hf-transformers-bot commited on
Commit
76c061d
·
verified ·
1 Parent(s): cd76e8d

Upload benchmark results for run 18137027745

Browse files
2025-09-30/18137027745/benchmark_results/Llama-2-7b-hf/Llama-2-7b-hf_benchmark_20250930_164929.json ADDED
@@ -0,0 +1,1175 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_name": "Llama-2-7b-hf",
3
+ "benchmark_scenarios": [
4
+ {
5
+ "scenario_name": "eager_eager_attn",
6
+ "metadata": {
7
+ "timestamp": "2025-09-30T16:43:25.046839",
8
+ "commit_id": "16a141765c3f44a77359482c54d2c6506847e834",
9
+ "hardware_info": {
10
+ "gpu_name": "NVIDIA A10G",
11
+ "gpu_memory_total_mb": 23028,
12
+ "cpu_count": 16,
13
+ "memory_total_mb": 63607,
14
+ "python_version": "3.10.12",
15
+ "torch_version": "2.8.0+cu126",
16
+ "cuda_version": "12.6"
17
+ },
18
+ "config": {
19
+ "name": "eager",
20
+ "model_id": "meta-llama/Llama-2-7b-hf",
21
+ "variant": "eager",
22
+ "warmup_iterations": 3,
23
+ "measurement_iterations": 5,
24
+ "num_tokens_to_generate": 100,
25
+ "device": "cuda",
26
+ "torch_dtype": "float16",
27
+ "compile_mode": null,
28
+ "compile_options": {},
29
+ "use_cache": true,
30
+ "batch_size": 1,
31
+ "sequence_length": null,
32
+ "attn_implementation": "eager",
33
+ "sdpa_backend": null,
34
+ "custom_params": {}
35
+ }
36
+ },
37
+ "measurements": {
38
+ "latency_seconds": {
39
+ "name": "latency_seconds",
40
+ "measurements": [
41
+ 3.683091064453125,
42
+ 3.67912255859375,
43
+ 3.686330322265625,
44
+ 3.69022021484375,
45
+ 3.682760986328125
46
+ ],
47
+ "mean": 3.6843050292968753,
48
+ "median": 3.683091064453125,
49
+ "std": 0.0037364960479889155,
50
+ "min": 3.67912255859375,
51
+ "max": 3.69022021484375,
52
+ "p25": 3.682760986328125,
53
+ "p75": 3.686330322265625,
54
+ "p90": 3.6886642578125,
55
+ "p95": 3.689442236328125,
56
+ "p99": 3.690064619140625,
57
+ "unit": "seconds"
58
+ },
59
+ "time_to_first_token_seconds": {
60
+ "name": "time_to_first_token_seconds",
61
+ "measurements": [
62
+ 0.03788246536254883,
63
+ 0.03844540786743164,
64
+ 0.039570079803466794,
65
+ 0.03852460861206055,
66
+ 0.03871273422241211
67
+ ],
68
+ "mean": 0.038627059173583986,
69
+ "median": 0.03852460861206055,
70
+ "std": 0.0005467250461413762,
71
+ "min": 0.03788246536254883,
72
+ "max": 0.039570079803466794,
73
+ "p25": 0.03844540786743164,
74
+ "p75": 0.03871273422241211,
75
+ "p90": 0.03922714157104492,
76
+ "p95": 0.03939861068725586,
77
+ "p99": 0.03953578598022461,
78
+ "unit": "seconds"
79
+ },
80
+ "tokens_per_second": {
81
+ "name": "tokens_per_second",
82
+ "measurements": [
83
+ 27.151107113570177,
84
+ 27.180393805152942,
85
+ 27.12724885124777,
86
+ 27.09865378704348,
87
+ 27.153540610221466
88
+ ],
89
+ "mean": 27.142188833447165,
90
+ "median": 27.151107113570177,
91
+ "std": 0.027519163333264256,
92
+ "min": 27.09865378704348,
93
+ "max": 27.180393805152942,
94
+ "p25": 27.12724885124777,
95
+ "p75": 27.153540610221466,
96
+ "p90": 27.16965252718035,
97
+ "p95": 27.175023166166646,
98
+ "p99": 27.179319677355682,
99
+ "unit": "tokens/sec"
100
+ },
101
+ "time_per_output_token_seconds": {
102
+ "name": "time_per_output_token_seconds",
103
+ "measurements": [
104
+ 0.03683091064453125,
105
+ 0.0367912255859375,
106
+ 0.036863303222656255,
107
+ 0.0369022021484375,
108
+ 0.03682760986328125
109
+ ],
110
+ "mean": 0.03684305029296875,
111
+ "median": 0.03683091064453125,
112
+ "std": 3.736496047989062e-05,
113
+ "min": 0.0367912255859375,
114
+ "max": 0.0369022021484375,
115
+ "p25": 0.03682760986328125,
116
+ "p75": 0.036863303222656255,
117
+ "p90": 0.036886642578125,
118
+ "p95": 0.03689442236328125,
119
+ "p99": 0.03690064619140625,
120
+ "unit": "seconds/token"
121
+ }
122
+ },
123
+ "gpu_metrics": {
124
+ "gpu_utilization_mean": 90.60215053763442,
125
+ "gpu_utilization_max": 93,
126
+ "gpu_utilization_min": 87,
127
+ "gpu_memory_used_mean": 13253,
128
+ "gpu_memory_used_max": 13253,
129
+ "gpu_memory_used_min": 13253,
130
+ "sample_count": 93,
131
+ "gpu_monitoring_status": "success"
132
+ }
133
+ },
134
+ {
135
+ "scenario_name": "compiled_compile_max-autotune_eager_attn",
136
+ "metadata": {
137
+ "timestamp": "2025-09-30T16:44:00.736913",
138
+ "commit_id": "16a141765c3f44a77359482c54d2c6506847e834",
139
+ "hardware_info": {
140
+ "gpu_name": "NVIDIA A10G",
141
