Enable CUDA tests for GQA attention tests

Co-authored-by: titaiwangms <18010845+titaiwangms@users.noreply.github.com>

Author: Copilot

onnxruntime/test/providers/cpu/llm/attention_op_test.cc

@@ -891,7 +891,7 @@ TEST(AttentionTest, Attention3DGqaAttn) {
             q, k, v, std::vector<float>(), std::initializer_list<bool>(), std::vector<float>(), std::vector<float>(),
             -1, -1, std::numeric_limits<float>::quiet_NaN(), std::numeric_limits<float>::quiet_NaN(), -1, TensorType::kFloat,  // is_causal, qk_matmul_output_mode, scale, softcap, softmax_precision, tensor_type
             y, std::vector<float>(), std::vector<float>(), std::vector<float>(),
-            false, true, true  // disable_cpu, disable_cuda, disable_dml
+            false, false, true  // disable_cpu, disable_cuda, disable_dml
   );
 }
 
@@ -926,7 +926,7 @@ TEST(AttentionTest, Attention4DGqaAttnMask) {
             q, k, v, m, std::initializer_list<bool>(), std::vector<float>(), std::vector<float>(),
             -1, -1, std::numeric_limits<float>::quiet_NaN(), std::numeric_limits<float>::quiet_NaN(), -1, TensorType::kFloat,  // is_causal, qk_matmul_output_mode, scale, softcap, softmax_precision, tensor_type
             y, std::vector<float>(), std::vector<float>(), std::vector<float>(),
-            false, true, true  // disable_cpu, disable_cuda, disable_dml
+            false, false, true  // disable_cpu, disable_cuda, disable_dml
   );
 }
 
@@ -973,7 +973,7 @@ TEST(AttentionTest, Attention4DGqaWithPastAndPresent) {
             q, k, v, m, std::initializer_list<bool>(), past_key, past_value,
             -1, -1, std::numeric_limits<float>::quiet_NaN(), std::numeric_limits<float>::quiet_NaN(), -1, TensorType::kFloat,  // is_causal, qk_matmul_output_mode, scale, softcap, softmax_precision, tensor_type
             y, present_key, present_value, std::vector<float>(),
-            false, true, true  // disable_cpu, disable_cuda, disable_dml
+            false, false, true  // disable_cpu, disable_cuda, disable_dml
   );
 }