Spaces:
Sleeping
Sleeping
Upload models/apm.py with huggingface_hub
Browse files- models/apm.py +26 -0
models/apm.py
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import torch
|
| 2 |
+
import torch.nn as nn
|
| 3 |
+
import torch.nn.functional as F
|
| 4 |
+
|
| 5 |
+
|
| 6 |
+
class APM(nn.Module):
|
| 7 |
+
"""Action Proposal Module: predicts the action bridging two latent states."""
|
| 8 |
+
|
| 9 |
+
def __init__(self, z_dim: int = 4, action_dim: int = 4, hidden_dim: int = 256):
|
| 10 |
+
super().__init__()
|
| 11 |
+
self.net = nn.Sequential(
|
| 12 |
+
nn.Linear(2 * z_dim, hidden_dim),
|
| 13 |
+
nn.ReLU(inplace=True),
|
| 14 |
+
nn.Linear(hidden_dim, hidden_dim),
|
| 15 |
+
nn.ReLU(inplace=True),
|
| 16 |
+
nn.Linear(hidden_dim, action_dim),
|
| 17 |
+
)
|
| 18 |
+
self.z_dim = z_dim
|
| 19 |
+
self.action_dim = action_dim
|
| 20 |
+
|
| 21 |
+
def forward(self, z_i: torch.Tensor, z_j: torch.Tensor) -> torch.Tensor:
|
| 22 |
+
return self.net(torch.cat([z_i, z_j], dim=-1))
|
| 23 |
+
|
| 24 |
+
|
| 25 |
+
def apm_loss(pred: torch.Tensor, target: torch.Tensor) -> torch.Tensor:
|
| 26 |
+
return F.mse_loss(pred, target)
|