+ "gpu_memory_total_mb": 23028,
142
+ "cpu_count": 16,
143
+ "memory_total_mb": 63607,
144
+ "python_version": "3.10.12",
145
+ "torch_version": "2.8.0+cu126",
146
+ "cuda_version": "12.6"
147
+ },
148
+ "config": {
149
+ "name": "compiled",
150
+ "model_id": "meta-llama/Llama-2-7b-hf",
151
+ "variant": "compiled",
152
+ "warmup_iterations": 3,
153
+ "measurement_iterations": 5,
154
+ "num_tokens_to_generate": 100,
155
+ "device": "cuda",
156
+ "torch_dtype": "float16",
157
+ "compile_mode": "max-autotune",
158
+ "compile_options": {},
159
+ "use_cache": true,
160
+ "batch_size": 1,
161
+ "sequence_length": null,
162
+ "attn_implementation": "eager",
163
+ "sdpa_backend": null,
164
+ "custom_params": {}
165
+ }
166
+ },
167
+ "measurements": {
168
+ "latency_seconds": {
169
+ "name": "latency_seconds",
170
+ "measurements": [
171
+ 7.89758251953125,
172
+ 7.88480615234375,
173
+ 7.91612158203125,
174
+ 7.94574609375,
175
+ 7.970818359375
176
+ ],
177
+ "mean": 7.923014941406249,
178
+ "median": 7.91612158203125,
179
+ "std": 0.0314836786411414,
180
+ "min": 7.88480615234375,
181
+ "max": 7.970818359375,
182
+ "p25": 7.89758251953125,
183
+ "p75": 7.94574609375,
184
+ "p90": 7.960789453125001,
185
+ "p95": 7.965803906250001,
186
+ "p99": 7.96981546875,
187
+ "unit": "seconds"
188
+ },
189
+ "time_to_first_token_seconds": {
190
+ "name": "time_to_first_token_seconds",
191
+ "measurements": [
192
+ 0.041134689331054686,
193
+ 0.038475841522216794,
194
+ 0.03858860778808594,
195
+ 0.039509311676025394,
196
+ 0.03838544082641602
197
+ ],
198
+ "mean": 0.039218778228759764,
199
+ "median": 0.03858860778808594,
200
+ "std": 0.001039100187301491,
201
+ "min": 0.03838544082641602,
202
+ "max": 0.041134689331054686,
203
+ "p25": 0.038475841522216794,
204
+ "p75": 0.039509311676025394,
205
+ "p90": 0.04048453826904297,
206
+ "p95": 0.04080961380004883,
207
+ "p99": 0.04106967422485351,
208
+ "unit": "seconds"
209
+ },
210
+ "tokens_per_second": {
211
+ "name": "tokens_per_second",
212
+ "measurements": [
213
+ 12.662102580466023,
214
+ 12.682619974148015,
215
+ 12.63244872678425,
216
+ 12.58535055363252,
217
+ 12.545763244295168
218
+ ],
219
+ "mean": 12.621657015865194,
220
+ "median": 12.63244872678425,
221
+ "std": 0.05009353404584092,
222
+ "min": 12.545763244295168,
223
+ "max": 12.682619974148015,
224
+ "p25": 12.58535055363252,
225
+ "p75": 12.662102580466023,
226
+ "p90": 12.674413016675217,
227
+ "p95": 12.678516495411616,
228
+ "p99": 12.681799278400735,
229
+ "unit": "tokens/sec"
230
+ },
231
+ "time_per_output_token_seconds": {
232
+ "name": "time_per_output_token_seconds",
233
+ "measurements": [
234
+ 0.0789758251953125,
235
+ 0.0788480615234375,
236
+ 0.0791612158203125,
237
+ 0.0794574609375,
238
+ 0.07970818359375001
239
+ ],
240
+ "mean": 0.0792301494140625,
241
+ "median": 0.0791612158203125,
242
+ "std": 0.00031483678641141573,
243
+ "min": 0.0788480615234375,
244
+ "max": 0.07970818359375001,
245
+ "p25": 0.0789758251953125,
246
+ "p75": 0.0794574609375,
247
+ "p90": 0.07960789453125,
248
+ "p95": 0.0796580390625,
249
+ "p99": 0.07969815468750001,
250
+ "unit": "seconds/token"
251
+ }
252
+ },
253
+ "gpu_metrics": {
254
+ "gpu_utilization_mean": 44.48235294117647,
255
+ "gpu_utilization_max": 91,
256
+ "gpu_utilization_min": 0,
257
+ "gpu_memory_used_mean": 13629.141176470588,
258
+ "gpu_memory_used_max": 13777,
259
+ "gpu_memory_used_min": 13415,
260
+ "sample_count": 170,
261
+ "gpu_monitoring_status": "success"
262
+ }
263
+ },
264
+ {
265
+ "scenario_name": "eager_sdpa_default",
266
+ "metadata": {
267
+ "timestamp": "2025-09-30T16:45:29.513837",
268
+ "commit_id": "16a141765c3f44a77359482c54d2c6506847e834",
269
+ "hardware_info": {
270
+ "gpu_name": "NVIDIA A10G",
271
+ "gpu_memory_total_mb": 23028,
272
+ "cpu_count": 16,
273
+ "memory_total_mb": 63607,
274
+ "python_version": "3.10.12",
275
+ "torch_version": "2.8.0+cu126",
276
+ "cuda_version": "12.6"
277
+ },
278
+ "config": {
279
+ "name": "eager",
280
+ "model_id": "meta-llama/Llama-2-7b-hf",
281
+ "variant": "eager",
282
+ "warmup_iterations": 3,
283
+ "measurement_iterations": 5,
284
+ "num_tokens_to_generate": 100,
285
+ "device": "cuda",
286
+ "torch_dtype": "float16",
287
+ "compile_mode": null,
288
+ "compile_options": {},
289
+ "use_cache": true,
