nle basic

Author: Joseph Suarez

pufferlib/config/nethack.ini

@@ -1,16 +1,31 @@
 [base]
 package = nethack
 env_name = nethack
+policy_name = Policy
+rnn_name = Recurrent
 
 [vec]
-num_envs = 128
+num_envs = 8192
 num_workers = 16
-batch_size = 64
+batch_size = 4096
 
 [train]
-batch_size = 8192
-minibatch_size = 2048
-update_epochs = 1
+total_timesteps = 90_000_000
+adam_beta1 = 0.8946507418260217
+adam_beta2 = 0.9
+adam_eps = 0.0001
+batch_size = auto
 bptt_horizon = 64
-total_timesteps = 10_000_000
-anneal_lr = False
+clip_coef = 0.19696765958267629
+ent_coef = 0.0005690816545012474
+gae_lambda = 0.747650023961198
+gamma = 0.9997053654668936
+learning_rate = 0.044482546441415506
+max_grad_norm = 2.2356112188495723
+minibatch_size = 32768
+prio_alpha = 0.98967001208896
+prio_beta0 = 0.09999999999999998
+vf_clip_coef = 2.178492167689251
+vf_coef = 1.6832989594296321
+vtrace_c_clip = 2.878171091654008
+vtrace_rho_clip = 0.7876748061547312