diff --git a/physics_inspired_models/models/autoregressive.py b/physics_inspired_models/models/autoregressive.py
index 0336b86..f34d3e8 100644
--- a/physics_inspired_models/models/autoregressive.py
+++ b/physics_inspired_models/models/autoregressive.py
@@ -220,7 +220,7 @@ class TeacherForcingAutoregressiveModel(base.SequenceModel):
                               is_training=is_training)
     return p_x, z0, decoder_z
 
-  def training_objectives(
+  def training_objectives(  # pytype: disable=signature-mismatch  # jax-ndarray
       self,
       params: hk.Params,
       state: hk.State,
@@ -300,7 +300,7 @@ class TeacherForcingAutoregressiveModel(base.SequenceModel):
         include_z0=False,
     )[0]
 
-  def gt_state_and_latents(
+  def gt_state_and_latents(  # pytype: disable=signature-mismatch  # jax-ndarray
       self,
       params: hk.Params,
       rng: jnp.ndarray,
@@ -336,7 +336,7 @@ class TeacherForcingAutoregressiveModel(base.SequenceModel):
   ) -> Tuple[Dict[str, jnp.ndarray], Dict[str, jnp.ndarray]]:
     return dict(), dict()
 
-  def _init_latent_system(
+  def _init_latent_system(  # pytype: disable=signature-mismatch  # jax-ndarray
       self,
       rng: jnp.ndarray,
       z: jnp.ndarray,
diff --git a/physics_inspired_models/models/base.py b/physics_inspired_models/models/base.py
index f987052..31b0a0f 100644
--- a/physics_inspired_models/models/base.py
+++ b/physics_inspired_models/models/base.py
@@ -341,7 +341,7 @@ class SequenceModel(abc.ABC, Generic[T]):
     params = hk.data_structures.to_immutable_dict(params)
     state = hk.data_structures.to_immutable_dict(state)
 
-    return params, state
+    return params, state  # pytype: disable=bad-return-type  # jax-ndarray
 
   def init(
       self,
diff --git a/physics_inspired_models/models/deterministic_vae.py b/physics_inspired_models/models/deterministic_vae.py
index a477b97..0aaecea 100644
--- a/physics_inspired_models/models/deterministic_vae.py
+++ b/physics_inspired_models/models/deterministic_vae.py
@@ -155,8 +155,8 @@ class DeterministicLatentsGenerativeModel(base.SequenceModel[_ArrayOrPhase]):
 
   def process_latents_for_decoder(self, z: _ArrayOrPhase) -> jnp.ndarray:
     if self.latent_dynamics_type == "Physics":
-      return z.q if self.render_from_q_only else z.single_state
-    return z
+      return z.q if self.render_from_q_only else z.single_state  # pytype: disable=attribute-error  # jax-ndarray
+    return z  # pytype: disable=bad-return-type  # jax-ndarray
 
   @property
   def inferred_index(self) -> int:
@@ -327,7 +327,7 @@ class DeterministicLatentsGenerativeModel(base.SequenceModel[_ArrayOrPhase]):
     if num_steps_backward > 0 and not self.can_run_backwards:
       raise ValueError("This model can not be unrolled backward in time.")
 
-  def unroll_latent_dynamics(
+  def unroll_latent_dynamics(  # pytype: disable=signature-mismatch  # jax-ndarray
       self,
       z: phase_space.PhaseSpace,
       params: hk.Params,
@@ -393,7 +393,7 @@ class DeterministicLatentsGenerativeModel(base.SequenceModel[_ArrayOrPhase]):
     z = z.single_state if isinstance(z, phase_space.PhaseSpace) else z
     return p_x, q_z, self.prior(), z0, z, dyn_stats
 
-  def training_objectives(
+  def training_objectives(  # pytype: disable=signature-mismatch  # jax-ndarray
       self,
       params: utils.Params,
       state: hk.State,
@@ -532,7 +532,7 @@ class DeterministicLatentsGenerativeModel(base.SequenceModel[_ArrayOrPhase]):
         include_z0=True,
     )[0]
 
-  def gt_state_and_latents(
+  def gt_state_and_latents(  # pytype: disable=signature-mismatch  # jax-ndarray
       self,
       params: hk.Params,
       rng: jnp.ndarray,
diff --git a/physics_inspired_models/models/dynamics.py b/physics_inspired_models/models/dynamics.py
index d5e08ca..d4f6b51 100644
--- a/physics_inspired_models/models/dynamics.py
+++ b/physics_inspired_models/models/dynamics.py
@@ -610,12 +610,12 @@ class PhysicsSimulationNetwork(hk.Module):
           y.q, y.p, **nets_kwargs)
       # Special Haiku magic to avoid tracer issues
       if hk.running_init():
-        return self.lagrangian(y0, **nets_kwargs)
+        return self.lagrangian(y0, **nets_kwargs)  # pytype: disable=bad-return-type  # jax-ndarray
     else:
       hamiltonian = lambda t_, y: self.hamiltonian(y, **nets_kwargs)
       dy_dt = phase_space.poisson_bracket_with_q_and_p(hamiltonian)
       if hk.running_init():
-        return self.hamiltonian(y0, **nets_kwargs)
+        return self.hamiltonian(y0, **nets_kwargs)  # pytype: disable=bad-return-type  # jax-ndarray
 
     # Optionally switch coordinate frame
     if self.input_space == "velocity" and self.simulation_space == "momentum":