242 lines
8.9 KiB
JSON
242 lines
8.9 KiB
JSON
[
|
|
{
|
|
"test_name": "test_sglang_qwen_image_single_device",
|
|
"description": "sglang: single-device baseline (no parallelism)",
|
|
"server_type": "sglang",
|
|
"server_params": {
|
|
"model": "Qwen/Qwen-Image",
|
|
"serve_args": {
|
|
"num-gpus": 1,
|
|
"dit-cpu-offload": false,
|
|
"text-encoder-cpu-offload": false,
|
|
"vae-cpu-offload": false
|
|
}
|
|
},
|
|
"benchmark_params": [
|
|
{
|
|
"name": "512x512_steps20",
|
|
"dataset": "random",
|
|
"task": "t2i",
|
|
"width": 512,
|
|
"height": 512,
|
|
"num-inference-steps": 20,
|
|
"num-prompts": 10,
|
|
"max-concurrency": 1,
|
|
"enable-negative-prompt": true,
|
|
"baseline": {
|
|
"throughput_qps": 0.001,
|
|
"latency_p99": 1000.0,
|
|
"peak_memory_mb_max": 400000,
|
|
"peak_memory_mb_mean": 400000
|
|
}
|
|
},
|
|
{
|
|
"name": "1536x1536_steps35",
|
|
"dataset": "random",
|
|
"task": "t2i",
|
|
"width": 1536,
|
|
"height": 1536,
|
|
"num-inference-steps": 35,
|
|
"num-prompts": 10,
|
|
"max-concurrency": 1,
|
|
"enable-negative-prompt": true,
|
|
"baseline": {
|
|
"throughput_qps": 0.001,
|
|
"latency_p99": 1000.0,
|
|
"peak_memory_mb_max": 400000,
|
|
"peak_memory_mb_mean": 400000
|
|
}
|
|
},
|
|
{
|
|
"name": "mixed_resolution",
|
|
"dataset": "random",
|
|
"task": "t2i",
|
|
"num-inference-steps": 20,
|
|
"num-prompts": 10,
|
|
"max-concurrency": 1,
|
|
"enable-negative-prompt": true,
|
|
"random-request-config": [
|
|
{"width": 512, "height": 512, "num_inference_steps": 20, "weight": 0.15},
|
|
{"width": 768, "height": 768, "num_inference_steps": 20, "weight": 0.25},
|
|
{"width": 1024, "height": 1024, "num_inference_steps": 25, "weight": 0.45},
|
|
{"width": 1536, "height": 1536, "num_inference_steps": 35, "weight": 0.15}
|
|
],
|
|
"baseline": {
|
|
"throughput_qps": 0.001,
|
|
"latency_p99": 1000.0,
|
|
"peak_memory_mb_max": 400000,
|
|
"peak_memory_mb_mean": 400000
|
|
}
|
|
}
|
|
]
|
|
},
|
|
|
|
{
|
|
"test_name": "test_sglang_qwen_image_ulysses2_cfg2_vae_patch4",
|
|
"description": "sglang: USP=2 + CFG-parallel=2 + VAE patch parallel via --use-parallel-tiling",
|
|
"server_type": "sglang",
|
|
"server_params": {
|
|
"model": "Qwen/Qwen-Image",
|
|
"serve_args": {
|
|
"num-gpus": 4,
|
|
"sp-degree": 2,
|
|
"ulysses-degree": 2,
|
|
"enable-cfg-parallel": true,
|
|
"use-parallel-tiling": true,
|
|
"dit-cpu-offload": false,
|
|
"text-encoder-cpu-offload": false,
|
|
"vae-cpu-offload": false
|
|
}
|
|
},
|
|
"benchmark_params": [
|
|
{
|
|
"name": "512x512_steps20",
|
|
"dataset": "random",
|
|
"task": "t2i",
|
|
"width": 512,
|
|
"height": 512,
|
|
"num-inference-steps": 20,
|
|
"num-prompts": 10,
|
|
"max-concurrency": 1,
|
|
"enable-negative-prompt": true,
|
|
"baseline": {
|
|
"throughput_qps": 0.001,
|
|
"latency_p99": 1000.0,
|
|
"peak_memory_mb_max": 400000,
|
|
"peak_memory_mb_mean": 400000
|
|
}
|
|
},
|
|
{
|
|
"name": "1536x1536_steps35",
|
|
"dataset": "random",
|
|
"task": "t2i",
|
|
"width": 1536,
|
|
"height": 1536,
|
|
"num-inference-steps": 35,
|
|
"num-prompts": 10,
|
|
"max-concurrency": 1,
|
|
"enable-negative-prompt": true,
|
|
"baseline": {
|
|
"throughput_qps": 0.001,
|
|
"latency_p99": 1000.0,
|
|
"peak_memory_mb_max": 400000,
|
