patrickbdevaney commited on
Commit
737f6bb
·
verified ·
1 Parent(s): 30f4987

archive auto-ace0p3: eval.log

Browse files
Files changed (1) hide show
  1. arms/auto-ace0p3/eval.log +1097 -0
arms/auto-ace0p3/eval.log ADDED
@@ -0,0 +1,1097 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [decode] loading /home/patrickd/models/DeepSeek-V4-Flash-0731-REAP ... s=6 NDEC=8 seqmax=297
2
+ [decode] loaded 100.40 GiB, 45821 tensors (weights in MANAGED device-preferred memory)
3
+ [decode] KV layout: fp32, 2048 B/row, 0.024 GiB for 297 rows x 43 layers
4
+ [decode] building 43 layer structs once (persistent)...
5
+ [decode] structs built. mem 106.7/122.8 GiB
6
+ [decode] prefill 5 positions...
7
+ [f32mk] tile 4x4 (DSV4_F32MK_TILE=unset)
8
+ [decode] PREFILL: 5 positions in 202.0 ms = 24.8 tok/s (40.403 ms/tok)
9
+ [decode] prefill done. caches populated. starting decode.
10
+ step 0 pos 5 -> token 11111 (69.5 ms warmup)
11
+ step 1 pos 6 -> token 16 (69.6 ms)
12
+ step 2 pos 7 -> token 455 (76.1 ms)
13
+ step 3 pos 8 -> token 6102 (69.2 ms)
14
+ step 4 pos 9 -> token 294 (70.0 ms)
15
+ step 5 pos 10 -> token 16603 (70.8 ms)
16
+ step 6 pos 11 -> token 344 (75.3 ms)
17
+ step 7 pos 12 -> token 29168 (69.5 ms)
18
+
19
+ [decode] first decoded token argmax = 11111 (expect 11111) -> GATE PASS
20
+ [decode] generated: 11111 16 455 6102 294 16603 344 29168
21
+ [decode] WARM decode: 71.5 ms/tok = 13.99 tok/s (M=1 steady state, 7-step avg)
22
+
23
+ [spec-verify] M=5 verify in ONE forward: 121.3 ms (= 24.3 ms/tok if all accepted vs 71.5 M=1 -> 2.95x)
24
+ [spec-verify] verify argmax: 11111 16 455 6102 294
25
+ [spec-verify] decode tokens: 11111 16 455 6102 294
26
+ [spec-verify] MATCH 5/5 -> PASS (M=K verify == K sequential decodes; diffs = MoE-atomic near-ties)
27
+
28
+ [spec] loading EXTERNAL DSpark head /home/patrickd/s5-capture/auto-ace0p3/head ...
29
+ [spec] head loaded 6.53 GiB, 2977 mtp tensors (SEPARATE store)
30
+ [spec] NSTAGE=3 head-experts=160 BLK sweep: 5 5 5 5 5 5 5 5 5
31
+ [spec] head built. mem 112.9/122.8 GiB
32
+ [decode] PREFILL: 5 positions in 141.0 ms = 35.5 tok/s (28.198 ms/tok)
33
+ [spec] decoding 200 tokens (block=5, draft passes=1, adaptK=1.50, prompt=0 s=6)...
34
+ [arena] first draft: draft footprint 210.72 MB, global hwm 210.72 MB / cap 512.00 MB (41.2%), draft path = ARENA
35
+ [margins] 1.93 0.92 0.64 0.66 0.65 K=2
36
+ verify 1: accepted 1/1 (K=2) + correction -> +2 tokens (108.3 ms) cpos=7
37
+ [margins] 0.05 0.67 0.35 0.66 0.11 K=2
38
+ verify 2: accepted 0/1 (K=2) + correction -> +1 tokens (114.9 ms) cpos=8
39
+ [margins] 1.48 0.41 1.95 0.46 0.13 K=2
40
+ verify 3: accepted 1/1 (K=2) + correction -> +2 tokens (109.4 ms) cpos=10
41
+ [margins] 2.02 0.33 3.81 0.58 0.78 K=2
42
+ verify 4: accepted 0/1 (K=2) + correction -> +1 tokens (102.2 ms) cpos=11
43
+ [margins] 6.43 0.16 0.79 1.83 2.42 K=2
44
+ verify 5: accepted 1/1 (K=2) + correction -> +2 tokens (102.9 ms) cpos=13
45
+ [margins] 4.25 2.68 5.44 2.00 0.37 K=5
46
+ verify 6: accepted 4/4 (K=5) + correction -> +5 tokens (140.4 ms) cpos=18
47
+ [margins] 4.69 1.79 2.77 0.68 0.60 K=4
48
+ verify 7: accepted 1/3 (K=4) + correction -> +2 tokens (126.5 ms) cpos=20
49
+ [margins] 2.63 2.42 5.03 1.52 1.11 K=5
50
+ verify 8: accepted 4/4 (K=5) + correction -> +5 tokens (139.4 ms) cpos=25
51
+ [margins] 6.57 1.87 2.30 2.74 3.31 K=6
52
+ verify 9: accepted 1/5 (K=6) + correction -> +2 tokens (157.9 ms) cpos=27
53
+ [margins] 3.73 3.49 6.10 2.42 0.32 K=5
54
+ verify 10: accepted 4/4 (K=5) + correction -> +5 tokens (145.0 ms) cpos=32
55
+ [margins] 6.85 1.06 3.63 1.80 2.37 K=2
56
+ verify 11: accepted 1/1 (K=2) + correction -> +2 tokens (96.2 ms) cpos=34
57
+ [margins] 2.16 4.09 4.96 1.59 0.31 K=5
58
+ verify 12: accepted 4/4 (K=5) + correction -> +5 tokens (140.8 ms) cpos=39
59
+ [margins] 4.35 0.52 4.48 3.03 3.05 K=2
60
+ verify 13: accepted 1/1 (K=2) + correction -> +2 tokens (101.2 ms) cpos=41
61
+ [margins] 3.73 3.93 6.36 1.35 1.67 K=4
62
+ verify 14: accepted 3/3 (K=4) + correction -> +4 tokens (126.7 ms) cpos=45
63
+ [margins] 2.82 0.52 6.46 1.69 4.32 K=2
64
+ verify 15: accepted 1/1 (K=2) + correction -> +2 tokens (97.0 ms) cpos=47
65
+ [margins] 0.93 4.06 0.13 0.11 0.23 K=3
66
+ verify 16: accepted 2/2 (K=3) + correction -> +3 tokens (112.5 ms) cpos=50
67
+ [margins] 4.61 3.94 4.57 0.85 2.11 K=4
68
+ verify 17: accepted 3/3 (K=4) + correction -> +4 tokens (131.1 ms) cpos=54
69
+ [margins] 2.60 1.52 6.38 0.04 2.93 K=4
70
+ verify 18: accepted 3/3 (K=4) + correction -> +4 tokens (128.7 ms) cpos=58
71
+ [margins] 0.19 5.65 5.71 6.51 1.08 K=5
72
+ verify 19: accepted 0/4 (K=5) + correction -> +1 tokens (141.9 ms) cpos=59
73
+ [margins] 5.04 6.74 6.69 0.35 4.25 K=4
74
+ verify 20: accepted 3/3 (K=4) + correction -> +4 tokens (129.8 ms) cpos=63
75
+ [margins] 0.51 0.79 6.53 0.27 3.63 K=2
76
+ verify 21: accepted 0/1 (K=2) + correction -> +1 tokens (102.5 ms) cpos=64
77
+ [margins] 6.57 0.72 2.93 2.08 4.32 K=2
78
+ verify 22: accepted 1/1 (K=2) + correction -> +2 tokens (98.3 ms) cpos=66
79
+ [margins] 4.62 5.43 5.69 2.10 0.40 K=5
80
+ verify 23: accepted 4/4 (K=5) + correction -> +5 tokens (142.8 ms) cpos=71
81
+ [margins] 6.85 1.66 2.43 1.03 2.12 K=4
82
+ verify 24: accepted 3/3 (K=4) + correction -> +4 tokens (130.0 ms) cpos=75
83
+ [margins] 8.05 6.01 2.54 0.45 0.76 K=4
84
+ verify 25: accepted 2/3 (K=4) + correction -> +3 tokens (128.2 ms) cpos=78
85
+ [margins] 5.87 0.10 1.20 4.63 2.70 K=2
86
+ verify 26: accepted 1/1 (K=2) + correction -> +2 tokens (103.8 ms) cpos=80
87
+ [margins] 5.01 6.46 5.65 2.71 0.20 K=5
88
+ verify 27: accepted 4/4 (K=5) + correction -> +5 tokens (144.0 ms) cpos=85
89
+ [margins] 7.40 1.13 1.23 3.62 4.00 K=2
90
+ verify 28: accepted 1/1 (K=2) + correction -> +2 tokens (99.7 ms) cpos=87
91
+ [margins] 4.98 6.44 7.88 1.72 2.67 K=6
92
+ verify 29: accepted 4/5 (K=6) + correction -> +5 tokens (166.6 ms) cpos=92
93
+ [margins] 6.92 0.53 0.52 3.07 4.43 K=2
94
+ verify 30: accepted 1/1 (K=2) + correction -> +2 tokens (101.2 ms) cpos=94
95
+ [margins] 5.15 7.90 5.26 3.04 2.19 K=6
96
+ verify 31: accepted 4/5 (K=6) + correction -> +5 tokens (167.9 ms) cpos=99
97
+ [margins] 5.84 2.05 0.90 2.60 4.06 K=3
98
+ verify 32: accepted 1/2 (K=3) + correction -> +2 tokens (119.0 ms) cpos=101
99
+ [margins] 6.20 6.01 5.59 1.14 2.39 K=4
100
+ verify 33: accepted 3/3 (K=4) + correction -> +4 tokens (131.4 ms) cpos=105
101
+ [margins] 1.38 6.86 0.38 5.61 3.63 K=3
102
+ verify 34: accepted 2/2 (K=3) + correction -> +3 tokens (118.1 ms) cpos=108
103
+ [margins] 6.78 5.31 6.11 2.91 1.73 K=6
104
+ verify 35: accepted 4/5 (K=6) + correction -> +5 tokens (162.8 ms) cpos=113
105
+ [margins] 7.25 1.26 0.66 3.71 2.64 K=2
106
+ verify 36: accepted 1/1 (K=2) + correction -> +2 tokens (102.1 ms) cpos=115
107
+ [margins] 4.18 3.23 7.28 1.93 0.92 K=5
108
+ verify 37: accepted 4/4 (K=5) + correction -> +5 tokens (149.7 ms) cpos=120
109
+ [margins] 6.96 1.26 0.44 2.16 2.45 K=2
110
+ verify 38: accepted 1/1 (K=2) + correction -> +2 tokens (101.9 ms) cpos=122
111
+ [margins] 5.72 5.36 5.66 4.14 1.10 K=5
112
+ verify 39: accepted 4/4 (K=5) + correction -> +5 tokens (145.4 ms) cpos=127
113
+ [margins] 8.02 0.83 0.86 4.16 4.46 K=2
114
+ verify 40: accepted 1/1 (K=2) + correction -> +2 tokens (119.4 ms) cpos=129
115
+ [margins] 7.47 6.61 7.95 1.66 1.77 K=6
116
+ verify 41: accepted 4/5 (K=6) + correction -> +5 tokens (163.4 ms) cpos=134
117
+ [margins] 7.72 1.35 0.93 2.07 3.64 K=2
118
+ verify 42: accepted 1/1 (K=2) + correction -> +2 tokens (107.9 ms) cpos=136
119
+ [margins] 7.30 6.74 6.18 2.21 0.63 K=5
120
+ verify 43: accepted 4/4 (K=5) + correction -> +5 tokens (144.8 ms) cpos=141
121
+ [margins] 5.16 7.66 3.96 2.09 3.82 K=6
122
+ verify 44: accepted 3/5 (K=6) + correction -> +4 tokens (158.9 ms) cpos=145
123
+ [margins] 6.87 6.13 7.75 1.75 1.49 K=5
124
+ verify 45: accepted 4/4 (K=5) + correction -> +5 tokens (144.3 ms) cpos=150
125
+ [margins] 6.68 0.42 1.62 5.83 5.12 K=2
126
+ verify 46: accepted 1/1 (K=2) + correction -> +2 tokens (105.3 ms) cpos=152
127
+ [margins] 15.19 2.99 6.72 7.16 9.43 K=6
128
+ verify 47: accepted 1/5 (K=6) + correction -> +2 tokens (161.7 ms) cpos=154
129
+ [margins] 3.83 8.25 7.69 4.74 2.41 K=6
130
+ verify 48: accepted 4/5 (K=6) + correction -> +5 tokens (166.5 ms) cpos=159
131
+ [margins] 4.17 0.64 3.32 2.54 5.93 K=2
132
+ verify 49: accepted 1/1 (K=2) + correction -> +2 tokens (105.6 ms) cpos=161
133
+ [margins] 9.98 6.07 8.08 7.74 3.44 K=6
134
+ verify 50: accepted 5/5 (K=6) + correction -> +6 tokens (163.6 ms) cpos=167
135
+ [margins] 6.60 0.93 2.96 4.83 4.44 K=2
136
+ verify 51: accepted 1/1 (K=2) + correction -> +2 tokens (105.4 ms) cpos=169
137
+ [margins] 2.65 5.60 6.49 3.43 2.64 K=6
138
+ verify 52: accepted 4/5 (K=6) + correction -> +5 tokens (160.9 ms) cpos=174
139
+ [margins] 4.21 1.72 1.07 4.00 5.38 K=3
140
+ verify 53: accepted 1/2 (K=3) + correction -> +2 tokens (117.6 ms) cpos=176
141
+ [margins] 4.27 5.34 7.41 2.14 0.20 K=5
142
+ verify 54: accepted 4/4 (K=5) + correction -> +5 tokens (144.8 ms) cpos=181
143
+ [margins] 6.34 1.36 0.18 1.07 6.11 K=2
144
+ verify 55: accepted 1/1 (K=2) + correction -> +2 tokens (102.4 ms) cpos=183
145
+ [margins] 3.64 6.53 8.43 3.05 0.91 K=5
146
+ verify 56: accepted 4/4 (K=5) + correction -> +5 tokens (151.5 ms) cpos=188
147
+ [margins] 1.25 0.14 2.21 0.20 6.22 K=2
148
+ verify 57: accepted 1/1 (K=2) + correction -> +2 tokens (102.6 ms) cpos=190
149
+ [margins] 8.40 2.21 0.37 3.60 2.92 K=3
150
+ verify 58: accepted 1/2 (K=3) + correction -> +2 tokens (118.7 ms) cpos=192
151
+ [margins] 5.61 2.78 5.98 5.72 8.63 K=6
152
+ verify 59: accepted 5/5 (K=6) + correction -> +6 tokens (162.6 ms) cpos=198
153
+ [margins] 1.03 5.61 0.78 1.26 3.18 K=3
154
+ verify 60: accepted 2/2 (K=3) + correction -> +3 tokens (119.6 ms) cpos=201
155
+ [margins] 6.20 7.65 8.81 7.77 1.15 K=5
156
+ verify 61: accepted 4/4 (K=5) + correction -> +5 tokens (144.9 ms) cpos=206
157
+
158
+ [spec] generated 201 tokens over 61 verifies: mean tokens/verify = 3.30 (block=5, max 6)
159
+ [spec] tokens: 11111 16 455 6102 294 16603 344 29168 16 455 6102 294 14251 344 16235 16 455 6102 294 29585 344 76405 16 455 6102 294 25062 344 40977 16 455 6102 294 10322 344 17575 16 455 6102 294
160
+ [spec] LOSSLESS GATE: first 8 tokens match base AR -> PASS
161
+ [spec] SPEC-DECODE: 38.9 ms/tok = 25.74 tok/s (vs base M=1 71.5 ms/tok = 13.99 tok/s -> 1.84x)
162
+ [decode] PREFILL: 34 positions in 605.2 ms = 56.2 tok/s (17.800 ms/tok)
163
+ [spec] decoding 200 tokens (block=5, draft passes=1, adaptK=1.50, prompt=1 s=35)...
