fix(inverse-grin): update regularization weights in loss calculation

Author: liblaf

exp/2025/09/24/inverse-grin/data/32-inverse.vtu.series

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fddfec6f322a3cbec7a1b3a65cac5103ee2566fa440e1072ec41f9bba2016da8
-size 22592
+oid sha256:b9ba7de5157ef92dccc218fec5d5bafda378347dfb89969c406a10f69a8a74c8
+size 115

exp/2025/09/24/inverse-grin/data/32-inverse.vtu/32-inverse_000000.vtu

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5f2d7537bc7844f6ff6e9c7c834a5c777b7367adaa52cad51c547fd2388b7068
+oid sha256:a9c078b86a0155a8b79d5fb1b91a2b53eb4453aaea188801b90c01ddb245d40f
 size 4978337

exp/2025/09/24/inverse-grin/src/32-inverse.py

@@ -275,9 +275,9 @@ def fun_and_jac(self, q: Array) -> tuple[Scalar, Array]:
 
     def loss(self, x: Vector, params: Params) -> tuple[Scalar, InverseLossAux]:
         loss_surface: Scalar = self.loss_surface(x)
-        reg_mean: Scalar = 1e3 * self.regularize_mean(params)
-        reg_shear: Scalar = 1e3 * self.regularize_shear(params)
-        reg_volume: Scalar = 1e3 * self.regularize_volume(params)
+        reg_mean: Scalar = self.reg_mean_weight * self.regularize_mean(params)
+        reg_shear: Scalar = self.reg_shear_weight * self.regularize_shear(params)
+        reg_volume: Scalar = self.reg_volume_weight * self.regularize_volume(params)
         # jax.debug.print("loss_surface = {}", loss_surface)
         # jax.debug.print("reg_mean = {}", reg_mean)
         # jax.debug.print("reg_shear = {}", reg_shear)

exp/2025/09/24/inverse-grin/src/33-inverse-no-reg.py

@@ -275,9 +275,9 @@ def fun_and_jac(self, q: Array) -> tuple[Scalar, Array]:
 
     def loss(self, x: Vector, params: Params) -> tuple[Scalar, InverseLossAux]:
         loss_surface: Scalar = self.loss_surface(x)
-        reg_mean: Scalar = 1e3 * self.regularize_mean(params)
-        reg_shear: Scalar = 1e3 * self.regularize_shear(params)
-        reg_volume: Scalar = 1e3 * self.regularize_volume(params)
+        reg_mean: Scalar = self.reg_mean_weight * self.regularize_mean(params)
+        reg_shear: Scalar = self.reg_shear_weight * self.regularize_shear(params)
+        reg_volume: Scalar = self.reg_volume_weight * self.regularize_volume(params)
         # jax.debug.print("loss_surface = {}", loss_surface)
         # jax.debug.print("reg_mean = {}", reg_mean)
         # jax.debug.print("reg_shear = {}", reg_shear)
@@ -341,7 +341,7 @@ def regularize_volume(self, params: Params) -> Scalar:
         return regularization
 
 
-def main(cfg: Config) -> None:
+def main(cfg: Config) -> None:  # noqa: PLR0915
     mesh: pv.UnstructuredGrid = melon.load_unstructured_grid(cfg.input)
     target: pv.UnstructuredGrid = melon.load_unstructured_grid(cfg.target)
 
@@ -395,16 +395,17 @@ def callback(intermediate_result: optim.Solution) -> None:
         # result: pv.UnstructuredGrid = mesh.warp_by_vector("solution")  # pyright: ignore[reportAssignmentType]
         writer.append(mesh)
 
+    inverse.reg_mean_weight = 0.0
+    inverse.reg_shear_weight = 0.0
+    inverse.reg_volume_weight = 0.0
+
     q_init: Float[Array, "ca 6"] = sim_jax.rest_activation(inverse.n_active_cells)
     # q_init = q_init.at[:, :3].set(jnp.log(1.0))
     callback(optim.Solution({"x": q_init}))
     optimizer = optim.MinimizerScipy(
         jit=False, method="L-BFGS-B", tol=1e-15, options={}
     )
     inverse.linear_solver = lx.CG(rtol=1e-3, atol=1e-30, max_steps=10000)
-    inverse.reg_mean_weight = 0.0
-    inverse.reg_shear_weight = 0.0
-    inverse.reg_volume_weight = 0.0
     solution: optim.Solution = optimizer.minimize(
         x0=q_init, fun_and_jac=inverse.fun_and_jac, callback=callback
     )