290
+ "batch_size": 1,
291
+ "sequence_length": null,
292
+ "attn_implementation": "sdpa",
293
+ "sdpa_backend": null,
294
+ "custom_params": {}
295
+ }
296
+ },
297
+ "measurements": {
298
+ "latency_seconds": {
299
+ "name": "latency_seconds",
300
+ "measurements": [
301
+ 3.428746337890625,
302
+ 3.43129443359375,
303
+ 3.430367431640625,
304
+ 3.42946044921875,
305
+ 3.43266845703125
306
+ ],
307
+ "mean": 3.4305074218750002,
308
+ "median": 3.430367431640625,
309
+ "std": 0.0013788855065359818,
310
+ "min": 3.428746337890625,
311
+ "max": 3.43266845703125,
312
+ "p25": 3.42946044921875,
313
+ "p75": 3.43129443359375,
314
+ "p90": 3.43211884765625,
315
+ "p95": 3.43239365234375,
316
+ "p99": 3.43261349609375,
317
+ "unit": "seconds"
318
+ },
319
+ "time_to_first_token_seconds": {
320
+ "name": "time_to_first_token_seconds",
321
+ "measurements": [
322
+ 0.03533564758300781,
323
+ 0.034872127532958985,
324
+ 0.03475302505493164,
325
+ 0.0348765754699707,
326
+ 0.03488902282714844
327
+ ],
328
+ "mean": 0.03494527969360352,
329
+ "median": 0.0348765754699707,
330
+ "std": 0.00020128848906190037,
331
+ "min": 0.03475302505493164,
332
+ "max": 0.03533564758300781,
333
+ "p25": 0.034872127532958985,
334
+ "p75": 0.03488902282714844,
335
+ "p90": 0.03515699768066406,
336
+ "p95": 0.03524632263183594,
337
+ "p99": 0.03531778259277344,
338
+ "unit": "seconds"
339
+ },
340
+ "tokens_per_second": {
341
+ "name": "tokens_per_second",
342
+ "measurements": [
343
+ 29.165178798709356,
344
+ 29.143520597055108,
345
+ 29.151396167545087,
346
+ 29.159105777930915,
347
+ 29.131855071865928
348
+ ],
349
+ "mean": 29.150211282621278,
350
+ "median": 29.151396167545087,
351
+ "std": 0.011715445220586842,
352
+ "min": 29.131855071865928,
353
+ "max": 29.165178798709356,
354
+ "p25": 29.143520597055108,
355
+ "p75": 29.159105777930915,
356
+ "p90": 29.16274959039798,
357
+ "p95": 29.163964194553667,
358
+ "p99": 29.164935877878218,
359
+ "unit": "tokens/sec"
360
+ },
361
+ "time_per_output_token_seconds": {
362
+ "name": "time_per_output_token_seconds",
363
+ "measurements": [
364
+ 0.03428746337890625,
365
+ 0.0343129443359375,
366
+ 0.03430367431640625,
367
+ 0.0342946044921875,
368
+ 0.034326684570312495
369
+ ],
370
+ "mean": 0.034305074218749995,
371
+ "median": 0.03430367431640625,
372
+ "std": 1.3788855065358781e-05,
373
+ "min": 0.03428746337890625,
374
+ "max": 0.034326684570312495,
375
+ "p25": 0.0342946044921875,
376
+ "p75": 0.0343129443359375,
377
+ "p90": 0.03432118847656249,
378
+ "p95": 0.034323936523437494,
379
+ "p99": 0.0343261349609375,
380
+ "unit": "seconds/token"
381
+ }
382
+ },
383
+ "gpu_metrics": {
384
+ "gpu_utilization_mean": 96.89655172413794,
385
+ "gpu_utilization_max": 98,
386
+ "gpu_utilization_min": 94,
387
+ "gpu_memory_used_mean": 13849,
388
+ "gpu_memory_used_max": 13849,
389
+ "gpu_memory_used_min": 13849,
390
+ "sample_count": 87,
391
+ "gpu_monitoring_status": "success"
392
+ }
393
+ },
394
+ {
395
+ "scenario_name": "eager_sdpa_math",
396
+ "metadata": {
397
+ "timestamp": "2025-09-30T16:46:02.389673",
398
+ "commit_id": "16a141765c3f44a77359482c54d2c6506847e834",
399
+ "hardware_info": {
400
+ "gpu_name": "NVIDIA A10G",
401
+ "gpu_memory_total_mb": 23028,
402
+ "cpu_count": 16,
403
+ "memory_total_mb": 63607,
404
+ "python_version": "3.10.12",
405
+ "torch_version": "2.8.0+cu126",
406
+ "cuda_version": "12.6"
407
+ },
408
+ "config": {
409
+ "name": "eager",
410
+ "model_id": "meta-llama/Llama-2-7b-hf",
411
+ "variant": "eager",
412
+ "warmup_iterations": 3,
413
+ "measurement_iterations": 5,
414
+ "num_tokens_to_generate": 100,
415
+ "device": "cuda",
416
+ "torch_dtype": "float16",
417
+ "compile_mode": null,
418
+ "compile_options": {},
419
+ "use_cache": true,
420
+ "batch_size": 1,
421
+ "sequence_length": null,
422
+ "attn_implementation": "sdpa",
423
+ "sdpa_backend": "math",
424
+ "custom_params": {}
425
+ }
426
+ },
427
+ "measurements": {
428
+ "latency_seconds": {
429
+ "name": "latency_seconds",
430
+ "measurements": [
431
+ 3.577140625,
432
+ 3.561303955078125,
433
+ 3.55283642578125,
434
+ 3.56051318359375,
435
+ 3.55671142578125
436
+ ],
437
+ "mean": 3.5617011230468747,
438
+ "median": 3.56051318359375,
439
+ "std": 0.008287666734540101,
440
+ "min": 3.55283642578125,
441
