env subclass with defined observation and action spaces enables training with stable-baselines3 PPO (or SAC, A2C). The wrapper handles state encoding, action translation, opponent management, and episode termination.
import gym
from gym import spaces
import numpy as np
from kaggle_environments import make
from stable_baselines3 import PPO
from stable_baselines3.common.monitor import Monitor
class KaggleGymEnv(gym.Env):
def __init__(self, opponent="random"):
super().__init__()
self.env = make("kore_fleets", debug=True)
self.opponent = opponent
self.observation_space = spaces.Box(-1, 1, shape=(21*21*4+3,), dtype=np.float32)
self.action_space = spaces.Box(-1, 1, shape=(3,), dtype=np.float32)
def reset(self):
self.trainer = self.env.train([None, self.opponent])
obs = self.trainer.reset()
return self._encode(obs)
def step(self, action):
game_action = self._decode(action)
obs, reward, done, info = self.trainer.step(game_action)
return self._encode(obs), reward, done, info
env = Monitor(KaggleGymEnv())
model = PPO("MlpPolicy", env, verbose=1)
model.learn(total_timesteps=100_000)
gym.Env, define observation and action spacesreset(), create a trainer via env.train([None, opponent]) — None marks the learning agentstep(), decode continuous actions to game actions, call trainer.step()Monitor for logging, train with PPO or similar algorithmstep()