edge results: cold-cache init profile, GPU hardware emulation, CPU thread sweep, edge_final benches

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
This commit is contained in:
tzhouamandClaude Fable 5.1 committed 2026-09-09 15:23:08 +08:00
1 parent 536866b916
commit a786a88919
42 files changed
+58321

No files matched your search

@@ -0,0 +1,99 @@
async_chunk: true
connectors:
connector_of_shared_memory:
name: SharedMemoryConnector
extra:
codec_streaming: true
connector_get_sleep_s: 0.01
connector_get_max_wait_first_chunk: 3000
connector_get_max_wait: 300
codec_chunk_frames: 25
codec_left_context_frames: 72
initial_codec_chunk_frames: 1
decode_batch_max_size: 4
stages:
- stage_id: 0
max_num_seqs: 64
gpu_memory_utilization: 0.3
trust_remote_code: true
enable_prefix_caching: false
async_scheduling: true
max_num_batched_tokens: 32768
max_model_len: 4096
devices: '0'
output_connectors:
to_stage_1: connector_of_shared_memory
default_sampling_params:
temperature: 0.9
top_k: 50
max_tokens: 4096
min_tokens: 2
repetition_penalty: 1.05
silence_ban_frames: 0
subtalker_sampling_params:
do_sample: true
temperature: 0.9
top_k: 50
top_p: 1.0
- stage_id: 1
max_num_seqs: 64
gpu_memory_utilization: 0.3
enforce_eager: false
trust_remote_code: true
enable_prefix_caching: false
async_scheduling: true
max_num_batched_tokens: 65536
max_model_len: 65536
devices: '0'
input_connectors:
from_stage_0: connector_of_shared_memory
default_sampling_params:
temperature: 0.0
top_p: 1.0
top_k: -1
max_tokens: 65536
repetition_penalty: 1.0
platforms:
rocm:
stages:
- stage_id: 1
enforce_eager: true
npu:
stages:
- stage_id: 0
enforce_eager: false
compilation_config:
cudagraph_mode: PIECEWISE
- stage_id: 1
compilation_config:
cudagraph_mode: PIECEWISE
cpu:
stages:
- stage_id: 0
engine_args:
enforce_eager: true
async_scheduling: false
max_num_seqs: 2
max_num_batched_tokens: 2048
max_model_len: 2048
gpu_memory_utilization: 0.05
dtype: bfloat16
kv_cache_memory_bytes: 4294967296
runtime:
devices: cpu
env:
VLLM_CPU_OMP_THREADS_BIND: 56-67
- stage_id: 1
engine_args:
enforce_eager: true
async_scheduling: false
max_num_seqs: 2
max_num_batched_tokens: 8192
max_model_len: 8192
gpu_memory_utilization: 0.05
dtype: bfloat16
kv_cache_memory_bytes: 2147483648
runtime:
devices: cpu
env:
VLLM_CPU_OMP_THREADS_BIND: 68-71
@@ -0,0 +1,7 @@
{
"role": "orchestrator",
"pid": 418532,
"started_at": 1788937436.3958519,
"dumped_at": 1788937445.0899134,
"counters": {}
}
@@ -0,0 +1,48 @@
{
"role": "stage0",
"pid": 424871,
"started_at": 1788937242.1795187,
"dumped_at": 1788937439.7330387,
"counters": {
"sched.schedule_ms": {
"count": 880,
"sum_ms": 104.70030456781387,
"mean_ms": 0.11897761882706122,
"p50_ms": 0.12095319107174873,
"p95_ms": 0.14224182814359665,
"max_ms": 0.48884283751249313
},
"sched.update_from_output_ms": {
"count": 880,
"sum_ms": 101.731535512954,
"mean_ms": 0.11560401762835681,
"p50_ms": 0.11708587408065796,
"p95_ms": 0.1456798054277897,
"max_ms": 0.36838511005043983
},
"core.step_ms": {
"count": 880,
"sum_ms": 128478.91991212964,
"mean_ms": 145.99877262742004,
"p50_ms": 124.99470030888915,
"p95_ms": 219.14790803566575,
"max_ms": 7732.182804029435
},
"hop.enqueue_to_put_ms": {
"count": 53,
"sum_ms": 42.715925723314285,
"mean_ms": 0.8059608627040431,
"p50_ms": 0.7813479751348495,
"p95_ms": 1.0402752086520195,
"max_ms": 1.1817519553005695
},
"hop.put_ms": {
"count": 53,
"sum_ms": 24.499356746673584,
"mean_ms": 0.46225201408818084,
"p50_ms": 0.4214141517877579,
"p95_ms": 0.8697109296917915,
"max_ms": 1.365570817142725
}
}
}
@@ -0,0 +1,48 @@
{
"role": "stage1",
"pid": 432911,
"started_at": 1788937295.357378,
"dumped_at": 1788937443.5552511,
"counters": {
"sched.schedule_ms": {
"count": 75484,
"sum_ms": 3421.6363141313195,
"mean_ms": 0.04532929248756451,
"p50_ms": 0.0394163653254509,
"p95_ms": 0.08744187653064728,
"max_ms": 2.5428608059883118
},
"hop.poll_miss_ms": {
"count": 90934,
"sum_ms": 3579.9412396736443,
"mean_ms": 0.03936856664914822,
"p50_ms": 0.010110903531312943,
"p95_ms": 0.04783039912581444,
"max_ms": 233.60275384038687
},
"sched.update_from_output_ms": {
"count": 75484,
"sum_ms": 572.5311180576682,
"mean_ms": 0.007584800991702456,
"p50_ms": 0.006647780537605286,
"p95_ms": 0.014132820069789886,
"max_ms": 0.2440027892589569
},
"core.step_ms": {
"count": 75484,
"sum_ms": 49971.80697135627,
"mean_ms": 0.6620185333495346,
"p50_ms": 0.1842789351940155,
"p95_ms": 0.3467649221420288,
"max_ms": 837.5107739120722
},
"hop.get_ms": {
"count": 53,
"sum_ms": 41.48649424314499,
"mean_ms": 0.7827640423234904,
"p50_ms": 0.2963901497423649,
"p95_ms": 1.0263491421937943,
"max_ms": 15.571518335491419
}
}
}
@@ -0,0 +1,243 @@
{
"schema": 1,
"t0": 1788937191.3843935,
"t_end": 1788937298.6831546,
"total_s": 107.29876112937927,
"phases": [
{
"name": "stages_init",
"start": 1788937191.3848991,
"end": 1788937298.6800125,
"stage_id": null,
"replica_id": 0,
"pid": 418532,
"role": "engine",
"extra": {},
"dur_s": 107.29511332511902
},
{
"name": "g3_device_lock",
"start": 1788937192.1774457,
"end": 1788937192.177517,
"stage_id": 0,
"replica_id": 0,
"pid": 418532,
"role": "engine",
"extra": {},
"dur_s": 7.128715515136719e-05
},
{
"name": "g2_launch_lock_wait",
"start": 1788937192.1775415,
"end": 1788937192.177544,
"stage_id": 0,
"replica_id": 0,
"pid": 418532,
"role": "engine",
"extra": {
"acquired": true
},
"dur_s": 2.6226043701171875e-06
},
{
"name": "g2_spawn_and_ready",
"start": 1788937192.1777198,
"end": 1788937242.1766372,
"stage_id": 0,
"replica_id": 0,
"pid": 418532,
"role": "engine",
"extra": {},
"dur_s": 49.9989173412323
},
{
"name": "proc_entry",
"start": 1788937214.0035586,
"end": 1788937214.0035586,
"stage_id": 0,
"replica_id": 0,
"pid": 424871,
"role": "worker",
"extra": {},
"dur_s": 0.0
},
{
"name": "engine_core_init",
"start": 1788937214.0103385,
"end": 1788937242.1774259,
"stage_id": 0,
"replica_id": 0,
"pid": 424871,
"role": "worker",
"extra": {},
"dur_s": 28.16708731651306
},
{
"name": "init_device",
"start": 1788937236.8496158,
"end": 1788937237.1621988,
"stage_id": 0,
"replica_id": 0,
"pid": 428096,
"role": "worker",
"extra": {},
"dur_s": 0.31258296966552734
},
{
"name": "load_model",
"start": 1788937237.1683319,
"end": 1788937241.1311529,
"stage_id": 0,
"replica_id": 0,
"pid": 428096,
"role": "worker",
"extra": {},
"dur_s": 3.9628210067749023
},
{
"name": "load_weights",
"start": 1788937237.1683376,
"end": 1788937241.1255398,
"stage_id": 0,
"replica_id": 0,
"pid": 428096,
"role": "worker",
"extra": {},
"dur_s": 3.957202196121216
},
{
"name": "talker_mtp_init",
"start": 1788937241.1274526,
"end": 1788937241.1276531,
"stage_id": 0,
"replica_id": 0,
"pid": 428096,
"role": "worker",
"extra": {},
"dur_s": 0.00020051002502441406
},
{
"name": "attn_prewarm",
"start": 1788937241.1294017,
"end": 1788937241.1294043,
"stage_id": 0,
"replica_id": 0,
"pid": 428096,
"role": "worker",
"extra": {},
"dur_s": 2.6226043701171875e-06
},
{
"name": "g3_device_lock",
"start": 1788937242.3604438,
"end": 1788937242.3605406,
"stage_id": 1,
"replica_id": 0,
"pid": 418532,
"role": "engine",
"extra": {},
"dur_s": 9.679794311523438e-05
},
{
"name": "g2_launch_lock_wait",
"start": 1788937242.3605652,
"end": 1788937242.3605673,
"stage_id": 1,
"replica_id": 0,
"pid": 418532,
"role": "engine",
"extra": {
"acquired": true
},
"dur_s": 2.1457672119140625e-06
},
{
"name": "g2_spawn_and_ready",
"start": 1788937242.3608234,
"end": 1788937295.3548195,
"stage_id": 1,
"replica_id": 0,
"pid": 418532,
"role": "engine",
"extra": {},
"dur_s": 52.993996143341064
},
{
"name": "proc_entry",
"start": 1788937269.3247695,
"end": 1788937269.3247695,
"stage_id": 1,
"replica_id": 0,
"pid": 432911,
"role": "worker",
"extra": {},
"dur_s": 0.0
},
{
"name": "engine_core_init",
"start": 1788937269.3295782,
"end": 1788937295.355413,
"stage_id": 1,
"replica_id": 0,
"pid": 432911,
"role": "worker",
"extra": {},
"dur_s": 26.025834798812866
},
{
"name": "init_device",
"start": 1788937292.4094946,
"end": 1788937292.6462276,
"stage_id": 1,
"replica_id": 0,
"pid": 436860,
"role": "worker",
"extra": {},
"dur_s": 0.23673295974731445
},
{
"name": "load_model",
"start": 1788937292.652116,
"end": 1788937294.7258286,
"stage_id": 1,
"replica_id": 0,
"pid": 436860,
"role": "worker",
"extra": {},
"dur_s": 2.0737125873565674
},
{
"name": "load_weights",
"start": 1788937292.6521232,
"end": 1788937294.7196088,
"stage_id": 1,
"replica_id": 0,
"pid": 436860,
"role": "worker",
"extra": {},
"dur_s": 2.0674855709075928
},
{
"name": "talker_mtp_init",
"start": 1788937294.7217488,
"end": 1788937294.721756,
"stage_id": 1,
"replica_id": 0,
"pid": 436860,
"role": "worker",
"extra": {},
"dur_s": 7.152557373046875e-06
},
{
"name": "attn_prewarm",
"start": 1788937294.7237742,
"end": 1788937294.7237816,
"stage_id": 1,
"replica_id": 0,
"pid": 436860,
"role": "worker",
"extra": {},
"dur_s": 7.3909759521484375e-06
}
]
}
@@ -0,0 +1,99 @@
async_chunk: true
connectors:
connector_of_shared_memory:
name: SharedMemoryConnector
extra:
codec_streaming: true
connector_get_sleep_s: 0.01
connector_get_max_wait_first_chunk: 3000
connector_get_max_wait: 300
codec_chunk_frames: 25
codec_left_context_frames: 72
initial_codec_chunk_frames: 1
decode_batch_max_size: 4
stages:
- stage_id: 0
max_num_seqs: 64
gpu_memory_utilization: 0.3
trust_remote_code: true
enable_prefix_caching: false
async_scheduling: true
max_num_batched_tokens: 32768
max_model_len: 4096
devices: '0'
output_connectors:
to_stage_1: connector_of_shared_memory
default_sampling_params:
temperature: 0.9
top_k: 50
max_tokens: 4096
min_tokens: 2
repetition_penalty: 1.05
silence_ban_frames: 0
subtalker_sampling_params:
do_sample: true
temperature: 0.9
top_k: 50
top_p: 1.0
- stage_id: 1
max_num_seqs: 64
gpu_memory_utilization: 0.3
enforce_eager: false
trust_remote_code: true
enable_prefix_caching: false
async_scheduling: true
max_num_batched_tokens: 65536
max_model_len: 65536
devices: '0'
input_connectors:
from_stage_0: connector_of_shared_memory
default_sampling_params:
temperature: 0.0
top_p: 1.0
top_k: -1
max_tokens: 65536
repetition_penalty: 1.0
platforms:
rocm:
stages:
