chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 14 Packages Projects Releases Wiki Activity

[diffusion] refactor: reduce redundancy and improve stage api (#19060)

Browse Source
This commit is contained in:
Mick
2026-02-21 16:35:47 +08:00
committed by GitHub
parent 33c33a7de9
commit b89ca65789
26 changed files with 444 additions and 1029 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/multimodal_gen/test/server/ascend/perf_baselines_npu.json
-1
View File
@@ -9,7 +9,6 @@
"stages_ms": {
"InputValidationStage": 0.1,
"TextEncodingStage": 1609.27,
"ConditioningStage": 0.02,
"TimestepPreparationStage": 3.46,
"LatentPreparationStage": 0.39,
"DenoisingStage": 26324.0,
python/sglang/multimodal_gen/test/server/perf_baselines.json
-101
View File
@@ -37,7 +37,6 @@
"qwen_image_t2i": {
"notes": "Single-image generation using the default prompt",
"stages_ms": {
"ConditioningStage": 0.02,
"DecodingStage": 51.86,
"TextEncodingStage": 611.83,
"InputValidationStage": 0.05,
@@ -105,7 +104,6 @@
"stages_ms": {
"InputValidationStage": 0.04,
"TextEncodingStage": 693.2,
"ConditioningStage": 0.02,
"TimestepPreparationStage": 2.84,
"LatentPreparationStage": 9.13,
"DenoisingStage": 24529.77,
@@ -169,7 +167,6 @@
},
"flux_image_t2i": {
"stages_ms": {
"ConditioningStage": 0.01,
"DecodingStage": 32.72,
"TextEncodingStage": 51.96,
"InputValidationStage": 0.03,
@@ -239,7 +236,6 @@
"TimestepPreparationStage": 2.91,
"TextEncodingStage": 518.54,
"ImageVAEEncodingStage": 0.0,
"ConditioningStage": 0.02,
"InputValidationStage": 0.05,
"DenoisingStage": 24901.97,
"DecodingStage": 8.98
@@ -302,7 +298,6 @@
},
"flux_2_klein_image_t2i": {
"stages_ms": {
"ConditioningStage": 0.01,
"DecodingStage": 9.27,
"TextEncodingStage": 92.17,
"InputValidationStage": 0.05,
@@ -326,7 +321,6 @@
"InputValidationStage": 0.06,
"TextEncodingStage": 513.58,
"ImageVAEEncodingStage": 0.0,
"ConditioningStage": 0.03,
"LatentPreparationStage": 0.46,
"TimestepPreparationStage": 2.38,
"DenoisingStage": 52187.62,
@@ -393,7 +387,6 @@
"InputValidationStage": 99.82,
"TextEncodingStage": 519.88,
"ImageVAEEncodingStage": 254.56,
"ConditioningStage": 0.01,
"LatentPreparationStage": 12.4,
"TimestepPreparationStage": 2.71,
"DenoisingStage": 54705.41,
@@ -459,7 +452,6 @@
"stages_ms": {
"InputValidationStage": 0.03,
"TextEncodingStage": 74.47,
"ConditioningStage": 0.01,
"TimestepPreparationStage": 2.23,
"LatentPreparationStage": 6.17,
"DenoisingStage": 8400.49,
@@ -525,7 +517,6 @@
"stages_ms": {
"InputValidationStage": 0.03,
"TextEncodingStage": 403.47,
"ConditioningStage": 0.01,
"TimestepPreparationStage": 1.41,
"LatentPreparationStage": 0.11,
"DenoisingStage": 756.21,
@@ -550,7 +541,6 @@
"stages_ms": {
"InputValidationStage": 0.04,
"TextEncodingStage": 413.69,
"ConditioningStage": 0.01,
"TimestepPreparationStage": 1.3,
"LatentPreparationStage": 0.11,
"DenoisingStage": 813.7,
@@ -575,7 +565,6 @@
"stages_ms": {
"InputValidationStage": 0.08,
"TextEncodingStage": 420.74,
"ConditioningStage": 0.01,
"TimestepPreparationStage": 1.5,
"LatentPreparationStage": 0.12,
"DenoisingStage": 1304.07,
@@ -602,7 +591,6 @@
"TimestepPreparationStage": 2.62,
"ImageEncodingStage": 1174.26,
"ImageVAEEncodingStage": 132.67,
"ConditioningStage": 0.01,
"InputValidationStage": 38.1,
"DenoisingStage": 38135.64,
"DecodingStage": 139.72