164
+ [margins] 1.60 1.07 1.36 2.10 0.21 K=2
165
+ verify 1: accepted 0/1 (K=2) + correction -> +1 tokens (104.2 ms) cpos=35
166
+ [margins] 3.64 2.99 1.47 1.93 4.56 K=3
167
+ verify 2: accepted 0/2 (K=3) + correction -> +1 tokens (111.8 ms) cpos=36
168
+ [margins] 1.06 1.31 0.28 3.26 0.88 K=2
169
+ verify 3: accepted 0/1 (K=2) + correction -> +1 tokens (95.5 ms) cpos=37
170
+ [margins] 1.74 2.40 1.85 1.57 0.67 K=5
171
+ verify 4: accepted 2/4 (K=5) + correction -> +3 tokens (135.9 ms) cpos=40
172
+ [margins] 0.09 0.42 1.15 0.33 0.80 K=2
173
+ verify 5: accepted 0/1 (K=2) + correction -> +1 tokens (97.8 ms) cpos=41
174
+ [margins] 1.14 4.48 0.14 0.47 0.49 K=3
175
+ verify 6: accepted 2/2 (K=3) + correction -> +3 tokens (112.3 ms) cpos=44
176
+ [margins] 0.13 5.55 10.60 2.42 0.25 K=5
177
+ verify 7: accepted 4/4 (K=5) + correction -> +5 tokens (138.0 ms) cpos=49
178
+ [margins] 5.97 1.34 1.61 0.47 0.51 K=2
179
+ verify 8: accepted 1/1 (K=2) + correction -> +2 tokens (95.7 ms) cpos=51
180
+ [margins] 0.96 1.11 3.81 4.85 3.42 K=2
181
+ verify 9: accepted 0/1 (K=2) + correction -> +1 tokens (101.4 ms) cpos=52
182
+ [margins] 3.89 4.60 6.66 1.39 1.12 K=4
183
+ verify 10: accepted 3/3 (K=4) + correction -> +4 tokens (127.5 ms) cpos=56
184
+ [margins] 3.45 1.31 0.47 1.57 2.10 K=2
185
+ verify 11: accepted 1/1 (K=2) + correction -> +2 tokens (102.0 ms) cpos=58
186
+ [margins] 2.47 3.06 3.27 0.14 3.52 K=4
187
+ verify 12: accepted 0/3 (K=4) + correction -> +1 tokens (116.1 ms) cpos=59
188
+ [margins] 0.75 0.49 3.11 0.24 1.30 K=2
189
+ verify 13: accepted 0/1 (K=2) + correction -> +1 tokens (102.0 ms) cpos=60
190
+ [margins] 0.77 1.68 5.36 0.59 1.53 K=4
191
+ verify 14: accepted 0/3 (K=4) + correction -> +1 tokens (126.5 ms) cpos=61
192
+ [margins] 0.92 3.89 0.44 2.35 0.05 K=3
193
+ verify 15: accepted 1/2 (K=3) + correction -> +2 tokens (114.7 ms) cpos=63
194
+ [margins] 1.94 4.97 0.08 1.67 0.90 K=3
195
+ verify 16: accepted 0/2 (K=3) + correction -> +1 tokens (115.9 ms) cpos=64
196
+ [margins] 4.68 0.31 0.60 6.94 0.57 K=2
197
+ verify 17: accepted 1/1 (K=2) + correction -> +2 tokens (97.9 ms) cpos=66
198
+ [margins] 1.32 1.46 3.17 4.97 0.32 K=2
199
+ verify 18: accepted 1/1 (K=2) + correction -> +2 tokens (104.4 ms) cpos=68
200
+ [margins] 0.47 6.48 0.16 0.70 1.07 K=3
201
+ verify 19: accepted 2/2 (K=3) + correction -> +3 tokens (112.0 ms) cpos=71
202
+ [margins] 2.69 1.67 1.20 4.73 0.73 K=3
203
+ verify 20: accepted 2/2 (K=3) + correction -> +3 tokens (115.1 ms) cpos=74
204
+ [margins] 7.06 1.81 3.10 2.73 3.06 K=6
205
+ verify 21: accepted 5/5 (K=6) + correction -> +6 tokens (161.1 ms) cpos=80
206
+ [margins] 5.36 5.12 4.45 2.26 6.47 K=6
207
+ verify 22: accepted 5/5 (K=6) + correction -> +6 tokens (160.3 ms) cpos=86
208
+ [margins] 7.59 4.01 2.39 5.32 0.11 K=5
209
+ verify 23: accepted 4/4 (K=5) + correction -> +5 tokens (144.2 ms) cpos=91
210
+ [margins] 6.64 4.12 3.50 8.32 1.08 K=5
211
+ verify 24: accepted 4/4 (K=5) + correction -> +5 tokens (146.7 ms) cpos=96
212
+ [margins] 8.29 4.87 4.10 8.52 2.45 K=6
213
+ verify 25: accepted 5/5 (K=6) + correction -> +6 tokens (158.8 ms) cpos=102
214
+ [margins] 9.99 3.21 6.98 7.27 5.90 K=6
215
+ verify 26: accepted 5/5 (K=6) + correction -> +6 tokens (165.7 ms) cpos=108
216
+ [margins] 2.56 1.49 0.25 6.66 3.55 K=2
217
+ verify 27: accepted 0/1 (K=2) + correction -> +1 tokens (102.4 ms) cpos=109
218
+ [margins] 8.92 5.87 5.12 6.71 5.54 K=6
219
+ verify 28: accepted 5/5 (K=6) + correction -> +6 tokens (160.0 ms) cpos=115
220
+ [margins] 8.98 8.48 8.71 2.75 6.27 K=6
221
+ verify 29: accepted 5/5 (K=6) + correction -> +6 tokens (167.5 ms) cpos=121
222
+ [margins] 8.40 4.15 5.24 7.55 0.58 K=5
223
+ verify 30: accepted 4/4 (K=5) + correction -> +5 tokens (147.6 ms) cpos=126
224
+ [margins] 7.45 5.62 6.15 1.06 0.92 K=4
225
+ verify 31: accepted 3/3 (K=4) + correction -> +4 tokens (148.9 ms) cpos=130
226
+ [margins] 8.17 7.08 4.84 2.92 8.96 K=6
227
+ verify 32: accepted 5/5 (K=6) + correction -> +6 tokens (166.6 ms) cpos=136
228
+ [margins] 6.98 4.75 6.45 7.84 2.48 K=6
229
+ verify 33: accepted 5/5 (K=6) + correction -> +6 tokens (162.5 ms) cpos=142
230
+ [margins] 12.01 5.59 9.42 10.23 8.38 K=6
231
+ verify 34: accepted 5/5 (K=6) + correction -> +6 tokens (168.7 ms) cpos=148
232
+ [margins] 2.09 0.06 7.71 4.32 4.21 K=2
233
+ verify 35: accepted 0/1 (K=2) + correction -> +1 tokens (103.7 ms) cpos=149
234
+ [margins] 8.92 9.79 7.62 8.30 4.38 K=6
235
+ verify 36: accepted 5/5 (K=6) + correction -> +6 tokens (161.1 ms) cpos=155
236
+ [margins] 9.86 7.86 8.09 2.48 7.58 K=6
237
+ verify 37: accepted 5/5 (K=6) + correction -> +6 tokens (168.0 ms) cpos=161
238
+ [margins] 8.89 6.57 5.07 8.25 6.92 K=6
239
+ verify 38: accepted 5/5 (K=6) + correction -> +6 tokens (162.0 ms) cpos=167
240
+ [margins] 13.32 5.42 10.41 9.78 10.07 K=6
241
+ verify 39: accepted 5/5 (K=6) + correction -> +6 tokens (168.6 ms) cpos=173
242
+ [margins] 2.97 6.92 9.69 6.66 6.12 K=6
243
+ verify 40: accepted 5/5 (K=6) + correction -> +6 tokens (160.6 ms) cpos=179
244
+ [margins] 10.23 10.13 6.61 8.05 5.89 K=6
245
+ verify 41: accepted 5/5 (K=6) + correction -> +6 tokens (165.8 ms) cpos=185
246
+ [margins] 10.29 9.82 9.46 7.67 8.46 K=6
247
+ verify 42: accepted 5/5 (K=6) + correction -> +6 tokens (161.8 ms) cpos=191
248
+ [margins] 11.29 8.92 5.59 8.81 0.86 K=5
249
+ verify 43: accepted 4/4 (K=5) + correction -> +5 tokens (152.2 ms) cpos=196
250
+ [margins] 11.22 6.95 9.87 9.22 5.71 K=6
251
+ verify 44: accepted 5/5 (K=6) + correction -> +6 tokens (163.2 ms) cpos=202
252
+ [margins] 13.47 6.09 8.01 6.30 6.39 K=6
253
+ verify 45: accepted 5/5 (K=6) + correction -> +6 tokens (169.1 ms) cpos=208
254
+ [margins] 3.58 1.04 0.21 0.73 0.35 K=2
255
+ verify 46: accepted 0/1 (K=2) + correction -> +1 tokens (104.2 ms) cpos=209
256
+ [margins] 8.19 6.21 6.88 8.44 8.08 K=6
257
+ verify 47: accepted 5/5 (K=6) + correction -> +6 tokens (162.7 ms) cpos=215
258
+ [margins] 8.26 11.85 8.64 8.71 8.11 K=6
259
+ verify 48: accepted 5/5 (K=6) + correction -> +6 tokens (169.9 ms) cpos=221
260
+ [margins] 9.15 3.50 9.40 9.32 0.85 K=5
261
+ verify 49: accepted 4/4 (K=5) + correction -> +5 tokens (148.9 ms) cpos=226
262
+ [margins] 12.10 4.67 6.82 6.47 5.41 K=6
263
+ verify 50: accepted 5/5 (K=6) + correction -> +6 tokens (168.6 ms) cpos=232
264
+ [margins] 12.88 5.61 10.76 8.86 8.30 K=6
265
+ verify 51: accepted 5/5 (K=6) + correction -> +6 tokens (165.5 ms) cpos=238
266
+
267
+ [spec] generated 204 tokens over 51 verifies: mean tokens/verify = 4.00 (block=5, max 6)
268
+ [spec] tokens: 4147 554 1926 270 4105 339 43 1114 1926 270 1178 65 50219 4105 418 270 9517 28 4593 1281 106239 3305 1 1 14 223 19 16 23 14 223 20 16 23 14 223 21 16 23 14
269
+ [spec] SPEC-DECODE: 34.2 ms/tok = 29.25 tok/s (vs base M=1 71.5 ms/tok = 13.99 tok/s -> 2.09x)
270
+ [decode] PREFILL: 41 positions in 700.1 ms = 58.6 tok/s (17.074 ms/tok)
271
+ [spec] decoding 200 tokens (block=5, draft passes=1, adaptK=1.50, prompt=2 s=42)...