- stage_id: 1
enforce_eager: true
npu:
stages:
- stage_id: 0
enforce_eager: false
compilation_config:
cudagraph_mode: PIECEWISE
- stage_id: 1
compilation_config:
cudagraph_mode: PIECEWISE
cpu:
stages:
- stage_id: 0
engine_args:
enforce_eager: true
async_scheduling: false
max_num_seqs: 2
max_num_batched_tokens: 2048
max_model_len: 2048
gpu_memory_utilization: 0.05
dtype: bfloat16
kv_cache_memory_bytes: 4294967296
runtime:
devices: cpu
env:
VLLM_CPU_OMP_THREADS_BIND: 56-67
- stage_id: 1
engine_args:
enforce_eager: true
async_scheduling: false
max_num_seqs: 2
max_num_batched_tokens: 8192
max_model_len: 8192
gpu_memory_utilization: 0.05
dtype: bfloat16
kv_cache_memory_bytes: 2147483648
runtime:
devices: cpu
env:
VLLM_CPU_OMP_THREADS_BIND: 68-71
@@ -0,0 +1,7 @@
{
"role": "orchestrator",
"pid": 464652,
"started_at": 1788937706.733928,
"dumped_at": 1788937714.6591258,
"counters": {}
}
@@ -0,0 +1,48 @@
{
"role": "stage0",
"pid": 468972,
"started_at": 1788937526.005224,
"dumped_at": 1788937709.6415462,
"counters": {
"sched.schedule_ms": {
"count": 873,
"sum_ms": 107.25896526128054,
"mean_ms": 0.12286250316297885,
"p50_ms": 0.12072501704096794,
"p95_ms": 0.1628398895263672,
"max_ms": 0.4603560082614422
},
"sched.update_from_output_ms": {
"count": 873,
"sum_ms": 102.26774914190173,
"mean_ms": 0.117145188020506,
"p50_ms": 0.11497782543301582,
"p95_ms": 0.15713786706328392,
"max_ms": 0.4851659759879112
},
"core.step_ms": {
"count": 873,
"sum_ms": 119137.0151345618,
"mean_ms": 136.46851676353015,
"p50_ms": 90.59954388067126,
"p95_ms": 178.77295799553394,
"max_ms": 22189.76076738909
},
"hop.enqueue_to_put_ms": {
"count": 53,
"sum_ms": 43.450802098959684,
"mean_ms": 0.8198264546973525,
"p50_ms": 0.7725879549980164,
"p95_ms": 1.120619010180235,
"max_ms": 1.4559072442352772
},
"hop.put_ms": {
"count": 53,
"sum_ms": 24.027912877500057,
"mean_ms": 0.4533568467452841,
"p50_ms": 0.3942418843507767,
"p95_ms": 0.986012164503336,
"max_ms": 1.529117114841938
}
}
}
@@ -0,0 +1,48 @@
{
"role": "stage1",
"pid": 475171,
"started_at": 1788937574.7616582,
"dumped_at": 1788937713.174787,
"counters": {
"sched.schedule_ms": {
"count": 70400,
"sum_ms": 3303.8047766312957,
"mean_ms": 0.04692904512260363,
"p50_ms": 0.045369379222393036,
"p95_ms": 0.16833608970046043,
"max_ms": 0.5868319422006607
},
"hop.poll_miss_ms": {
"count": 84571,
"sum_ms": 3332.7820263803005,
"mean_ms": 0.03940809528538507,
"p50_ms": 0.01399405300617218,
"p95_ms": 0.08834479376673698,
"max_ms": 199.43405222147703
},
"sched.update_from_output_ms": {
"count": 70400,
"sum_ms": 551.4583084732294,
"mean_ms": 0.007833214608994736,
"p50_ms": 0.007598660886287689,
"p95_ms": 0.028978101909160614,
"max_ms": 0.19824272021651268
},
"core.step_ms": {
"count": 70400,
"sum_ms": 47886.25876745209,
"mean_ms": 0.680202539310399,
"p50_ms": 0.20564114674925804,
"p95_ms": 0.6246641278266907,
"max_ms": 708.3039227873087
},
"hop.get_ms": {
"count": 53,
"sum_ms": 47.45385330170393,
"mean_ms": 0.895355722673659,
"p50_ms": 0.2906098961830139,
"p95_ms": 6.81871036067605,
"max_ms": 7.4676391668617725
}
}
}
@@ -0,0 +1,243 @@
{
"schema": 1,
"t0": 1788937476.7571697,
"t_end": 1788937577.983278,
"total_s": 101.22610831260681,
"phases": [
{
"name": "stages_init",
"start": 1788937476.757685,
"end": 1788937577.9063513,
"stage_id": null,
"replica_id": 0,
"pid": 464652,
"role": "engine",
"extra": {},
"dur_s": 101.14866638183594
},
{
"name": "g3_device_lock",
"start": 1788937477.460587,
"end": 1788937477.460645,
"stage_id": 0,
"replica_id": 0,
"pid": 464652,
"role": "engine",
"extra": {},
"dur_s": 5.793571472167969e-05
},
{
"name": "g2_launch_lock_wait",
"start": 1788937477.4606662,
"end": 1788937477.460668,
"stage_id": 0,
"replica_id": 0,
"pid": 464652,
"role": "engine",
"extra": {
"acquired": true
},
"dur_s": 1.9073486328125e-06
},
{
"name": "g2_spawn_and_ready",
"start": 1788937477.4608014,
"end": 1788937526.0026684,
"stage_id": 0,
"replica_id": 0,
"pid": 464652,
"role": "engine",
"extra": {},
"dur_s": 48.54186701774597
},
{
"name": "proc_entry",
"start": 1788937498.8621824,
"end": 1788937498.8621824,
"stage_id": 0,
"replica_id": 0,
"pid": 468972,
"role": "worker",
"extra": {},
"dur_s": 0.0
},
{
"name": "engine_core_init",
"start": 1788937498.8673916,
"end": 1788937526.0034242,
"stage_id": 0,
"replica_id": 0,
"pid": 468972,
"role": "worker",
"extra": {},
"dur_s": 27.136032581329346
},
{
"name": "init_device",
"start": 1788937520.8931713,
"end": 1788937521.1476817,
"stage_id": 0,
"replica_id": 0,
"pid": 471869,
"role": "worker",
"extra": {},
"dur_s": 0.25451040267944336
},
{
"name": "load_model",
"start": 1788937521.1529937,
"end": 1788937525.0946898,
"stage_id": 0,
"replica_id": 0,
"pid": 471869,
"role": "worker",
"extra": {},
"dur_s": 3.9416961669921875
},
{
"name": "load_weights",
"start": 1788937521.152999,
"end": 1788937525.0873435,
"stage_id": 0,
"replica_id": 0,
"pid": 471869,
"role": "worker",
"extra": {},
"dur_s": 3.934344530105591
},
{
"name": "talker_mtp_init",
"start": 1788937525.0892782,
"end": 1788937525.0894785,
"stage_id": 0,
"replica_id": 0,
"pid": 471869,
"role": "worker",
"extra": {},
"dur_s": 0.0002002716064453125
},
{
"name": "attn_prewarm",
"start": 1788937525.091496,
"end": 1788937525.0914981,
"stage_id": 0,
"replica_id": 0,
"pid": 471869,
"role": "worker",
"extra": {},
"dur_s": 2.1457672119140625e-06
},
{
"name": "g3_device_lock",
"start": 1788937526.186592,
"end": 1788937526.1866658,
"stage_id": 1,
"replica_id": 0,
"pid": 464652,
"role": "engine",
"extra": {},
"dur_s": 7.367134094238281e-05
},
{
"name": "g2_launch_lock_wait",
"start": 1788937526.186685,
"end": 1788937526.1866865,
"stage_id": 1,
"replica_id": 0,
"pid": 464652,
"role": "engine",
"extra": {
"acquired": true
},
"dur_s": 1.430511474609375e-06
},
{
"name": "g2_spawn_and_ready",
"start": 1788937526.1868832,
"end": 1788937574.7588286,
"stage_id": 1,
"replica_id": 0,
"pid": 464652,
"role": "engine",
"extra": {},
"dur_s": 48.57194542884827
},
{
"name": "proc_entry",
"start": 1788937551.5488408,
"end": 1788937551.5488408,
"stage_id": 1,
"replica_id": 0,
"pid": 475171,
"role": "worker",
"extra": {},
"dur_s": 0.0
},
{
"name": "engine_core_init",
"start": 1788937551.5539594,
"end": 1788937574.7597423,
"stage_id": 1,
"replica_id": 0,
"pid": 475171,
"role": "worker",
"extra": {},
"dur_s": 23.205782890319824
},
{
"name": "init_device",
"start": 1788937572.3821156,
"end": 1788937572.5746398,
"stage_id": 1,
"replica_id": 0,
"pid": 478288,
"role": "worker",
"extra": {},
"dur_s": 0.19252419471740723
},
{
"name": "load_model",
"start": 1788937572.5795212,
"end": 1788937573.9114175,
"stage_id": 1,
"replica_id": 0,
"pid": 478288,
"role": "worker",
"extra": {},
"dur_s": 1.3318963050842285
},
{
"name": "load_weights",
"start": 1788937572.5795257,
"end": 1788937573.9038074,
"stage_id": 1,
"replica_id": 0,
"pid": 478288,
"role": "worker",
"extra": {},
"dur_s": 1.3242816925048828
},
{
"name": "talker_mtp_init",
"start": 1788937573.9056342,
"end": 1788937573.905638,
"stage_id": 1,
"replica_id": 0,
"pid": 478288,
"role": "worker",
"extra": {},
"dur_s": 3.814697265625e-06
},
{
"name": "attn_prewarm",
"start": 1788937573.9096172,
"end": 1788937573.9096208,
"stage_id": 1,
"replica_id": 0,
"pid": 478288,
"role": "worker",
"extra": {},
"dur_s": 3.5762786865234375e-06
}
]
}
@@ -0,0 +1,99 @@
async_chunk: true
connectors:
connector_of_shared_memory:
name: SharedMemoryConnector
extra:
codec_streaming: true
connector_get_sleep_s: 0.01
connector_get_max_wait_first_chunk: 3000
connector_get_max_wait: 300
codec_chunk_frames: 25
codec_left_context_frames: 72
initial_codec_chunk_frames: 1
decode_batch_max_size: 4
stages:
- stage_id: 0
max_num_seqs: 64
gpu_memory_utilization: 0.3
trust_remote_code: true
enable_prefix_caching: false
async_scheduling: true
max_num_batched_tokens: 32768
max_model_len: 4096
devices: '0'
output_connectors:
to_stage_1: connector_of_shared_memory
default_sampling_params:
temperature: 0.9
top_k: 50
max_tokens: 4096
min_tokens: 2
repetition_penalty: 1.05
silence_ban_frames: 0
subtalker_sampling_params:
do_sample: true
temperature: 0.9
top_k: 50
top_p: 1.0
- stage_id: 1
max_num_seqs: 64
gpu_memory_utilization: 0.3
enforce_eager: false
trust_remote_code: true
enable_prefix_caching: false
async_scheduling: true
max_num_batched_tokens: 65536
max_model_len: 65536
devices: '0'
input_connectors:
from_stage_0: connector_of_shared_memory
default_sampling_params:
temperature: 0.0
top_p: 1.0
top_k: -1
max_tokens: 65536
repetition_penalty: 1.0
platforms:
rocm:
stages:
- stage_id: 1
enforce_eager: true
npu:
stages:
- stage_id: 0
enforce_eager: false
compilation_config:
cudagraph_mode: PIECEWISE
- stage_id: 1
compilation_config:
cudagraph_mode: PIECEWISE
cpu:
stages:
- stage_id: 0
engine_args:
enforce_eager: true
async_scheduling: false
max_num_seqs: 2
max_num_batched_tokens: 2048
max_model_len: 2048
gpu_memory_utilization: 0.05
dtype: bfloat16
kv_cache_memory_bytes: 4294967296
runtime:
devices: cpu
env:
VLLM_CPU_OMP_THREADS_BIND: 56-79
- stage_id: 1
engine_args:
enforce_eager: true
async_scheduling: false
max_num_seqs: 2
max_num_batched_tokens: 8192
max_model_len: 8192
gpu_memory_utilization: 0.05
dtype: bfloat16
kv_cache_memory_bytes: 2147483648
runtime:
devices: cpu
env:
VLLM_CPU_OMP_THREADS_BIND: 80-87
@@ -0,0 +1,7 @@
{
"role": "orchestrator",
"pid": 504842,
"started_at": 1788938041.6438563,
"dumped_at": 1788938049.569157,
"counters": {}
}
@@ -0,0 +1,48 @@
{
"role": "stage0",
"pid": 511598,
"started_at": 1788937814.4889548,
"dumped_at": 1788938044.505461,
"counters": {
"sched.schedule_ms": {
"count": 900,
"sum_ms": 98.5742025077343,
"mean_ms": 0.10952689167526033,
"p50_ms": 0.11252798140048981,
"p95_ms": 0.13857567682862282,
"max_ms": 0.3441832959651947
},
"sched.update_from_output_ms": {
"count": 900,
"sum_ms": 99.51333235949278,
"mean_ms": 0.11057036928832531,
"p50_ms": 0.11337967589497566,
"p95_ms": 0.14364393427968025,
"max_ms": 0.3099888563156128
},
"core.step_ms": {
"count": 900,
"sum_ms": 122386.67163113132,
"mean_ms": 135.985190701257,
"p50_ms": 121.72688776627183,
"p95_ms": 172.62085806578398,
"max_ms": 7528.453212697059
},
"hop.enqueue_to_put_ms": {
"count": 54,
"sum_ms": 46.5872585773468,
"mean_ms": 0.8627270106916074,
"p50_ms": 0.8627558127045631,
"p95_ms": 1.1294898577034473,
"max_ms": 1.1827964335680008
},
"hop.put_ms": {
"count": 54,
"sum_ms": 34.14346417412162,
"mean_ms": 0.6322863735948447,
"p50_ms": 0.509193167090416,
"p95_ms": 1.3038450852036476,
"max_ms": 2.899942919611931
}
}
}
@@ -0,0 +1,48 @@
{
"role": "stage1",
"pid": 522100,
"started_at": 1788937911.2442284,
"dumped_at": 1788938048.1289275,
"counters": {
"sched.schedule_ms": {
"count": 75821,
"sum_ms": 3223.3080160804093,
"mean_ms": 0.04251207470331978,
"p50_ms": 0.03743031993508339,
"p95_ms": 0.11504581198096275,
"max_ms": 0.834654550999403
},
"hop.poll_miss_ms": {
"count": 93129,
"sum_ms": 3384.352858643979,
"mean_ms": 0.036340483186160906,
"p50_ms": 0.009352806955575943,
"p95_ms": 0.07634004577994347,
"max_ms": 191.62802305072546
},
"sched.update_from_output_ms": {
"count": 75821,
"sum_ms": 499.78485191240907,
"mean_ms": 0.006591641522960777,
"p50_ms": 0.00578584149479866,
"p95_ms": 0.01600617542862892,
"max_ms": 2.6703840121626854
},
"core.step_ms": {
"count": 75821,
"sum_ms": 40393.862828146666,
"mean_ms": 0.5327529685462691,
"p50_ms": 0.17837760969996452,
"p95_ms": 0.4669683985412121,
"max_ms": 479.6791849657893
},
"hop.get_ms": {
"count": 54,
"sum_ms": 29.41588358953595,
"mean_ms": 0.5447385849914065,
"p50_ms": 0.29903603717684746,
"p95_ms": 0.8803871460258961,
"max_ms": 7.884413935244083
}
}
}
@@ -0,0 +1,243 @@
{
"schema": 1,
"t0": 1788937760.9075842,
"t_end": 1788937914.4210823,
"total_s": 153.51349806785583,
"phases": [
{
"name": "stages_init",
"start": 1788937760.9080393,
"end": 1788937914.3464415,
"stage_id": null,
"replica_id": 0,
"pid": 504842,
"role": "engine",
"extra": {},
"dur_s": 153.43840217590332
},
{
"name": "g3_device_lock",
"start": 1788937761.5689328,
"end": 1788937761.5689902,
"stage_id": 0,
"replica_id": 0,
"pid": 504842,
"role": "engine",
"extra": {},
"dur_s": 5.745887756347656e-05
},
{
"name": "g2_launch_lock_wait",
"start": 1788937761.569012,
"end": 1788937761.569014,
"stage_id": 0,
"replica_id": 0,
"pid": 504842,
"role": "engine",
"extra": {
"acquired": true
},
"dur_s": 2.1457672119140625e-06
},
{
"name": "g2_spawn_and_ready",
"start": 1788937761.5691514,
"end": 1788937814.4864094,
"stage_id": 0,
"replica_id": 0,
"pid": 504842,
"role": "engine",
"extra": {},
"dur_s": 52.9172580242157
},
{
"name": "proc_entry",
"start": 1788937783.0036478,
"end": 1788937783.0036478,
"stage_id": 0,
"replica_id": 0,
"pid": 511598,
"role": "worker",
"extra": {},
"dur_s": 0.0
},
{
"name": "engine_core_init",
"start": 1788937783.0094001,
"end": 1788937814.4868238,
"stage_id": 0,
"replica_id": 0,
"pid": 511598,
"role": "worker",
"extra": {},
"dur_s": 31.477423667907715
},
{
"name": "init_device",
"start": 1788937809.3742902,
"end": 1788937809.604538,
"stage_id": 0,
"replica_id": 0,
"pid": 516643,
"role": "worker",
"extra": {},
"dur_s": 0.23024773597717285
},
{
"name": "load_model",
"start": 1788937809.6125417,
"end": 1788937813.601946,
"stage_id": 0,
"replica_id": 0,
"pid": 516643,
"role": "worker",
"extra": {},
"dur_s": 3.9894044399261475
},
{
"name": "load_weights",
"start": 1788937809.6125474,
"end": 1788937813.5952604,
"stage_id": 0,
"replica_id": 0,
"pid": 516643,
"role": "worker",
"extra": {},
"dur_s": 3.982712984085083
},
{
"name": "talker_mtp_init",
"start": 1788937813.5975797,
"end": 1788937813.5977764,
"stage_id": 0,
"replica_id": 0,
"pid": 516643,
"role": "worker",
"extra": {},
"dur_s": 0.00019669532775878906
},
{
"name": "attn_prewarm",
"start": 1788937813.5999393,
"end": 1788937813.599942,
"stage_id": 0,
"replica_id": 0,
"pid": 516643,
"role": "worker",
"extra": {},
"dur_s": 2.6226043701171875e-06
},
{
"name": "g3_device_lock",
"start": 1788937814.669887,
"end": 1788937814.6699686,
"stage_id": 1,
"replica_id": 0,
"pid": 504842,
"role": "engine",
"extra": {},
"dur_s": 8.153915405273438e-05
},
{
"name": "g2_launch_lock_wait",
"start": 1788937814.6699884,
"end": 1788937814.66999,
"stage_id": 1,
"replica_id": 0,
"pid": 504842,
"role": "engine",
"extra": {
"acquired": true
},
"dur_s": 1.6689300537109375e-06
},
{
"name": "g2_spawn_and_ready",
"start": 1788937814.6701822,
"end": 1788937911.2417424,
"stage_id": 1,
"replica_id": 0,
"pid": 504842,
"role": "engine",
"extra": {},
"dur_s": 96.57156014442444
},
{
"name": "proc_entry",
"start": 1788937861.6442833,
"end": 1788937861.6442833,
"stage_id": 1,
"replica_id": 0,
"pid": 522100,
"role": "worker",
"extra": {},
"dur_s": 0.0
},
{
"name": "engine_core_init",
"start": 1788937861.6744096,
"end": 1788937911.242202,
"stage_id": 1,
"replica_id": 0,
"pid": 522100,
"role": "worker",
"extra": {},
"dur_s": 49.5677924156189
},
{
"name": "init_device",
"start": 1788937909.126535,
"end": 1788937909.3243504,
"stage_id": 1,
"replica_id": 0,
"pid": 528373,
"role": "worker",
"extra": {},
"dur_s": 0.1978154182434082
},
{
"name": "load_model",
"start": 1788937909.3314798,
"end": 1788937910.6644092,
"stage_id": 1,
"replica_id": 0,
"pid": 528373,
"role": "worker",
"extra": {},
"dur_s": 1.3329293727874756
},
{
"name": "load_weights",
"start": 1788937909.331484,
"end": 1788937910.6586757,
"stage_id": 1,
"replica_id": 0,
"pid": 528373,
"role": "worker",
"extra": {},
"dur_s": 1.3271915912628174
},
{
"name": "talker_mtp_init",
"start": 1788937910.6605468,
"end": 1788937910.6605513,
"stage_id": 1,
"replica_id": 0,
"pid": 528373,
"role": "worker",
"extra": {},
"dur_s": 4.5299530029296875e-06
},
{
"name": "attn_prewarm",
"start": 1788937910.6625953,
"end": 1788937910.6625993,
"stage_id": 1,
"replica_id": 0,
"pid": 528373,
"role": "worker",
"extra": {},
"dur_s": 4.0531158447265625e-06
}
]
}
@@ -0,0 +1,99 @@
async_chunk: true
connectors:
connector_of_shared_memory:
name: SharedMemoryConnector
extra:
codec_streaming: true
connector_get_sleep_s: 0.01
connector_get_max_wait_first_chunk: 3000
connector_get_max_wait: 300
codec_chunk_frames: 25
codec_left_context_frames: 72
initial_codec_chunk_frames: 1
decode_batch_max_size: 4
stages:
- stage_id: 0
max_num_seqs: 64
gpu_memory_utilization: 0.3
trust_remote_code: true
enable_prefix_caching: false
async_scheduling: true
max_num_batched_tokens: 32768
max_model_len: 4096
devices: '0'
output_connectors:
to_stage_1: connector_of_shared_memory
default_sampling_params:
temperature: 0.9
top_k: 50
max_tokens: 4096
min_tokens: 2
repetition_penalty: 1.05
silence_ban_frames: 0
subtalker_sampling_params:
do_sample: true
temperature: 0.9
top_k: 50
top_p: 1.0
- stage_id: 1
max_num_seqs: 64
gpu_memory_utilization: 0.3
enforce_eager: false
trust_remote_code: true
enable_prefix_caching: false
async_scheduling: true
max_num_batched_tokens: 65536
max_model_len: 65536
devices: '0'
input_connectors:
from_stage_0: connector_of_shared_memory
default_sampling_params:
temperature: 0.0
top_p: 1.0
top_k: -1
max_tokens: 65536
repetition_penalty: 1.0
platforms:
rocm:
stages:
- stage_id: 1
enforce_eager: true
npu:
stages:
- stage_id: 0
enforce_eager: false
compilation_config:
cudagraph_mode: PIECEWISE
- stage_id: 1
compilation_config:
cudagraph_mode: PIECEWISE
cpu:
stages:
- stage_id: 0
engine_args:
enforce_eager: true
async_scheduling: false
max_num_seqs: 2
max_num_batched_tokens: 2048
max_model_len: 2048
gpu_memory_utilization: 0.05
dtype: bfloat16
kv_cache_memory_bytes: 4294967296
runtime:
devices: cpu
env:
VLLM_CPU_OMP_THREADS_BIND: 56-61
- stage_id: 1
engine_args:
enforce_eager: true
async_scheduling: false
max_num_seqs: 2
max_num_batched_tokens: 8192
max_model_len: 8192
gpu_memory_utilization: 0.05
dtype: bfloat16
kv_cache_memory_bytes: 2147483648
runtime:
devices: cpu
env:
VLLM_CPU_OMP_THREADS_BIND: 62-63
@@ -0,0 +1,7 @@
{
"role": "orchestrator",
"pid": 350505,
"started_at": 1788937149.2157419,
"dumped_at": 1788937157.8900526,
"counters": {}
}
@@ -0,0 +1,48 @@
{
"role": "stage0",
"pid": 354824,
"started_at": 1788936858.899882,
"dumped_at": 1788937152.1771297,
"counters": {
"sched.schedule_ms": {
"count": 880,
"sum_ms": 113.85760828852654,
"mean_ms": 0.1293836457824165,
"p50_ms": 0.12300070375204086,
"p95_ms": 0.15993323177099228,
"max_ms": 0.5592219531536102
},
"sched.update_from_output_ms": {
"count": 880,
"sum_ms": 111.58864293247461,
"mean_ms": 0.12680527605963024,
"p50_ms": 0.11766469106078148,
"p95_ms": 0.1712641678750515,
"max_ms": 0.4406590014696121
},
"core.step_ms": {
"count": 880,
"sum_ms": 218658.92411675304,
"mean_ms": 248.47605013267392,
"p50_ms": 189.374472014606,
"p95_ms": 495.7753666676581,
"max_ms": 9211.024974938482
},
"hop.enqueue_to_put_ms": {
"count": 53,
"sum_ms": 69.13604401051998,
"mean_ms": 1.3044536605758488,
"p50_ms": 0.566443894058466,
"p95_ms": 7.188800722360611,
"max_ms": 14.182327780872583
},
"hop.put_ms": {
"count": 53,
"sum_ms": 15.770158264786005,
"mean_ms": 0.2975501559393586,
"p50_ms": 0.23616105318069458,
"p95_ms": 0.7588141597807407,
"max_ms": 1.000771764665842
}
}
}
@@ -0,0 +1,48 @@
{
"role": "stage1",
"pid": 363878,
"started_at": 1788936909.8249059,
"dumped_at": 1788937156.2564445,
"counters": {
"sched.schedule_ms": {
"count": 48578,
"sum_ms": 2433.9906335808337,
"mean_ms": 0.050104792984084025,
"p50_ms": 0.041312072426080704,
"p95_ms": 0.09867222979664803,
"max_ms": 1.761691179126501
},
"hop.poll_miss_ms": {
"count": 65753,
"sum_ms": 2836.558175738901,
"mean_ms": 0.04313960086595138,
"p50_ms": 0.010591000318527222,
"p95_ms": 0.03549410030245781,
"max_ms": 168.18306082859635
},
"sched.update_from_output_ms": {
"count": 48578,
"sum_ms": 399.6013831347227,
"mean_ms": 0.008225974373887824,
"p50_ms": 0.006903894245624542,
"p95_ms": 0.015707220882177353,
"max_ms": 0.28642406687140465
},
"core.step_ms": {
"count": 48578,
"sum_ms": 82254.69604181126,
"mean_ms": 1.693249949397078,
"p50_ms": 0.21502003073692322,
"p95_ms": 5.378467030823231,
"max_ms": 2009.8296799696982
},
"hop.get_ms": {
"count": 53,
"sum_ms": 49.13898929953575,
"mean_ms": 0.9271507415006746,
"p50_ms": 0.28171995654702187,
"p95_ms": 4.237850196659565,
"max_ms": 11.287613306194544
}
}
}
@@ -0,0 +1,243 @@
{
"schema": 1,
"t0": 1788936802.6801364,
"t_end": 1788936913.186558,
"total_s": 110.50642156600952,
"phases": [
{
"name": "stages_init",
"start": 1788936802.6806157,
"end": 1788936913.1836624,
"stage_id": null,
"replica_id": 0,
"pid": 350505,
"role": "engine",
"extra": {},
"dur_s": 110.50304675102234
},
{
"name": "g3_device_lock",
"start": 1788936803.5798295,
"end": 1788936803.5798979,
"stage_id": 0,
"replica_id": 0,
"pid": 350505,
"role": "engine",
"extra": {},
"dur_s": 6.842613220214844e-05
},
{
"name": "g2_launch_lock_wait",
"start": 1788936803.5799193,
"end": 1788936803.5799212,
"stage_id": 0,
"replica_id": 0,
"pid": 350505,
"role": "engine",
"extra": {
"acquired": true
},
"dur_s": 1.9073486328125e-06
},
{
"name": "g2_spawn_and_ready",
"start": 1788936803.5800498,
"end": 1788936858.896481,
"stage_id": 0,
"replica_id": 0,
"pid": 350505,
"role": "engine",
"extra": {},
"dur_s": 55.316431283950806
},
{
"name": "proc_entry",
"start": 1788936826.8371336,
"end": 1788936826.8371336,
"stage_id": 0,
"replica_id": 0,
"pid": 354824,
"role": "worker",
"extra": {},
"dur_s": 0.0
},
{
"name": "engine_core_init",
"start": 1788936826.8422565,
"end": 1788936858.8969872,
"stage_id": 0,
"replica_id": 0,
"pid": 354824,
"role": "worker",
"extra": {},
"dur_s": 32.05473065376282
},
{
"name": "init_device",
"start": 1788936853.3436105,
"end": 1788936853.7399018,
"stage_id": 0,
"replica_id": 0,
"pid": 358566,
"role": "worker",
"extra": {},
"dur_s": 0.39629125595092773
},
{
"name": "load_model",
"start": 1788936853.7478306,
"end": 1788936857.920361,
"stage_id": 0,
"replica_id": 0,
"pid": 358566,
"role": "worker",
"extra": {},
"dur_s": 4.17253041267395
},
{
"name": "load_weights",
"start": 1788936853.747836,
"end": 1788936857.9141886,
"stage_id": 0,
"replica_id": 0,
"pid": 358566,
"role": "worker",
"extra": {},
"dur_s": 4.1663525104522705
},
{
"name": "talker_mtp_init",
"start": 1788936857.9163105,
"end": 1788936857.916512,
"stage_id": 0,
"replica_id": 0,
"pid": 358566,
"role": "worker",
"extra": {},
"dur_s": 0.0002014636993408203
},
{
"name": "attn_prewarm",
"start": 1788936857.9184418,
"end": 1788936857.918444,
"stage_id": 0,
"replica_id": 0,
"pid": 358566,
"role": "worker",
"extra": {},
"dur_s": 2.1457672119140625e-06
},
{
"name": "g3_device_lock",
"start": 1788936859.0799925,
"end": 1788936859.0800705,
"stage_id": 1,
"replica_id": 0,
"pid": 350505,
"role": "engine",
"extra": {},
"dur_s": 7.796287536621094e-05
},
{
"name": "g2_launch_lock_wait",
"start": 1788936859.08009,
"end": 1788936859.0800924,
"stage_id": 1,
"replica_id": 0,
"pid": 350505,
"role": "engine",
"extra": {
"acquired": true
},
"dur_s": 2.384185791015625e-06
},
{
"name": "g2_spawn_and_ready",
"start": 1788936859.080314,
"end": 1788936909.8221347,
"stage_id": 1,
"replica_id": 0,
"pid": 350505,
"role": "engine",
"extra": {},
"dur_s": 50.74182081222534
},
{
"name": "proc_entry",
"start": 1788936883.5199432,
"end": 1788936883.5199432,
"stage_id": 1,
"replica_id": 0,
"pid": 363878,
"role": "worker",
"extra": {},
"dur_s": 0.0
},
{
"name": "engine_core_init",
"start": 1788936883.5268247,
"end": 1788936909.8229566,
"stage_id": 1,
"replica_id": 0,
"pid": 363878,
"role": "worker",
"extra": {},
"dur_s": 26.29613184928894
},
{
"name": "init_device",
"start": 1788936907.4546945,
"end": 1788936907.6483772,
"stage_id": 1,
"replica_id": 0,
"pid": 368929,
"role": "worker",
"extra": {},
"dur_s": 0.19368267059326172
},
{
"name": "load_model",
"start": 1788936907.6537755,
"end": 1788936909.2131197,
"stage_id": 1,
"replica_id": 0,
"pid": 368929,
"role": "worker",
"extra": {},
"dur_s": 1.5593442916870117
},
{
"name": "load_weights",
"start": 1788936907.6537797,
"end": 1788936909.2066395,
"stage_id": 1,
"replica_id": 0,
"pid": 368929,
"role": "worker",
"extra": {},
"dur_s": 1.5528597831726074
},
{
"name": "talker_mtp_init",
"start": 1788936909.208844,
"end": 1788936909.2088485,
"stage_id": 1,
"replica_id": 0,
"pid": 368929,
"role": "worker",
"extra": {},
"dur_s": 4.5299530029296875e-06
},
{
"name": "attn_prewarm",
"start": 1788936909.2109854,
"end": 1788936909.2109895,
"stage_id": 1,
"replica_id": 0,
"pid": 368929,
"role": "worker",
"extra": {},
"dur_s": 4.0531158447265625e-06
}
]
}
@@ -0,0 +1,435 @@
{
"model": "Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice",
"timestamp": "20260909-145257",
"host_cpu_base": 56,
"rows": [
{
"case": "cpu8",
"threads": 8,
"stage0_cpus": "56-61",
"stage1_cpus": "62-63",
"inductor": false,
"returncode": 0,
"cmd": [
"taskset",
"-c",
"56-63",
"/data/zhoutaichang/embedding_infer/.venvs/omni-cpu/bin/python",
"/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/stream_latency_bench.py",
"--model",
"Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice",
"--deploy-config",
"/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/edge_results/cpu_sweep/20260909-145257/cpu8/cpu8.yaml",
"--query-type",
"CustomVoice",
"--txt-prompts",
"/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/prompts_12.txt",
"--warmup",
"1",
"--repeat",
"1",
"--output-dir",
"/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/edge_results/cpu_sweep/20260909-145257/cpu8/bench",
"--tag",
"cpu8",
"--step-stats",
"--init-timeout",
"1500",
"--stage-init-timeout",
"1200"
],
"init_s": 112.65361651126295,
"ttfa_p50": 426.0161090642214,
"playback_start_p50": 9438.19383399561,
"rtf_total_p50": 2.780552295553207,
"peak_rss_gib": 13.310592651367188,
"step_stats": [
{
"role": "stage0",
"counter": "core.step_ms",
"count": 801,
"mean_ms": 250.51031519918146,
"p95_ms": 499.69329591840506,
"share_of_frame": 3.1313789399897685
},
{
"role": "stage1",
"counter": "core.step_ms",
"count": 48571,
"mean_ms": 1.6700252576988601,
"p95_ms": 5.3447638638317585,
"share_of_frame": 0.02087531572123575
},
{
"role": "stage0",
"counter": "hop.enqueue_to_put_ms",
"count": 48,
"mean_ms": 1.3707510370295495,
"p95_ms": 7.9169743694365025,
"share_of_frame": 0.01713438796286937
},
{
"role": "stage1",
"counter": "hop.get_ms",
"count": 52,
"mean_ms": 0.939564027178746,
"p95_ms": 4.237850196659565,
"share_of_frame": 0.011744550339734325
},
{
"role": "stage0",
"counter": "hop.put_ms",
"count": 48,
"mean_ms": 0.30433394325276214,
"p95_ms": 0.7615480571985245,
"share_of_frame": 0.003804174290659527
},
{
"role": "stage0",
"counter": "sched.schedule_ms",
"count": 802,
"mean_ms": 0.12998572409041503,
"p95_ms": 0.16232719644904137,
"share_of_frame": 0.001624821551130188
},
{
"role": "stage0",
"counter": "sched.update_from_output_ms",
"count": 801,
"mean_ms": 0.12777436329296019,
"p95_ms": 0.17483020201325417,
"share_of_frame": 0.0015971795411620024
},
{
"role": "stage1",
"counter": "sched.schedule_ms",
"count": 48572,
"mean_ms": 0.050099294604372346,
"p95_ms": 0.0986647792160511,
"share_of_frame": 0.0006262411825546544
}
]
},
{
"case": "cpu16",
"threads": 16,
"stage0_cpus": "56-67",
"stage1_cpus": "68-71",
"inductor": false,
"returncode": 0,
"cmd": [
"taskset",
"-c",
"56-71",
"/data/zhoutaichang/embedding_infer/.venvs/omni-cpu/bin/python",
"/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/stream_latency_bench.py",
"--model",
"Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice",
"--deploy-config",
"/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/edge_results/cpu_sweep/20260909-145257/cpu16/cpu16.yaml",
"--query-type",
"CustomVoice",
"--txt-prompts",
"/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/prompts_12.txt",
"--warmup",
"1",
"--repeat",
"1",
"--output-dir",
"/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/edge_results/cpu_sweep/20260909-145257/cpu16/bench",
"--tag",
"cpu16",
"--step-stats",
"--init-timeout",
"1500",
"--stage-init-timeout",
"1200"
],
"init_s": 109.45583191420883,
"ttfa_p50": 292.3040329478681,
"playback_start_p50": 5346.291882991791,
"rtf_total_p50": 1.8194366359597887,
"peak_rss_gib": 13.319080352783203,
"step_stats": [
{
"role": "stage0",
"counter": "core.step_ms",
"count": 801,
"mean_ms": 147.36169048314088,
"p95_ms": 219.59225786849856,
"share_of_frame": 1.842021131039261
},
{
"role": "stage0",
"counter": "hop.enqueue_to_put_ms",
"count": 48,
"mean_ms": 0.8175775195316722,
"p95_ms": 1.054082065820694,
"share_of_frame": 0.010219718994145904
},
{
"role": "stage1",
"counter": "hop.get_ms",
"count": 52,
"mean_ms": 0.7918375956181151,
"p95_ms": 1.0263491421937943,
"share_of_frame": 0.009897969945226438
},
{
"role": "stage1",
"counter": "core.step_ms",
"count": 75381,
"mean_ms": 0.6547511186745705,
"p95_ms": 0.3465181216597557,
"share_of_frame": 0.008184388983432132
},
{
"role": "stage0",
"counter": "hop.put_ms",
"count": 48,
"mean_ms": 0.4695340370138486,
"p95_ms": 1.1264756321907043,
"share_of_frame": 0.005869175462673107
},
{
"role": "stage0",
"counter": "sched.schedule_ms",
"count": 802,
"mean_ms": 0.11892363596289532,
"p95_ms": 0.1422208733856678,
"share_of_frame": 0.0014865454495361914
},
{
"role": "stage0",
"counter": "sched.update_from_output_ms",
"count": 801,
"mean_ms": 0.1157221533172065,
"p95_ms": 0.14588981866836548,
"share_of_frame": 0.0014465269164650814
},
{
"role": "stage1",
"counter": "sched.schedule_ms",
"count": 75382,
"mean_ms": 0.045330475403062113,
"p95_ms": 0.08748378604650497,
"share_of_frame": 0.0005666309425382764
}
]
},
{
"case": "cpu16_inductor",
"threads": 16,
"stage0_cpus": "56-67",
"stage1_cpus": "68-71",
"inductor": true,
"returncode": 0,
"cmd": [
"taskset",
"-c",
"56-71",
"/data/zhoutaichang/embedding_infer/.venvs/omni-cpu/bin/python",
"/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/stream_latency_bench.py",
"--model",
"Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice",
"--deploy-config",
"/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/edge_results/cpu_sweep/20260909-145257/cpu16_inductor/cpu16_inductor.yaml",
"--query-type",
"CustomVoice",
"--txt-prompts",
"/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/prompts_12.txt",
"--warmup",
"1",
"--repeat",
"1",
"--output-dir",
"/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/edge_results/cpu_sweep/20260909-145257/cpu16_inductor/bench",
"--tag",
"cpu16_inductor",
"--step-stats",
"--init-timeout",
"1500",
"--stage-init-timeout",
"1200"
],
"init_s": 103.38288806378841,
"ttfa_p50": 233.9726546779275,