|
"peak_memory_mb_mean": 400000
|
|
}
|
|
},
|
|
{
|
|
"name": "mixed_resolution",
|
|
"dataset": "random",
|
|
"task": "t2i",
|
|
"num-inference-steps": 20,
|
|
"num-prompts": 10,
|
|
"max-concurrency": 1,
|
|
"enable-negative-prompt": true,
|
|
"random-request-config": [
|
|
{"width": 512, "height": 512, "num_inference_steps": 20, "weight": 0.15},
|
|
{"width": 768, "height": 768, "num_inference_steps": 20, "weight": 0.25},
|
|
{"width": 1024, "height": 1024, "num_inference_steps": 25, "weight": 0.45},
|
|
{"width": 1536, "height": 1536, "num_inference_steps": 35, "weight": 0.15}
|
|
],
|
|
"baseline": {
|
|
"throughput_qps": 0.001,
|
|
"latency_p99": 1000.0,
|
|
"peak_memory_mb_max": 400000,
|
|
"peak_memory_mb_mean": 400000
|
|
}
|
|
}
|
|
]
|
|
},
|
|
|
|
{
|
|
"test_name": "test_sglang_qwen_image_ulysses2_cfg2_cache_dit",
|
|
"description": "sglang: USP=2 + CFG-parallel=2 + Cache-DiT via YAML config (--cache-dit-config)",
|
|
"server_type": "sglang",
|
|
"server_params": {
|
|
"model": "Qwen/Qwen-Image",
|
|
"cache_dit_config": {
|
|
"cache_config": {
|
|
"Fn_compute_blocks": 1,
|
|
"Bn_compute_blocks": 0,
|
|
"max_warmup_steps": 4,
|
|
"residual_diff_threshold": 0.24,
|
|
"max_continuous_cached_steps": 3,
|
|
"enable_taylorseer": false,
|
|
"taylorseer_order": 1,
|
|
"scm_steps_mask_policy": "none",
|
|
"scm_steps_policy": "dynamic"
|
|
},
|
|
"parallelism_config": {
|
|
"ulysses_size": 2,
|
|
"attention_backend": "native"
|
|
}
|
|
},
|
|
"serve_args": {
|
|
"num-gpus": 4,
|
|
"sp-degree": 2,
|
|
"ulysses-degree": 2,
|
|
"enable-cfg-parallel": true,
|
|
"dit-cpu-offload": false,
|
|
"text-encoder-cpu-offload": false,
|
|
"vae-cpu-offload": false
|
|
}
|
|
},
|
|
"benchmark_params": [
|
|
{
|
|
"name": "512x512_steps20",
|
|
"dataset": "random",
|
|
"task": "t2i",
|
|
"width": 512,
|
|
"height": 512,
|
|
"num-inference-steps": 20,
|
|
"num-prompts": 10,
|
|
"max-concurrency": 1,
|
|
"enable-negative-prompt": true,
|
|
"baseline": {
|
|
"throughput_qps": 0.001,
|
|
"latency_p99": 1000.0,
|
|
"peak_memory_mb_max": 400000,
|
|
"peak_memory_mb_mean": 400000
|
|
}
|
|
},
|
|
{
|
|
"name": "1536x1536_steps35",
|
|
"dataset": "random",
|
|
"task": "t2i",
|
|
"width": 1536,
|
|
"height": 1536,
|
|
"num-inference-steps": 35,
|
|
"num-prompts": 10,
|
|
"max-concurrency": 1,
|
|
"enable-negative-prompt": true,
|
|
"baseline": {
|
|
"throughput_qps": 0.001,
|
|
"latency_p99": 1000.0,
|
|
"peak_memory_mb_max": 400000,
|
|
"peak_memory_mb_mean": 400000
|
|
}
|
|
},
|
|
{
|
|
"name": "mixed_resolution",
|
|
"dataset": "random",
|
|
"task": "t2i",
|
|
"num-inference-steps": 20,
|
|
"num-prompts": 10,
|
|
"max-concurrency": 1,
|
|
"enable-negative-prompt": true,
|
|
"random-request-config": [
|
|
{"width": 512, "height": 512, "num_inference_steps": 20, "weight": 0.15},
|
|
{"width": 768, "height": 768, "num_inference_steps": 20, "weight": 0.25},
|
|
{"width": 1024, "height": 1024, "num_inference_steps": 25, "weight": 0.45},
|
|
{"width": 1536, "height": 1536, "num_inference_steps": 35, "weight": 0.15}
|
|
],
|
|
"baseline": {
|
|
"throughput_qps": 0.001,
|
|
"latency_p99": 1000.0,
|
|
"peak_memory_mb_max": 400000,
|
|
"peak_memory_mb_mean": 400000
|
|
}
|
|
}
|
|
]
|
|
}
|
|
]
|