Add cost

cap_metrics_gsm8k.json

@@ -11,7 +11,7 @@
     "exact_match": 0.8112206216830933,
     "correct": 1070,
     "total": 1319,
-    "cost": 15342.20,
+
     "model_name": "Qwen/Qwen3-30B-A3B",
     "method": "sglang",
     "precision": "bfloat16",

sglang/Qwen/Qwen1.5-MoE-A2.7B-Chat/cap_metrics_gsm8k_20251120_204353.json

@@ -1,23 +0,0 @@
-{
-    "prefill_smbu": 0.6161871789135226,
-    "prefill_smfu": 0.04090522068016735,
-    "decoding_smbu": 0.671695399609032,
-    "decoding_smfu": 0.027067809730019755,
-    "kv_size": 967.951456653987,
-    "decoding_throughput": 599.3539979814984,
-    "prefill_tp": 2176.3331070143954,
-    "ttft": 0.05797824891629358,
-    "tpot": 0.029830356329599007,
-    "exact_match": 0.45716451857467777,
-    "correct": 603,
-    "total": 1319,
-    "cost": 7158.92,
-    "model_name": "Qwen/Qwen1.5-MoE-A2.7B-Chat",
-    "method": "sglang",
-    "precision": "bfloat16",
-    "e2e_s": 295.91,
-    "batch_size": null,
-    "gpu_type": "1xRTX A6000",
-    "dataset": "gsm8k",
-    "model_type": "instruct"
-}

sglang/Qwen/Qwen3-235B-A22B-Thinking-2507-FP8/cap_metrics_gsm8k_20251121_012619.json

@@ -1,23 +0,0 @@
-{
-    "prefill_smbu": 0.14092614495475234,
-    "prefill_smfu": 0.0014371285385894436,
-    "decoding_smbu": 0.3203664805789544,
-    "decoding_smfu": 0.005892042790709394,
-    "kv_size": 4033.749704013076,
-    "decoding_throughput": 1136.774328309585,
-    "prefill_tp": 1102.0549270552817,
-    "ttft": 0.09287328138655972,
-    "tpot": 0.02449785142839567,
-    "exact_match": 0.6884003032600455,
-    "correct": 908,
-    "total": 1319,
-    "model_name": "Qwen/Qwen3-235B-A22B-Thinking-2507-FP8",
-    "method": "sglang",
-    "precision": "fp8",
-    "e2e_s": 728.81,
-    "batch_size": null,
-    "cost": 104052.07,
-    "gpu_type": "2xH200",
-    "dataset": "gsm8k",
-    "model_type": "thinking"
-}

sglang/Qwen/Qwen3-235B-A22B-Thinking-2507/cap_metrics_gsm8k_20251121_003244.json

@@ -1,23 +0,0 @@
-{
-    "prefill_smbu": 0.17849474014488356,
-    "prefill_smfu": 0.001752245254156507,
-    "decoding_smbu": 0.3381575818423562,
-    "decoding_smfu": 0.006024281466006331,
-    "kv_size": 3627.284702586172,
-    "decoding_throughput": 1206.3211462911283,
-    "prefill_tp": 1344.0795393815022,
-    "ttft": 0.08147775903823071,
-    "tpot": 0.022421781783965894,
-    "exact_match": 0.7028051554207733,
-    "correct": 927,
-    "total": 1319,
-    "model_name": "Qwen/Qwen3-235B-A22B-Thinking-2507",
-    "method": "sglang",
-    "precision": "bfloat16",
-    "e2e_s": 654.28,
-    "batch_size": null,
-    "cost": 195252.11,
-    "gpu_type": "4xH200",
-    "dataset": "gsm8k",
-    "model_type": "thinking"
-}

sglang/Qwen/Qwen3-235B-A22B/cap_metrics_gsm8k_20251122_004149.json

@@ -1,23 +0,0 @@
-{
-    "prefill_smbu": 0.0500890302554711,
-    "prefill_smfu": 0.0003749803951029884,
-    "decoding_smbu": 0.3837587550954451,
-    "decoding_smfu": 0.006175928749469473,
-    "kv_size": 9510.041884336408,
-    "decoding_throughput": 1694.3033755455053,
-    "prefill_tp": 575.3677685986171,
-    "ttft": 0.21382533143595106,
-    "tpot": 0.026343578621429595,
-    "exact_match": 0.7119029567854435,
-    "correct": 939,
-    "total": 1319,
-    "cost": 344657.14,
-    "model_name": "Qwen/Qwen3-235B-A22B",
-    "method": "sglang",
-    "precision": "bfloat16",
-    "e2e_s": 1099.04,
-    "batch_size": null,
-    "gpu_type": "8xH100",
-    "dataset": "gsm8k",
-    "model_type": "thinking"
-}

sglang/Qwen/Qwen3-30B-A3B-Instruct-2507/cap_metrics_gsm8k_20251120_180756.json

@@ -11,7 +11,6 @@
     "exact_match": 0.5329795299469295,
     "correct": 703,
     "total": 1319,
-    "cost": 21600.27,
     "model_name": "Qwen/Qwen3-30B-A3B-Instruct-2507",
     "method": "sglang",
     "precision": "bfloat16",

sglang/Qwen/Qwen3-30B-A3B-Thinking-2507/cap_metrics_gsm8k.json

@@ -11,7 +11,6 @@
     "exact_match": 0.6929492039423806,
     "correct": 914,
     "total": 1319,
-    "cost": 21600.54,
     "model_name": "Qwen/Qwen3-30B-A3B-Thinking-2507",
     "method": "sglang",
     "precision": "bfloat16",

sglang/Qwen/Qwen3-30B-A3B/cap_metrics_gsm8k.json

@@ -11,7 +11,6 @@
     "exact_match": 0.8066717210007581,
     "correct": 1064,
     "total": 1319,
-    "cost": 21600.69,
     "model_name": "Qwen/Qwen3-30B-A3B",
     "method": "sglang",
     "precision": "bfloat16",

sglang/Qwen/Qwen3-30B-A3B/cap_metrics_gsm8k_20251121_224947.json

@@ -1,23 +0,0 @@
-{
-    "prefill_smbu": 0.06985210551743316,
-    "prefill_smfu": 0.0011168447824740323,
-    "decoding_smbu": 0.31406183034677493,
-    "decoding_smfu": 0.024780736967378484,
-    "kv_size": 1794.9723102102703,
-    "decoding_throughput": 1806.0915757549812,
-    "prefill_tp": 800.7711146275104,
-    "ttft": 0.1253664656356315,
-    "tpot": 0.01482144513337436,
-    "exact_match": 0.8097043214556482,
-    "cost": 54380.91,
-    "correct": 1068,
-    "total": 1319,
-    "model_name": "Qwen/Qwen3-30B-A3B",
-    "method": "sglang",
-    "precision": "bfloat16",
-    "e2e_s": 575.59,
-    "batch_size": null,
-    "gpu_type": "2xA100",
-    "dataset": "gsm8k",
-    "model_type": "thinking"
-}