272
+ [margins] 7.63 0.52 0.05 1.04 2.60 K=2
273
+ verify 1: accepted 1/1 (K=2) + correction -> +2 tokens (100.3 ms) cpos=43
274
+ [margins] 0.94 3.85 1.60 1.39 0.76 K=4
275
+ verify 2: accepted 3/3 (K=4) + correction -> +4 tokens (115.0 ms) cpos=47
276
+ [margins] 0.04 7.27 0.83 0.01 0.52 K=3
277
+ verify 3: accepted 0/2 (K=3) + correction -> +1 tokens (109.6 ms) cpos=48
278
+ [margins] 3.93 5.88 0.22 0.46 1.38 K=3
279
+ verify 4: accepted 0/2 (K=3) + correction -> +1 tokens (109.5 ms) cpos=49
280
+ [margins] 3.34 8.00 0.04 0.29 0.19 K=3
281
+ verify 5: accepted 0/2 (K=3) + correction -> +1 tokens (114.7 ms) cpos=50
282
+ [margins] 1.99 4.83 2.74 0.50 0.89 K=4
283
+ verify 6: accepted 0/3 (K=4) + correction -> +1 tokens (115.0 ms) cpos=51
284
+ [margins] 7.10 4.78 1.16 0.14 0.65 K=3
285
+ verify 7: accepted 1/2 (K=3) + correction -> +2 tokens (110.4 ms) cpos=53
286
+ [margins] 0.61 2.78 6.75 1.36 0.45 K=4
287
+ verify 8: accepted 0/3 (K=4) + correction -> +1 tokens (126.8 ms) cpos=54
288
+ [margins] 2.22 7.66 2.58 1.83 0.46 K=5
289
+ verify 9: accepted 2/4 (K=5) + correction -> +3 tokens (133.2 ms) cpos=57
290
+ [margins] 0.59 2.33 9.00 0.47 0.09 K=4
291
+ verify 10: accepted 0/3 (K=4) + correction -> +1 tokens (126.8 ms) cpos=58
292
+ [margins] 4.30 5.35 2.99 5.67 0.03 K=5
293
+ verify 11: accepted 2/4 (K=5) + correction -> +3 tokens (133.9 ms) cpos=61
294
+ [margins] 4.39 3.48 7.82 2.79 6.80 K=6
295
+ verify 12: accepted 3/5 (K=6) + correction -> +4 tokens (153.2 ms) cpos=65
296
+ [margins] 2.03 2.56 5.44 2.58 7.31 K=6
297
+ verify 13: accepted 3/5 (K=6) + correction -> +4 tokens (152.6 ms) cpos=69
298
+ [margins] 5.39 5.48 4.44 0.55 4.38 K=4
299
+ verify 14: accepted 2/3 (K=4) + correction -> +3 tokens (129.4 ms) cpos=72
300
+ [margins] 5.65 0.38 0.94 3.60 1.82 K=2
301
+ verify 15: accepted 1/1 (K=2) + correction -> +2 tokens (97.8 ms) cpos=74
302
+ [margins] 3.57 5.26 2.81 0.19 4.74 K=4
303
+ verify 16: accepted 1/3 (K=4) + correction -> +2 tokens (126.3 ms) cpos=76
304
+ [margins] 9.91 4.64 5.23 4.47 0.08 K=5
305
+ verify 17: accepted 2/4 (K=5) + correction -> +3 tokens (140.5 ms) cpos=79
306
+ [margins] 15.17 10.36 6.06 4.71 0.15 K=5
307
+ verify 18: accepted 4/4 (K=5) + correction -> +5 tokens (145.6 ms) cpos=84
308
+ [margins] 18.66 8.90 2.63 0.52 1.52 K=4
309
+ verify 19: accepted 2/3 (K=4) + correction -> +3 tokens (130.5 ms) cpos=87
310
+ [margins] 11.69 14.22 13.29 2.52 9.96 K=6
311
+ verify 20: accepted 5/5 (K=6) + correction -> +6 tokens (161.9 ms) cpos=93
312
+ [margins] 9.08 9.86 11.82 8.41 2.16 K=6
313
+ verify 21: accepted 4/5 (K=6) + correction -> +5 tokens (157.2 ms) cpos=98
314
+ [margins] 9.33 6.98 5.51 3.39 0.55 K=5
315
+ verify 22: accepted 4/4 (K=5) + correction -> +5 tokens (141.1 ms) cpos=103
316
+ [margins] 15.71 13.14 10.74 2.43 2.50 K=6
317
+ verify 23: accepted 5/5 (K=6) + correction -> +6 tokens (164.6 ms) cpos=109
318
+ [margins] 10.64 2.88 5.11 7.86 4.22 K=6
319
+ verify 24: accepted 5/5 (K=6) + correction -> +6 tokens (154.7 ms) cpos=115
320
+ [margins] 10.91 9.05 10.73 0.49 2.78 K=4
321
+ verify 25: accepted 3/3 (K=4) + correction -> +4 tokens (130.9 ms) cpos=119
322
+ [margins] 4.35 7.66 0.18 3.26 1.24 K=3
323
+ verify 26: accepted 2/2 (K=3) + correction -> +3 tokens (116.7 ms) cpos=122
324
+ [margins] 14.70 4.43 2.79 9.44 0.99 K=5
325
+ verify 27: accepted 4/4 (K=5) + correction -> +5 tokens (143.7 ms) cpos=127
326
+ [margins] 17.83 11.05 9.43 8.30 9.28 K=6
327
+ verify 28: accepted 5/5 (K=6) + correction -> +6 tokens (181.9 ms) cpos=133
328
+ [margins] 8.59 6.01 3.38 2.32 0.13 K=5
329
+ verify 29: accepted 4/4 (K=5) + correction -> +5 tokens (142.2 ms) cpos=138
330
+ [margins] 15.82 11.90 11.61 5.37 0.48 K=5
331
+ verify 30: accepted 4/4 (K=5) + correction -> +5 tokens (143.7 ms) cpos=143
332
+ [margins] 16.19 9.83 5.73 3.33 2.50 K=6
333
+ verify 31: accepted 5/5 (K=6) + correction -> +6 tokens (166.3 ms) cpos=149
334
+ [margins] 13.36 6.37 6.06 0.48 1.07 K=4
335
+ verify 32: accepted 3/3 (K=4) + correction -> +4 tokens (133.2 ms) cpos=153
336
+ [margins] 1.77 8.37 12.09 5.87 5.35 K=6
337
+ verify 33: accepted 4/5 (K=6) + correction -> +5 tokens (161.2 ms) cpos=158
338
+ [margins] 11.71 6.01 8.84 0.26 0.14 K=4
339
+ verify 34: accepted 3/3 (K=4) + correction -> +4 tokens (129.9 ms) cpos=162
340
+ [margins] 14.07 0.50 2.52 4.30 2.38 K=2
341
+ verify 35: accepted 1/1 (K=2) + correction -> +2 tokens (103.9 ms) cpos=164
342
+ [margins] 11.92 8.27 1.67 1.43 0.27 K=4
343
+ verify 36: accepted 2/3 (K=4) + correction -> +3 tokens (132.3 ms) cpos=167
344
+ [margins] 6.88 12.56 10.51 6.56 2.40 K=6
345
+ verify 37: accepted 4/5 (K=6) + correction -> +5 tokens (165.7 ms) cpos=172
346
+ [margins] 15.01 6.26 10.90 6.92 1.10 K=5
347
+ verify 38: accepted 4/4 (K=5) + correction -> +5 tokens (145.0 ms) cpos=177
348
+ [margins] 14.06 14.79 12.82 9.07 3.73 K=6
349
+ verify 39: accepted 5/5 (K=6) + correction -> +6 tokens (160.9 ms) cpos=183
350
+ [margins] 13.52 9.32 8.85 0.23 5.70 K=4
351
+ verify 40: accepted 3/3 (K=4) + correction -> +4 tokens (133.0 ms) cpos=187
352
+ [margins] 9.98 12.12 9.52 5.36 0.18 K=5
353
+ verify 41: accepted 4/4 (K=5) + correction -> +5 tokens (153.0 ms) cpos=192
354
+ [margins] 12.44 6.62 3.37 6.49 0.89 K=5
355
+ verify 42: accepted 4/4 (K=5) + correction -> +5 tokens (146.7 ms) cpos=197
356
+ [margins] 16.41 12.71 1.04 6.22 3.33 K=3
357
+ verify 43: accepted 2/2 (K=3) + correction -> +3 tokens (121.8 ms) cpos=200
358
+ [margins] 14.67 10.67 11.10 5.62 6.81 K=6
359
+ verify 44: accepted 5/5 (K=6) + correction -> +6 tokens (163.6 ms) cpos=206
360
+ [margins] 3.76 14.05 0.74 3.92 2.47 K=3
361
+ verify 45: accepted 2/2 (K=3) + correction -> +3 tokens (118.7 ms) cpos=209
362
+ [margins] 15.82 3.46 12.17 6.55 4.42 K=6
363
+ verify 46: accepted 3/5 (K=6) + correction -> +4 tokens (157.3 ms) cpos=213
364
+ [margins] 20.32 11.75 6.56 12.19 2.57 K=6
365
+ verify 47: accepted 5/5 (K=6) + correction -> +6 tokens (160.2 ms) cpos=219
366
+ [margins] 20.76 9.90 20.83 5.92 13.90 K=6
367
+ verify 48: accepted 5/5 (K=6) + correction -> +6 tokens (170.2 ms) cpos=225
368
+ [margins] 7.58 24.62 6.60 14.43 10.47 K=6
369
+ verify 49: accepted 5/5 (K=6) + correction -> +6 tokens (163.3 ms) cpos=231
370
+ [margins] 27.65 11.21 14.92 2.21 6.42 K=6
371
+ verify 50: accepted 5/5 (K=6) + correction -> +6 tokens (169.6 ms) cpos=237
372
+ [margins] 12.75 11.89 5.14 1.96 6.77 K=6
373
+ verify 51: accepted 5/5 (K=6) + correction -> +6 tokens (164.2 ms) cpos=243
374
+
375
+ [spec] generated 202 tokens over 51 verifies: mean tokens/verify = 3.96 (block=5, max 6)
376
+ [spec] tokens: 1 1 1 1 1 1 50 1697 1 16 1 50 1697 16 1 50 1697 16 1 50 1697 16 1 50 1697 16 1 50 1697 16 20692 50 1697 16 791 50 1697 4394 1015 50
377
+ [spec] SPEC-DECODE: 35.1 ms/tok = 28.45 tok/s (vs base M=1 71.5 ms/tok = 13.99 tok/s -> 2.03x)
378
+ [decode] PREFILL: 26 positions in 453.8 ms = 57.3 tok/s (17.455 ms/tok)
379
+ [spec] decoding 200 tokens (block=5, draft passes=1, adaptK=1.50, prompt=3 s=27)...
