diff --git a/submissions/self_tuning/muon_torch/submission.py b/submissions/self_tuning/muon_torch/submission.py
index 47d262da..9919b1cd 100644
--- a/submissions/self_tuning/muon_torch/submission.py
+++ b/submissions/self_tuning/muon_torch/submission.py
@@ -38,7 +38,7 @@
   "warmup_factor": 0.05,
   "step_reduce": 1.0
 }
-hyperparameters = SimpleNamespace(**HPARAMS)
+HPARAMS = SimpleNamespace(**HPARAMS)
 
 
 def _pytorch_cosine_warmup(step_hint: int, hyperparameters, optimizer):
@@ -65,6 +65,8 @@ def init_optimizer_state(
   del model_state
   del rng
 
+  hyperparameters = HPARAMS
+
   muon_params, adam_params = split_params_muon_adam(model_params)
 
   optimizer_state = {
@@ -119,6 +121,8 @@ def update_params(
   del train_state
   del eval_results
 
+  hyperparameters = HPARAMS
+
   reduced_steps = int(workload.step_hint * getattr(hyperparameters, "step_reduce", 1.0))
   if global_step >= reduced_steps:
     raise spec.TrainingCompleteError(