[bugfix] - fix L1 and L2 regularisation gradients in grad_compute

Previous code computed scalar norms and multiplied by W, giving gradients
that scaled with total weight magnitude rather than the correct per-element
derivatives:
  L1: gradient is λ·sign(W), not λ·‖W‖₁·W
  L2: gradient is 2λ·W, not λ·‖W‖₂²·W
Also removed the combined (l1+l2)*W term which incorrectly mixed the two.
weight_decay (correct L2 form λW) is kept as a separate term.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

src/rbm/entity.py

@@ -97,15 +97,11 @@ class Entity:
 		self.grad.b_v = params.momentum * self.grad.b_v + lr * d_bv
 		self.grad.b_h = params.momentum * self.grad.b_h + lr * d_bh
 
-		# compute L1 term and penalize cost function (d_whv)
-		l1_norm = np.sum(np.abs(self.state.w_hv))
-		l1_term = params.l1_lambda*l1_norm
-
-		# compute L2 term and penalize cost function (d_whv)
-		l2_norm = np.sum(np.square(self.state.w_hv))
-		l2_term = params.l2_lambda*l2_norm
-
-		self.grad.w_hv = params.momentum * self.grad.w_hv + lr * (d_whv - self.state.w_hv*(l1_term+l2_term)) - lr * params.weight_decay * self.state.w_hv
+		self.grad.w_hv = (params.momentum * self.grad.w_hv
+		                  + lr * d_whv
+		                  - lr * params.l1_lambda    * np.sign(self.state.w_hv)
+		                  - lr * 2 * params.l2_lambda * self.state.w_hv
+		                  - lr * params.weight_decay  * self.state.w_hv)
 
 		return self.grad