Skip to content

Commit 065f04b

Browse files
kengzkengz
committed
split ac roboschool specs to new files
1 parent a447ad8 commit 065f04b

8 files changed

+253
-247
lines changed

Diff for: ‎job/cont_benchmark.json

-14
This file was deleted.

Diff for: ‎job/roboschool_benchmark.json

+14
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,14 @@
1+
{
2+
"slm_lab/spec/benchmark/a2c/a2c_nstep_roboschool.json": {
3+
"a2c_nstep_roboschool": "train",
4+
},
5+
"slm_lab/spec/benchmark/a2c/a2c_gae_roboschool.json": {
6+
"a2c_gae_roboschool": "train",
7+
},
8+
"slm_lab/spec/benchmark/ppo/ppo_roboschool.json": {
9+
"ppo_roboschool": "train",
10+
},
11+
"slm_lab/spec/benchmark/sac/sac_roboschool.json": {
12+
"sac_roboschool": "train",
13+
}
14+
}

Diff for: ‎slm_lab/spec/benchmark/a2c/a2c_gae_cont.json

-76
Original file line numberDiff line numberDiff line change
@@ -142,81 +142,5 @@
142142
"max_session": 4,
143143
"max_trial": 1
144144
}
145-
},
146-
"a2c_gae_roboschool": {
147-
"agent": [{
148-
"name": "A2C",
149-
"algorithm": {
150-
"name": "ActorCritic",
151-
"action_pdtype": "default",
152-
"action_policy": "default",
153-
"explore_var_spec": null,
154-
"gamma": 0.99,
155-
"lam": 0.95,
156-
"num_step_returns": null,
157-
"entropy_coef_spec": {
158-
"name": "no_decay",
159-
"start_val": 0.0,
160-
"end_val": 0.0,
161-
"start_step": 0,
162-
"end_step": 0
163-
},
164-
"val_loss_coef": 0.5,
165-
"training_frequency": 2048
166-
},
167-
"memory": {
168-
"name": "OnPolicyBatchReplay",
169-
},
170-
"net": {
171-
"type": "MLPNet",
172-
"shared": false,
173-
"hid_layers": [64, 64],
174-
"hid_layers_activation": "tanh",
175-
"init_fn": "orthogonal_",
176-
"normalize": false,
177-
"batch_norm": false,
178-
"clip_grad_val": 0.5,
179-
"use_same_optim": false,
180-
"loss_spec": {
181-
"name": "MSELoss"
182-
},
183-
"actor_optim_spec": {
184-
"name": "Adam",
185-
"lr": 3e-4,
186-
},
187-
"critic_optim_spec": {
188-
"name": "Adam",
189-
"lr": 3e-4,
190-
},
191-
"lr_scheduler_spec": {
192-
"name": "LinearToZero",
193-
"frame": 1e6
194-
},
195-
"gpu": true
196-
}
197-
}],
198-
"env": [{
199-
"name": "${env}",
200-
"num_envs": 32,
201-
"max_t": null,
202-
"max_frame": 1e6
203-
}],
204-
"body": {
205-
"product": "outer",
206-
"num": 1
207-
},
208-
"meta": {
209-
"distributed": false,
210-
"log_frequency": 1000,
211-
"eval_frequency": 1000,
212-
"rigorous_eval": 0,
213-
"max_session": 4,
214-
"max_trial": 1
215-
},
216-
"spec_params": {
217-
"env": [
218-
"RoboschoolAnt-v1", "RoboschoolHalfCheetah-v1", "RoboschoolHopper-v1", "RoboschoolWalker2d-v1"
219-
]
220-
}
221145
}
222146
}

Diff for: ‎slm_lab/spec/benchmark/a2c/a2c_gae_roboschool.json

+78
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,78 @@
1+
{
2+
"a2c_gae_roboschool": {
3+
"agent": [{
4+
"name": "A2C",
5+
"algorithm": {
6+
"name": "ActorCritic",
7+
"action_pdtype": "default",
8+
"action_policy": "default",
9+
"explore_var_spec": null,
10+
"gamma": 0.99,
11+
"lam": 0.95,
12+
"num_step_returns": null,
13+
"entropy_coef_spec": {
14+
"name": "no_decay",
15+
"start_val": 0.0,
16+
"end_val": 0.0,
17+
"start_step": 0,
18+
"end_step": 0
19+
},
20+
"val_loss_coef": 0.5,
21+
"training_frequency": 2048
22+
},
23+
"memory": {
24+
"name": "OnPolicyBatchReplay",
25+
},
26+
"net": {
27+
"type": "MLPNet",
28+
"shared": false,
29+
"hid_layers": [64, 64],
30+
"hid_layers_activation": "tanh",
31+
"init_fn": "orthogonal_",
32+
"normalize": false,
33+
"batch_norm": false,
34+
"clip_grad_val": 0.5,
35+
"use_same_optim": false,
36+
"loss_spec": {
37+
"name": "MSELoss"
38+
},
39+
"actor_optim_spec": {
40+
"name": "Adam",
41+
"lr": 3e-4,
42+
},
43+
"critic_optim_spec": {
44+
"name": "Adam",
45+
"lr": 3e-4,
46+
},
47+
"lr_scheduler_spec": {
48+
"name": "LinearToZero",
49+
"frame": 1e6
50+
},
51+
"gpu": true
52+
}
53+
}],
54+
"env": [{
55+
"name": "${env}",
56+
"num_envs": 32,
57+
"max_t": null,
58+
"max_frame": 1e6
59+
}],
60+
"body": {
61+
"product": "outer",
62+
"num": 1
63+
},
64+
"meta": {
65+
"distributed": false,
66+
"log_frequency": 1000,
67+
"eval_frequency": 1000,
68+
"rigorous_eval": 0,
69+
"max_session": 4,
70+
"max_trial": 1
71+
},
72+
"spec_params": {
73+
"env": [
74+
"RoboschoolAnt-v1", "RoboschoolHalfCheetah-v1", "RoboschoolHopper-v1", "RoboschoolWalker2d-v1"
75+
]
76+
}
77+
}
78+
}

