From 3bbf1cea6fa3e43268efeda9da4b6c75cde77936 Mon Sep 17 00:00:00 2001
From: Erik Nygren <erik.nygren@sbb.ch>
Date: Tue, 23 Apr 2019 17:09:30 +0200
Subject: [PATCH] updated level parameters for training
---
examples/training_navigation.py | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/examples/training_navigation.py b/examples/training_navigation.py
index 18bb6356..681e8b78 100644
--- a/examples/training_navigation.py
+++ b/examples/training_navigation.py
@@ -74,7 +74,7 @@ for trials in range(1, n_trials + 1):
#print(step)
# Action
for a in range(env.number_of_agents):
- action = agent.act(np.array(obs[a]), eps=0)
+ action = agent.act(np.array(obs[a]), eps=eps)
action_prob[action] += 1
action_dict.update({a: action})
@@ -110,7 +110,7 @@ for trials in range(1, n_trials + 1):
eps, action_prob/np.sum(action_prob)),
end=" ")
if trials % 100 == 0:
- action_prob = [1]*4
+
print(
'\rTraining {} Agents.\tEpisode {}\tAverage Score: {:.0f}\tDones: {:.2f}%\tEpsilon: {:.2f} \t Action Probabilities: \t {}'.format(
env.number_of_agents,
@@ -122,4 +122,4 @@ for trials in range(1, n_trials + 1):
eps, action_prob / np.sum(action_prob)))
torch.save(agent.qnetwork_local.state_dict(),
'../flatland/baselines/Nets/avoid_checkpoint' + str(trials) + '.pth')
-
+ action_prob = [1]*4
--
GitLab