@@ -667,7 +655,6 @@
"stages_ms": {
"InputValidationStage": 0.05,
"TextEncodingStage": 675.95,
"ConditioningStage": 0.02,
"TimestepPreparationStage": 3.21,
"LatentPreparationStage": 0.2,
"DenoisingStage": 5248.83,
@@ -733,74 +720,6 @@
"stages_ms": {
"InputValidationStage": 0.07,
"TextEncodingStage": 2237.78,
"ConditioningStage": 0.01,
"TimestepPreparationStage": 2.1,
"LatentPreparationStage": 0.84,
"DenoisingStage": 13041.23,
"DecodingStage": 1274.63,
"per_frame_generation": null
},
"denoise_step_ms": {
"0": 240.71,
"1": 248.13,
"2": 246.48,
"3": 247.87,
"4": 249.38,
"5": 246.76,
"6": 250.42,
"7": 250.81,
"8": 250.98,
"9": 249.9,
"10": 246.72,
"11": 249.79,
"12": 250.46,
"13": 249.19,
"14": 247.55,
"15": 250.12,
"16": 247.57,
"17": 247.21,
"18": 247.32,
"19": 247.42,
"20": 248.21,
"21": 247.19,
"22": 247.72,
"23": 247.45,
"24": 247.9,
"25": 247.87,
"26": 247.18,
"27": 247.65,
"28": 246.91,
"29": 248.26,
"30": 247.82,
"31": 247.73,
"32": 247.38,
"33": 247.84,
"34": 247.46,
"35": 247.52,
"36": 247.94,
"37": 248.76,
"38": 248.01,
"39": 247.45,
"40": 247.84,
"41": 248.33,
"42": 247.41,
"43": 248.16,
"44": 248.18,
"45": 248.44,
"46": 248.65,
"47": 247.73,
"48": 247.48,
"49": 247.54
},
"expected_e2e_ms": 18382.19,
"expected_avg_denoise_ms": 260.76,
"expected_median_denoise_ms": 247.84
},
"wan2_1_t2v_1.3b_teacache_enabled": {
"stages_ms": {
"InputValidationStage": 0.07,
"TextEncodingStage": 2237.78,
"ConditioningStage": 0.01,
"TimestepPreparationStage": 2.1,
"LatentPreparationStage": 0.84,
"DenoisingStage": 13041.23,
@@ -867,7 +786,6 @@
"stages_ms": {
"InputValidationStage": 0.07,
"TextEncodingStage": 2237.78,
"ConditioningStage": 0.01,
"TimestepPreparationStage": 2.1,
"LatentPreparationStage": 0.84,
"DenoisingStage": 13041.23,
@@ -934,7 +852,6 @@
"stages_ms": {
"InputValidationStage": 0.09,
"TextEncodingStage": 2480.54,
"ConditioningStage": 0.07,
"TimestepPreparationStage": 3.73,
"LatentPreparationStage": 1.34,
"DenoisingStage": 12514.88,
@@ -1001,7 +918,6 @@
"stages_ms": {
"InputValidationStage": 0.08,
"TextEncodingStage": 2700.44,
"ConditioningStage": 0.02,
"TimestepPreparationStage": 2.82,
"LatentPreparationStage": 2.0,
"DenoisingStage": 11640.75,
@@ -1068,7 +984,6 @@
"stages_ms": {
"InputValidationStage": 0.06,
"TextEncodingStage": 2508.95,
"ConditioningStage": 0.04,
"TimestepPreparationStage": 73.51,
"LatentPreparationStage": 1.34,
"DmdDenoisingStage": 1285.25,
@@ -1089,7 +1004,6 @@
"stages_ms": {
"InputValidationStage": 96.27,
"TextEncodingStage": 2238.81,
"ConditioningStage": 0.02,
"TimestepPreparationStage": 2.39,
"LatentPreparationStage": 27.62,
"DenoisingStage": 134069.79,
@@ -1159,7 +1073,6 @@
"ImageVAEEncodingStage": 304.56,
"TimestepPreparationStage": 2.94,
"LatentPreparationStage": 0.2,
"ConditioningStage": 0.01,
"DenoisingStage": 50724.5,
"DecodingStage": 601.02
},
@@ -1276,7 +1189,6 @@
"stages_ms": {
"InputValidationStage": 300.00,
"TextEncodingStage": 843.86,
"ConditioningStage": 0.01,
"TimestepPreparationStage": 58.66,
"LatentPreparationStage": 28.55,
"DmdDenoisingStage": 499.34,
@@ -1296,7 +1208,6 @@
"stages_ms": {
"InputValidationStage": 0.34,
"TextEncodingStage": 550.63,
"ConditioningStage": 0.02,
"TimestepPreparationStage": 44.28,
"LatentPreparationStage": 0.29,
"DenoisingStage": 9154.39,
@@ -1319,7 +1230,6 @@