+ "max": 3.577140625,
442
+ "p25": 3.55671142578125,
443
+ "p75": 3.561303955078125,
444
+ "p90": 3.5708059570312503,
445
+ "p95": 3.573973291015625,
446
+ "p99": 3.576507158203125,
447
+ "unit": "seconds"
448
+ },
449
+ "time_to_first_token_seconds": {
450
+ "name": "time_to_first_token_seconds",
451
+ "measurements": [
452
+ 0.04070230484008789,
453
+ 0.03977475357055664,
454
+ 0.03802313613891602,
455
+ 0.03785820770263672,
456
+ 0.03798422241210937
457
+ ],
458
+ "mean": 0.038868524932861324,
459
+ "median": 0.03802313613891602,
460
+ "std": 0.0011577057621651762,
461
+ "min": 0.03785820770263672,
462
+ "max": 0.04070230484008789,
463
+ "p25": 0.03798422241210937,
464
+ "p75": 0.03977475357055664,
465
+ "p90": 0.04033128433227539,
466
+ "p95": 0.04051679458618164,
467
+ "p99": 0.04066520278930664,
468
+ "unit": "seconds"
469
+ },
470
+ "tokens_per_second": {
471
+ "name": "tokens_per_second",
472
+ "measurements": [
473
+ 27.955289009640204,
474
+ 28.07960265716951,
475
+ 28.146525202890682,
476
+ 28.085838991071093,
477
+ 28.11585985726533
478
+ ],
479
+ "mean": 28.07662314360736,
480
+ "median": 28.085838991071093,
481
+ "std": 0.06517913623519167,
482
+ "min": 27.955289009640204,
483
+ "max": 28.146525202890682,
484
+ "p25": 28.07960265716951,
485
+ "p75": 28.11585985726533,
486
+ "p90": 28.134259064640542,
487
+ "p95": 28.140392133765612,
488
+ "p99": 28.145298589065668,
489
+ "unit": "tokens/sec"
490
+ },
491
+ "time_per_output_token_seconds": {
492
+ "name": "time_per_output_token_seconds",
493
+ "measurements": [
494
+ 0.03577140625,
495
+ 0.03561303955078125,
496
+ 0.0355283642578125,
497
+ 0.0356051318359375,
498
+ 0.035567114257812504
499
+ ],
500
+ "mean": 0.03561701123046875,
501
+ "median": 0.0356051318359375,
502
+ "std": 8.287666734539926e-05,
503
+ "min": 0.0355283642578125,
504
+ "max": 0.03577140625,
505
+ "p25": 0.035567114257812504,
506
+ "p75": 0.03561303955078125,
507
+ "p90": 0.0357080595703125,
508
+ "p95": 0.03573973291015625,
509
+ "p99": 0.03576507158203125,
510
+ "unit": "seconds/token"
511
+ }
512
+ },
513
+ "gpu_metrics": {
514
+ "gpu_utilization_mean": 96.17777777777778,
515
+ "gpu_utilization_max": 98,
516
+ "gpu_utilization_min": 93,
517
+ "gpu_memory_used_mean": 13849,
518
+ "gpu_memory_used_max": 13849,
519
+ "gpu_memory_used_min": 13849,
520
+ "sample_count": 90,
521
+ "gpu_monitoring_status": "success"
522
+ }
523
+ },
524
+ {
525
+ "scenario_name": "eager_sdpa_flash_attention",
526
+ "metadata": {
527
+ "timestamp": "2025-09-30T16:46:36.258443",
528
+ "commit_id": "16a141765c3f44a77359482c54d2c6506847e834",
529
+ "hardware_info": {
530
+ "gpu_name": "NVIDIA A10G",
531
+ "gpu_memory_total_mb": 23028,
532
+ "cpu_count": 16,
533
+ "memory_total_mb": 63607,
534
+ "python_version": "3.10.12",
535
+ "torch_version": "2.8.0+cu126",
536
+ "cuda_version": "12.6"
537
+ },
538
+ "config": {
539
+ "name": "eager",
540
+ "model_id": "meta-llama/Llama-2-7b-hf",
541
+ "variant": "eager",
542
+ "warmup_iterations": 3,
543
+ "measurement_iterations": 5,
544
+ "num_tokens_to_generate": 100,
545
+ "device": "cuda",
546
+ "torch_dtype": "float16",
547
+ "compile_mode": null,
548
+ "compile_options": {},
549
+ "use_cache": true,
550
+ "batch_size": 1,
551
+ "sequence_length": null,
552
+ "attn_implementation": "sdpa",
553
+ "sdpa_backend": "flash_attention",
554
+ "custom_params": {}
555
+ }
556
+ },
557
+ "measurements": {
558
+ "latency_seconds": {
559
+ "name": "latency_seconds",
560
+ "measurements": [
561
+ 3.429301025390625,
562
+ 3.427078857421875,
563
+ 3.428075439453125,
564
+ 3.42877001953125,
565
+ 3.426885498046875
566
+ ],
567
+ "mean": 3.42802216796875,
568
+ "median": 3.428075439453125,
569
+ "std": 0.0009358898883102497,
570
+ "min": 3.426885498046875,
571
+ "max": 3.429301025390625,
572
+ "p25": 3.427078857421875,
573
+ "p75": 3.42877001953125,
574
+ "p90": 3.429088623046875,
575
+ "p95": 3.42919482421875,
576
+ "p99": 3.42927978515625,
577
+ "unit": "seconds"
578
+ },
579
+ "time_to_first_token_seconds": {
580
+ "name": "time_to_first_token_seconds",
581
+ "measurements": [
582
+ 0.035518016815185544,
583
+ 0.03478976058959961,
584
+ 0.034790912628173826,
585
+ 0.03490806579589844,
586
+ 0.0348454704284668
587
+ ],
588
+ "mean": 0.034970445251464845,
589
+ "median": 0.0348454704284668,
590
+ "std": 0.0002772087408848411,
591
+ "min": 0.03478976058959961,
592
+ "max": 0.035518016815185544,
593
+ "p25": 0.034790912628173826,
594
+ "p75": 0.03490806579589844,
595
+ "p90": 0.035274036407470705,
596
+ "p95": 0.035396026611328124,
597
+ "p99": 0.03549361877441406,
598
+ "unit": "seconds"
599
+ },
600
+ "tokens_per_second": {
601
+ "name": "tokens_per_second",
602
+ "measurements": [
603
+ 29.160461347545073,
604
+ 29.17936941644467,
605
+ 29.170886629015616,
606
+ 29.1649773622528,
607
+ 29.181015839891405
608
+ ],
609
+ "mean": 29.171342119029912,
610
+ "median": 29.170886629015616,
611
+ "std": 0.007964008628589571,
612
+ "min": 29.160461347545073,
613
+ "max": 29.181015839891405,
614
+ "p25": 29.1649773622528,
615
+ "p75": 29.17936941644467,
616
+ "p90": 29.18035727051271,
617
+ "p95": 29.18068655520206,
618
+ "p99": 29.180949982953535,
619
+ "unit": "tokens/sec"
620
+ },
621
+ "time_per_output_token_seconds": {
622
+ "name": "time_per_output_token_seconds",
623
+ "measurements": [
624
+ 0.03429301025390625,
625
+ 0.03427078857421875,
626
+ 0.03428075439453125,
627
+ 0.0342877001953125,
628
+ 0.03426885498046875
629
+ ],
630
+ "mean": 0.0342802216796875,
631
+ "median": 0.03428075439453125,
632
+ "std": 9.358898883101938e-06,
633
+ "min": 0.03426885498046875,
634
+ "max": 0.03429301025390625,
635
+ "p25": 0.03427078857421875,
636
+ "p75": 0.0342877001953125,
637
+ "p90": 0.03429088623046875,
638
+ "p95": 0.0342919482421875,
639
+ "p99": 0.0342927978515625,
640
+ "unit": "seconds/token"
641
+ }
642
+ },
643
+ "gpu_metrics": {
644
+ "gpu_utilization_mean": 96.88372093023256,
645
+ "gpu_utilization_max": 98,
646
+ "gpu_utilization_min": 95,
647
+ "gpu_memory_used_mean": 13849,
648
+ "gpu_memory_used_max": 13849,
649
+ "gpu_memory_used_min": 13849,
650
+ "sample_count": 86,
651
+ "gpu_monitoring_status": "success"
652
+ }
653
+ },
654
+ {
655
+ "scenario_name": "eager_sdpa_efficient_attention",
656
+ "metadata": {
657
+ "timestamp": "2025-09-30T16:47:09.004081",
658
+ "commit_id": "16a141765c3f44a77359482c54d2c6506847e834",
659
+ "hardware_info": {
660
+ "gpu_name": "NVIDIA A10G",
661
+ "gpu_memory_total_mb": 23028,
662
+ "cpu_count": 16,
663
+ "memory_total_mb": 63607,
664
+ "python_version": "3.10.12",
665
+ "torch_version": "2.8.0+cu126",
666
+ "cuda_version": "12.6"
667
+ },
668
+ "config": {
669
+ "name": "eager",
670
+ "model_id": "meta-llama/Llama-2-7b-hf",
671
+ "variant": "eager",
672
+ "warmup_iterations": 3,
673
+ "measurement_iterations": 5,
674
+ "num_tokens_to_generate": 100,
675
+ "device": "cuda",
676
+ "torch_dtype": "float16",
677
+ "compile_mode": null,
678
+ "compile_options": {},
679
+ "use_cache": true,
680
+ "batch_size": 1,
681
+ "sequence_length": null,
682
+ "attn_implementation": "sdpa",
683
+ "sdpa_backend": "efficient_attention",
684
+ "custom_params": {}
685
+ }
686
+ },
687
+ "measurements": {
688
+ "latency_seconds": {
689
+ "name": "latency_seconds",
690
+ "measurements": [
691
+ 3.41038916015625,
692
+ 3.410630859375,
693
+ 3.413573486328125,
694
+ 3.40968212890625,
695
+ 3.41161669921875
696
+ ],
697
+ "mean": 3.4111784667968754,
698
+ "median": 3.410630859375,
699
+ "std": 0.0013483376546694554,
700
+ "min": 3.40968212890625,
701
+ "max": 3.413573486328125,
702
+ "p25": 3.41038916015625,
703
+ "p75": 3.41161669921875,
704
+ "p90": 3.412790771484375,
705
+ "p95": 3.41318212890625,
706
+ "p99": 3.41349521484375,
707
+ "unit": "seconds"
708
+ },
709
+ "time_to_first_token_seconds": {
710
+ "name": "time_to_first_token_seconds",
711
+ "measurements": [
712
+ 0.035351360321044925,
713
+ 0.03479539108276367,
714
+ 0.0348675537109375,
715
+ 0.034801727294921876,
716
+ 0.0347694091796875
717
+ ],
718
+ "mean": 0.03491708831787109,
719
+ "median": 0.034801727294921876,
720
+ "std": 0.00021953284971867288,
721
+ "min": 0.0347694091796875,
722
+ "max": 0.035351360321044925,
723
+ "p25": 0.03479539108276367,
724
+ "p75": 0.0348675537109375,
725
+ "p90": 0.03515783767700195,
726
+ "p95": 0.03525459899902344,
727
+ "p99": 0.03533200805664063,
728
+ "unit": "seconds"
729
+ },
730
+ "tokens_per_second": {
731
+ "name": "tokens_per_second",
732
+ "measurements": [
733
+ 29.32216685658783,
734
+ 29.320088899425798,
735
+ 29.29481389532554,
736
+ 29.3282470973556,
737
+ 29.311616402540093
738
+ ],
739
+ "mean": 29.31538663024697,
740
+ "median": 29.320088899425798,
741
+ "std": 0.011583837516339509,
742
+ "min": 29.29481389532554,
743
+ "max": 29.3282470973556,
744
+ "p25": 29.311616402540093,
745
+ "p75": 29.32216685658783,
746
+ "p90": 29.325815001048493,
747
+ "p95": 29.327031049202045,
748
+ "p99": 29.32800388772489,
749
+ "unit": "tokens/sec"
750
+ },
751
+ "time_per_output_token_seconds": {
752
+ "name": "time_per_output_token_seconds",
753
+ "measurements": [
754
+ 0.0341038916015625,
755
+ 0.034106308593749995,
756
+ 0.034135734863281246,
757
+ 0.0340968212890625,
758
+ 0.0341161669921875
759
+ ],
760
+ "mean": 0.03411178466796875,
761
+ "median": 0.034106308593749995,
762
+ "std": 1.3483376546694119e-05,
763
+ "min": 0.0340968212890625,
764
+ "max": 0.034135734863281246,
765
+ "p25": 0.0341038916015625,
766
+ "p75": 0.0341161669921875,
767
+ "p90": 0.034127907714843746,
768
+ "p95": 0.034131821289062496,
769
+ "p99": 0.034134952148437495,
770
+ "unit": "seconds/token"
771
+ }
772
+ },
773
+ "gpu_metrics": {
774
+ "gpu_utilization_mean": 96.94186046511628,
775
+ "gpu_utilization_max": 98,
776
+ "gpu_utilization_min": 95,
777
+ "gpu_memory_used_mean": 13849,
778
+ "gpu_memory_used_max": 13849,
779
+ "gpu_memory_used_min": 13849,
780
+ "sample_count": 86,
781
+ "gpu_monitoring_status": "success"
782
+ }
783
+ },
784
+ {
785
+ "scenario_name": "compiled_compile_max-autotune_sdpa_default",
786
+ "metadata": {
787
+ "timestamp": "2025-09-30T16:47:41.505623",
788
+ "commit_id": "16a141765c3f44a77359482c54d2c6506847e834",
789
+ "hardware_info": {
790
+ "gpu_name": "NVIDIA A10G",
791
+ "gpu_memory_total_mb": 23028,
792
+ "cpu_count": 16,
793
+ "memory_total_mb": 63607,
794
+ "python_version": "3.10.12",
795
+ "torch_version": "2.8.0+cu126",
796
+ "cuda_version": "12.6"
797
+ },
798
+ "config": {
799
+ "name": "compiled",
800
+ "model_id": "meta-llama/Llama-2-7b-hf",
801
+ "variant": "compiled",
802
+ "warmup_iterations": 3,
803
+ "measurement_iterations": 5,
804
+ "num_tokens_to_generate": 100,
805
+ "device": "cuda",
806
+ "torch_dtype": "float16",
807
+ "compile_mode": "max-autotune",
808
+ "compile_options": {},
809
+ "use_cache": true,
810
+ "batch_size": 1,
811
+ "sequence_length": null,
812
+ "attn_implementation": "sdpa",
813
+ "sdpa_backend": null,
814
+ "custom_params": {}
815
+ }
816
+ },
817
+ "measurements": {
818
+ "latency_seconds": {
819
+ "name": "latency_seconds",
820
+ "measurements": [
821
+ 3.7046064453125,
822
+ 3.70480859375,
823
+ 3.707125,
824
+ 3.711868408203125,
825
+ 3.70238427734375
826
+ ],
827
+ "mean": 3.7061585449218755,
828
+ "median": 3.70480859375,
829
+ "std": 0.0032253077751977295,
830
+ "min": 3.70238427734375,
831
+ "max": 3.711868408203125,
832
+ "p25": 3.7046064453125,
833
+ "p75": 3.707125,
834
+ "p90": 3.709971044921875,
835
+ "p95": 3.7109197265625,
836
+ "p99": 3.711678671875,
837
+ "unit": "seconds"
838
+ },
839
+ "time_to_first_token_seconds": {
840
+ "name": "time_to_first_token_seconds",
841
+ "measurements": [
842
+ 0.03805187225341797,
843
+ 0.03744492721557617,
844
+ 0.03763449478149414,
845
+ 0.03801891326904297,
846
+ 0.037760929107666014
847
+ ],
848
+ "mean": 0.03778222732543945,
849
+ "median": 0.037760929107666014,
850
+ "std": 0.00023012105762780033,
851
+ "min": 0.03744492721557617,
852
+ "max": 0.03805187225341797,
853
+ "p25": 0.03763449478149414,
854
+ "p75": 0.03801891326904297,
855
+ "p90": 0.03803868865966797,
856
+ "p95": 0.03804528045654297,
857
+ "p99": 0.038050553894042966,
858
+ "unit": "seconds"
859
+ },
860
+ "tokens_per_second": {
861
+ "name": "tokens_per_second",
862
+ "measurements": [
863
+ 26.99342061733215,
864
+ 26.991947753711127,
865
+ 26.97508176821661,
866
+ 26.940610227184457,
867
+ 27.009622045970957
868
+ ],
869
+ "mean": 26.98213648248306,
870
+ "median": 26.991947753711127,
871
+ "std": 0.02346592344296844,
872
+ "min": 26.940610227184457,
873
+ "max": 27.009622045970957,
874
+ "p25": 26.97508176821661,
875
+ "p75": 26.99342061733215,
876
+ "p90": 27.003141474515434,
877
+ "p95": 27.006381760243194,
878
+ "p99": 27.008973988825403,
879
+ "unit": "tokens/sec"
880
+ },
881
+ "time_per_output_token_seconds": {
882
+ "name": "time_per_output_token_seconds",
883
+ "measurements": [
884
+ 0.037046064453125,
885
+ 0.0370480859375,
886
+ 0.03707125,
887
+ 0.03711868408203125,
888
+ 0.0370238427734375
889
+ ],
890
+ "mean": 0.037061585449218754,
891
+ "median": 0.0370480859375,
892
+ "std": 3.225307775197615e-05,
893
+ "min": 0.0370238427734375,
894
+ "max": 0.03711868408203125,
895
+ "p25": 0.037046064453125,
896
+ "p75": 0.03707125,
897
+ "p90": 0.03709971044921875,
898
+ "p95": 0.037109197265625,
899
+ "p99": 0.037116786718749996,
900
+ "unit": "seconds/token"
901
+ }
902
+ },
903
+ "gpu_metrics": {
904
+ "gpu_utilization_mean": 90.23655913978494,
905
+ "gpu_utilization_max": 93,
906
+ "gpu_utilization_min": 87,
907
+ "gpu_memory_used_mean": 13845,
908
+ "gpu_memory_used_max": 13845,
909
+ "gpu_memory_used_min": 13845,
910
+ "sample_count": 93,
911
+ "gpu_monitoring_status": "success"
912
+ }
913
+ },
914
+ {
915
+ "scenario_name": "compiled_compile_max-autotune_sdpa_math",
916
+ "metadata": {
917
+ "timestamp": "2025-09-30T16:48:16.319708",
918
+ "commit_id": "16a141765c3f44a77359482c54d2c6506847e834",
919
+ "hardware_info": {
920
+ "gpu_name": "NVIDIA A10G",
921
+ "gpu_memory_total_mb": 23028,
922
+ "cpu_count": 16,
923
+ "memory_total_mb": 63607,
924
+ "python_version": "3.10.12",
925
+ "torch_version": "2.8.0+cu126",
926
+ "cuda_version": "12.6"
927
+ },
928
+ "config": {
929
+ "name": "compiled",
930
+ "model_id": "meta-llama/Llama-2-7b-hf",
931
+ "variant": "compiled",
932
+ "warmup_iterations": 3,
933
+ "measurement_iterations": 5,
934
+ "num_tokens_to_generate": 100,
935
+ "device": "cuda",
936
+ "torch_dtype": "float16",
937
+ "compile_mode": "max-autotune",
938
+ "compile_options": {},
939
+ "use_cache": true,
940
+ "batch_size": 1,
941
+ "sequence_length": null,
942
+ "attn_implementation": "sdpa",
943
+ "sdpa_backend": "math",
944
+ "custom_params": {}
945
+ }
946
+ },
947
+ "measurements": {
948
+ "latency_seconds": {
949
+ "name": "latency_seconds",
950
+ "measurements": [
951
+ 3.91871630859375,
952
+ 3.922378173828125,
953
+ 3.86250146484375,
954
+ 3.98304638671875,
955
+ 4.043645751953125
956
+ ],
957
+ "mean": 3.9460576171875004,
958
+ "median": 3.922378173828125,
959
+ "std": 0.061937883248283906,
960
+ "min": 3.86250146484375,
961
+ "max": 4.043645751953125,
962
+ "p25": 3.91871630859375,
963
+ "p75": 3.98304638671875,
964
+ "p90": 4.019406005859375,
965
+ "p95": 4.03152587890625,
966
+ "p99": 4.04122177734375,
967
+ "unit": "seconds"
968
+ },
969
+ "time_to_first_token_seconds": {
970
+ "name": "time_to_first_token_seconds",
971
+ "measurements": [
972
+ 0.040916576385498046,
973
+ 0.04021180725097656,
974
+ 0.040242145538330075,
975
+ 0.04011174392700195,
976
+ 0.04037337493896485
977
+ ],
978
+ "mean": 0.040371129608154296,
979
+ "median": 0.040242145538330075,
980
+ "std": 0.00028524328695942553,
981
+ "min": 0.04011174392700195,
982
+ "max": 0.040916576385498046,
983
+ "p25": 0.04021180725097656,
984
+ "p75": 0.04037337493896485,
985
+ "p90": 0.040699295806884764,
986
+ "p95": 0.040807936096191405,
987
+ "p99": 0.04089484832763672,
988
+ "unit": "seconds"
989
+ },
990
+ "tokens_per_second": {
991
+ "name": "tokens_per_second",
992
+ "measurements": [
993
+ 25.518560703335393,
994
+ 25.494737011144174,
995
+ 25.889957818836788,
996
+ 25.106411096151056,
997
+ 24.730158410068167
998
+ ],
999
+ "mean": 25.347965007907113,
1000
+ "median": 25.494737011144174,
1001
+ "std": 0.39608948583629777,
1002
+ "min": 24.730158410068167,
1003
+ "max": 25.889957818836788,
1004
+ "p25": 25.106411096151056,
1005
+ "p75": 25.518560703335393,
1006
+ "p90": 25.74139897263623,
1007
+ "p95": 25.81567839573651,
1008
+ "p99": 25.875101934216733,
1009
+ "unit": "tokens/sec"
1010
+ },
1011
+ "time_per_output_token_seconds": {
1012
+ "name": "time_per_output_token_seconds",
1013
+ "measurements": [
1014
+ 0.0391871630859375,
1015
+ 0.03922378173828125,
1016
+ 0.038625014648437504,
1017
+ 0.0398304638671875,
1018
+ 0.04043645751953125
1019
+ ],
1020
+ "mean": 0.039460576171875,
1021
+ "median": 0.03922378173828125,
1022
+ "std": 0.0006193788324828384,
1023
+ "min": 0.038625014648437504,
1024
+ "max": 0.04043645751953125,
1025
+ "p25": 0.0391871630859375,
1026
+ "p75": 0.0398304638671875,
1027
+ "p90": 0.04019406005859375,
1028
+ "p95": 0.0403152587890625,
1029
+ "p99": 0.0404122177734375,
1030
+ "unit": "seconds/token"
1031
+ }
1032
+ },
1033
