diff --git a/c_learning/README.md b/c_learning/README.md index 9af9fa9ac96..b215b771f2c 100644 --- a/c_learning/README.md +++ b/c_learning/README.md @@ -41,7 +41,7 @@ echo "export LD_LIBRARY_PATH=\$LD_LIBRARY_PATH:$HOME/.mujoco/mujoco200/bin" >> ~ ``` wget https://repo.anaconda.com/miniconda/Miniconda2-latest-Linux-x86_64.sh chmod +x Miniconda2-latest-Linux-x86_64.sh -chmod +x ./Miniconda2-latest-Linux-x86_64.sh +./Miniconda2-latest-Linux-x86_64.sh ``` Restart your terminal so the changes take effect. diff --git a/c_learning/c_learning_agent.py b/c_learning/c_learning_agent.py index 26f0bea416b..cec0964e10b 100644 --- a/c_learning/c_learning_agent.py +++ b/c_learning/c_learning_agent.py @@ -370,7 +370,7 @@ def critic_loss(self, gamma = 1.0, weights = None, training = False, - w_clipping = None, + w_clipping = 20.0, self_normalized = False, lambda_fix = False, ): diff --git a/darc/README.md b/darc/README.md index cd8ee45f6aa..7eb229d71cd 100644 --- a/darc/README.md +++ b/darc/README.md @@ -42,7 +42,7 @@ echo "export LD_LIBRARY_PATH=\$LD_LIBRARY_PATH:$HOME/.mujoco/mujoco200/bin" >> ~ ``` wget https://repo.anaconda.com/miniconda/Miniconda2-latest-Linux-x86_64.sh chmod +x Miniconda2-latest-Linux-x86_64.sh -chmod +x ./Miniconda2-latest-Linux-x86_64.sh +./Miniconda2-latest-Linux-x86_64.sh ``` Restart your terminal so the changes take effect. diff --git a/rce/README.md b/rce/README.md index b3da431e47d..c87f5983985 100644 --- a/rce/README.md +++ b/rce/README.md @@ -41,7 +41,7 @@ echo "export LD_LIBRARY_PATH=\$LD_LIBRARY_PATH:$HOME/.mujoco/mujoco200/bin" >> ~ ``` wget https://repo.anaconda.com/miniconda/Miniconda2-latest-Linux-x86_64.sh chmod +x Miniconda2-latest-Linux-x86_64.sh -chmod +x ./Miniconda2-latest-Linux-x86_64.sh +./Miniconda2-latest-Linux-x86_64.sh ``` Restart your terminal so the changes take effect. @@ -69,19 +69,19 @@ The following lines replicate the RCE experiments on the Sawyer tasks and Adept ``` -python train_eval.py --root_dir=~/c_learning/sawyer_drawer_open --gin_bindings='train_eval.env_name="sawyer_drawer_open"' +python train_eval.py --root_dir=~/rce/sawyer_drawer_open --gin_bindings='train_eval.env_name="sawyer_drawer_open"' -python train_eval.py --root_dir=~/c_learning/sawyer_push --gin_bindings='train_eval.env_name="sawyer_push"' +python train_eval.py --root_dir=~/rce/sawyer_push --gin_bindings='train_eval.env_name="sawyer_push"' -python train_eval.py --root_dir=~/c_learning/sawyer_lift --gin_bindings='train_eval.env_name="sawyer_lift"' +python train_eval.py --root_dir=~/rce/sawyer_lift --gin_bindings='train_eval.env_name="sawyer_lift"' -python train_eval.py --root_dir=~/c_learning/door --gin_bindings='train_eval.env_name="door-human-v0"' +python train_eval.py --root_dir=~/rce/door --gin_bindings='train_eval.env_name="door-human-v0"' -python train_eval.py --root_dir=~/c_learning/sawyer_box_close --gin_bindings='train_eval.env_name="sawyer_box_close"' +python train_eval.py --root_dir=~/rce/sawyer_box_close --gin_bindings='train_eval.env_name="sawyer_box_close"' -python train_eval.py --root_dir=~/c_learning/sawyer_bin_picking --gin_bindings='train_eval.env_name="sawyer_bin_picking"' --gin_bindings='critic_loss.q_combinator="max"' --gin_bindings='actor_loss.q_combinator="max"' +python train_eval.py --root_dir=~/rce/sawyer_bin_picking --gin_bindings='train_eval.env_name="sawyer_bin_picking"' --gin_bindings='critic_loss.q_combinator="max"' --gin_bindings='actor_loss.q_combinator="max"' -python train_eval.py --root_dir=~/c_learning/hammer --gin_bindings='train_eval.env_name="hammer-human-v0"' +python train_eval.py --root_dir=~/rce/hammer --gin_bindings='train_eval.env_name="hammer-human-v0"' ``` To run SQIL, add the additional arguments: