[CI] Append test variant name to markdown report header in nightly test (#16166)

test/manual/nightly/test_deepseek_v31_perf.py

@@ -72,7 +72,7 @@ class TestNightlyDeepseekV31Performance(unittest.TestCase):
                     if not success:
                         failed_variants.append(variant_config["name"])
 
-                    self.runner.add_report(results)
+                    self.runner.add_report(results, variant=variant_config["name"])
         finally:
             self.runner.write_final_report()
 

test/manual/nightly/test_deepseek_v32_perf.py

@@ -113,7 +113,7 @@ class TestNightlyDeepseekV32Performance(unittest.TestCase):
                     if not success:
                         failed_variants.append(variant_config["name"])
 
-                    self.runner.add_report(results)
+                    self.runner.add_report(results, variant=variant_config["name"])
         finally:
             self.runner.write_final_report()
 

test/registered/8-gpu-models/test_deepseek_v31.py

@@ -48,12 +48,14 @@ class TestDeepseekV31Unified(unittest.TestCase):
                 DEEPSEEK_V31_MODEL_PATH,
                 tp_size=8,
                 extra_args=base_args,
+                variant="TP8",
             ),
             # Variant: "mtp" - TP=8 + EAGLE speculative decoding
             ModelLaunchSettings(
                 DEEPSEEK_V31_MODEL_PATH,
                 tp_size=8,
                 extra_args=base_args + mtp_args,
+                variant="TP8+MTP",
             ),
         ]
 

test/registered/8-gpu-models/test_deepseek_v32.py

@@ -55,24 +55,28 @@ class TestDeepseekV32Unified(unittest.TestCase):
                 DEEPSEEK_V32_MODEL_PATH,
                 tp_size=8,
                 extra_args=BASE_ARGS + DP_ARGS,
+                variant="DP8",
             ),
             # Variant: "dp+mtp" - DP + EAGLE speculative decoding
             ModelLaunchSettings(
                 DEEPSEEK_V32_MODEL_PATH,
                 tp_size=8,
                 extra_args=BASE_ARGS + DP_ARGS + MTP_ARGS,
+                variant="DP8+MTP",
             ),
             # Variant: "tp" - Pure TP=8 only
             ModelLaunchSettings(
                 DEEPSEEK_V32_MODEL_PATH,
                 tp_size=8,
                 extra_args=BASE_ARGS + TP_ARGS,
+                variant="TP8",
             ),
             # Variant: "tp+mtp" - Pure TP=8 + EAGLE speculative decoding
             ModelLaunchSettings(
                 DEEPSEEK_V32_MODEL_PATH,
                 tp_size=8,
                 extra_args=BASE_ARGS + TP_ARGS + MTP_ARGS,
+                variant="TP8+MTP",
             ),
         ]
 

test/registered/8-gpu-models/test_glm_46.py

@@ -34,6 +34,7 @@ class TestGLM46Unified(unittest.TestCase):
                 GLM_4_6_MODEL_PATH,
                 tp_size=8,
                 extra_args=base_args,
+                variant="TP8",
             ),
         ]
 

test/registered/8-gpu-models/test_glm_46_fp8.py

@@ -39,11 +39,13 @@ class TestGLM46FP8Unified(unittest.TestCase):
                 GLM_4_6_FP8_MODEL_PATH,
                 tp_size=8,
                 extra_args=base_args,
+                variant="TP8",
             ),
             ModelLaunchSettings(
                 GLM_4_6_FP8_MODEL_PATH,
                 tp_size=8,
                 extra_args=base_args + mtp_args,
+                variant="TP8+MTP",
             ),
         ]
 

test/registered/8-gpu-models/test_kimi_k2.py

@@ -35,6 +35,7 @@ class TestKimiK2Unified(unittest.TestCase):
                 KIMI_K2_THINKING_MODEL_PATH,
                 tp_size=8,
                 extra_args=base_args,
+                variant="TP8",
             ),
         ]
 

test/registered/8-gpu-models/test_minimax_m2.py

@@ -37,6 +37,7 @@ class TestMiniMaxM2Unified(unittest.TestCase):
                 MINIMAX_M2_MODEL_PATH,
                 tp_size=8,
                 extra_args=base_args,
+                variant="TP8+EP8",
             ),
         ]
 

test/registered/8-gpu-models/test_mistral_large3.py

@@ -63,12 +63,14 @@ class TestMistralLarge3Unified(unittest.TestCase):
                 MISTRAL_LARGE3_MODEL_PATH,
                 tp_size=8,
                 extra_args=base_args,
+                variant="TP8",
             ),
             # Variant: "eagle" - TP=8 + trtllm_mla + EAGLE with draft model
             ModelLaunchSettings(
                 MISTRAL_LARGE3_MODEL_PATH,
                 tp_size=8,
                 extra_args=base_args + eagle_args,
+                variant="TP8+MTP",
             ),
         ]
 

test/registered/8-gpu-models/test_qwen3_235b.py

@@ -34,6 +34,7 @@ class TestQwen3235BUnified(unittest.TestCase):
                 QWEN3_235B_MODEL_PATH,
                 tp_size=8,
                 extra_args=base_args,
+                variant="TP8",
             ),
         ]