+ "gpu_metrics": {
1034
+ "gpu_utilization_mean": 87.79797979797979,
1035
+ "gpu_utilization_max": 93,
1036
+ "gpu_utilization_min": 77,
1037
+ "gpu_memory_used_mean": 13845,
1038
+ "gpu_memory_used_max": 13845,
1039
+ "gpu_memory_used_min": 13845,
1040
+ "sample_count": 99,
1041
+ "gpu_monitoring_status": "success"
1042
+ }
1043
+ },
1044
+ {
1045
+ "scenario_name": "compiled_compile_max-autotune_sdpa_efficient_attention",
1046
+ "metadata": {
1047
+ "timestamp": "2025-09-30T16:48:57.651107",
1048
+ "commit_id": "16a141765c3f44a77359482c54d2c6506847e834",
1049
+ "hardware_info": {
1050
+ "gpu_name": "NVIDIA A10G",
1051
+ "gpu_memory_total_mb": 23028,
1052
+ "cpu_count": 16,
1053
+ "memory_total_mb": 63607,
1054
+ "python_version": "3.10.12",
1055
+ "torch_version": "2.8.0+cu126",
1056
+ "cuda_version": "12.6"
1057
+ },
1058
+ "config": {
1059
+ "name": "compiled",
1060
+ "model_id": "meta-llama/Llama-2-7b-hf",
1061
+ "variant": "compiled",
1062
+ "warmup_iterations": 3,
1063
+ "measurement_iterations": 5,
1064
+ "num_tokens_to_generate": 100,
1065
+ "device": "cuda",
1066
+ "torch_dtype": "float16",
1067
+ "compile_mode": "max-autotune",
1068
+ "compile_options": {},
1069
+ "use_cache": true,
1070
+ "batch_size": 1,
1071
+ "sequence_length": null,
1072
+ "attn_implementation": "sdpa",
1073
+ "sdpa_backend": "efficient_attention",
1074
+ "custom_params": {}
1075
+ }
1076
+ },
1077
+ "measurements": {
1078
+ "latency_seconds": {
1079
+ "name": "latency_seconds",
1080
+ "measurements": [
1081
+ 3.703684814453125,
1082
+ 3.707244873046875,
1083
+ 3.70141748046875,
1084
+ 3.691416015625,
1085
+ 3.697431884765625
1086
+ ],
1087
+ "mean": 3.700239013671875,
1088
+ "median": 3.70141748046875,
1089
+ "std": 0.005441868425993674,
1090
+ "min": 3.691416015625,
1091
+ "max": 3.707244873046875,
1092
+ "p25": 3.697431884765625,
1093
+ "p75": 3.703684814453125,
1094
+ "p90": 3.705820849609375,
1095
+ "p95": 3.706532861328125,
1096
+ "p99": 3.707102470703125,
1097
+ "unit": "seconds"
1098
+ },
1099
+ "time_to_first_token_seconds": {
1100
+ "name": "time_to_first_token_seconds",
1101
+ "measurements": [
1102
+ 0.03815523147583008,
1103
+ 0.03766099166870117,
1104
+ 0.03858988952636719,
1105
+ 0.037560638427734376,
1106
+ 0.03743907165527344
1107
+ ],
1108
+ "mean": 0.03788116455078125,
1109
+ "median": 0.03766099166870117,
1110
+ "std": 0.0004298983582760934,
1111
+ "min": 0.03743907165527344,
1112
+ "max": 0.03858988952636719,
1113
+ "p25": 0.037560638427734376,
1114
+ "p75": 0.03815523147583008,
1115
+ "p90": 0.038416026306152345,
1116
+ "p95": 0.03850295791625977,
1117
+ "p99": 0.0385725032043457,
1118
+ "unit": "seconds"
1119
+ },
1120
+ "tokens_per_second": {
1121
+ "name": "tokens_per_second",
1122
+ "measurements": [
1123
+ 27.00013770333902,
1124
+ 26.974209534158167,
1125
+ 27.01667686168055,
1126
+ 27.08987542360999,
1127
+ 27.04579911587441
1128
+ ],
1129
+ "mean": 27.025339727732426,
1130
+ "median": 27.01667686168055,
1131
+ "std": 0.039769150310876265,
1132
+ "min": 26.974209534158167,
1133
+ "max": 27.08987542360999,
1134
+ "p25": 27.00013770333902,
1135
+ "p75": 27.04579911587441,
1136
+ "p90": 27.072244900515756,
1137
+ "p95": 27.081060162062872,
1138
+ "p99": 27.088112371300564,
1139
+ "unit": "tokens/sec"
1140
+ },
1141
+ "time_per_output_token_seconds": {
1142
+ "name": "time_per_output_token_seconds",
1143
+ "measurements": [
1144
+ 0.03703684814453125,
1145
+ 0.03707244873046875,
1146
+ 0.0370141748046875,
1147
+ 0.03691416015625,
1148
+ 0.03697431884765625
1149
+ ],
1150
+ "mean": 0.03700239013671875,
1151
+ "median": 0.0370141748046875,
1152
+ "std": 5.441868425993764e-05,
1153
+ "min": 0.03691416015625,
1154
+ "max": 0.03707244873046875,
1155
+ "p25": 0.03697431884765625,
1156
+ "p75": 0.03703684814453125,
1157
+ "p90": 0.03705820849609375,
1158
+ "p95": 0.03706532861328125,
1159
+ "p99": 0.03707102470703125,
1160
+ "unit": "seconds/token"
1161
+ }
1162
+ },
1163
+ "gpu_metrics": {
1164
+ "gpu_utilization_mean": 90.31182795698925,
1165
+ "gpu_utilization_max": 94,
1166
+ "gpu_utilization_min": 86,
1167
+ "gpu_memory_used_mean": 13845,
1168
+ "gpu_memory_used_max": 13845,
1169
+ "gpu_memory_used_min": 13845,
1170
+ "sample_count": 93,
1171
+ "gpu_monitoring_status": "success"
1172
+ }
1173
+ }
1174
+ ]
1175
+ }