Update README.md
Browse files
README.md
CHANGED
|
@@ -32,4 +32,49 @@ tags:
|
|
| 32 |
2. Step 1: Find your model_id: lambdavi/ppo-SnowballTarget
|
| 33 |
3. Step 2: Select your *.nn /*.onnx file
|
| 34 |
4. Click on Watch the agent play 👀
|
| 35 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 32 |
2. Step 1: Find your model_id: lambdavi/ppo-SnowballTarget
|
| 33 |
3. Step 2: Select your *.nn /*.onnx file
|
| 34 |
4. Click on Watch the agent play 👀
|
| 35 |
+
|
| 36 |
+
### Hyperparams used:
|
| 37 |
+
SnowballTarget:
|
| 38 |
+
trainer_type: ppo
|
| 39 |
+
hyperparameters:
|
| 40 |
+
batch_size: 128
|
| 41 |
+
buffer_size: 2048
|
| 42 |
+
learning_rate: 0.005
|
| 43 |
+
beta: 0.005
|
| 44 |
+
epsilon: 0.2
|
| 45 |
+
lambd: 0.95
|
| 46 |
+
num_epoch: 5
|
| 47 |
+
shared_critic: False
|
| 48 |
+
learning_rate_schedule: linear
|
| 49 |
+
beta_schedule: linear
|
| 50 |
+
epsilon_schedule: linear
|
| 51 |
+
checkpoint_interval: 50000
|
| 52 |
+
network_settings:
|
| 53 |
+
normalize: False
|
| 54 |
+
hidden_units: 256
|
| 55 |
+
num_layers: 2
|
| 56 |
+
vis_encode_type: simple
|
| 57 |
+
memory: None
|
| 58 |
+
goal_conditioning_type: hyper
|
| 59 |
+
deterministic: False
|
| 60 |
+
reward_signals:
|
| 61 |
+
extrinsic:
|
| 62 |
+
gamma: 0.99
|
| 63 |
+
strength: 1.0
|
| 64 |
+
network_settings:
|
| 65 |
+
normalize: False
|
| 66 |
+
hidden_units: 128
|
| 67 |
+
num_layers: 2
|
| 68 |
+
vis_encode_type: simple
|
| 69 |
+
memory: None
|
| 70 |
+
goal_conditioning_type: hyper
|
| 71 |
+
deterministic: False
|
| 72 |
+
init_path: None
|
| 73 |
+
keep_checkpoints: 10
|
| 74 |
+
even_checkpoints: False
|
| 75 |
+
max_steps: 500000
|
| 76 |
+
time_horizon: 64
|
| 77 |
+
summary_freq: 10000
|
| 78 |
+
threaded: True
|
| 79 |
+
self_play: None
|
| 80 |
+
behavioral_cloning: None
|