From 2e6f49db85fcdf29a3a7f5929e6b9e28ecd43696 Mon Sep 17 00:00:00 2001 From: Shangtong Zhang Date: Sat, 22 Jul 2017 10:07:30 -0600 Subject: [PATCH] Fix failed softmax layer test --- DDPG_agent.py | 2 ++ main.py | 5 ++--- 2 files changed, 4 insertions(+), 3 deletions(-) diff --git a/DDPG_agent.py b/DDPG_agent.py index 7dc41be..db97f40 100644 --- a/DDPG_agent.py +++ b/DDPG_agent.py @@ -61,11 +61,13 @@ class DDPGAgent: total_reward = 0.0 while not self.step_limit or steps < self.step_limit: action = self.actor.predict(np.stack([state])).flatten() + self.logger.histo_summary('action', action, self.total_steps) if not deterministic: if self.total_steps < self.exploration_steps: action = self.task.random_action() else: action += self.random_process.sample() + self.logger.histo_summary('noised action', action, self.total_steps) next_state, reward, done, info = self.task.step(action) if not deterministic: self.replay.feed([state, action, reward, next_state, int(done)]) diff --git a/main.py b/main.py index 42280e8..09ee9b6 100644 --- a/main.py +++ b/main.py @@ -3,7 +3,6 @@ from DQN_agent import * from DDPG_agent import * from logger import * import logging -import traceback from random_process import * def dqn_cart_pole(): @@ -211,5 +210,5 @@ if __name__ == '__main__': # async_pixel_atari('BreakoutNoFrameskip-v3') # a3c_pixel_atari('BreakoutNoFrameskip-v3') - ddpg_pendulum() - # ddpg_bipedal_walker() \ No newline at end of file + # ddpg_pendulum() + ddpg_bipedal_walker() \ No newline at end of file