380
+ [margins] 1.79 2.75 1.49 0.66 0.50 K=3
381
+ verify 1: accepted 1/2 (K=3) + correction -> +2 tokens (114.9 ms) cpos=28
382
+ [margins] 3.74 1.23 0.13 0.66 2.13 K=2
383
+ verify 2: accepted 1/1 (K=2) + correction -> +2 tokens (99.3 ms) cpos=30
384
+ [margins] 3.62 1.83 0.08 0.07 2.93 K=3
385
+ verify 3: accepted 1/2 (K=3) + correction -> +2 tokens (109.3 ms) cpos=32
386
+ [margins] 3.17 3.22 2.26 0.88 0.74 K=4
387
+ verify 4: accepted 3/3 (K=4) + correction -> +4 tokens (118.3 ms) cpos=36
388
+ [margins] 3.79 5.76 2.68 1.46 1.96 K=4
389
+ verify 5: accepted 3/3 (K=4) + correction -> +4 tokens (118.5 ms) cpos=40
390
+ [margins] 3.22 7.73 2.69 3.79 3.04 K=6
391
+ verify 6: accepted 5/5 (K=6) + correction -> +6 tokens (143.4 ms) cpos=46
392
+ [margins] 2.65 9.55 3.42 6.34 3.56 K=6
393
+ verify 7: accepted 5/5 (K=6) + correction -> +6 tokens (149.8 ms) cpos=52
394
+ [margins] 0.98 9.53 3.01 7.38 4.31 K=6
395
+ verify 8: accepted 5/5 (K=6) + correction -> +6 tokens (145.0 ms) cpos=58
396
+ [margins] 1.00 8.86 2.51 6.99 3.98 K=6
397
+ verify 9: accepted 5/5 (K=6) + correction -> +6 tokens (151.0 ms) cpos=64
398
+ [margins] 2.13 8.05 3.42 6.61 4.97 K=6
399
+ verify 10: accepted 5/5 (K=6) + correction -> +6 tokens (144.6 ms) cpos=70
400
+ [margins] 1.22 8.63 2.62 6.97 4.58 K=6
401
+ verify 11: accepted 4/5 (K=6) + correction -> +5 tokens (151.6 ms) cpos=75
402
+ [margins] 7.82 2.72 5.71 2.65 3.82 K=6
403
+ verify 12: accepted 0/5 (K=6) + correction -> +1 tokens (155.7 ms) cpos=76
404
+ [margins] 2.37 6.39 0.37 1.88 0.05 K=3
405
+ verify 13: accepted 0/2 (K=3) + correction -> +1 tokens (110.4 ms) cpos=77
406
+ [margins] 1.66 3.84 0.22 1.02 0.39 K=3
407
+ verify 14: accepted 2/2 (K=3) + correction -> +3 tokens (114.0 ms) cpos=80
408
+ [margins] 1.28 2.18 1.73 3.62 1.13 K=5
409
+ verify 15: accepted 0/4 (K=5) + correction -> +1 tokens (136.9 ms) cpos=81
410
+ [margins] 1.55 0.44 0.32 0.47 0.97 K=2
411
+ verify 16: accepted 1/1 (K=2) + correction -> +2 tokens (97.9 ms) cpos=83
412
+ [margins] 0.96 0.39 0.12 0.40 0.03 K=2
413
+ verify 17: accepted 1/1 (K=2) + correction -> +2 tokens (101.3 ms) cpos=85
414
+ [margins] 4.39 1.53 0.24 0.56 1.45 K=3
415
+ verify 18: accepted 1/2 (K=3) + correction -> +2 tokens (112.7 ms) cpos=87
416
+ [margins] 3.13 2.36 1.07 2.49 0.41 K=3
417
+ verify 19: accepted 1/2 (K=3) + correction -> +2 tokens (115.2 ms) cpos=89
418
+ [margins] 1.62 1.85 0.36 0.77 0.63 K=3
419
+ verify 20: accepted 2/2 (K=3) + correction -> +3 tokens (113.7 ms) cpos=92
420
+ [margins] 3.10 0.76 0.25 1.72 3.15 K=2
421
+ verify 21: accepted 1/1 (K=2) + correction -> +2 tokens (99.0 ms) cpos=94
422
+ [margins] 0.21 8.48 0.83 1.32 2.56 K=3
423
+ verify 22: accepted 2/2 (K=3) + correction -> +3 tokens (112.7 ms) cpos=97
424
+ [margins] 2.90 2.35 0.20 0.83 0.38 K=3
425
+ verify 23: accepted 2/2 (K=3) + correction -> +3 tokens (117.2 ms) cpos=100
426
+ [margins] 0.01 0.95 3.14 0.13 0.51 K=2
427
+ verify 24: accepted 1/1 (K=2) + correction -> +2 tokens (97.5 ms) cpos=102
428
+ [margins] 1.38 0.36 0.44 3.15 7.98 K=2
429
+ verify 25: accepted 0/1 (K=2) + correction -> +1 tokens (102.4 ms) cpos=103
430
+ [margins] 5.69 6.77 0.90 3.98 6.55 K=3
431
+ verify 26: accepted 2/2 (K=3) + correction -> +3 tokens (112.7 ms) cpos=106
432
+ [margins] 6.52 5.25 0.31 4.93 1.52 K=3
433
+ verify 27: accepted 2/2 (K=3) + correction -> +3 tokens (113.3 ms) cpos=109
434
+ [margins] 8.83 1.54 1.46 1.41 0.81 K=3
435
+ verify 28: accepted 2/2 (K=3) + correction -> +3 tokens (117.3 ms) cpos=112
436
+ [margins] 9.79 0.08 2.14 0.93 0.31 K=2
437
+ verify 29: accepted 1/1 (K=2) + correction -> +2 tokens (103.5 ms) cpos=114
438
+ [margins] 0.59 9.04 1.74 1.69 2.27 K=6
439
+ verify 30: accepted 2/5 (K=6) + correction -> +3 tokens (161.0 ms) cpos=117
440
+ [margins] 0.76 5.37 3.75 3.18 7.97 K=6
441
+ verify 31: accepted 5/5 (K=6) + correction -> +6 tokens (161.4 ms) cpos=123
442
+ [margins] 9.57 10.96 3.62 11.97 4.36 K=6
443
+ verify 32: accepted 5/5 (K=6) + correction -> +6 tokens (177.8 ms) cpos=129
444
+ [margins] 10.62 10.95 11.11 6.57 4.14 K=6
445
+ verify 33: accepted 5/5 (K=6) + correction -> +6 tokens (159.9 ms) cpos=135
446
+ [margins] 3.79 1.24 0.93 2.04 0.55 K=2
447
+ verify 34: accepted 1/1 (K=2) + correction -> +2 tokens (107.1 ms) cpos=137
448
+ [margins] 1.21 0.10 0.49 7.82 1.59 K=2
449
+ verify 35: accepted 0/1 (K=2) + correction -> +1 tokens (101.7 ms) cpos=138
450
+ [margins] 0.78 2.01 1.04 1.17 0.43 K=3
451
+ verify 36: accepted 1/2 (K=3) + correction -> +2 tokens (115.4 ms) cpos=140
452
+ [margins] 3.13 1.68 0.79 0.05 0.52 K=3
453
+ verify 37: accepted 0/2 (K=3) + correction -> +1 tokens (112.9 ms) cpos=141
454
+ [margins] 1.78 0.20 0.73 0.19 2.75 K=2
455
+ verify 38: accepted 1/1 (K=2) + correction -> +2 tokens (99.5 ms) cpos=143
456
+ [margins] 0.81 0.24 0.06 0.47 0.74 K=2
457
+ verify 39: accepted 0/1 (K=2) + correction -> +1 tokens (105.5 ms) cpos=144
458
+ [margins] 1.49 3.29 1.73 2.33 0.38 K=5
459
+ verify 40: accepted 1/4 (K=5) + correction -> +2 tokens (141.2 ms) cpos=146
460
+ [margins] 7.74 1.67 1.64 1.61 0.30 K=5
461
+ verify 41: accepted 2/4 (K=5) + correction -> +3 tokens (140.2 ms) cpos=149
462
+ [margins] 1.56 2.00 3.04 0.62 1.31 K=4
463
+ verify 42: accepted 0/3 (K=4) + correction -> +1 tokens (128.7 ms) cpos=150
464
+ [margins] 1.75 0.74 0.36 1.87 0.39 K=2
465
+ verify 43: accepted 1/1 (K=2) + correction -> +2 tokens (102.1 ms) cpos=152
466
+ [margins] 3.88 0.45 1.40 1.13 1.00 K=2
467
+ verify 44: accepted 1/1 (K=2) + correction -> +2 tokens (99.5 ms) cpos=154
468
+ [margins] 0.73 1.19 3.43 0.24 0.44 K=2
469
+ verify 45: accepted 1/1 (K=2) + correction -> +2 tokens (105.1 ms) cpos=156
470
+ [margins] 5.10 2.09 0.69 1.20 0.55 K=3
471
+ verify 46: accepted 1/2 (K=3) + correction -> +2 tokens (110.9 ms) cpos=158
472
+ [margins] 2.59 1.76 1.65 0.11 1.24 K=4
473
+ verify 47: accepted 3/3 (K=4) + correction -> +4 tokens (131.6 ms) cpos=162
474
+ [margins] 5.79 3.09 0.21 1.07 0.42 K=3
475
+ verify 48: accepted 2/2 (K=3) + correction -> +3 tokens (114.7 ms) cpos=165
476
+ [margins] 3.42 1.65 1.39 1.53 1.60 K=3
477
+ verify 49: accepted 2/2 (K=3) + correction -> +3 tokens (113.9 ms) cpos=168
478
+ [margins] 4.78 1.84 7.44 3.02 1.28 K=5
479
+ verify 50: accepted 4/4 (K=5) + correction -> +5 tokens (143.9 ms) cpos=173
480
+ [margins] 0.27 0.11 1.50 0.41 0.42 K=2
481
+ verify 51: accepted 0/1 (K=2) + correction -> +1 tokens (102.0 ms) cpos=174
482
+ [margins] 0.08 2.10 0.60 3.90 2.81 K=3
483
+ verify 52: accepted 0/2 (K=3) + correction -> +1 tokens (118.7 ms) cpos=175
484
+ [margins] 0.32 1.19 0.85 0.25 1.94 K=2
485
+ verify 53: accepted 1/1 (K=2) + correction -> +2 tokens (106.9 ms) cpos=177
486
+ [margins] 6.45 1.88 1.49 1.86 2.29 K=3
487
+ verify 54: accepted 2/2 (K=3) + correction -> +3 tokens (116.5 ms) cpos=180
488
+ [margins] 0.49 2.15 2.02 0.62 0.10 K=4
489
+ verify 55: accepted 0/3 (K=4) + correction -> +1 tokens (129.4 ms) cpos=181
490
+ [margins] 5.35 0.98 0.87 1.81 1.48 K=2
491
+ verify 56: accepted 1/1 (K=2) + correction -> +2 tokens (100.0 ms) cpos=183
492
+ [margins] 3.24 2.84 5.55 2.43 2.63 K=6
493
+ verify 57: accepted 2/5 (K=6) + correction -> +3 tokens (160.4 ms) cpos=186
494
+ [margins] 2.61 2.32 0.08 0.08 0.84 K=3
495
+ verify 58: accepted 2/2 (K=3) + correction -> +3 tokens (118.1 ms) cpos=189
496
+ [margins] 3.13 1.70 0.20 0.16 0.54 K=3
497
+ verify 59: accepted 2/2 (K=3) + correction -> +3 tokens (119.7 ms) cpos=192
498
+ [margins] 7.13 0.32 0.19 1.06 1.00 K=2
499
+ verify 60: accepted 1/1 (K=2) + correction -> +2 tokens (100.1 ms) cpos=194
500
+ [margins] 0.22 1.63 3.16 0.96 0.44 K=4
501
+ verify 61: accepted 0/3 (K=4) + correction -> +1 tokens (131.4 ms) cpos=195
502
+ [margins] 1.84 0.77 2.14 3.54 0.30 K=2
503
+ verify 62: accepted 1/1 (K=2) + correction -> +2 tokens (104.0 ms) cpos=197
504
+ [margins] 0.68 0.89 6.72 0.79 2.59 K=2
505
+ verify 63: accepted 0/1 (K=2) + correction -> +1 tokens (100.0 ms) cpos=198
506
+ [margins] 2.04 0.56 1.17 4.66 0.22 K=2
507
+ verify 64: accepted 0/1 (K=2) + correction -> +1 tokens (106.0 ms) cpos=199
508
+ [margins] 4.37 6.44 6.80 0.89 4.84 K=4
509
+ verify 65: accepted 0/3 (K=4) + correction -> +1 tokens (131.0 ms) cpos=200
510
+ [margins] 7.29 3.89 0.09 2.84 2.98 K=3
511
+ verify 66: accepted 2/2 (K=3) + correction -> +3 tokens (113.2 ms) cpos=203
512
+ [margins] 4.21 3.76 2.14 2.26 1.14 K=5
513
+ verify 67: accepted 1/4 (K=5) + correction -> +2 tokens (143.5 ms) cpos=205
514
+ [margins] 1.79 0.27 0.33 0.42 1.28 K=2
515
+ verify 68: accepted 1/1 (K=2) + correction -> +2 tokens (99.5 ms) cpos=207
516
+ [margins] 0.03 2.25 1.28 3.17 3.50 K=3
517
+ verify 69: accepted 0/2 (K=3) + correction -> +1 tokens (118.7 ms) cpos=208
518
+ [margins] 10.48 0.94 1.65 4.53 0.97 K=2
519
+ verify 70: accepted 1/1 (K=2) + correction -> +2 tokens (101.8 ms) cpos=210
520
+ [margins] 2.31 5.03 0.65 1.42 4.38 K=3
521
+ verify 71: accepted 2/2 (K=3) + correction -> +3 tokens (118.3 ms) cpos=213
522
+ [margins] 0.88 0.03 0.85 0.84 0.26 K=2
523
+ verify 72: accepted 1/1 (K=2) + correction -> +2 tokens (99.6 ms) cpos=215
524
+ [margins] 3.27 2.22 4.43 4.44 2.45 K=6
525
+ verify 73: accepted 4/5 (K=6) + correction -> +5 tokens (166.5 ms) cpos=220
526
+ [margins] 4.09 2.30 1.90 1.05 1.56 K=4
527
+ verify 74: accepted 0/3 (K=4) + correction -> +1 tokens (133.5 ms) cpos=221
528
+ [margins] 10.80 2.31 2.11 0.28 1.20 K=4
529
+ verify 75: accepted 3/3 (K=4) + correction -> +4 tokens (131.4 ms) cpos=225
530
+ [margins] 12.90 5.49 1.36 1.40 0.29 K=3
531
+ verify 76: accepted 2/2 (K=3) + correction -> +3 tokens (119.9 ms) cpos=228
532
+
533
+ [spec] generated 202 tokens over 76 verifies: mean tokens/verify = 2.66 (block=5, max 6)
534
+ [spec] tokens: 1 16 1 16 1 16 1 16 1 16 1 16 1 16 1 16 1 16 1 16 1 16 1 16 1 16 1 16 1 16 1 16 1 16 1 16 1 16 1 16
535
+ [spec] SPEC-DECODE: 45.4 ms/tok = 22.01 tok/s (vs base M=1 71.5 ms/tok = 13.99 tok/s -> 1.57x)
536
+ [decode] PREFILL: 14 positions in 281.3 ms = 49.8 tok/s (20.091 ms/tok)
537
+ [spec] decoding 200 tokens (block=5, draft passes=1, adaptK=1.50, prompt=4 s=15)...
538
+ [margins] 0.19 8.68 2.54 3.60 4.40 K=6
539
+ verify 1: accepted 0/5 (K=6) + correction -> +1 tokens (158.8 ms) cpos=15
540
+ [margins] 0.72 1.75 0.05 1.23 0.18 K=3
541
+ verify 2: accepted 1/2 (K=3) + correction -> +2 tokens (111.9 ms) cpos=17
542
+ [margins] 0.27 0.67 0.09 1.01 0.96 K=2
543
+ verify 3: accepted 0/1 (K=2) + correction -> +1 tokens (95.9 ms) cpos=18
544
+ [margins] 1.47 0.74 0.47 0.07 1.55 K=2
545
+ verify 4: accepted 1/1 (K=2) + correction -> +2 tokens (99.5 ms) cpos=20
546
+ [margins] 0.82 1.16 0.33 0.62 1.40 K=2
547
+ verify 5: accepted 1/1 (K=2) + correction -> +2 tokens (94.5 ms) cpos=22
548
+ [margins] 2.14 1.02 0.38 0.83 0.28 K=2
549
+ verify 6: accepted 1/1 (K=2) + correction -> +2 tokens (101.1 ms) cpos=24
550
+ [margins] 0.01 0.63 0.00 1.42 3.99 K=2
551
+ verify 7: accepted 1/1 (K=2) + correction -> +2 tokens (96.7 ms) cpos=26
552
+ [margins] 0.01 0.00 0.60 0.59 0.30 K=2
553
+ verify 8: accepted 0/1 (K=2) + correction -> +1 tokens (101.0 ms) cpos=27
554
+ [margins] 0.04 0.25 0.13 0.33 0.93 K=2
555
+ verify 9: accepted 1/1 (K=2) + correction -> +2 tokens (103.7 ms) cpos=29
556
+ [margins] 3.78 2.73 0.53 3.17 0.87 K=3
557
+ verify 10: accepted 2/2 (K=3) + correction -> +3 tokens (118.2 ms) cpos=32
558
+ [margins] 1.23 0.93 0.02 2.07 0.20 K=2
559
+ verify 11: accepted 1/1 (K=2) + correction -> +2 tokens (95.9 ms) cpos=34
560
+ [margins] 1.20 0.57 0.67 1.04 0.34 K=2
561
+ verify 12: accepted 1/1 (K=2) + correction -> +2 tokens (99.5 ms) cpos=36
562
+ [margins] 0.51 1.50 11.90 1.07 2.15 K=2
563
+ verify 13: accepted 0/1 (K=2) + correction -> +1 tokens (96.4 ms) cpos=37
564
+ [margins] 0.27 2.73 0.05 0.31 0.10 K=3
565
+ verify 14: accepted 1/2 (K=3) + correction -> +2 tokens (113.5 ms) cpos=39
566
+ [margins] 2.77 0.14 0.06 2.03 0.15 K=2
567
+ verify 15: accepted 0/1 (K=2) + correction -> +1 tokens (99.5 ms) cpos=40
568
+ [margins] 3.25 3.29 3.23 1.29 1.76 K=4
569
+ verify 16: accepted 1/3 (K=4) + correction -> +2 tokens (127.3 ms) cpos=42
570
+ [margins] 0.43 0.44 0.68 1.02 0.27 K=2
571
+ verify 17: accepted 1/1 (K=2) + correction -> +2 tokens (101.7 ms) cpos=44
572
+ [margins] 0.95 7.32 0.57 1.28 2.12 K=3
573
+ verify 18: accepted 2/2 (K=3) + correction -> +3 tokens (109.9 ms) cpos=47
574
+ [margins] 3.48 6.79 2.35 1.42 2.99 K=4
575
+ verify 19: accepted 3/3 (K=4) + correction -> +4 tokens (125.7 ms) cpos=51
576
+ [margins] 2.58 0.44 0.63 2.03 0.14 K=2
577
+ verify 20: accepted 0/1 (K=2) + correction -> +1 tokens (101.1 ms) cpos=52
578
+ [margins] 1.89 1.90 1.58 1.23 3.08 K=4
579
+ verify 21: accepted 0/3 (K=4) + correction -> +1 tokens (125.1 ms) cpos=53
580
+ [margins] 7.25 2.71 1.85 0.51 2.47 K=4
581
+ verify 22: accepted 2/3 (K=4) + correction -> +3 tokens (129.5 ms) cpos=56
582
+ [margins] 4.25 1.35 4.48 3.10 0.92 K=2
583
+ verify 23: accepted 1/1 (K=2) + correction -> +2 tokens (97.7 ms) cpos=58
584
+ [margins] 4.77 10.16 1.59 0.54 0.16 K=4
585
+ verify 24: accepted 2/3 (K=4) + correction -> +3 tokens (123.7 ms) cpos=61
586
+ [margins] 1.68 1.63 0.56 1.03 0.11 K=3
587
+ verify 25: accepted 0/2 (K=3) + correction -> +1 tokens (112.7 ms) cpos=62
588
+ [margins] 1.45 0.54 1.49 1.59 2.52 K=2
589
+ verify 26: accepted 0/1 (K=2) + correction -> +1 tokens (99.7 ms) cpos=63
590
+ [margins] 0.59 1.33 0.56 0.76 0.21 K=2
591
+ verify 27: accepted 1/1 (K=2) + correction -> +2 tokens (101.9 ms) cpos=65
592
+ [margins] 1.86 3.65 0.06 1.05 0.93 K=3
593
+ verify 28: accepted 0/2 (K=3) + correction -> +1 tokens (113.6 ms) cpos=66
594
+ [margins] 0.69 3.87 0.53 0.06 0.15 K=3
595
+ verify 29: accepted 0/2 (K=3) + correction -> +1 tokens (113.3 ms) cpos=67
596
+ [margins] 1.59 0.38 0.94 1.26 1.35 K=2
597
+ verify 30: accepted 1/1 (K=2) + correction -> +2 tokens (102.8 ms) cpos=69
598
+ [margins] 0.07 2.91 0.69 2.07 0.07 K=3
599
+ verify 31: accepted 0/2 (K=3) + correction -> +1 tokens (115.7 ms) cpos=70
600
+ [margins] 1.97 1.18 0.46 0.99 0.28 K=2
601
+ verify 32: accepted 0/1 (K=2) + correction -> +1 tokens (102.3 ms) cpos=71
602
+ [margins] 0.19 0.06 1.94 0.82 0.02 K=2
603
+ verify 33: accepted 0/1 (K=2) + correction -> +1 tokens (103.0 ms) cpos=72
604
+ [margins] 0.05 1.34 4.03 0.62 0.76 K=2
605
+ verify 34: accepted 0/1 (K=2) + correction -> +1 tokens (97.3 ms) cpos=73
606
+ [margins] 5.32 1.49 1.44 0.85 0.60 K=2
607
+ verify 35: accepted 1/1 (K=2) + correction -> +2 tokens (98.1 ms) cpos=75
608
+ [margins] 0.56 1.10 2.12 2.52 0.69 K=2
609
+ verify 36: accepted 1/1 (K=2) + correction -> +2 tokens (101.3 ms) cpos=77
610
+ [margins] 4.48 1.63 0.56 1.57 0.98 K=3
611
+ verify 37: accepted 2/2 (K=3) + correction -> +3 tokens (115.9 ms) cpos=80
612
+ [margins] 1.85 4.93 1.85 2.35 8.44 K=6
613
+ verify 38: accepted 5/5 (K=6) + correction -> +6 tokens (162.9 ms) cpos=86
614
+ [margins] 4.94 5.13 3.55 3.67 1.50 K=6
615
+ verify 39: accepted 2/5 (K=6) + correction -> +3 tokens (150.1 ms) cpos=89
616
+ [margins] 8.76 4.19 2.52 1.05 0.28 K=4
617
+ verify 40: accepted 1/3 (K=4) + correction -> +2 tokens (122.5 ms) cpos=91
618
+ [margins] 6.76 3.89 1.49 0.65 1.85 K=3
619
+ verify 41: accepted 1/2 (K=3) + correction -> +2 tokens (111.7 ms) cpos=93
620
+ [margins] 4.97 0.23 5.48 4.19 1.35 K=2
621
+ verify 42: accepted 1/1 (K=2) + correction -> +2 tokens (100.9 ms) cpos=95
622
+ [margins] 5.32 1.67 3.15 1.22 4.49 K=4
623
+ verify 43: accepted 3/3 (K=4) + correction -> +4 tokens (121.9 ms) cpos=99
624
+ [margins] 3.71 6.00 3.75 2.44 3.79 K=6
625
+ verify 44: accepted 5/5 (K=6) + correction -> +6 tokens (152.6 ms) cpos=105
626
+ [margins] 3.67 8.36 3.77 5.07 3.66 K=6
627
+ verify 45: accepted 5/5 (K=6) + correction -> +6 tokens (148.7 ms) cpos=111
628
+ [margins] 4.11 8.18 4.31 5.20 4.21 K=6
629
+ verify 46: accepted 5/5 (K=6) + correction -> +6 tokens (154.2 ms) cpos=117
630
+ [margins] 4.71 7.62 4.10 5.50 4.61 K=6
631
+ verify 47: accepted 5/5 (K=6) + correction -> +6 tokens (150.4 ms) cpos=123
632
+ [margins] 3.41 8.17 4.74 5.84 4.85 K=6
633
+ verify 48: accepted 5/5 (K=6) + correction -> +6 tokens (170.1 ms) cpos=129
634
+ [margins] 4.34 7.98 5.14 5.56 5.05 K=6
635
+ verify 49: accepted 5/5 (K=6) + correction -> +6 tokens (150.7 ms) cpos=135
636
+ [margins] 4.91 8.80 3.50 5.30 3.71 K=6
637
+ verify 50: accepted 5/5 (K=6) + correction -> +6 tokens (154.9 ms) cpos=141
638
+ [margins] 5.42 8.47 5.06 4.87 4.40 K=6
639
+ verify 51: accepted 5/5 (K=6) + correction -> +6 tokens (151.0 ms) cpos=147
640
+ [margins] 6.47 7.95 4.66 4.06 4.26 K=6
641
+ verify 52: accepted 5/5 (K=6) + correction -> +6 tokens (157.6 ms) cpos=153
642
+ [margins] 4.65 8.77 3.36 5.98 4.05 K=6
643
+ verify 53: accepted 2/5 (K=6) + correction -> +3 tokens (150.2 ms) cpos=156
644
+ [margins] 4.31 3.83 5.85 2.81 1.17 K=5
645
+ verify 54: accepted 1/4 (K=5) + correction -> +2 tokens (138.8 ms) cpos=158
646
+ [margins] 3.84 0.41 1.87 0.13 6.51 K=2
647
+ verify 55: accepted 0/1 (K=2) + correction -> +1 tokens (104.1 ms) cpos=159
648
+ [margins] 7.31 0.45 1.96 1.51 0.18 K=2
649
+ verify 56: accepted 1/1 (K=2) + correction -> +2 tokens (104.4 ms) cpos=161
650
+ [margins] 7.40 4.33 1.00 3.91 1.41 K=3
651
+ verify 57: accepted 2/2 (K=3) + correction -> +3 tokens (116.3 ms) cpos=164
652
+ [margins] 6.18 4.67 2.61 3.69 2.28 K=6
653
+ verify 58: accepted 2/5 (K=6) + correction -> +3 tokens (156.8 ms) cpos=167
654
+ [margins] 9.39 4.18 7.67 6.68 4.20 K=6
655
+ verify 59: accepted 5/5 (K=6) + correction -> +6 tokens (165.7 ms) cpos=173
656
+ [margins] 5.01 1.84 1.33 2.34 0.46 K=3
657
+ verify 60: accepted 1/2 (K=3) + correction -> +2 tokens (118.9 ms) cpos=175
658
+ [margins] 6.92 8.15 1.93 0.16 4.62 K=4
659
+ verify 61: accepted 3/3 (K=4) + correction -> +4 tokens (132.4 ms) cpos=179
660
+ [margins] 9.61 9.55 7.38 9.65 0.27 K=5
661
+ verify 62: accepted 4/4 (K=5) + correction -> +5 tokens (149.3 ms) cpos=184
662
+ [margins] 6.28 6.89 10.03 11.90 5.12 K=6
663
+ verify 63: accepted 5/5 (K=6) + correction -> +6 tokens (162.3 ms) cpos=190
664
+ [margins] 10.17 5.67 8.71 6.43 9.56 K=6
665
+ verify 64: accepted 5/5 (K=6) + correction -> +6 tokens (169.4 ms) cpos=196
666
+ [margins] 12.06 7.83 7.20 7.93 0.95 K=5
667
+ verify 65: accepted 4/4 (K=5) + correction -> +5 tokens (147.1 ms) cpos=201
668
+ [margins] 6.28 8.69 9.74 14.95 7.73 K=6
669
+ verify 66: accepted 5/5 (K=6) + correction -> +6 tokens (162.7 ms) cpos=207
670
+ [margins] 10.33 6.07 8.98 9.53 9.79 K=6
671
+ verify 67: accepted 5/5 (K=6) + correction -> +6 tokens (167.3 ms) cpos=213
672
+ [margins] 11.45 14.09 11.74 7.94 5.63 K=6
673
+ verify 68: accepted 4/5 (K=6) + correction -> +5 tokens (164.1 ms) cpos=218
674
+
675
+ [spec] generated 204 tokens over 68 verifies: mean tokens/verify = 3.00 (block=5, max 6)
676
+ [spec] tokens: 5715 14 295 834 10175 14 8470 3939 436 344 2239 362 4197 339 35 126664 23809 16765 270 3077 305 1990 23714 538 9479 21121 35977 2184 47270 34788 14 832 851 588 366 99782 1082 20351 14319 35977
677
+ [spec] SPEC-DECODE: 41.0 ms/tok = 24.41 tok/s (vs base M=1 71.5 ms/tok = 13.99 tok/s -> 1.75x)
678
+ [decode] PREFILL: 82 positions in 1338.2 ms = 61.3 tok/s (16.320 ms/tok)
679
+ [spec] decoding 200 tokens (block=5, draft passes=1, adaptK=1.50, prompt=5 s=83)...