"stages_ms": {
"InputValidationStage": 18.45,
"TextEncodingStage": 3337.77,
"ConditioningStage": 0.03,
"TimestepPreparationStage": 2.9,
"LatentPreparationStage": 1.25,
"ImageVAEEncodingStage": 1655.89,
@@ -1377,7 +1287,6 @@
"stages_ms": {
"InputValidationStage": 25.01,
"TextEncodingStage": 5198.6,
"ConditioningStage": 0.04,
"TimestepPreparationStage": 56.26,
"LatentPreparationStage": 1.4,
"ImageVAEEncodingStage": 1001.89,
@@ -1400,7 +1309,6 @@
"InputValidationStage": 38.23,
"TextEncodingStage": 3550.36,
"ImageEncodingStage": 3462.55,
"ConditioningStage": 0.01,
"TimestepPreparationStage": 2.6,
"LatentPreparationStage": 9.73,
"ImageVAEEncodingStage": 2290.98,
@@ -1469,7 +1377,6 @@
"InputValidationStage": 53.67,
"TextEncodingStage": 2838,
"ImageEncodingStage": 3123.99,
"ConditioningStage": 0.01,
"TimestepPreparationStage": 3.39,
"LatentPreparationStage": 8.41,
"ImageVAEEncodingStage": 2261.05,
@@ -1536,7 +1443,6 @@
"stages_ms": {
"InputValidationStage": 0.07,
"TextEncodingStage": 2575.3,
"ConditioningStage": 0.01,
"TimestepPreparationStage": 1.99,
"LatentPreparationStage": 1.26,
"DenoisingStage": 156678.8406,
@@ -1593,7 +1499,6 @@
"stages_ms": {
"InputValidationStage": 0.05,
"TextEncodingStage": 2310.34,
"ConditioningStage": 0.02,
"TimestepPreparationStage": 2.42,
"LatentPreparationStage": 27.7,
"DenoisingStage": 803631.52,
@@ -1660,7 +1565,6 @@
"stages_ms": {
"InputValidationStage": 0.09,
"TextEncodingStage": 2552.97,
"ConditioningStage": 0.03,
"TimestepPreparationStage": 1.99,
"LatentPreparationStage": 1.29,
"DenoisingStage": 154340.69,
@@ -1717,7 +1621,6 @@
"stages_ms": {
"InputValidationStage": 0.06,
"TextEncodingStage": 2467.44,
"ConditioningStage": 0.02,
"TimestepPreparationStage": 2.96,
"LatentPreparationStage": 1.87,
"DenoisingStage": 14859.47,
@@ -1785,7 +1688,6 @@
"InputValidationStage": 23.97,
"TextEncodingStage": 2485.39,
"ImageEncodingStage": 2372.07,
"ConditioningStage": 0.01,
"TimestepPreparationStage": 2.6,
"LatentPreparationStage": 0.18,
"ImageVAEEncodingStage": 2500.13,
@@ -1854,7 +1756,6 @@
"InputValidationStage": 0.05,
"TextEncodingStage": 518.88,
"ImageVAEEncodingStage": 0.0,
"ConditioningStage": 0.01,
"LatentPreparationStage": 0.45,
"TimestepPreparationStage": 3.41,
"DenoisingStage": 26377.63,
@@ -1923,7 +1824,6 @@
"ImageVAEEncodingStage": 88.06,
"TimestepPreparationStage": 2.12,
"LatentPreparationStage": 0.14,
"ConditioningStage": 0.01,
"DenoisingStage": 23869.32,
"DecodingStage": 108.23
},
@@ -1977,7 +1877,6 @@
"stages_ms": {
"InputValidationStage": 0.04,
"TextEncodingStage": 411.12,
"ConditioningStage": 0.01,
"TimestepPreparationStage": 1.44,
"LatentPreparationStage": 0.1,
"DenoisingStage": 1569.61,
python/sglang/multimodal_gen/test/server/testcase_configs.py
+1 -1
View File
@@ -10,7 +10,7 @@ pytest python/sglang/multimodal_gen/test/server/test_server_a.py -k qwen_image_t
To add a new testcase:
1. add your testcase with case-id: `my_new_test_case_id` to DIFFUSION_CASES
2. run `SGLANG_GEN_BASELINE=1 pytest -s python/sglang/multimodal_gen/test/server/test_server_a.py -k my_new_test_case_id`
2. run `SGLANG_GEN_BASELINE=1 pytest -s python/sglang/multimodal_gen/test/server/ -k my_new_test_case_id`
3. insert or override the corresponding scenario in `scenarios` section of perf_baselines.json with the output baseline of step-2