"playback_start_p50": 3642.257673740387,
"rtf_total_p50": 1.4156768089402225,
"peak_rss_gib": 13.337944030761719,
"step_stats": [
{
"role": "stage0",
"counter": "core.step_ms",
"count": 801,
"mean_ms": 138.89987101794665,
"p95_ms": 179.95364079251885,
"share_of_frame": 1.736248387724333
},
{
"role": "stage1",
"counter": "hop.get_ms",
"count": 52,
"mean_ms": 0.9077971783251717,
"p95_ms": 6.81871036067605,
"share_of_frame": 0.011347464729064647
},
{
"role": "stage0",
"counter": "hop.enqueue_to_put_ms",
"count": 48,
"mean_ms": 0.8297054834353427,
"p95_ms": 1.1348621919751167,
"share_of_frame": 0.010371318542941783
},
{
"role": "stage1",
"counter": "core.step_ms",
"count": 70325,
"mean_ms": 0.672337398657405,
"p95_ms": 0.6240149959921837,
"share_of_frame": 0.008404217483217563
},
{
"role": "stage0",
"counter": "hop.put_ms",
"count": 48,
"mean_ms": 0.4615955598031481,
"p95_ms": 1.0002129711210728,
"share_of_frame": 0.005769944497539351
},
{
"role": "stage0",
"counter": "sched.schedule_ms",
"count": 802,
"mean_ms": 0.12305161237549454,
"p95_ms": 0.16312114894390106,
"share_of_frame": 0.0015381451546936818
},
{
"role": "stage0",
"counter": "sched.update_from_output_ms",
"count": 801,
"mean_ms": 0.11718296474973658,
"p95_ms": 0.15717511996626854,
"share_of_frame": 0.0014647870593717072
},
{
"role": "stage1",
"counter": "sched.schedule_ms",
"count": 70326,
"mean_ms": 0.04691729749081007,
"p95_ms": 0.16828998923301697,
"share_of_frame": 0.0005864662186351259
}
]
},
{
"case": "cpu32",
"threads": 32,
"stage0_cpus": "56-79",
"stage1_cpus": "80-87",
"inductor": false,
"returncode": 0,
"cmd": [
"taskset",
"-c",
"56-87",
"/data/zhoutaichang/embedding_infer/.venvs/omni-cpu/bin/python",
"/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/stream_latency_bench.py",
"--model",
"Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice",
"--deploy-config",
"/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/edge_results/cpu_sweep/20260909-145257/cpu32/cpu32.yaml",
"--query-type",
"CustomVoice",
"--txt-prompts",
"/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/prompts_12.txt",
"--warmup",
"1",
"--repeat",
"1",
"--output-dir",
"/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/edge_results/cpu_sweep/20260909-145257/cpu32/bench",
"--tag",
"cpu32",
"--step-stats",
"--init-timeout",
"1500",
"--stage-init-timeout",
"1200"
],
"init_s": 155.67474265303463,
"ttfa_p50": 251.5225620009005,
"playback_start_p50": 4995.450793839991,
"rtf_total_p50": 1.6667016652018691,
"peak_rss_gib": 13.523807525634766,
"step_stats": [
{
"role": "stage0",
"counter": "core.step_ms",
"count": 801,
"mean_ms": 136.15325865011368,
"p95_ms": 168.9455881714821,
"share_of_frame": 1.701915733126421
},
{
"role": "stage0",
"counter": "hop.enqueue_to_put_ms",
"count": 48,
"mean_ms": 0.8607769462590417,
"p95_ms": 1.1321199126541615,
"share_of_frame": 0.010759711828238021
},
{
"role": "stage0",
"counter": "hop.put_ms",
"count": 48,
"mean_ms": 0.648114801151678,
"p95_ms": 1.8421937711536884,
"share_of_frame": 0.008101435014395975
},
{
"role": "stage1",
"counter": "hop.get_ms",
"count": 53,
"mean_ms": 0.5498284843029841,
"p95_ms": 0.8803871460258961,
"share_of_frame": 0.006872856053787302
},
{
"role": "stage1",
"counter": "core.step_ms",
"count": 75788,
"mean_ms": 0.528387298923063,
"p95_ms": 0.46643195673823357,
"share_of_frame": 0.006604841236538287
},
{
"role": "stage0",
"counter": "sched.update_from_output_ms",
"count": 801,
"mean_ms": 0.11039824781941712,
"p95_ms": 0.14352239668369293,
"share_of_frame": 0.0013799780977427141
},
{
"role": "stage0",
"counter": "sched.schedule_ms",
"count": 802,
"mean_ms": 0.1092211127355509,
"p95_ms": 0.13850722461938858,
"share_of_frame": 0.0013652639091943863
},
{
"role": "stage1",
"counter": "sched.schedule_ms",
"count": 75789,
"mean_ms": 0.042512980838002465,
"p95_ms": 0.11504581198096275,
"share_of_frame": 0.0005314122604750309
}
]
}
]
}
@@ -0,0 +1,6 @@
| case | threads | inductor | rc | init_s | ttfa_p50 | playback_start_p50 | rtf_total_p50 | peak_rss_gib |
|---|---|---|---|---|---|---|---|---|
| cpu8 | 8 | False | 0 | 112.65 | 426.02 | 9438.19 | 2.78 | 13.31 |
| cpu16 | 16 | False | 0 | 109.46 | 292.30 | 5346.29 | 1.82 | 13.32 |
| cpu16_inductor | 16 | True | 0 | 103.38 | 233.97 | 3642.26 | 1.42 | 13.34 |
| cpu32 | 32 | False | 0 | 155.67 | 251.52 | 4995.45 | 1.67 | 13.52 |
@@ -0,0 +1,40 @@
{
"arch": "x86_64",
"cpu_model": "emulated small discrete GPU (emulated) host",
"cpu_flags": [
"avx2",
"avx512"
],
"n_logical": 8,
"n_physical": 8,
"clusters": [
{
"max_khz": 2200000,
"cpus": [
0,
1,
2,
3,
4,
5,
6,
7
]
}
],
"numa_nodes": 2,
"ram_total_bytes": 17179869184,
"ram_available_bytes": 12884901888,
"accelerator": "cuda_discrete",
"gpu_name": "small discrete GPU (emulated)",
"gpu_mem_bytes": 8589934592,
"gpu_sm_count": 16,
"gpu_capability": [
8,
6
],
"supports_bf16": true,
"supports_fp16": true,
"thermal_zones": 2,
"source": "emulation"
}
@@ -0,0 +1,353 @@
{
"schema": 1,
"t0": 1788937450.8325427,
"t_end": 1788937635.7648778,
"total_s": 184.93233513832092,
"phases": [
{
"name": "stages_init",
"start": 1788937450.8331246,
"end": 1788937635.7607691,
"stage_id": null,
"replica_id": 0,
"pid": 455757,
"role": "engine",
"extra": {},
"dur_s": 184.92764449119568
},
{
"name": "g3_device_lock",
"start": 1788937452.3881056,
"end": 1788937452.388198,
"stage_id": 0,
"replica_id": 0,
"pid": 455757,
"role": "engine",
"extra": {},
"dur_s": 9.226799011230469e-05
},
{
"name": "g2_launch_lock_wait",
"start": 1788937452.3882296,
"end": 1788937452.3882322,
"stage_id": 0,
"replica_id": 0,
"pid": 455757,
"role": "engine",
"extra": {
"acquired": true
},
"dur_s": 2.6226043701171875e-06
},
{
"name": "g2_spawn_and_ready",
"start": 1788937452.3883023,
"end": 1788937569.3442385,
"stage_id": 0,
"replica_id": 0,
"pid": 455757,
"role": "engine",
"extra": {},
"dur_s": 116.95593619346619
},
{
"name": "proc_entry",
"start": 1788937481.8358915,
"end": 1788937481.8358915,
"stage_id": 0,
"replica_id": 0,
"pid": 465618,
"role": "worker",
"extra": {},
"dur_s": 0.0
},
{
"name": "engine_core_init",
"start": 1788937481.8420215,
"end": 1788937569.6319711,
"stage_id": 0,
"replica_id": 0,
"pid": 465618,
"role": "worker",
"extra": {},
"dur_s": 87.78994965553284
},
{
"name": "init_device",
"start": 1788937482.7140331,
"end": 1788937486.262075,
"stage_id": 0,
"replica_id": 0,
"pid": 465618,
"role": "worker",
"extra": {},
"dur_s": 3.548041820526123
},
{
"name": "load_model",
"start": 1788937486.268097,
"end": 1788937489.9686635,
"stage_id": 0,
"replica_id": 0,
"pid": 465618,
"role": "worker",
"extra": {},
"dur_s": 3.700566530227661
},
{
"name": "load_weights",
"start": 1788937486.2681026,
"end": 1788937489.946917,
"stage_id": 0,
"replica_id": 0,
"pid": 465618,
"role": "worker",
"extra": {},
"dur_s": 3.67881441116333
},
{
"name": "talker_mtp_init",
"start": 1788937489.9508278,
"end": 1788937489.9514086,
"stage_id": 0,
"replica_id": 0,
"pid": 465618,
"role": "worker",
"extra": {},
"dur_s": 0.0005807876586914062
},
{
"name": "attn_prewarm",
"start": 1788937489.9555438,
"end": 1788937489.966001,
"stage_id": 0,
"replica_id": 0,
"pid": 465618,
"role": "worker",
"extra": {},
"dur_s": 0.010457277297973633
},
{
"name": "cudagraph_capture",
"start": 1788937531.3516045,
"end": 1788937533.0407588,
"stage_id": 0,
"replica_id": 0,
"pid": 465618,
"role": "worker",
"extra": {},
"dur_s": 1.6891543865203857
},
{
"name": "talker_mtp_capture",
"start": 1788937533.042705,
"end": 1788937565.1452994,
"stage_id": 0,
"replica_id": 0,
"pid": 465618,
"role": "worker",
"extra": {},
"dur_s": 32.10259437561035
},
{
"name": "predictor_setup_compile",
"start": 1788937533.0475538,
"end": 1788937564.9444778,
"stage_id": 0,
"replica_id": 0,
"pid": 465618,
"role": "worker",
"extra": {},
"dur_s": 31.896924018859863
},
{
"name": "predictor_warmup",
"start": 1788937533.051145,
"end": 1788937564.9411893,
"stage_id": 0,
"replica_id": 0,
"pid": 465618,
"role": "worker",
"extra": {},
"dur_s": 31.89004421234131
},
{
"name": "predictor_setup_compile",
"start": 1788937565.01186,
"end": 1788937565.011862,
"stage_id": 0,
"replica_id": 0,
"pid": 465618,
"role": "worker",
"extra": {},
"dur_s": 2.1457672119140625e-06
},
{
"name": "predictor_setup_compile",
"start": 1788937565.0462866,
"end": 1788937565.0462885,
"stage_id": 0,
"replica_id": 0,
"pid": 465618,
"role": "worker",
"extra": {},
"dur_s": 1.9073486328125e-06
},
{
"name": "predictor_setup_compile",
"start": 1788937565.0681021,
"end": 1788937565.0681038,
"stage_id": 0,
"replica_id": 0,
"pid": 465618,
"role": "worker",
"extra": {},
"dur_s": 1.6689300537109375e-06
},
{
"name": "predictor_setup_compile",
"start": 1788937565.0940685,
"end": 1788937565.09407,
"stage_id": 0,
"replica_id": 0,
"pid": 465618,
"role": "worker",
"extra": {},
"dur_s": 1.430511474609375e-06
},
{
"name": "predictor_setup_compile",
"start": 1788937565.1154912,
"end": 1788937565.1154923,
"stage_id": 0,
"replica_id": 0,
"pid": 465618,
"role": "worker",
"extra": {},
"dur_s": 1.1920928955078125e-06
},
{
"name": "g3_device_lock",
"start": 1788937569.5289836,
"end": 1788937569.5291307,
"stage_id": 1,
"replica_id": 0,
"pid": 455757,
"role": "engine",
"extra": {},
"dur_s": 0.00014710426330566406
},
{
"name": "g2_launch_lock_wait",
"start": 1788937569.529175,
"end": 1788937569.529181,
"stage_id": 1,
"replica_id": 0,
"pid": 455757,
"role": "engine",
"extra": {
"acquired": true
},
"dur_s": 5.9604644775390625e-06
},
{
"name": "g2_spawn_and_ready",
"start": 1788937569.529404,
"end": 1788937632.0512912,
"stage_id": 1,
"replica_id": 0,
"pid": 455757,
"role": "engine",
"extra": {},
"dur_s": 62.52188730239868
},
{
"name": "proc_entry",
"start": 1788937615.6697354,
"end": 1788937615.6697354,
"stage_id": 1,
"replica_id": 0,
"pid": 480675,
"role": "worker",
"extra": {},
"dur_s": 0.0
},
{
"name": "engine_core_init",
"start": 1788937615.6755059,
"end": 1788937632.371459,
"stage_id": 1,
"replica_id": 0,
"pid": 480675,
"role": "worker",
"extra": {},
"dur_s": 16.695953130722046
},
{
"name": "init_device",
"start": 1788937616.7243662,
"end": 1788937619.6206415,
"stage_id": 1,
