Files
llm_speed_test_app/results/run_20260826-004328.json

268 lines
7.5 KiB
JSON

{
"version": "1.0",
"timestamp": "2026-08-26T00:43:28.377694",
"run_id": "20260826-004328",
"environment": {
"python_version": "3.12.0",
"platform": "win32",
"model": "Q3.6-35B-A3B-Orig-Thi",
"backend": "openai"
},
"config": {
"cases_requested": [
"case_01_generation",
"case_08_concurrency"
],
"repeats": 1,
"skip_heavy": false,
"total_cases": 2
},
"summary": {
"total_cases": 2,
"passed": 2,
"failed": 0,
"stopped": false,
"total_elapsed_seconds": 16.39
},
"results": [
{
"case_id": "case_01_generation",
"case_name": "纯文本生成基准",
"difficulty": "简单",
"status": "passed",
"raw_data": {
"case_id": "case_01_generation",
"name": "纯文本生成基准",
"description": "基础生成速度基准测试",
"difficulty": "简单",
"estimated_seconds": 5,
"results": [
{
"iteration": 1,
"ttft_ms": 1469.4,
"prefill_ms": 176.92,
"decode_speed_tok_s": 56.05,
"total_tokens": 331,
"prompt_tokens": 31,
"completion_tokens": 300,
"e2e_ms": 5781.87,
"elapsed_ms": 5781.87,
"response_length": 931
}
],
"prompt_used": "请用200字左右概述人工智能从诞生至今的关键发展阶段,每段不超过50字。"
},
"statistics": {
"count": 1,
"mean_ms": 5781.87,
"std_ms": 0.0,
"min_ms": 5781.87,
"max_ms": 5781.87,
"p50_ms": 5781.87,
"p95_ms": 5781.87,
"p99_ms": 5781.87
},
"inference_metrics": {
"ttft_ms_mean": 1469.4,
"ttft_ms_p95": 1469.4,
"prefill_ms_mean": 176.92,
"prefill_ms_p95": 176.92,
"e2e_ms_mean": 5781.87,
"e2e_ms_p95": 5781.87,
"decode_speed_tok_s_mean": 56.05,
"decode_speed_tok_s_p95": 56.05,
"total_tokens_mean": 331.0,
"total_tokens_p95": 331
},
"elapsed_seconds": 5.78
},
{
"case_id": "case_08_concurrency",
"case_name": "并发压力测试",
"difficulty": "压力",
"status": "passed",
"raw_data": {
"case_id": "case_08_concurrency",
"name": "并发压力测试",
"description": "高并发下的 TPS 保持率、P95/P99 延迟、错误率",
"difficulty": "压力",
"estimated_seconds": 30,
"results": [
{
"iteration": 1,
"ttft_ms": 47.68,
"prefill_ms": 47.68,
"decode_speed_tok_s": 39.99,
"total_tokens": 36,
"prompt_tokens": 4,
"completion_tokens": 32,
"e2e_ms": 1737.06,
"elapsed_ms": 1737.06,
"response_length": 91,
"batch": "concurrent",
"error": ""
},
{
"iteration": 2,
"ttft_ms": 68.83,
"prefill_ms": 68.83,
"decode_speed_tok_s": 36.31,
"total_tokens": 36,
"prompt_tokens": 4,
"completion_tokens": 32,
"e2e_ms": 1984.56,
"elapsed_ms": 1984.56,
"response_length": 91,
"batch": "concurrent",
"error": ""
},
{
"iteration": 3,
"ttft_ms": 66.03,
"prefill_ms": 66.03,
"decode_speed_tok_s": 39.33,
"total_tokens": 36,
"prompt_tokens": 4,
"completion_tokens": 32,
"e2e_ms": 2687.98,
"elapsed_ms": 2687.98,
"response_length": 91,
"batch": "concurrent",
"error": ""
},
{
"iteration": 4,
"ttft_ms": 67.81,
"prefill_ms": 67.81,
"decode_speed_tok_s": 39.4,
"total_tokens": 36,
"prompt_tokens": 4,
"completion_tokens": 32,
"e2e_ms": 2873.81,
"elapsed_ms": 2873.81,
"response_length": 91,
"batch": "concurrent",
"error": ""
},
{
"iteration": 5,
"ttft_ms": 66.02,
"prefill_ms": 66.02,
"decode_speed_tok_s": 39.32,
"total_tokens": 36,
"prompt_tokens": 4,
"completion_tokens": 32,
"e2e_ms": 3567.66,
"elapsed_ms": 3567.66,
"response_length": 91,
"batch": "concurrent",
"error": ""
},
{
"iteration": 6,
"ttft_ms": 67.78,
"prefill_ms": 67.78,
"decode_speed_tok_s": 39.37,
"total_tokens": 36,
"prompt_tokens": 4,
"completion_tokens": 32,
"e2e_ms": 3751.12,
"elapsed_ms": 3751.12,
"response_length": 91,
"batch": "concurrent",
"error": ""
},
{
"iteration": 7,
"ttft_ms": 66.06,
"prefill_ms": 66.06,
"decode_speed_tok_s": 39.32,
"total_tokens": 36,
"prompt_tokens": 4,
"completion_tokens": 32,
"e2e_ms": 4435.97,
"elapsed_ms": 4435.97,
"response_length": 91,
"batch": "concurrent",
"error": ""
},
{
"iteration": 8,
"ttft_ms": 67.75,
"prefill_ms": 67.75,
"decode_speed_tok_s": 39.41,
"total_tokens": 36,
"prompt_tokens": 4,
"completion_tokens": 32,
"e2e_ms": 4628.97,
"elapsed_ms": 4628.97,
"response_length": 91,
"batch": "concurrent",
"error": ""
},
{
"iteration": 9,
"ttft_ms": 65.95,
"prefill_ms": 65.95,
"decode_speed_tok_s": 39.36,
"total_tokens": 36,
"prompt_tokens": 4,
"completion_tokens": 32,
"e2e_ms": 5330.84,
"elapsed_ms": 5330.84,
"response_length": 91,
"batch": "concurrent",
"error": ""
},
{
"iteration": 10,
"ttft_ms": 67.72,
"prefill_ms": 67.72,
"decode_speed_tok_s": 43.93,
"total_tokens": 36,
"prompt_tokens": 4,
"completion_tokens": 32,
"e2e_ms": 5430.27,
"elapsed_ms": 5430.27,
"response_length": 91,
"batch": "concurrent",
"error": ""
}
],
"concurrency": 10,
"baseline_tps": 57.41,
"concurrent_tps": 39.57,
"tps_keep_rate_pct": 68.93,
"e2e_p50_ms": 3659.39,
"e2e_p95_ms": 5385.53,
"e2e_p99_ms": 5421.32,
"success_count": 10,
"failed_count": 0,
"error_rate_pct": 0.0
},
"statistics": {
"count": 10,
"mean_ms": 3642.82,
"std_ms": 1314.37,
"min_ms": 1737.06,
"max_ms": 5430.27,
"p50_ms": 3659.39,
"p95_ms": 5385.53,
"p99_ms": 5421.32
},
"inference_metrics": {
"ttft_ms_mean": 65.16,
"ttft_ms_p95": 68.37,
"prefill_ms_mean": 65.16,
"prefill_ms_p95": 68.37,
"e2e_ms_mean": 3642.82,
"e2e_ms_p95": 5385.53,
"decode_speed_tok_s_mean": 39.57,
"decode_speed_tok_s_p95": 42.16,
"total_tokens_mean": 36.0,
"total_tokens_p95": 36.0
},
"elapsed_seconds": 10.61
}
]
}