DQN: Make the soft_update_model_params method static

alessiodm · May 23, 2024 · e2fe543 · e2fe543
1 parent 52d37a3
commit e2fe543
Show file tree

Hide file tree

Showing 2 changed files with 4 additions and 2 deletions.
diff --git a/03_DQN.ipynb b/03_DQN.ipynb
@@ -478,6 +478,7 @@
  "\n",
  " # TODO: Minimize the loss. Hint: zero_grad the optim, backward on loss, step the optim. \n",
  "\n",
+ " @staticmethod\n",
  " def soft_update_model_params(src: torch.nn.Module, dest: torch.nn.Module, tau=1e-3):\n",
  " \"\"\"Soft updates model parameters (θ_dest = τ * θ_src + (1 - τ) * θ_src).\"\"\"\n",
  " # TODO: For each dest parameter (get them via the parameters() function), update it with\n",

diff --git a/solution/03_DQN.ipynb b/solution/03_DQN.ipynb
@@ -448,7 +448,7 @@
  " self.learn()\n",
  " self.t_update_target_step = (self.t_update_target_step + 1) % self.update_target_every\n",
  " if self.t_update_target_step == 0:\n",
- " self.soft_update_model_params(self.qnetwork_local, self.qnetwork_target, self.tau)\n",
+ " Agent.soft_update_model_params(self.qnetwork_local, self.qnetwork_target, self.tau)\n",
  "\n",
  " def act(self, state: np.array, eps=0.):\n",
  " \"\"\"Makes the agent take an action for the state passed as input.\"\"\"\n",
@@ -480,7 +480,8 @@
  " loss.backward()\n",
  " self.optimizer.step() \n",
  "\n",
- " def soft_update_model_params(self, src: torch.nn.Module, dest: torch.nn.Module, tau=1e-3):\n",
+ " @staticmethod\n",
+ " def soft_update_model_params(src: torch.nn.Module, dest: torch.nn.Module, tau=1e-3):\n",
  " \"\"\"Soft updates model parameters (θ_dest = τ * θ_src + (1 - τ) * θ_src).\"\"\"\n",
  " for dest_param, src_param in zip(dest.parameters(), src.parameters()):\n",
  " dest_param.data.copy_(tau * src_param.data + (1.0 - tau) * dest_param.data)\n",