"replica_id": 0,
"pid": 480675,
"role": "worker",
"extra": {},
"dur_s": 2.896275281906128
},
{
"name": "load_model",
"start": 1788937619.623499,
"end": 1788937626.310504,
"stage_id": 1,
"replica_id": 0,
"pid": 480675,
"role": "worker",
"extra": {},
"dur_s": 6.687005043029785
},
{
"name": "load_weights",
"start": 1788937619.6235042,
"end": 1788937626.288639,
"stage_id": 1,
"replica_id": 0,
"pid": 480675,
"role": "worker",
"extra": {},
"dur_s": 6.665134906768799
},
{
"name": "code2wav_cudagraph",
"start": 1788937621.158521,
"end": 1788937625.3600535,
"stage_id": 1,
"replica_id": 0,
"pid": 480675,
"role": "worker",
"extra": {},
"dur_s": 4.201532602310181
},
{
"name": "talker_mtp_init",
"start": 1788937626.291026,
"end": 1788937626.2910318,
"stage_id": 1,
"replica_id": 0,
"pid": 480675,
"role": "worker",
"extra": {},
"dur_s": 5.7220458984375e-06
},
{
"name": "attn_prewarm",
"start": 1788937626.2930505,
"end": 1788937626.3084538,
"stage_id": 1,
"replica_id": 0,
"pid": 480675,
"role": "worker",
"extra": {},
"dur_s": 0.015403270721435547
}
]
}
@@ -0,0 +1,41 @@
{
"arch": "aarch64",
"cpu_model": "emulated Jetson Orin 32GB (emulated) host",
"cpu_flags": [
"neon",
"dotprod",
"fp16_arith"
],
"n_logical": 8,
"n_physical": 8,
"clusters": [
{
"max_khz": 2200000,
"cpus": [
0,
1,
2,
3,
4,
5,
6,
7
]
}
],
"numa_nodes": 2,
"ram_total_bytes": 34359738368,
"ram_available_bytes": 25769803776,
"accelerator": "cuda_unified",
"gpu_name": "Jetson Orin 32GB (emulated)",
"gpu_mem_bytes": 34359738368,
"gpu_sm_count": 16,
"gpu_capability": [
8,
7
],
"supports_bf16": true,
"supports_fp16": true,
"thermal_zones": 2,
"source": "emulation"
}
@@ -0,0 +1,331 @@
{
"schema": 1,
"t0": 1788937196.095764,
"t_end": 1788937375.0144205,
"total_s": 178.9186565876007,
"phases": [
{
"name": "stages_init",
"start": 1788937196.0966666,
"end": 1788937375.010322,
"stage_id": null,
"replica_id": 0,
"pid": 418683,
"role": "engine",
"extra": {},
"dur_s": 178.91365551948547
},
{
"name": "g3_device_lock",
"start": 1788937197.701576,
"end": 1788937197.7016745,
"stage_id": 0,
"replica_id": 0,
"pid": 418683,
"role": "engine",
"extra": {},
"dur_s": 9.846687316894531e-05
},
{
"name": "g2_launch_lock_wait",
"start": 1788937197.7017043,
"end": 1788937197.7017064,
"stage_id": 0,
"replica_id": 0,
"pid": 418683,
"role": "engine",
"extra": {
"acquired": true
},
"dur_s": 2.1457672119140625e-06
},
{
"name": "g2_spawn_and_ready",
"start": 1788937197.7017646,
"end": 1788937306.7017794,
"stage_id": 0,
"replica_id": 0,
"pid": 418683,
"role": "engine",
"extra": {},
"dur_s": 109.0000147819519
},
{
"name": "proc_entry",
"start": 1788937227.1890504,
"end": 1788937227.1890504,
"stage_id": 0,
"replica_id": 0,
"pid": 425541,
"role": "worker",
"extra": {},
"dur_s": 0.0
},
{
"name": "engine_core_init",
"start": 1788937227.1953025,
"end": 1788937307.0603929,
"stage_id": 0,
"replica_id": 0,
"pid": 425541,
"role": "worker",
"extra": {},
"dur_s": 79.86509037017822
},
{
"name": "init_device",
"start": 1788937228.1301286,
"end": 1788937231.9570189,
"stage_id": 0,
"replica_id": 0,
"pid": 425541,
"role": "worker",
"extra": {},
"dur_s": 3.826890230178833
},
{
"name": "load_model",
"start": 1788937231.9600875,
"end": 1788937235.1971068,
"stage_id": 0,
"replica_id": 0,
"pid": 425541,
"role": "worker",
"extra": {},
"dur_s": 3.2370193004608154
},
{
"name": "load_weights",
"start": 1788937231.9600928,
"end": 1788937235.1799903,
"stage_id": 0,
"replica_id": 0,
"pid": 425541,
"role": "worker",
"extra": {},
"dur_s": 3.219897508621216
},
{
"name": "talker_mtp_init",
"start": 1788937235.1823616,
"end": 1788937235.1829107,
"stage_id": 0,
"replica_id": 0,
"pid": 425541,
"role": "worker",
"extra": {},
"dur_s": 0.0005490779876708984
},
{
"name": "attn_prewarm",
"start": 1788937235.1850374,
"end": 1788937235.1949809,
"stage_id": 0,
"replica_id": 0,
"pid": 425541,
"role": "worker",
"extra": {},
"dur_s": 0.009943485260009766
},
{
"name": "cudagraph_capture",
"start": 1788937278.018602,
"end": 1788937279.9298906,
"stage_id": 0,
"replica_id": 0,
"pid": 425541,
"role": "worker",
"extra": {},
"dur_s": 1.9112887382507324
},
{
"name": "talker_mtp_capture",
"start": 1788937279.932117,
"end": 1788937303.1939397,
"stage_id": 0,
"replica_id": 0,
"pid": 425541,
"role": "worker",
"extra": {},
"dur_s": 23.26182270050049
},
{
"name": "predictor_setup_compile",
"start": 1788937279.9345863,
"end": 1788937303.0044587,
"stage_id": 0,
"replica_id": 0,
"pid": 425541,
"role": "worker",
"extra": {},
"dur_s": 23.06987237930298
},
{
"name": "predictor_warmup",
"start": 1788937279.9400873,
"end": 1788937302.996208,
"stage_id": 0,
"replica_id": 0,
"pid": 425541,
"role": "worker",
"extra": {},
"dur_s": 23.05612063407898
},
{
"name": "predictor_setup_compile",
"start": 1788937303.0989366,
"end": 1788937303.0989392,
"stage_id": 0,
"replica_id": 0,
"pid": 425541,
"role": "worker",
"extra": {},
"dur_s": 2.6226043701171875e-06
},
{
"name": "predictor_setup_compile",
"start": 1788937303.1291308,
"end": 1788937303.1291327,
"stage_id": 0,
"replica_id": 0,
"pid": 425541,
"role": "worker",
"extra": {},
"dur_s": 1.9073486328125e-06
},
{
"name": "predictor_setup_compile",
"start": 1788937303.1514804,
"end": 1788937303.151482,
"stage_id": 0,
"replica_id": 0,
"pid": 425541,
"role": "worker",
"extra": {},
"dur_s": 1.6689300537109375e-06
},
{
"name": "g3_device_lock",
"start": 1788937306.8852873,
"end": 1788937306.8854792,
"stage_id": 1,
"replica_id": 0,
"pid": 418683,
"role": "engine",
"extra": {},
"dur_s": 0.0001919269561767578
},
{
"name": "g2_launch_lock_wait",
"start": 1788937306.8855052,
"end": 1788937306.885508,
"stage_id": 1,
"replica_id": 0,
"pid": 418683,
"role": "engine",
"extra": {
"acquired": true
},
"dur_s": 2.86102294921875e-06
},
{
"name": "g2_spawn_and_ready",
"start": 1788937306.8857636,
"end": 1788937371.2813861,
"stage_id": 1,
"replica_id": 0,
"pid": 418683,
"role": "engine",
"extra": {},
"dur_s": 64.39562249183655
},
{
"name": "proc_entry",
"start": 1788937359.3308196,
"end": 1788937359.3308196,
"stage_id": 1,
"replica_id": 0,
"pid": 442113,
"role": "worker",
"extra": {},
"dur_s": 0.0
},
{
"name": "engine_core_init",
"start": 1788937359.3371582,
"end": 1788937371.5819116,
"stage_id": 1,
"replica_id": 0,
"pid": 442113,
"role": "worker",
"extra": {},
"dur_s": 12.244753360748291
},
{
"name": "init_device",
"start": 1788937360.346136,
"end": 1788937364.6208138,
"stage_id": 1,
"replica_id": 0,
"pid": 442113,
"role": "worker",
"extra": {},
"dur_s": 4.274677753448486
},
{
"name": "load_model",
"start": 1788937364.6241953,
"end": 1788937366.8248458,
"stage_id": 1,
"replica_id": 0,
"pid": 442113,
"role": "worker",
"extra": {},
"dur_s": 2.200650453567505
},
{
"name": "load_weights",
"start": 1788937364.624201,
"end": 1788937366.8182845,
"stage_id": 1,
"replica_id": 0,
"pid": 442113,
"role": "worker",
"extra": {},
"dur_s": 2.1940834522247314
},
{
"name": "code2wav_cudagraph",
"start": 1788937365.7831078,
"end": 1788937365.783183,
"stage_id": 1,
"replica_id": 0,
"pid": 442113,
"role": "worker",
"extra": {},
"dur_s": 7.534027099609375e-05
},
{
"name": "talker_mtp_init",
"start": 1788937366.8206184,
"end": 1788937366.8206284,
"stage_id": 1,
"replica_id": 0,
"pid": 442113,
"role": "worker",
"extra": {},
"dur_s": 1.0013580322265625e-05
},
{
"name": "attn_prewarm",
"start": 1788937366.8227258,
"end": 1788937366.8227358,
"stage_id": 1,
"replica_id": 0,
"pid": 442113,
"role": "worker",
"extra": {},
"dur_s": 1.0013580322265625e-05
}
]
}
@@ -0,0 +1,41 @@
{
"arch": "aarch64",
"cpu_model": "emulated Jetson Orin 8GB (emulated) host",
"cpu_flags": [
"neon",
"dotprod",
"fp16_arith"
],
"n_logical": 8,
"n_physical": 8,
"clusters": [
{
"max_khz": 2200000,
"cpus": [
0,
1,
2,
3,
4,
5,
6,
7
]
}
],
"numa_nodes": 2,
"ram_total_bytes": 8589934592,
"ram_available_bytes": 6442450944,
"accelerator": "cuda_unified",
"gpu_name": "Jetson Orin 8GB (emulated)",
"gpu_mem_bytes": 8589934592,
"gpu_sm_count": 16,
"gpu_capability": [
8,
7
],
"supports_bf16": true,
"supports_fp16": true,
"thermal_zones": 2,
"source": "emulation"
}
@@ -0,0 +1,254 @@
{
"schema": 1,
"t0": 1788936976.2017155,
"t_end": 1788937100.9199085,
"total_s": 124.71819305419922,
"phases": [
{
"name": "stages_init",
"start": 1788936976.2023551,
"end": 1788937100.9153888,
"stage_id": null,
"replica_id": 0,
"pid": 382030,
"role": "engine",
"extra": {},
"dur_s": 124.71303367614746
},
{
"name": "g3_device_lock",
"start": 1788936977.6465156,
"end": 1788936977.6466358,
"stage_id": 0,
"replica_id": 0,
"pid": 382030,
"role": "engine",
"extra": {},
"dur_s": 0.0001201629638671875
},
{
"name": "g2_launch_lock_wait",
"start": 1788936977.6466682,
"end": 1788936977.6466706,
"stage_id": 0,
"replica_id": 0,
"pid": 382030,
"role": "engine",
"extra": {
"acquired": true
},
"dur_s": 2.384185791015625e-06
},
{
"name": "g2_spawn_and_ready",
"start": 1788936977.6467502,
"end": 1788937036.7105439,
"stage_id": 0,
"replica_id": 0,
"pid": 382030,
"role": "engine",
"extra": {},
"dur_s": 59.063793659210205
},
{
"name": "proc_entry",
"start": 1788937021.92302,
"end": 1788937021.92302,
"stage_id": 0,
"replica_id": 0,
"pid": 389077,
"role": "worker",
"extra": {},
"dur_s": 0.0
},
{
"name": "engine_core_init",
"start": 1788937021.9341168,
"end": 1788937037.0617135,
"stage_id": 0,
"replica_id": 0,
"pid": 389077,
"role": "worker",
"extra": {},
"dur_s": 15.127596616744995
},
{
"name": "init_device",
"start": 1788937022.9382017,
"end": 1788937026.638886,
"stage_id": 0,
"replica_id": 0,
"pid": 389077,
"role": "worker",
"extra": {},
"dur_s": 3.7006843090057373
},
{
"name": "load_model",
"start": 1788937026.6422162,
"end": 1788937029.5129046,
"stage_id": 0,
"replica_id": 0,
"pid": 389077,
"role": "worker",
"extra": {},
"dur_s": 2.8706884384155273
},
{
"name": "load_weights",
"start": 1788937026.6422246,
"end": 1788937029.5050254,
"stage_id": 0,
"replica_id": 0,
"pid": 389077,
"role": "worker",
"extra": {},
"dur_s": 2.8628008365631104
},
{
"name": "talker_mtp_init",
"start": 1788937029.5073655,
"end": 1788937029.5087717,
"stage_id": 0,
"replica_id": 0,
"pid": 389077,
"role": "worker",
"extra": {},
"dur_s": 0.0014061927795410156