680
+ [margins] 8.06 3.64 1.20 0.79 2.93 K=3
681
+ verify 1: accepted 1/2 (K=3) + correction -> +2 tokens (118.6 ms) cpos=84
682
+ [margins] 0.04 0.53 0.14 0.12 0.84 K=2
683
+ verify 2: accepted 0/1 (K=2) + correction -> +1 tokens (101.3 ms) cpos=85
684
+ [margins] 0.03 0.34 0.66 1.29 0.46 K=2
685
+ verify 3: accepted 0/1 (K=2) + correction -> +1 tokens (97.1 ms) cpos=86
686
+ [margins] 4.70 4.46 0.23 0.29 1.48 K=3
687
+ verify 4: accepted 2/2 (K=3) + correction -> +3 tokens (115.3 ms) cpos=89
688
+ [margins] 4.70 5.48 2.47 1.64 4.18 K=6
689
+ verify 5: accepted 5/5 (K=6) + correction -> +6 tokens (156.3 ms) cpos=95
690
+ [margins] 6.94 1.78 3.47 6.15 1.81 K=6
691
+ verify 6: accepted 5/5 (K=6) + correction -> +6 tokens (162.6 ms) cpos=101
692
+ [margins] 4.88 1.00 0.67 2.09 0.21 K=2
693
+ verify 7: accepted 1/1 (K=2) + correction -> +2 tokens (102.2 ms) cpos=103
694
+ [margins] 2.01 0.47 2.13 1.42 1.72 K=2
695
+ verify 8: accepted 1/1 (K=2) + correction -> +2 tokens (104.1 ms) cpos=105
696
+ [margins] 8.79 7.59 6.26 5.95 1.51 K=6
697
+ verify 9: accepted 5/5 (K=6) + correction -> +6 tokens (159.6 ms) cpos=111
698
+ [margins] 6.74 7.68 5.77 9.72 1.91 K=6
699
+ verify 10: accepted 5/5 (K=6) + correction -> +6 tokens (166.9 ms) cpos=117
700
+ [margins] 13.68 10.26 1.38 1.61 1.56 K=3
701
+ verify 11: accepted 2/2 (K=3) + correction -> +3 tokens (118.8 ms) cpos=120
702
+ [margins] 3.49 0.79 1.66 0.04 0.83 K=2
703
+ verify 12: accepted 1/1 (K=2) + correction -> +2 tokens (103.3 ms) cpos=122
704
+ [margins] 7.64 3.38 7.01 1.86 2.70 K=6
705
+ verify 13: accepted 3/5 (K=6) + correction -> +4 tokens (181.4 ms) cpos=126
706
+ [margins] 11.15 6.24 5.38 10.00 7.97 K=6
707
+ verify 14: accepted 5/5 (K=6) + correction -> +6 tokens (181.4 ms) cpos=132
708
+ [margins] 11.30 12.87 1.84 14.53 10.03 K=6
709
+ verify 15: accepted 5/5 (K=6) + correction -> +6 tokens (160.5 ms) cpos=138
710
+ [margins] 7.88 1.84 1.91 2.81 4.62 K=6
711
+ verify 16: accepted 1/5 (K=6) + correction -> +2 tokens (166.4 ms) cpos=140
712
+ [margins] 7.95 1.10 5.85 0.24 4.15 K=2
713
+ verify 17: accepted 0/1 (K=2) + correction -> +1 tokens (100.8 ms) cpos=141
714
+ [margins] 7.77 5.28 2.71 0.13 0.58 K=4
715
+ verify 18: accepted 2/3 (K=4) + correction -> +3 tokens (128.6 ms) cpos=144
716
+ [margins] 10.19 5.01 4.45 5.58 0.32 K=5
717
+ verify 19: accepted 4/4 (K=5) + correction -> +5 tokens (145.4 ms) cpos=149
718
+ [margins] 14.87 10.57 6.46 5.74 11.06 K=6
719
+ verify 20: accepted 5/5 (K=6) + correction -> +6 tokens (160.6 ms) cpos=155
720
+ [margins] 14.67 10.82 10.49 8.55 1.97 K=6
721
+ verify 21: accepted 4/5 (K=6) + correction -> +5 tokens (166.9 ms) cpos=160
722
+ [margins] 9.28 0.74 10.50 5.60 4.63 K=2
723
+ verify 22: accepted 0/1 (K=2) + correction -> +1 tokens (101.1 ms) cpos=161
724
+ [margins] 9.36 5.67 1.05 4.87 0.98 K=3
725
+ verify 23: accepted 2/2 (K=3) + correction -> +3 tokens (116.7 ms) cpos=164
726
+ [margins] 9.75 8.31 3.37 9.73 6.23 K=6
727
+ verify 24: accepted 5/5 (K=6) + correction -> +6 tokens (161.3 ms) cpos=170
728
+ [margins] 2.59 7.34 10.23 8.86 0.35 K=5
729
+ verify 25: accepted 4/4 (K=5) + correction -> +5 tokens (145.7 ms) cpos=175
730
+ [margins] 16.11 12.83 10.96 7.59 0.81 K=5
731
+ verify 26: accepted 4/4 (K=5) + correction -> +5 tokens (149.1 ms) cpos=180
732
+ [margins] 14.87 14.86 5.04 1.73 7.75 K=6
733
+ verify 27: accepted 5/5 (K=6) + correction -> +6 tokens (160.7 ms) cpos=186
734
+ [margins] 6.19 9.10 3.70 5.96 2.30 K=6
735
+ verify 28: accepted 5/5 (K=6) + correction -> +6 tokens (167.1 ms) cpos=192
736
+ [margins] 12.31 14.42 10.34 8.64 11.03 K=6
737
+ verify 29: accepted 5/5 (K=6) + correction -> +6 tokens (161.4 ms) cpos=198
738
+ [margins] 8.13 11.15 10.75 12.43 3.06 K=6
739
+ verify 30: accepted 5/5 (K=6) + correction -> +6 tokens (163.7 ms) cpos=204
740
+ [margins] 9.40 4.83 3.16 6.50 3.16 K=6
741
+ verify 31: accepted 5/5 (K=6) + correction -> +6 tokens (160.8 ms) cpos=210
742
+ [margins] 8.83 8.57 7.23 8.44 0.37 K=5
743
+ verify 32: accepted 4/4 (K=5) + correction -> +5 tokens (144.9 ms) cpos=215
744
+ [margins] 14.63 12.40 9.41 8.43 0.16 K=5
745
+ verify 33: accepted 4/4 (K=5) + correction -> +5 tokens (151.0 ms) cpos=220
746
+ [margins] 17.46 12.69 1.26 0.01 0.67 K=3
747
+ verify 34: accepted 2/2 (K=3) + correction -> +3 tokens (112.3 ms) cpos=223
748
+ [margins] 2.60 6.39 1.59 5.72 3.25 K=6
749
+ verify 35: accepted 5/5 (K=6) + correction -> +6 tokens (168.1 ms) cpos=229
750
+ [margins] 18.71 9.60 8.81 9.68 9.67 K=6
751
+ verify 36: accepted 5/5 (K=6) + correction -> +6 tokens (165.8 ms) cpos=235
752
+ [margins] 7.27 10.56 7.04 6.56 0.10 K=5
753
+ verify 37: accepted 4/4 (K=5) + correction -> +5 tokens (151.7 ms) cpos=240
754
+ [margins] 18.01 13.21 4.02 1.45 6.55 K=4
755
+ verify 38: accepted 3/3 (K=4) + correction -> +4 tokens (130.3 ms) cpos=244
756
+ [margins] 9.80 5.24 6.07 4.29 0.22 K=5
757
+ verify 39: accepted 4/4 (K=5) + correction -> +5 tokens (145.6 ms) cpos=249
758
+ [margins] 15.49 11.07 7.06 11.24 10.93 K=6
759
+ verify 40: accepted 5/5 (K=6) + correction -> +6 tokens (161.4 ms) cpos=255
760
+ [margins] 14.04 12.09 10.84 6.90 6.53 K=6
761
+ verify 41: accepted 5/5 (K=6) + correction -> +6 tokens (183.3 ms) cpos=261
762
+ [margins] 14.39 8.51 6.01 7.52 0.41 K=5
763
+ verify 42: accepted 4/4 (K=5) + correction -> +5 tokens (146.4 ms) cpos=266
764
+ [margins] 11.12 10.64 0.15 4.55 5.12 K=3
765
+ verify 43: accepted 2/2 (K=3) + correction -> +3 tokens (120.5 ms) cpos=269
766
+ [margins] 14.58 11.53 7.02 9.45 7.94 K=6
767
+ verify 44: accepted 5/5 (K=6) + correction -> +6 tokens (165.8 ms) cpos=275
768
+ [margins] 13.62 14.08 10.46 8.72 8.78 K=6
769
+ verify 45: accepted 5/5 (K=6) + correction -> +6 tokens (168.7 ms) cpos=281
770
+ [margins] 13.75 5.82 5.30 7.33 0.42 K=5
771
+ verify 46: accepted 4/4 (K=5) + correction -> +5 tokens (145.7 ms) cpos=286
772
+
773
+ [spec] generated 204 tokens over 46 verifies: mean tokens/verify = 4.43 (block=5, max 6)
774
+ [spec] tokens: 1 50 201 671 15020 4181 362 2775 14817 14922 344 223 25 22765 7975 55 15 3180 16 1 50 201 671 15020 4181 362 2775 14817 14922 344 223 25 22765 7975 55 15 3180 16 1 50
775
+ [spec] SPEC-DECODE: 32.3 ms/tok = 30.94 tok/s (vs base M=1 71.5 ms/tok = 13.99 tok/s -> 2.21x)
776
+ [decode] PREFILL: 14 positions in 284.7 ms = 49.2 tok/s (20.338 ms/tok)
777
+ [spec] decoding 200 tokens (block=5, draft passes=1, adaptK=1.50, prompt=6 s=15)...
778
+ [margins] 0.53 3.09 0.45 0.94 0.17 K=3
779
+ verify 1: accepted 1/2 (K=3) + correction -> +2 tokens (115.6 ms) cpos=16
780
+ [margins] 3.55 0.19 2.32 0.84 1.15 K=2
781
+ verify 2: accepted 1/1 (K=2) + correction -> +2 tokens (99.2 ms) cpos=18
782
+ [margins] 2.62 0.03 1.78 1.22 2.10 K=2
783
+ verify 3: accepted 1/1 (K=2) + correction -> +2 tokens (103.2 ms) cpos=20
784
+ [margins] 2.21 1.22 1.78 2.74 2.29 K=2
785
+ verify 4: accepted 1/1 (K=2) + correction -> +2 tokens (98.4 ms) cpos=22
786
+ [margins] 4.08 2.60 2.14 2.81 3.27 K=6
787
+ verify 5: accepted 3/5 (K=6) + correction -> +4 tokens (152.1 ms) cpos=26
788
+ [margins] 2.31 4.44 3.74 2.08 2.98 K=6
789
+ verify 6: accepted 4/5 (K=6) + correction -> +5 tokens (149.1 ms) cpos=31
790
+ [margins] 0.46 9.02 3.74 1.60 4.18 K=6
791
+ verify 7: accepted 0/5 (K=6) + correction -> +1 tokens (154.6 ms) cpos=32
792
+ [margins] 7.64 1.53 0.15 6.23 3.21 K=3
793
+ verify 8: accepted 1/2 (K=3) + correction -> +2 tokens (109.2 ms) cpos=34
794
+ [margins] 2.19 5.70 1.40 0.06 3.33 K=3
795
+ verify 9: accepted 2/2 (K=3) + correction -> +3 tokens (110.5 ms) cpos=37
796
+ [margins] 11.21 6.26 2.86 4.53 6.73 K=6
797
+ verify 10: accepted 5/5 (K=6) + correction -> +6 tokens (148.6 ms) cpos=43
798
+ [margins] 12.94 7.70 8.10 10.64 7.32 K=6
799
+ verify 11: accepted 5/5 (K=6) + correction -> +6 tokens (156.3 ms) cpos=49
800
+ [margins] 15.18 10.16 9.12 8.98 10.96 K=6
801
+ verify 12: accepted 5/5 (K=6) + correction -> +6 tokens (152.0 ms) cpos=55
802
+ [margins] 12.79 9.70 9.87 7.93 9.89 K=6
803
+ verify 13: accepted 5/5 (K=6) + correction -> +6 tokens (155.2 ms) cpos=61
804
+ [margins] 17.09 9.66 10.66 9.59 9.44 K=6
805
+ verify 14: accepted 5/5 (K=6) + correction -> +6 tokens (152.3 ms) cpos=67
806
+ [margins] 15.90 9.77 8.41 9.38 8.92 K=6
807
+ verify 15: accepted 5/5 (K=6) + correction -> +6 tokens (159.1 ms) cpos=73
808
+ [margins] 14.72 10.35 11.26 7.45 9.76 K=6
809
+ verify 16: accepted 5/5 (K=6) + correction -> +6 tokens (153.1 ms) cpos=79
810
+ [margins] 16.00 9.95 11.11 12.39 11.24 K=6
811
+ verify 17: accepted 5/5 (K=6) + correction -> +6 tokens (157.9 ms) cpos=85
812
+ [margins] 15.20 10.54 9.15 11.01 9.73 K=6
813
+ verify 18: accepted 5/5 (K=6) + correction -> +6 tokens (155.8 ms) cpos=91
814
+ [margins] 15.15 10.39 9.93 9.33 11.53 K=6
815
+ verify 19: accepted 5/5 (K=6) + correction -> +6 tokens (162.0 ms) cpos=97
816
+ [margins] 16.14 10.55 10.85 10.70 11.82 K=6
817
+ verify 20: accepted 5/5 (K=6) + correction -> +6 tokens (153.6 ms) cpos=103
818
+ [margins] 14.03 10.13 11.72 11.84 10.00 K=6
819
+ verify 21: accepted 5/5 (K=6) + correction -> +6 tokens (162.4 ms) cpos=109
820
+ [margins] 15.96 10.19 11.97 12.93 9.87 K=6
821
+ verify 22: accepted 5/5 (K=6) + correction -> +6 tokens (156.5 ms) cpos=115
822
+ [margins] 13.38 10.17 11.46 11.99 10.90 K=6
823
+ verify 23: accepted 5/5 (K=6) + correction -> +6 tokens (160.5 ms) cpos=121
824
+ [margins] 14.36 10.18 12.57 9.23 10.39 K=6
825
+ verify 24: accepted 5/5 (K=6) + correction -> +6 tokens (157.7 ms) cpos=127
826
+ [margins] 15.53 9.88 11.45 10.98 11.81 K=6
827
+ verify 25: accepted 5/5 (K=6) + correction -> +6 tokens (175.9 ms) cpos=133
828
+ [margins] 15.87 9.15 11.39 8.47 10.39 K=6
829
+ verify 26: accepted 5/5 (K=6) + correction -> +6 tokens (153.8 ms) cpos=139
