fix(tests): skip CPU-replay DQN tests on CUDA builds

7 tests call train_step() via CPU experience replay, but with the cuda
feature GPU PER is mandatory (no CPU path). Mark them
#[cfg_attr(feature = "cuda", ignore)] — the GPU training pipeline
integration tests cover this path properly on H100.

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

crates/ml-dqn/src/dqn.rs

@@ -4748,6 +4748,7 @@ mod tests {
     }
 
     #[test]
+    #[cfg_attr(feature = "cuda", ignore)] // GPU PER mandatory on CUDA — CPU replay path unavailable
     fn test_training_step_with_data() -> anyhow::Result<()> {
         let mut config = DQNConfig::emergency_safe_defaults();
         config.min_replay_size = 4;
@@ -4808,6 +4809,7 @@ mod tests {
     }
 
     #[test]
+    #[cfg_attr(feature = "cuda", ignore)] // GPU PER mandatory on CUDA — CPU replay path unavailable
     fn test_cql_regularization() -> anyhow::Result<()> {
         let mut config = DQNConfig::emergency_safe_defaults();
         config.state_dim = 8;
@@ -5041,6 +5043,7 @@ mod tests {
     /// With weight_decay > 0, training a few steps should produce different final
     /// weights than weight_decay == 0, proving L2 regularization is active.
     #[test]
+    #[cfg_attr(feature = "cuda", ignore)] // GPU PER mandatory on CUDA — CPU replay path unavailable
     fn test_weight_decay_wired_to_optimizer() -> anyhow::Result<()> {
         // Helper: run a few training steps and return the first Q-network parameter tensor
         fn train_and_get_params(weight_decay: f64) -> anyhow::Result<Vec<f32>> {
@@ -5212,6 +5215,7 @@ mod tests {
     }
 
     #[test]
+    #[cfg_attr(feature = "cuda", ignore)] // GPU PER mandatory on CUDA — CPU replay path unavailable
     fn test_branching_dqn_end_to_end() -> anyhow::Result<()> {
         // Phase C+: Verify branching DQN creates, selects actions, and trains
         let mut config = DQNConfig::emergency_safe_defaults();
@@ -5261,6 +5265,7 @@ mod tests {
     }
 
     #[test]
+    #[cfg_attr(feature = "cuda", ignore)] // GPU PER mandatory on CUDA — CPU replay path unavailable
     fn test_branching_regime_conditioning_affects_loss() -> anyhow::Result<()> {
         // Verify that regime-conditional IS weighting produces different losses
         // than the no-regime path when states contain regime features.

crates/ml/src/validation/adapters.rs

@@ -222,6 +222,7 @@ mod tests {
     }
 
     #[test]
+    #[cfg_attr(feature = "cuda", ignore)] // GPU PER mandatory on CUDA — CPU replay path unavailable
     fn test_dqn_strategy_train_and_evaluate() {
         let config = make_test_config();
         let mut strategy = DqnStrategy::new(config)
@@ -242,6 +243,7 @@ mod tests {
     }
 
     #[test]
+    #[cfg_attr(feature = "cuda", ignore)] // GPU PER mandatory on CUDA — CPU replay path unavailable
     fn test_dqn_strategy_reset() {
         let config = make_test_config();
         let mut strategy = DqnStrategy::new(config)