shivakanthsujit
commited on
Commit
·
8cc1b26
1
Parent(s):
42c1704
Init Commit
Browse files- README.md +3 -2
- config.yaml +97 -15
- model/env_stats.pickle +2 -2
- model/model.pth +2 -2
README.md
CHANGED
@@ -6,8 +6,9 @@ tags:
|
|
6 |
- reinforcement-learning
|
7 |
- mbrl-lib
|
8 |
---
|
9 |
-
# **OneDTransitionRewardModel** Agent playing **mbrl-continuous-cartpole**
|
10 |
-
This is a trained model of a **OneDTransitionRewardModel** agent
|
|
|
11 |
using [MBRL-Lib](https://github.com/facebookresearch/mbrl-lib).
|
12 |
|
13 |
## Usage (with MBRL-Lib)
|
|
|
6 |
- reinforcement-learning
|
7 |
- mbrl-lib
|
8 |
---
|
9 |
+
# **OneDTransitionRewardModel w/ SACAgent** Agent playing **mbrl-continuous-cartpole**
|
10 |
+
This is a trained model of a **OneDTransitionRewardModel w/ SACAgent** agent
|
11 |
+
playing **mbrl-continuous-cartpole**
|
12 |
using [MBRL-Lib](https://github.com/facebookresearch/mbrl-lib).
|
13 |
|
14 |
## Usage (with MBRL-Lib)
|
config.yaml
CHANGED
@@ -1,20 +1,102 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
dynamics_model:
|
2 |
_target_: mbrl.models.GaussianMLP
|
3 |
-
|
4 |
-
|
5 |
-
|
|
|
|
|
6 |
hid_size: 200
|
7 |
in_size: 5
|
8 |
-
|
9 |
-
|
10 |
-
|
11 |
-
|
12 |
-
|
13 |
-
|
14 |
-
algorithm:
|
15 |
-
learned_rewards: false
|
16 |
-
target_is_delta: true
|
17 |
-
normalize: true
|
18 |
overrides:
|
19 |
-
|
20 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
action_optimizer:
|
2 |
+
_target_: mbrl.planning.CEMOptimizer
|
3 |
+
alpha: 0.1
|
4 |
+
clipped_normal: false
|
5 |
+
device: cpu:0
|
6 |
+
elite_ratio: 0.1
|
7 |
+
lower_bound: ???
|
8 |
+
num_iterations: 5
|
9 |
+
population_size: 350
|
10 |
+
return_mean_elites: true
|
11 |
+
upper_bound: ???
|
12 |
+
algorithm:
|
13 |
+
agent:
|
14 |
+
_target_: mbrl.third_party.pytorch_sac_pranz24.sac.SAC
|
15 |
+
action_space:
|
16 |
+
_target_: gym.env.Box
|
17 |
+
high:
|
18 |
+
- 1.0
|
19 |
+
low:
|
20 |
+
- -1.0
|
21 |
+
shape:
|
22 |
+
- 1
|
23 |
+
args:
|
24 |
+
alpha: 0.2
|
25 |
+
automatic_entropy_tuning: true
|
26 |
+
device: cpu:0
|
27 |
+
gamma: 0.99
|
28 |
+
hidden_size: 256
|
29 |
+
lr: 0.0003
|
30 |
+
policy: Gaussian
|
31 |
+
target_entropy: -0.05
|
32 |
+
target_update_interval: 4
|
33 |
+
tau: 0.005
|
34 |
+
num_inputs: 4
|
35 |
+
freq_train_model: 200
|
36 |
+
initial_exploration_steps: 5000
|
37 |
+
learned_rewards: true
|
38 |
+
name: mbpo
|
39 |
+
normalize: true
|
40 |
+
normalize_double_precision: true
|
41 |
+
num_eval_episodes: 1
|
42 |
+
random_initial_explore: false
|
43 |
+
real_data_ratio: 0.0
|
44 |
+
sac_samples_action: true
|
45 |
+
target_is_delta: true
|
46 |
+
debug_mode: false
|
47 |
+
device: cpu:0
|
48 |
dynamics_model:
|
49 |
_target_: mbrl.models.GaussianMLP
|
50 |
+
activation_fn_cfg:
|
51 |
+
_target_: torch.nn.SiLU
|
52 |
+
deterministic: false
|
53 |
+
device: cpu:0
|
54 |
+
ensemble_size: 7
|
55 |
hid_size: 200
|
56 |
in_size: 5
|
57 |
+
learn_logvar_bounds: false
|
58 |
+
num_layers: 4
|
59 |
+
out_size: 5
|
60 |
+
propagation_method: random_model
|
61 |
+
experiment: default
|
62 |
+
log_frequency_agent: 1000
|
|
|
|
|
|
|
|
|
63 |
overrides:
|
64 |
+
cem_alpha: 0.1
|
65 |
+
cem_clipped_normal: false
|
66 |
+
cem_elite_ratio: 0.1
|
67 |
+
cem_num_iters: 5
|
68 |
+
cem_population_size: 350
|
69 |
+
effective_model_rollouts_per_step: 400
|
70 |
+
env: cartpole_continuous
|
71 |
+
epoch_length: 200
|
72 |
+
freq_train_model: 200
|
73 |
+
model_batch_size: 256
|
74 |
+
model_lr: 0.001
|
75 |
+
model_wd: 5.0e-05
|
76 |
+
num_elites: 5
|
77 |
+
num_epochs_to_retain_sac_buffer: 1
|
78 |
+
num_sac_updates_per_step: 20
|
79 |
+
num_steps: 5000
|
80 |
+
patience: 5
|
81 |
+
planning_horizon: 15
|
82 |
+
rollout_schedule:
|
83 |
+
- 1
|
84 |
+
- 15
|
85 |
+
- 1
|
86 |
+
- 1
|
87 |
+
sac_alpha: 0.2
|
88 |
+
sac_automatic_entropy_tuning: true
|
89 |
+
sac_batch_size: 256
|
90 |
+
sac_gamma: 0.99
|
91 |
+
sac_hidden_size: 256
|
92 |
+
sac_lr: 0.0003
|
93 |
+
sac_policy: Gaussian
|
94 |
+
sac_target_entropy: -0.05
|
95 |
+
sac_target_update_interval: 4
|
96 |
+
sac_tau: 0.005
|
97 |
+
sac_updates_every_steps: 1
|
98 |
+
trial_length: 200
|
99 |
+
validation_ratio: 0.2
|
100 |
+
root_dir: ./logs
|
101 |
+
save_video: false
|
102 |
+
seed: 0
|
model/env_stats.pickle
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:41c1ac53edc417a20114e12671aea2434e2e9b5125ebfef999e87c267d9fb5c8
|
3 |
+
size 278
|
model/model.pth
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a51a85c2bd7f4b05886da8820eb36cc4032084cd2acf14c0ae1d579ccfa9b2dc
|
3 |
+
size 3470565
|