Diff for: ‎slm_lab/spec/benchmark/a2c/a2c_nstep_cont.json

-76
Original file line numberDiff line numberDiff line change
@@ -135,81 +135,5 @@
135135
"max_session": 4,
136136
"max_trial": 1
137137
}
138-
},
139-
"a2c_nstep_roboschool": {
140-
"agent": [{
141-
"name": "A2C",
142-
"algorithm": {
143-
"name": "ActorCritic",
144-
"action_pdtype": "default",
145-
"action_policy": "default",
146-
"explore_var_spec": null,
147-
"gamma": 0.99,
148-
"lam": null,
149-
"num_step_returns": 5,
150-
"entropy_coef_spec": {
151-
"name": "no_decay",
152-
"start_val": 0.0,
153-
"end_val": 0.0,
154-
"start_step": 0,
155-
"end_step": 0
156-
},
157-
"val_loss_coef": 0.5,
158-
"training_frequency": 2048
159-
},
160-
"memory": {
161-
"name": "OnPolicyBatchReplay",
162-
},
163-
"net": {
164-
"type": "MLPNet",
165-
"shared": false,
166-
"hid_layers": [64, 64],
167-
"hid_layers_activation": "tanh",
168-
"init_fn": "orthogonal_",
169-
"normalize": false,
170-
"batch_norm": false,
171-
"clip_grad_val": 0.5,
172-
"use_same_optim": false,
173-
"loss_spec": {
174-
"name": "MSELoss"
175-
},
176-
"actor_optim_spec": {
177-
"name": "Adam",
178-
"lr": 3e-4,
179-
},
180-
"critic_optim_spec": {
181-
"name": "Adam",
182-
"lr": 3e-4,
183-
},
184-
"lr_scheduler_spec": {
185-
"name": "LinearToZero",
186-
"frame": 1e6
187-
},
188-
"gpu": true
189-
}
190-
}],
191-
"env": [{
192-
"name": "${env}",
193-
"num_envs": 32,
194-
"max_t": null,
195-
"max_frame": 1e6
196-
}],
197-
"body": {
198-
"product": "outer",
199-
"num": 1
200-
},
201-
"meta": {
202-
"distributed": false,
203-
"log_frequency": 1000,
204-
"eval_frequency": 1000,
205-
"rigorous_eval": 0,
206-
"max_session": 4,
207-
"max_trial": 1
208-
},
209-
"spec_params": {
210-
"env": [
211-
"RoboschoolAnt-v1", "RoboschoolHalfCheetah-v1", "RoboschoolHopper-v1", "RoboschoolWalker2d-v1"
212-
]
213-
}
214138
}
215139
}

Diff for: ‎slm_lab/spec/benchmark/a2c/a2c_nstep_roboschool.json

+78
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,78 @@
1+
{
2+
"a2c_nstep_roboschool": {
3+
"agent": [{
4+
"name": "A2C",
5+
"algorithm": {
6+
"name": "ActorCritic",
7+
"action_pdtype": "default",
8+
"action_policy": "default",
9+
"explore_var_spec": null,
10+
"gamma": 0.99,
11+
"lam": null,
12+
"num_step_returns": 5,
13+
"entropy_coef_spec": {
14+
"name": "no_decay",
15+
"start_val": 0.0,
16+
"end_val": 0.0,
17+
"start_step": 0,
18+
"end_step": 0
19+
},
20+
"val_loss_coef": 0.5,
21+
"training_frequency": 2048
22+
},
23+
"memory": {
24+
"name": "OnPolicyBatchReplay",
25+
},
26+
"net": {
27+
"type": "MLPNet",
28+
"shared": false,
29+
"hid_layers": [64, 64],
30+
"hid_layers_activation": "tanh",
31+
"init_fn": "orthogonal_",
32+
"normalize": false,
33+
"batch_norm": false,
34+
"clip_grad_val": 0.5,
35+
"use_same_optim": false,
36+
"loss_spec": {
37+
"name": "MSELoss"
38+
},
39+
"actor_optim_spec": {
40+
"name": "Adam",
41+
"lr": 3e-4,
42+
},
43+
"critic_optim_spec": {
44+
"name": "Adam",
45+
"lr": 3e-4,
46+
},
47+
"lr_scheduler_spec": {
48+
"name": "LinearToZero",
49+
"frame": 1e6
50+
},
51+
"gpu": true
52+
}
53+
}],
54+
"env": [{
55+
"name": "${env}",
56+
"num_envs": 32,
57+
"max_t": null,
58+
"max_frame": 1e6
59+
}],
60+
"body": {
61+
"product": "outer",
62+
"num": 1
63+
},
64+
"meta": {
65+
"distributed": false,
66+
"log_frequency": 1000,
67+
"eval_frequency": 1000,
68+
"rigorous_eval": 0,
69+
"max_session": 4,
70+
"max_trial": 1
71+
},
72+
"spec_params": {
73+
"env": [
74+
"RoboschoolAnt-v1", "RoboschoolHalfCheetah-v1", "RoboschoolHopper-v1", "RoboschoolWalker2d-v1"
75+
]
76+
}
77+
}
78+
}

0 commit comments

Comments
 (0)