added ROCm/Triton attention comparison

benchmarks/benchmark_results_rocm_attn/benchmark_results/Qwen_Qwen3-14B-AWQ_tp1_throughput.json

@@ -0,0 +1,7 @@
+{
+    "elapsed_time": 1237.550695703001,
+    "num_requests": 200,
+    "total_num_tokens": 146805,
+    "requests_per_second": 0.16160954108339642,
+    "tokens_per_second": 118.62544339374007
+}

benchmarks/benchmark_results_rocm_attn/benchmark_results/google_gemma-3-12b-it_tp1_throughput.json

@@ -0,0 +1,7 @@
+{
+    "elapsed_time": 540.6128817510034,
+    "num_requests": 200,
+    "total_num_tokens": 148857,
+    "requests_per_second": 0.36995048906754757,
+    "tokens_per_second": 275.34859975563967
+}

benchmarks/benchmark_results_rocm_attn/benchmark_results/meta-llama_Meta-Llama-3.1-8B-Instruct_tp1_throughput.json

@@ -0,0 +1,7 @@
+{
+    "elapsed_time": 455.23138687500614,
+    "num_requests": 200,
+    "total_num_tokens": 145877,
+    "requests_per_second": 0.43933701797875907,
+    "tokens_per_second": 320.4458308584372
+}

benchmarks/benchmark_results_rocm_attn/benchmark_results/openai_gpt-oss-120b_tp1_throughput.json

@@ -0,0 +1,7 @@
+{
+    "elapsed_time": 1279.5375675789983,
+    "num_requests": 200,
+    "total_num_tokens": 147036,
+    "requests_per_second": 0.15630646967124087,
+    "tokens_per_second": 114.91339037290285
+}

benchmarks/benchmark_results_rocm_attn/benchmark_results/openai_gpt-oss-20b_tp1_throughput.json

@@ -0,0 +1,7 @@
+{
+    "elapsed_time": 460.97370730798866,
+    "num_requests": 200,
+    "total_num_tokens": 147036,
+    "requests_per_second": 0.43386422442175154,
+    "tokens_per_second": 318.9683005103833
+}