AirSim-v0 | prabhasak Workspace – Weights & Biases

Skip to main content

Prabhasak's group workspace

Group: AirSim-v0

1-11

of 11

Tags

Notes

Author

State

Finished

Start time

October 1st, 2020 5:49:35 PM

Runtime

20h 7m 10s

Tracked hours

20h 7m 10s

Run path

prabhasak/masters-thesis/3hd23ggi

OS

Windows-10-10.0.18362-SP0

Python version

3.7.9

Command

train.py --env AirSim-v0 --algo sac --exp-id 2 -il -eval -best -tb -wandb --env-kwargs name:'Drone3' rew_complexity:'complex' --num-trajs 20

System Hardware

CPU count	12
GPU count	1
GPU type	GeForce GTX 1660 Ti

W&B CLI Version

0.10.2

Group

Config parameters are your model's inputs. Learn more

▶
Config parameters:{} 30 keys
- algo:
  "sac"
- cg_damping:
  0.1
- cg_iters:
  10
- check_callback:
  false
- device:
  "gpu"
- entcoeff:
  0
- env:
  "AirSim-v0"
- ▶
  env_kwargs:{} 3 keys
  - expert:
    "prabhasa"
  - name:
    "Drone3"
  - rew_complexity:
    "complex"
- eval_callback:
  true
- exp_id:
  2
- gamma:
  0.99
- human_expert:
  "sac"
- hyperparams_IL:{} 0 keys
- hyperparams_RL:{} 0 keys
- lam:
  0.98
- max_kl:
  0.01
- num_trajs:
  20
- policy:
  "MlpPolicy"
- save_best_model:
  true
- seed:
  42
- tensorboard:
  true
- timesteps_IL:
  0
- timesteps_per_batch:
  1,024
- timesteps_RL:
  0
- train_IL:
  true
- train_RL:
  false
- verbose:
  0
- vf_iters:
  5
- vf_stepsize:
  0.001
- wandb_log:
  true

Summary metrics are your model's outputs. Learn more

▶
Summary metrics:{} 26 keys
- ▶
  advantage.bins:[] 33 items
- ▶
  advantage.values:[] 32 items
- entloss:
  0
- entropy:
  1.0163768529891968
- episodes:
  3,647
- expert episode length (mean):
  94.07
- expert episode length (std):
  0
- expert score (mean):
  1,265.4131816387176
- expert score (std):
  225.7059747942512
- iterations:
  489
- learning rate:
  0.001
- mean episode length:
  298.075
- mean true reward (IL):
  349.736770675
- meankl:
  0.010927680879831314
- num traj:
  20
- optimgain:
  0.024516791105270382
- seed:
  42
- surrgain:
  0.024516791105270382
- test episode length (mean):
  380.16
- test episode length (std):
  86.48175761396155
- test score (mean):
  385.0349731445313
- test score (std):
  91.41716003417967
- time elapsed:
  71,380.24983739853
- time per step:
  0.07512456667609513
- total loss:
  1.0763381160795689
- total steps:
  500,736

This run produced these artifacts as outputs. Total: 1. Learn more

run-3hd23ggi-history:v0

Loading...