},
{
"name": "attn_prewarm",
"start": 1788937029.5109634,
"end": 1788937029.5109673,
"stage_id": 0,
"replica_id": 0,
"pid": 389077,
"role": "worker",
"extra": {},
"dur_s": 3.814697265625e-06
},
{
"name": "g3_device_lock",
"start": 1788937036.8933508,
"end": 1788937036.8934786,
"stage_id": 1,
"replica_id": 0,
"pid": 382030,
"role": "engine",
"extra": {},
"dur_s": 0.0001277923583984375
},
{
"name": "g2_launch_lock_wait",
"start": 1788937036.8935075,
"end": 1788937036.8935106,
"stage_id": 1,
"replica_id": 0,
"pid": 382030,
"role": "engine",
"extra": {
"acquired": true
},
"dur_s": 3.0994415283203125e-06
},
{
"name": "g2_spawn_and_ready",
"start": 1788937036.8936777,
"end": 1788937097.130456,
"stage_id": 1,
"replica_id": 0,
"pid": 382030,
"role": "engine",
"extra": {},
"dur_s": 60.236778259277344
},
{
"name": "proc_entry",
"start": 1788937083.2225428,
"end": 1788937083.2225428,
"stage_id": 1,
"replica_id": 0,
"pid": 398558,
"role": "worker",
"extra": {},
"dur_s": 0.0
},
{
"name": "engine_core_init",
"start": 1788937083.228271,
"end": 1788937097.4330206,
"stage_id": 1,
"replica_id": 0,
"pid": 398558,
"role": "worker",
"extra": {},
"dur_s": 14.204749584197998
},
{
"name": "init_device",
"start": 1788937084.2443407,
"end": 1788937088.6246407,
"stage_id": 1,
"replica_id": 0,
"pid": 398558,
"role": "worker",
"extra": {},
"dur_s": 4.380300045013428
},
{
"name": "load_model",
"start": 1788937088.6272058,
"end": 1788937093.026989,
"stage_id": 1,
"replica_id": 0,
"pid": 398558,
"role": "worker",
"extra": {},
"dur_s": 4.399783134460449
},
{
"name": "load_weights",
"start": 1788937088.6272104,
"end": 1788937093.0111985,
"stage_id": 1,
"replica_id": 0,
"pid": 398558,
"role": "worker",
"extra": {},
"dur_s": 4.38398814201355
},
{
"name": "code2wav_cudagraph",
"start": 1788937092.0205486,
"end": 1788937092.0206237,
"stage_id": 1,
"replica_id": 0,
"pid": 398558,
"role": "worker",
"extra": {},
"dur_s": 7.510185241699219e-05
},
{
"name": "talker_mtp_init",
"start": 1788937093.0180583,
"end": 1788937093.0180619,
"stage_id": 1,
"replica_id": 0,
"pid": 398558,
"role": "worker",
"extra": {},
"dur_s": 3.5762786865234375e-06
},
{
"name": "attn_prewarm",
"start": 1788937093.0213637,
"end": 1788937093.0213668,
"stage_id": 1,
"replica_id": 0,
"pid": 398558,
"role": "worker",
"extra": {},
"dur_s": 3.0994415283203125e-06
}
]
}
@@ -0,0 +1,118 @@
{
"model": "Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice",
"timestamp": "20260909-145540",
"rows": [
{
"case": "jetson_orin_8g",
"note": "unified 8 GiB -> tight budgets, eager",
"returncode": 0,
"cmd": [
"/data/zhoutaichang/embedding_infer/.venvs/omni-cuda/bin/python",
"/data/zhoutaichang/embedding_infer/vllm-omni-run/benchmarks/tts/stream_latency_bench.py",
"--model",
"Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice",
"--deploy-profile",
"auto",
"--query-type",
"CustomVoice",
"--txt-prompts",
"/data/zhoutaichang/embedding_infer/vllm-omni-run/benchmarks/tts/prompts_12.txt",
"--warmup",
"1",
"--repeat",
"1",
"--output-dir",
"/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/edge_results/hw_emulation/20260909-145540/jetson_orin_8g/bench",
"--tag",
"jetson_orin_8g",
"--init-timeout",
"1500",
"--stage-init-timeout",
"1200"
],
"hardware_class": "jetson",
"init_s": 129.0033661830239,
"ttfa_p50": 180.74133479967713,
"playback_start_p50": 196.47769132629037,
"stall_p50": 12.012018039822578,
"rtf_total_p50": 0.5722792122162298,
"peak_rss_gib": 8.40072250366211,
"peak_gpu_mib": 6236
},
{
"case": "jetson_orin_32g",
"note": "unified 32 GiB",
"returncode": 0,
"cmd": [
"/data/zhoutaichang/embedding_infer/.venvs/omni-cuda/bin/python",
"/data/zhoutaichang/embedding_infer/vllm-omni-run/benchmarks/tts/stream_latency_bench.py",
"--model",
"Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice",
"--deploy-profile",
"auto",
"--query-type",
"CustomVoice",
"--txt-prompts",
"/data/zhoutaichang/embedding_infer/vllm-omni-run/benchmarks/tts/prompts_12.txt",
"--warmup",
"1",
"--repeat",
"1",
"--output-dir",
"/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/edge_results/hw_emulation/20260909-145540/jetson_orin_32g/bench",
"--tag",
"jetson_orin_32g",
"--init-timeout",
"1500",
"--stage-init-timeout",
"1200"
],
"hardware_class": "jetson",
"init_s": 183.16458623297513,
"ttfa_p50": 68.33451613783836,
"playback_start_p50": 68.33451613783836,
"stall_p50": 0.0,
"rtf_total_p50": 0.09134745778283103,
"peak_rss_gib": 8.63232421875,
"peak_gpu_mib": 23376
},
{
"case": "cuda_discrete_8g",
"note": "8 GiB VRAM -> eager + small KV",
"returncode": 0,
"cmd": [
"/data/zhoutaichang/embedding_infer/.venvs/omni-cuda/bin/python",
"/data/zhoutaichang/embedding_infer/vllm-omni-run/benchmarks/tts/stream_latency_bench.py",
"--model",
"Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice",
"--deploy-profile",
"auto",
"--query-type",
"CustomVoice",
"--txt-prompts",
"/data/zhoutaichang/embedding_infer/vllm-omni-run/benchmarks/tts/prompts_12.txt",
"--warmup",
"1",
"--repeat",
"1",
"--output-dir",
"/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/edge_results/hw_emulation/20260909-145540/cuda_discrete_8g/bench",
"--tag",
"cuda_discrete_8g",
"--init-timeout",
"1500",
"--stage-init-timeout",
"1200"
],
"hardware_class": "cuda_discrete",
"init_s": 189.15159784117714,
"ttfa_p50": 47.47185343876481,
"playback_start_p50": 47.47185343876481,
"stall_p50": 0.0,
"rtf_total_p50": 0.08802489839940213,
"peak_rss_gib": 8.674686431884766,
"peak_gpu_mib": 9680
}
],
"table": "| case | rc | init_s | ttfa_p50 | playback_start_p50 | stall_p50 | rtf_total_p50 | peak_rss_gib | peak_gpu_mib | note |\n|---|---|---|---|---|---|---|---|---|---|\n| jetson_orin_8g | 0 | 129.0 | 180.7 | 196.5 | 12.0 | 0.572 | 8.4 | 6236 | unified 8 GiB -> tight budgets, eager |\n| jetson_orin_32g | 0 | 183.2 | 68.3 | 68.3 | 0.0 | 0.091 | 8.6 | 23376 | unified 32 GiB |\n| cuda_discrete_8g | 0 | 189.2 | 47.5 | 47.5 | 0.0 | 0.088 | 8.7 | 9680 | 8 GiB VRAM -> eager + small KV |"
}
@@ -0,0 +1,705 @@
{
"schema": 1,
"t0": 1788936130.2227666,
"t_end": 1788936483.5290287,
"total_s": 353.3062620162964,
"phases": [
{
"name": "stages_init",
"start": 1788936130.2236135,
"end": 1788936483.523794,
"stage_id": null,
"replica_id": 0,
"pid": 233495,
"role": "engine",
"extra": {},
"dur_s": 353.30018043518066
},
{
"name": "g3_device_lock",
"start": 1788936213.8167253,
"end": 1788936213.8169532,
"stage_id": 0,
"replica_id": 0,
"pid": 233495,
"role": "engine",
"extra": {},
"dur_s": 0.00022792816162109375
},
{
"name": "g2_launch_lock_wait",
"start": 1788936213.8170276,
"end": 1788936213.817033,
"stage_id": 0,
"replica_id": 0,
"pid": 233495,
"role": "engine",
"extra": {
"acquired": true
},
"dur_s": 5.4836273193359375e-06
},
{
"name": "g2_spawn_and_ready",
"start": 1788936213.8171694,
"end": 1788936428.8296494,
"stage_id": 0,
"replica_id": 0,
"pid": 233495,
"role": "engine",
"extra": {},
"dur_s": 215.01248002052307
},
{
"name": "proc_entry",
"start": 1788936261.3698716,
"end": 1788936261.3698716,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 0.0
},
{
"name": "engine_core_init",
"start": 1788936261.3749473,
"end": 1788936429.1718779,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 167.79693055152893
},
{
"name": "init_device",
"start": 1788936262.3672578,
"end": 1788936266.9339638,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 4.566705942153931
},
{
"name": "load_model",
"start": 1788936266.9370952,
"end": 1788936271.971528,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 5.034432888031006
},
{
"name": "load_weights",
"start": 1788936266.9371006,
"end": 1788936271.8710577,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 4.933957099914551
},
{
"name": "talker_mtp_init",
"start": 1788936271.876301,
"end": 1788936271.878915,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 0.0026140213012695312
},
{
"name": "attn_prewarm",
"start": 1788936271.8815432,
"end": 1788936271.9681308,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 0.08658766746520996
},
{
"name": "cudagraph_capture",
"start": 1788936319.5136025,
"end": 1788936323.0069542,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 3.493351697921753
},
{
"name": "talker_mtp_capture",
"start": 1788936323.010157,
"end": 1788936424.7693493,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 101.75919222831726
},
{
"name": "predictor_setup_compile",
"start": 1788936323.0162032,
"end": 1788936370.2061796,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 47.18997645378113
},
{
"name": "predictor_warmup",
"start": 1788936323.0185823,
"end": 1788936370.2018816,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 47.18329930305481
},
{
"name": "predictor_setup_compile",
"start": 1788936376.4055586,
"end": 1788936376.4055622,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 3.5762786865234375e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936376.5447793,
"end": 1788936376.544781,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 1.6689300537109375e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936383.0447352,
"end": 1788936383.0447385,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 3.337860107421875e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936383.0775263,
"end": 1788936383.0775282,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 1.9073486328125e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936390.216489,
"end": 1788936390.2164912,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 2.1457672119140625e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936390.2486777,
"end": 1788936390.24868,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 2.384185791015625e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936396.4024348,
"end": 1788936396.4024384,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 3.5762786865234375e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936396.437745,
"end": 1788936396.4377475,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 2.384185791015625e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936403.0829825,