830
+ [margins] 12.29 9.83 11.75 11.52 11.08 K=6
831
+ verify 27: accepted 5/5 (K=6) + correction -> +6 tokens (160.9 ms) cpos=145
832
+ [margins] 15.57 9.50 11.03 10.35 9.40 K=6
833
+ verify 28: accepted 5/5 (K=6) + correction -> +6 tokens (156.2 ms) cpos=151
834
+ [margins] 13.57 9.86 12.59 9.42 11.59 K=6
835
+ verify 29: accepted 5/5 (K=6) + correction -> +6 tokens (160.1 ms) cpos=157
836
+ [margins] 14.57 9.25 9.94 13.17 10.43 K=6
837
+ verify 30: accepted 5/5 (K=6) + correction -> +6 tokens (159.4 ms) cpos=163
838
+ [margins] 18.14 8.91 10.71 9.71 9.47 K=6
839
+ verify 31: accepted 5/5 (K=6) + correction -> +6 tokens (161.1 ms) cpos=169
840
+ [margins] 15.19 10.42 11.53 9.09 12.17 K=6
841
+ verify 32: accepted 5/5 (K=6) + correction -> +6 tokens (158.4 ms) cpos=175
842
+ [margins] 16.20 8.96 10.60 8.10 10.14 K=6
843
+ verify 33: accepted 5/5 (K=6) + correction -> +6 tokens (162.2 ms) cpos=181
844
+ [margins] 16.98 9.84 10.48 9.08 10.75 K=6
845
+ verify 34: accepted 5/5 (K=6) + correction -> +6 tokens (159.3 ms) cpos=187
846
+ [margins] 15.42 9.92 12.21 10.69 7.97 K=6
847
+ verify 35: accepted 5/5 (K=6) + correction -> +6 tokens (163.8 ms) cpos=193
848
+ [margins] 15.54 9.81 10.89 9.71 9.03 K=6
849
+ verify 36: accepted 5/5 (K=6) + correction -> +6 tokens (158.1 ms) cpos=199
850
+ [margins] 18.59 9.65 11.31 11.17 8.07 K=6
851
+ verify 37: accepted 5/5 (K=6) + correction -> +6 tokens (162.9 ms) cpos=205
852
+ [margins] 14.82 9.94 10.97 10.41 9.52 K=6
853
+ verify 38: accepted 5/5 (K=6) + correction -> +6 tokens (158.2 ms) cpos=211
854
+ [margins] 17.07 9.32 11.00 7.70 10.56 K=6
855
+ verify 39: accepted 5/5 (K=6) + correction -> +6 tokens (163.2 ms) cpos=217
856
+
857
+ [spec] generated 203 tokens over 39 verifies: mean tokens/verify = 5.21 (block=5, max 6)
858
+ [spec] tokens: 1 16 1 16 1 16 1 16 1 16 1 16 1 16 1 16 403 31310 16 403 31310 16 403 31310 16 403 31310 16 403 31310 16 403 31310 16 403 31310 16 403 31310 16
859
+ [spec] SPEC-DECODE: 28.5 ms/tok = 35.06 tok/s (vs base M=1 71.5 ms/tok = 13.99 tok/s -> 2.51x)
860
+ [decode] PREFILL: 37 positions in 663.6 ms = 55.8 tok/s (17.934 ms/tok)
861
+ [spec] decoding 200 tokens (block=5, draft passes=1, adaptK=1.50, prompt=7 s=38)...
862
+ [margins] 0.70 1.59 0.03 0.61 1.27 K=3
863
+ verify 1: accepted 0/2 (K=3) + correction -> +1 tokens (116.6 ms) cpos=38
864
+ [margins] 0.02 5.81 0.07 0.02 0.02 K=3
865
+ verify 2: accepted 0/2 (K=3) + correction -> +1 tokens (111.4 ms) cpos=39
866
+ [margins] 1.63 0.36 5.16 0.50 1.26 K=2
867
+ verify 3: accepted 1/1 (K=2) + correction -> +2 tokens (99.7 ms) cpos=41
868
+ [margins] 2.81 0.33 0.50 5.77 0.68 K=2
869
+ verify 4: accepted 1/1 (K=2) + correction -> +2 tokens (94.9 ms) cpos=43
870
+ [margins] 1.20 1.88 0.17 0.05 0.09 K=3
871
+ verify 5: accepted 0/2 (K=3) + correction -> +1 tokens (111.5 ms) cpos=44
872
+ [margins] 0.05 0.09 0.50 0.04 0.10 K=2
873
+ verify 6: accepted 1/1 (K=2) + correction -> +2 tokens (95.5 ms) cpos=46
874
+ [margins] 0.97 1.13 1.88 0.01 0.20 K=2
875
+ verify 7: accepted 0/1 (K=2) + correction -> +1 tokens (101.2 ms) cpos=47
876
+ [margins] 1.39 1.85 0.43 2.56 0.73 K=3
877
+ verify 8: accepted 1/2 (K=3) + correction -> +2 tokens (113.8 ms) cpos=49
878
+ [margins] 7.49 6.73 6.03 1.93 0.11 K=5
879
+ verify 9: accepted 4/4 (K=5) + correction -> +5 tokens (138.4 ms) cpos=54
880
+ [margins] 0.25 2.35 2.80 3.25 1.16 K=5
881
+ verify 10: accepted 2/4 (K=5) + correction -> +3 tokens (138.8 ms) cpos=57
882
+ [margins] 2.84 0.96 0.73 1.62 0.22 K=2
883
+ verify 11: accepted 1/1 (K=2) + correction -> +2 tokens (97.4 ms) cpos=59
884
+ [margins] 3.71 2.44 0.77 0.30 0.66 K=3
885
+ verify 12: accepted 2/2 (K=3) + correction -> +3 tokens (113.7 ms) cpos=62
886
+ [margins] 0.99 1.63 2.25 0.32 2.71 K=4
887
+ verify 13: accepted 3/3 (K=4) + correction -> +4 tokens (125.9 ms) cpos=66
888
+ [margins] 13.30 2.58 5.99 0.98 1.24 K=4
889
+ verify 14: accepted 3/3 (K=4) + correction -> +4 tokens (128.7 ms) cpos=70
890
+ [margins] 1.38 0.74 0.77 2.65 0.63 K=2
891
+ verify 15: accepted 1/1 (K=2) + correction -> +2 tokens (101.9 ms) cpos=72
892
+ [margins] 2.42 6.28 0.11 2.17 0.67 K=3
893
+ verify 16: accepted 2/2 (K=3) + correction -> +3 tokens (108.3 ms) cpos=75
894
+ [margins] 2.54 8.84 0.86 3.09 0.26 K=3
895
+ verify 17: accepted 2/2 (K=3) + correction -> +3 tokens (111.4 ms) cpos=78
896
+ [margins] 7.93 7.63 6.44 7.42 9.10 K=6
897
+ verify 18: accepted 5/5 (K=6) + correction -> +6 tokens (160.2 ms) cpos=84
898
+ [margins] 7.65 8.49 5.89 3.00 0.63 K=5
899
+ verify 19: accepted 4/4 (K=5) + correction -> +5 tokens (141.4 ms) cpos=89
900
+ [margins] 7.26 6.42 4.10 0.58 0.53 K=4
901
+ verify 20: accepted 3/3 (K=4) + correction -> +4 tokens (128.2 ms) cpos=93
902
+ [margins] 4.87 3.32 2.36 1.10 0.17 K=4
903
+ verify 21: accepted 3/3 (K=4) + correction -> +4 tokens (126.3 ms) cpos=97
904
+ [margins] 9.72 12.06 10.28 9.68 3.82 K=6
905
+ verify 22: accepted 5/5 (K=6) + correction -> +6 tokens (159.6 ms) cpos=103
906
+ [margins] 9.84 4.14 2.91 0.98 9.67 K=4
907
+ verify 23: accepted 3/3 (K=4) + correction -> +4 tokens (130.5 ms) cpos=107
908
+ [margins] 4.92 3.14 10.27 1.89 3.97 K=6
909
+ verify 24: accepted 3/5 (K=6) + correction -> +4 tokens (167.0 ms) cpos=111
910
+ [margins] 10.05 5.75 7.36 4.34 6.65 K=6
911
+ verify 25: accepted 5/5 (K=6) + correction -> +6 tokens (167.2 ms) cpos=117
912
+ [margins] 10.51 9.03 3.71 5.42 4.64 K=6
913
+ verify 26: accepted 5/5 (K=6) + correction -> +6 tokens (161.4 ms) cpos=123
914
+ [margins] 13.39 5.92 8.62 7.32 0.04 K=5
915
+ verify 27: accepted 4/4 (K=5) + correction -> +5 tokens (163.1 ms) cpos=128
916
+ [margins] 5.40 5.12 4.08 2.71 3.11 K=6
917
+ verify 28: accepted 3/5 (K=6) + correction -> +4 tokens (160.2 ms) cpos=132
918
+ [margins] 5.05 2.19 3.59 2.25 0.85 K=5
919
+ verify 29: accepted 4/4 (K=5) + correction -> +5 tokens (142.1 ms) cpos=137
920
+ [margins] 3.27 0.09 0.24 0.24 0.60 K=2
921
+ verify 30: accepted 1/1 (K=2) + correction -> +2 tokens (102.0 ms) cpos=139
922
+ [margins] 5.66 0.71 5.56 6.67 5.08 K=2
923
+ verify 31: accepted 1/1 (K=2) + correction -> +2 tokens (105.2 ms) cpos=141
924
+ [margins] 9.18 7.93 5.62 1.54 0.48 K=5
925
+ verify 32: accepted 4/4 (K=5) + correction -> +5 tokens (145.0 ms) cpos=146
926
+ [margins] 7.84 9.14 9.91 11.09 1.70 K=6
927
+ verify 33: accepted 4/5 (K=6) + correction -> +5 tokens (168.2 ms) cpos=151
928
+ [margins] 2.10 1.29 0.49 7.61 0.17 K=2
929
+ verify 34: accepted 1/1 (K=2) + correction -> +2 tokens (109.9 ms) cpos=153
930
+ [margins] 1.31 2.69 0.77 1.10 3.79 K=3
931
+ verify 35: accepted 0/2 (K=3) + correction -> +1 tokens (109.5 ms) cpos=154
932
+ [margins] 6.53 2.50 1.28 2.10 1.42 K=3
933
+ verify 36: accepted 1/2 (K=3) + correction -> +2 tokens (115.7 ms) cpos=156
934
+ [margins] 6.30 3.19 0.44 1.70 2.19 K=3
935
+ verify 37: accepted 1/2 (K=3) + correction -> +2 tokens (110.5 ms) cpos=158
936
+ [margins] 4.70 0.85 3.48 1.28 1.42 K=2
937
+ verify 38: accepted 1/1 (K=2) + correction -> +2 tokens (107.4 ms) cpos=160
938
+ [margins] 5.44 3.82 4.37 0.40 3.31 K=4
939
+ verify 39: accepted 3/3 (K=4) + correction -> +4 tokens (124.0 ms) cpos=164
940
+ [margins] 5.02 7.90 6.10 3.80 4.65 K=6
941
+ verify 40: accepted 5/5 (K=6) + correction -> +6 tokens (149.4 ms) cpos=170
942
+ [margins] 3.93 9.37 3.85 5.33 4.60 K=6
943
+ verify 41: accepted 5/5 (K=6) + correction -> +6 tokens (154.8 ms) cpos=176
944
+ [margins] 3.99 8.09 3.92 6.60 5.39 K=6
945
+ verify 42: accepted 5/5 (K=6) + correction -> +6 tokens (150.4 ms) cpos=182
946
+ [margins] 3.36 8.74 4.97 7.39 5.80 K=6
947
+ verify 43: accepted 5/5 (K=6) + correction -> +6 tokens (155.4 ms) cpos=188
948
+ [margins] 2.67 8.65 4.97 6.18 5.71 K=6
949
+ verify 44: accepted 5/5 (K=6) + correction -> +6 tokens (150.6 ms) cpos=194
950
+ [margins] 2.54 8.14 4.54 6.41 5.21 K=6
951
+ verify 45: accepted 5/5 (K=6) + correction -> +6 tokens (155.2 ms) cpos=200
952
+ [margins] 3.45 8.92 4.69 5.34 4.57 K=6
953
+ verify 46: accepted 5/5 (K=6) + correction -> +6 tokens (151.2 ms) cpos=206
954
+ [margins] 3.91 8.12 4.03 4.74 5.54 K=6
955
+ verify 47: accepted 5/5 (K=6) + correction -> +6 tokens (156.4 ms) cpos=212
956
+ [margins] 3.34 6.86 4.37 4.16 5.20 K=6
957
+ verify 48: accepted 5/5 (K=6) + correction -> +6 tokens (149.2 ms) cpos=218
958
+ [margins] 3.90 7.05 4.89 3.78 5.48 K=6
959
+ verify 49: accepted 5/5 (K=6) + correction -> +6 tokens (156.1 ms) cpos=224
960
+ [margins] 4.14 7.19 4.34 3.83 5.06 K=6
961
+ verify 50: accepted 5/5 (K=6) + correction -> +6 tokens (150.0 ms) cpos=230
962
+ [margins] 4.07 6.49 4.86 2.97 5.66 K=6
963
+ verify 51: accepted 0/5 (K=6) + correction -> +1 tokens (156.6 ms) cpos=231
964
+ [margins] 9.25 1.93 3.54 5.73 1.32 K=5
965
+ verify 52: accepted 0/4 (K=5) + correction -> +1 tokens (143.5 ms) cpos=232
966
+ [margins] 5.83 2.47 7.65 1.78 0.59 K=5
967
+ verify 53: accepted 1/4 (K=5) + correction -> +2 tokens (140.6 ms) cpos=234
968
+ [margins] 2.68 1.47 2.85 1.57 1.17 K=2
969
+ verify 54: accepted 1/1 (K=2) + correction -> +2 tokens (104.3 ms) cpos=236
970
+ [margins] 4.04 0.04 6.10 0.59 0.14 K=2
971
+ verify 55: accepted 0/1 (K=2) + correction -> +1 tokens (100.8 ms) cpos=237
972
+
973
+ [spec] generated 200 tokens over 55 verifies: mean tokens/verify = 3.64 (block=5, max 6)
974
+ [spec] tokens: 343 12888 28 566 344 260 19075 3417 9508 5718 270 2753 294 270 6616 366 1512 1527 27819 13459 16 5715 270 12919 6329 1512 1527 940 223 19 27819 13459 16 455 3452 2753 344 979 7490 201
975
+ [spec] SPEC-DECODE: 35.8 ms/tok = 27.94 tok/s (vs base M=1 71.5 ms/tok = 13.99 tok/s -> 2.00x)
976
+ [decode] PREFILL: 12 positions in 250.9 ms = 47.8 tok/s (20.907 ms/tok)
977
+ [spec] decoding 200 tokens (block=5, draft passes=1, adaptK=1.50, prompt=8 s=13)...
