alexwengg commited on
Commit
ed61931
·
verified ·
1 Parent(s): de76e2e

benchmark-coreml: post tokenizer fix

Browse files
Files changed (1) hide show
  1. reports/benchmark-coreml.json +29 -29
reports/benchmark-coreml.json CHANGED
@@ -1,9 +1,9 @@
1
  {
2
  "chip" : "Apple M5 Pro",
3
- "elapsed_s" : 22.913529992103577,
4
  "latency_ms" : {
5
- "p50" : 5.2010420000000002,
6
- "p95" : 17.967292
7
  },
8
  "lengths" : [
9
  128,
@@ -11,7 +11,7 @@
11
  512,
12
  1024
13
  ],
14
- "load_s" : 2.2602130174636841,
15
  "questions" : 3899,
16
  "suites" : {
17
  "app.email_spam" : {
@@ -24,8 +24,8 @@
24
  },
25
  "dropped" : 0,
26
  "latency_ms" : {
27
- "p50" : 5.8082500000000001,
28
- "p95" : 18.628333000000001
29
  },
30
  "max_probability_delta_vs_reference" : 0.0070431232452392578,
31
  "n" : 400,
@@ -34,7 +34,7 @@
34
  "state_truncated" : 9
35
  },
36
  "app.guardrails_jailbreak" : {
37
- "accuracy" : 0.80500000000000005,
38
  "buckets" : {
39
  "L128" : 285,
40
  "L256" : 59,
@@ -43,12 +43,12 @@
43
  },
44
  "dropped" : 0,
45
  "latency_ms" : {
46
- "p50" : 3.8389169999999999,
47
- "p95" : 9.5504999999999995
48
  },
49
- "max_probability_delta_vs_reference" : 0.80200904607772827,
50
  "n" : 400,
51
- "reference_argmax_agreement" : 0.995,
52
  "reference_compared" : 400,
53
  "state_truncated" : 0
54
  },
@@ -61,10 +61,10 @@
61
  },
62
  "dropped" : 0,
63
  "latency_ms" : {
64
- "p50" : 5.2591669999999997,
65
- "p95" : 5.8784999999999998
66
  },
67
- "max_probability_delta_vs_reference" : 0.027976274490356445,
68
  "n" : 399,
69
  "reference_argmax_agreement" : 1,
70
  "reference_compared" : 399,
@@ -80,10 +80,10 @@
80
  },
81
  "dropped" : 0,
82
  "latency_ms" : {
83
- "p50" : 3.8188749999999998,
84
- "p95" : 9.387041
85
  },
86
- "max_probability_delta_vs_reference" : 0.0043558478355407715,
87
  "n" : 400,
88
  "reference_argmax_agreement" : 1,
89
  "reference_compared" : 400,
@@ -99,8 +99,8 @@
99
  },
100
  "dropped" : 0,
101
  "latency_ms" : {
102
- "p50" : 9.0027089999999994,
103
- "p95" : 19.462250000000001
104
  },
105
  "max_probability_delta_vs_reference" : 0.003923460841178894,
106
  "n" : 400,
@@ -117,8 +117,8 @@
117
  },
118
  "dropped" : 0,
119
  "latency_ms" : {
120
- "p50" : 5.3082500000000001,
121
- "p95" : 6.2931670000000004
122
  },
123
  "max_probability_delta_vs_reference" : 0.0088013410568237305,
124
  "n" : 400,
@@ -134,8 +134,8 @@
134
  },
135
  "dropped" : 0,
136
  "latency_ms" : {
137
- "p50" : 5.297625,
138
- "p95" : 8.8684999999999992
139
  },
140
  "max_probability_delta_vs_reference" : 0.0048961937427520752,
141
  "n" : 400,
@@ -151,8 +151,8 @@
151
  },
152
  "dropped" : 0,
153
  "latency_ms" : {
154
- "p50" : 3.7891249999999999,
155
- "p95" : 6.5880419999999997
156
  },
157
  "max_probability_delta_vs_reference" : 0.010490596294403076,
158
  "n" : 400,
@@ -167,8 +167,8 @@
167
  },
168
  "dropped" : 0,
169
  "latency_ms" : {
170
- "p50" : 3.7059579999999999,
171
- "p95" : 3.8140000000000001
172
  },
173
  "max_probability_delta_vs_reference" : 0.020582765340805054,
174
  "n" : 400,
@@ -183,8 +183,8 @@
183
  },
184
  "dropped" : 0,
185
  "latency_ms" : {
186
- "p50" : 5.2149999999999999,
187
- "p95" : 5.4784579999999998
188
  },
189
  "max_probability_delta_vs_reference" : 0.019814342260360718,
190
  "n" : 300,
 
1
  {
2
  "chip" : "Apple M5 Pro",
3
+ "elapsed_s" : 23.682847023010254,
4
  "latency_ms" : {
5
+ "p50" : 5.3144580000000001,
6
+ "p95" : 18.081707999999999
7
  },
8
  "lengths" : [
9
  128,
 
11
  512,
12
  1024
13
  ],
14
+ "load_s" : 2.8048709630966187,
15
  "questions" : 3899,
16
  "suites" : {
17
  "app.email_spam" : {
 
24
  },
25
  "dropped" : 0,
26
  "latency_ms" : {
27
+ "p50" : 6.297625,
28
+ "p95" : 18.979500000000002
29
  },
30
  "max_probability_delta_vs_reference" : 0.0070431232452392578,
31
  "n" : 400,
 
34
  "state_truncated" : 9
35
  },
36
  "app.guardrails_jailbreak" : {
37
+ "accuracy" : 0.8075,
38
  "buckets" : {
39
  "L128" : 285,
40
  "L256" : 59,
 
43
  },
44
  "dropped" : 0,
45
  "latency_ms" : {
46
+ "p50" : 4.1935419999999999,
47
+ "p95" : 9.8876670000000004
48
  },
49
+ "max_probability_delta_vs_reference" : 0.01729881763458252,
50
  "n" : 400,
51
+ "reference_argmax_agreement" : 0.99750000000000005,
52
  "reference_compared" : 400,
53
  "state_truncated" : 0
54
  },
 
61
  },
62
  "dropped" : 0,
63
  "latency_ms" : {
64
+ "p50" : 5.391,
65
+ "p95" : 6.2202919999999997
66
  },
67
+ "max_probability_delta_vs_reference" : 0.014412820339202881,
68
  "n" : 399,
69
  "reference_argmax_agreement" : 1,
70
  "reference_compared" : 399,
 
80
  },
81
  "dropped" : 0,
82
  "latency_ms" : {
83
+ "p50" : 3.8999999999999999,
84
+ "p95" : 9.4122920000000008
85
  },
86
+ "max_probability_delta_vs_reference" : 0.0036969184875488281,
87
  "n" : 400,
88
  "reference_argmax_agreement" : 1,
89
  "reference_compared" : 400,
 
99
  },
100
  "dropped" : 0,
101
  "latency_ms" : {
102
+ "p50" : 8.9892920000000007,
103
+ "p95" : 18.856666000000001
104
  },
105
  "max_probability_delta_vs_reference" : 0.003923460841178894,
106
  "n" : 400,
 
117
  },
118
  "dropped" : 0,
119
  "latency_ms" : {
120
+ "p50" : 5.4204169999999996,
121
+ "p95" : 6.3289999999999997
122
  },
123
  "max_probability_delta_vs_reference" : 0.0088013410568237305,
124
  "n" : 400,
 
134
  },
135
  "dropped" : 0,
136
  "latency_ms" : {
137
+ "p50" : 5.5369999999999999,
138
+ "p95" : 9.0276250000000005
139
  },
140
  "max_probability_delta_vs_reference" : 0.0048961937427520752,
141
  "n" : 400,
 
151
  },
152
  "dropped" : 0,
153
  "latency_ms" : {
154
+ "p50" : 4.6144170000000004,
155
+ "p95" : 7.2706249999999999
156
  },
157
  "max_probability_delta_vs_reference" : 0.010490596294403076,
158
  "n" : 400,
 
167
  },
168
  "dropped" : 0,
169
  "latency_ms" : {
170
+ "p50" : 4.0142920000000002,
171
+ "p95" : 4.3915829999999998
172
  },
173
  "max_probability_delta_vs_reference" : 0.020582765340805054,
174
  "n" : 400,
 
183
  },
184
  "dropped" : 0,
185
  "latency_ms" : {
186
+ "p50" : 5.3079169999999998,
187
+ "p95" : 6.1710409999999998
188
  },
189
  "max_probability_delta_vs_reference" : 0.019814342260360718,
190
  "n" : 300,