"end": 1788936403.0829859,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 3.337860107421875e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936403.1123192,
"end": 1788936403.1123214,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 2.1457672119140625e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936409.6557791,
"end": 1788936409.6557827,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 3.5762786865234375e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936409.712581,
"end": 1788936409.7125828,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 1.9073486328125e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936417.135123,
"end": 1788936417.135126,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 3.0994415283203125e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936417.1716454,
"end": 1788936417.1716475,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 2.1457672119140625e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936423.2379582,
"end": 1788936423.2379615,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 3.337860107421875e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936423.2684662,
"end": 1788936423.2684677,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 1.430511474609375e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936423.3225656,
"end": 1788936423.3225675,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 1.9073486328125e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936423.3649695,
"end": 1788936423.3649712,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 1.6689300537109375e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936423.4603908,
"end": 1788936423.4603932,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 2.384185791015625e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936423.4930012,
"end": 1788936423.4930036,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 2.384185791015625e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936423.560433,
"end": 1788936423.560435,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 2.1457672119140625e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936423.5957077,
"end": 1788936423.5957098,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 2.1457672119140625e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936423.623222,
"end": 1788936423.6232238,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 1.6689300537109375e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936423.663991,
"end": 1788936423.6639931,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 2.1457672119140625e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936423.6956782,
"end": 1788936423.6956801,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 1.9073486328125e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936423.7243166,
"end": 1788936423.7243187,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 2.1457672119140625e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936423.7642605,
"end": 1788936423.7642632,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 2.6226043701171875e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936423.7954204,
"end": 1788936423.7954223,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 1.9073486328125e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936423.8316667,
"end": 1788936423.831669,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 2.384185791015625e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936423.8622332,
"end": 1788936423.8622346,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 1.430511474609375e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936424.025228,
"end": 1788936424.025231,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 2.86102294921875e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936424.233902,
"end": 1788936424.233906,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 4.0531158447265625e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936424.4851813,
"end": 1788936424.4851854,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 4.0531158447265625e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936424.5221987,
"end": 1788936424.5222006,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 1.9073486328125e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936424.5575473,
"end": 1788936424.5575492,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 1.9073486328125e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936424.7090418,
"end": 1788936424.7090445,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 2.6226043701171875e-06
},
{
"name": "predictor_setup_compile",
"start": 1788936424.7333796,
"end": 1788936424.733381,
"stage_id": 0,
"replica_id": 0,
"pid": 257105,
"role": "worker",
"extra": {},
"dur_s": 1.430511474609375e-06
},
{
"name": "g3_device_lock",
"start": 1788936429.013076,
"end": 1788936429.0133293,
"stage_id": 1,
"replica_id": 0,
"pid": 233495,
"role": "engine",
"extra": {},
"dur_s": 0.0002532005310058594
},
{
"name": "g2_launch_lock_wait",
"start": 1788936429.0133696,
"end": 1788936429.0133734,
"stage_id": 1,
"replica_id": 0,
"pid": 233495,
"role": "engine",
"extra": {
"acquired": true
},
"dur_s": 3.814697265625e-06
},
{
"name": "g2_spawn_and_ready",
"start": 1788936429.01368,
"end": 1788936479.8813705,
"stage_id": 1,
"replica_id": 0,
"pid": 233495,
"role": "engine",
"extra": {},
"dur_s": 50.867690563201904
},
{
"name": "proc_entry",
"start": 1788936458.5271602,
"end": 1788936458.5271602,
"stage_id": 1,
"replica_id": 0,
"pid": 288575,
"role": "worker",
"extra": {},
"dur_s": 0.0
},
{
"name": "engine_core_init",
"start": 1788936458.5328898,
"end": 1788936480.2452476,
"stage_id": 1,
"replica_id": 0,
"pid": 288575,
"role": "worker",
"extra": {},
"dur_s": 21.712357759475708
},
{
"name": "init_device",
"start": 1788936459.4907556,
"end": 1788936462.8935797,
"stage_id": 1,
"replica_id": 0,
"pid": 288575,
"role": "worker",
"extra": {},
"dur_s": 3.4028241634368896
},
{
"name": "load_model",
"start": 1788936462.898718,
"end": 1788936472.3945782,
"stage_id": 1,
"replica_id": 0,
"pid": 288575,
"role": "worker",
"extra": {},
"dur_s": 9.49586009979248
},
{
"name": "load_weights",
"start": 1788936462.8987236,
"end": 1788936472.378269,
"stage_id": 1,
"replica_id": 0,
"pid": 288575,
"role": "worker",
"extra": {},
"dur_s": 9.47954535484314
},
{
"name": "code2wav_cudagraph",
"start": 1788936464.6127064,
"end": 1788936471.3545356,
"stage_id": 1,
"replica_id": 0,
"pid": 288575,
"role": "worker",
"extra": {},
"dur_s": 6.74182915687561
},
{
"name": "talker_mtp_init",
"start": 1788936472.3804116,
"end": 1788936472.380416,
"stage_id": 1,
"replica_id": 0,
"pid": 288575,
"role": "worker",
"extra": {},
"dur_s": 4.291534423828125e-06
},
{
"name": "attn_prewarm",
"start": 1788936472.3823671,
"end": 1788936472.3926113,
"stage_id": 1,
"replica_id": 0,
"pid": 288575,
"role": "worker",
"extra": {},
"dur_s": 0.010244131088256836
}
]
}
@@ -0,0 +1,88 @@
{
"model": "Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice",
"timestamp": "20260909-144054",
"rows": [
{
"config": "cold_cache",
"returncode": 0,
"wall_s": 443.2846204279922,
"cmd": [
"/data/zhoutaichang/embedding_infer/.venvs/omni-cuda/bin/python",
"/data/zhoutaichang/embedding_infer/vllm-omni-run/benchmarks/tts/stream_latency_bench.py",
"--init-only",
"--warmup",
"1",
"--model",
"Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice",
"--deploy-config",
"/data/zhoutaichang/embedding_infer/vllm-omni-run/vllm_omni/deploy/qwen3_tts.yaml",
"--output-dir",
"/data/zhoutaichang/embedding_infer/vllm-omni/benchmarks/tts/edge_results/init_profile/20260909-144054/cold_cache/bench",
"--tag",
"cold_cache",
"--init-timeout",
"1500",
"--stage-init-timeout",
"1200"
],
"deploy_config": "/data/zhoutaichang/embedding_infer/vllm-omni-run/vllm_omni/deploy/qwen3_tts.yaml",
"init_s": 355.5421999040991,
"ttfa_ms": 464.5279091782868,
"gpu_mib": 46640,
"phases": {
"stages_init": 353.30018043518066,
"g3_device_lock": 0.0004811286926269531,
"g2_launch_lock_wait": 9.298324584960938e-06,
"g2_spawn_and_ready": 265.880170583725,
"proc_entry": 0.0,
"engine_core_init": 189.50928831100464,
"init_device": 7.96953010559082,
"load_model": 14.530292987823486,
"load_weights": 14.41350245475769,
"talker_mtp_init": 0.0026183128356933594,
"attn_prewarm": 0.0968317985534668,
"cudagraph_capture": 3.493351697921753,
"talker_mtp_capture": 101.75919222831726,
"predictor_setup_compile": 47.19006609916687,
"predictor_warmup": 47.18329930305481,
"code2wav_cudagraph": 6.74182915687561
},
"phases_by_stage": {
"-": {
"stages_init": 353.30018043518066
},
"0": {
"g3_device_lock": 0.00022792816162109375,
"g2_launch_lock_wait": 5.4836273193359375e-06,
"g2_spawn_and_ready": 215.01248002052307,
"proc_entry": 0.0,
"engine_core_init": 167.79693055152893,
"init_device": 4.566705942153931,
"load_model": 5.034432888031006,
"load_weights": 4.933957099914551,
"talker_mtp_init": 0.0026140213012695312,
"attn_prewarm": 0.08658766746520996,
"cudagraph_capture": 3.493351697921753,
"talker_mtp_capture": 101.75919222831726,
"predictor_setup_compile": 47.19006609916687,
"predictor_warmup": 47.18329930305481
},
"1": {
"g3_device_lock": 0.0002532005310058594,
"g2_launch_lock_wait": 3.814697265625e-06,
"g2_spawn_and_ready": 50.867690563201904,
"proc_entry": 0.0,
"engine_core_init": 21.712357759475708,
"init_device": 3.4028241634368896,
"load_model": 9.49586009979248,
"load_weights": 9.47954535484314,
"code2wav_cudagraph": 6.74182915687561,
"talker_mtp_init": 4.291534423828125e-06,
"attn_prewarm": 0.010244131088256836
}
},
"total_s": 353.3062620162964
}
],
"table": "| config | init_s | ttfa_ms | gpu_mib | stages_init | g2_spawn_and_ready | engine_core_init | init_device | load_model | load_weights | predictor_setup_compile | predictor_warmup | cudagraph_capture | talker_mtp_capture | code2wav_cudagraph |\n|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n| cold_cache | 355.5 | 464.5 | 46640 | 353.3 | 265.9 | 189.5 | 8.0 | 14.5 | 14.4 | 47.2 | 47.2 | 3.5 | 101.8 | 6.7 |"
}
File diff suppressed because it is too large. Load diff
File diff suppressed because it is too large. Load diff