978
+ [margins] 2.75 1.00 0.20 0.17 0.38 K=2
979
+ verify 1: accepted 1/1 (K=2) + correction -> +2 tokens (99.2 ms) cpos=14
980
+ [margins] 0.33 6.66 2.87 0.68 6.36 K=4
981
+ verify 2: accepted 0/3 (K=4) + correction -> +1 tokens (127.8 ms) cpos=15
982
+ [margins] 0.37 0.83 1.89 0.16 3.54 K=2
983
+ verify 3: accepted 1/1 (K=2) + correction -> +2 tokens (102.6 ms) cpos=17
984
+ [margins] 0.57 0.05 0.82 2.51 0.04 K=2
985
+ verify 4: accepted 1/1 (K=2) + correction -> +2 tokens (98.3 ms) cpos=19
986
+ [margins] 2.30 0.09 2.86 0.34 2.94 K=2
987
+ verify 5: accepted 1/1 (K=2) + correction -> +2 tokens (102.4 ms) cpos=21
988
+ [margins] 1.66 0.27 3.32 1.05 3.32 K=2
989
+ verify 6: accepted 1/1 (K=2) + correction -> +2 tokens (98.4 ms) cpos=23
990
+ [margins] 2.41 0.04 1.43 0.74 3.43 K=2
991
+ verify 7: accepted 1/1 (K=2) + correction -> +2 tokens (102.7 ms) cpos=25
992
+ [margins] 3.08 0.53 5.08 0.59 4.15 K=2
993
+ verify 8: accepted 1/1 (K=2) + correction -> +2 tokens (98.7 ms) cpos=27
994
+ [margins] 3.73 1.07 5.31 1.34 4.79 K=2
995
+ verify 9: accepted 1/1 (K=2) + correction -> +2 tokens (102.6 ms) cpos=29
996
+ [margins] 5.15 0.66 4.81 1.01 4.48 K=2
997
+ verify 10: accepted 1/1 (K=2) + correction -> +2 tokens (98.7 ms) cpos=31
998
+ [margins] 5.79 1.31 4.87 1.54 4.35 K=2
999
+ verify 11: accepted 1/1 (K=2) + correction -> +2 tokens (102.7 ms) cpos=33
1000
+ [margins] 5.70 0.75 4.75 1.40 3.95 K=2
1001
+ verify 12: accepted 1/1 (K=2) + correction -> +2 tokens (99.2 ms) cpos=35
1002
+ [margins] 6.22 0.87 5.74 1.41 4.68 K=2
1003
+ verify 13: accepted 1/1 (K=2) + correction -> +2 tokens (102.5 ms) cpos=37
1004
+ [margins] 5.69 0.72 5.02 1.42 3.92 K=2
1005
+ verify 14: accepted 1/1 (K=2) + correction -> +2 tokens (99.0 ms) cpos=39
1006
+ [margins] 6.48 0.46 5.96 1.14 4.82 K=2
1007
+ verify 15: accepted 1/1 (K=2) + correction -> +2 tokens (102.6 ms) cpos=41
1008
+ [margins] 6.34 0.80 5.22 1.51 4.02 K=2
1009
+ verify 16: accepted 1/1 (K=2) + correction -> +2 tokens (99.3 ms) cpos=43
1010
+ [margins] 6.79 0.99 5.89 2.04 4.43 K=2
1011
+ verify 17: accepted 1/1 (K=2) + correction -> +2 tokens (103.4 ms) cpos=45
1012
+ [margins] 5.97 1.49 4.39 1.29 3.44 K=2
1013
+ verify 18: accepted 0/1 (K=2) + correction -> +1 tokens (96.1 ms) cpos=46
1014
+ [margins] 5.34 3.56 5.86 0.35 0.41 K=4
1015
+ verify 19: accepted 3/3 (K=4) + correction -> +4 tokens (126.4 ms) cpos=50
1016
+ [margins] 2.53 2.71 0.34 2.28 2.94 K=3
1017
+ verify 20: accepted 2/2 (K=3) + correction -> +3 tokens (113.2 ms) cpos=53
1018
+ [margins] 7.98 8.65 0.17 3.71 0.28 K=3
1019
+ verify 21: accepted 2/2 (K=3) + correction -> +3 tokens (112.0 ms) cpos=56
1020
+ [margins] 2.45 0.09 1.72 2.65 2.79 K=2
1021
+ verify 22: accepted 1/1 (K=2) + correction -> +2 tokens (100.2 ms) cpos=58
1022
+ [margins] 0.33 6.60 1.43 3.30 2.30 K=3
1023
+ verify 23: accepted 0/2 (K=3) + correction -> +1 tokens (110.4 ms) cpos=59
1024
+ [margins] 0.36 2.40 0.54 6.22 0.05 K=3
1025
+ verify 24: accepted 1/2 (K=3) + correction -> +2 tokens (113.3 ms) cpos=61
1026
+ [margins] 5.48 7.69 6.32 4.11 8.12 K=6
1027
+ verify 25: accepted 5/5 (K=6) + correction -> +6 tokens (156.1 ms) cpos=67
1028
+ [margins] 7.96 10.33 0.33 1.45 0.95 K=3
1029
+ verify 26: accepted 2/2 (K=3) + correction -> +3 tokens (114.5 ms) cpos=70
1030
+ [margins] 2.79 2.36 0.47 4.09 1.44 K=3
1031
+ verify 27: accepted 0/2 (K=3) + correction -> +1 tokens (115.0 ms) cpos=71
1032
+ [margins] 1.71 2.67 3.89 1.96 2.32 K=6
1033
+ verify 28: accepted 0/5 (K=6) + correction -> +1 tokens (163.3 ms) cpos=72
1034
+ [margins] 3.51 3.68 3.83 6.94 3.07 K=6
1035
+ verify 29: accepted 5/5 (K=6) + correction -> +6 tokens (157.4 ms) cpos=78
1036
+ [margins] 7.05 10.29 8.38 0.90 0.68 K=4
1037
+ verify 30: accepted 3/3 (K=4) + correction -> +4 tokens (128.7 ms) cpos=82
1038
+ [margins] 3.83 7.11 4.08 5.29 1.12 K=5
1039
+ verify 31: accepted 4/4 (K=5) + correction -> +5 tokens (139.7 ms) cpos=87
1040
+ [margins] 7.94 4.92 7.43 4.79 9.97 K=6
1041
+ verify 32: accepted 5/5 (K=6) + correction -> +6 tokens (163.1 ms) cpos=93
1042
+ [margins] 2.41 4.76 6.92 4.63 5.86 K=6
1043
+ verify 33: accepted 5/5 (K=6) + correction -> +6 tokens (157.2 ms) cpos=99
1044
+ [margins] 7.86 4.07 6.68 5.78 8.18 K=6
1045
+ verify 34: accepted 5/5 (K=6) + correction -> +6 tokens (165.8 ms) cpos=105
1046
+ [margins] 2.49 4.71 5.00 6.56 6.64 K=6
1047
+ verify 35: accepted 5/5 (K=6) + correction -> +6 tokens (159.0 ms) cpos=111
1048
+ [margins] 7.11 4.43 3.55 2.72 6.24 K=6
1049
+ verify 36: accepted 5/5 (K=6) + correction -> +6 tokens (164.8 ms) cpos=117
1050
+ [margins] 3.31 6.43 8.01 8.34 7.77 K=6
1051
+ verify 37: accepted 5/5 (K=6) + correction -> +6 tokens (158.5 ms) cpos=123
1052
+ [margins] 10.91 4.84 9.61 5.38 12.63 K=6
1053
+ verify 38: accepted 5/5 (K=6) + correction -> +6 tokens (179.6 ms) cpos=129
1054
+ [margins] 3.89 5.10 9.12 7.38 5.05 K=6
1055
+ verify 39: accepted 5/5 (K=6) + correction -> +6 tokens (162.0 ms) cpos=135
1056
+ [margins] 10.03 6.48 9.81 5.03 10.76 K=6
1057
+ verify 40: accepted 5/5 (K=6) + correction -> +6 tokens (168.7 ms) cpos=141
1058
+ [margins] 3.76 6.69 6.47 8.40 7.78 K=6
1059
+ verify 41: accepted 5/5 (K=6) + correction -> +6 tokens (161.3 ms) cpos=147
1060
+ [margins] 8.99 3.35 6.89 4.19 8.54 K=6
1061
+ verify 42: accepted 5/5 (K=6) + correction -> +6 tokens (167.2 ms) cpos=153
1062
+ [margins] 2.88 7.04 6.51 4.95 5.13 K=6
1063
+ verify 43: accepted 5/5 (K=6) + correction -> +6 tokens (162.5 ms) cpos=159
1064
+ [margins] 7.32 4.28 3.54 3.00 8.26 K=6
1065
+ verify 44: accepted 5/5 (K=6) + correction -> +6 tokens (166.1 ms) cpos=165
1066
+ [margins] 3.38 5.89 4.80 7.02 6.18 K=6
1067
+ verify 45: accepted 5/5 (K=6) + correction -> +6 tokens (160.3 ms) cpos=171
1068
+ [margins] 10.74 6.38 9.90 5.03 10.46 K=6
1069
+ verify 46: accepted 5/5 (K=6) + correction -> +6 tokens (168.0 ms) cpos=177
1070
+ [margins] 4.34 4.82 7.48 6.66 8.75 K=6
1071
+ verify 47: accepted 5/5 (K=6) + correction -> +6 tokens (160.2 ms) cpos=183
1072
+ [margins] 9.50 6.65 9.90 3.48 12.31 K=6
1073
+ verify 48: accepted 5/5 (K=6) + correction -> +6 tokens (168.3 ms) cpos=189
1074
+ [margins] 4.26 5.76 8.87 7.82 8.80 K=6
1075
+ verify 49: accepted 5/5 (K=6) + correction -> +6 tokens (160.6 ms) cpos=195
1076
+ [margins] 9.91 6.53 10.76 6.32 9.55 K=6
1077
+ verify 50: accepted 5/5 (K=6) + correction -> +6 tokens (168.1 ms) cpos=201
1078
+ [margins] 4.20 5.62 7.81 7.26 5.96 K=6
1079
+ verify 51: accepted 5/5 (K=6) + correction -> +6 tokens (161.6 ms) cpos=207
1080
+ [margins] 11.14 6.88 10.22 7.88 10.18 K=6
1081
+ verify 52: accepted 5/5 (K=6) + correction -> +6 tokens (167.3 ms) cpos=213
1082
+
1083
+ [spec] generated 201 tokens over 52 verifies: mean tokens/verify = 3.87 (block=5, max 6)
1084
+ [spec] tokens: 1 201 1 201 1 201 1 201 1 201 1 201 1 201 1 201 1 201 1 201 1 201 1 201 1 201 1 201 1 201 1 201 3085 344 270 6102 294 8760 33 9361
1085
+ [spec] SPEC-DECODE: 34.4 ms/tok = 29.10 tok/s (vs base M=1 71.5 ms/tok = 13.99 tok/s -> 2.08x)
1086
+
1087
+ [blksweep] BLK passes adaptK->eff prompt | mean tok/verify | ms/tok | tok/s | vs base 13.99 tok/s
1088
+ [blksweep] 5 1 1.50 1.50 0 | 3.30 | 38.9 | 25.74 | 1.84x
1089
+ [blksweep] 5 1 1.50 1.50 1 | 4.00 | 34.2 | 29.25 | 2.09x
1090
+ [blksweep] 5 1 1.50 1.50 2 | 3.96 | 35.1 | 28.45 | 2.03x
1091
+ [blksweep] 5 1 1.50 1.50 3 | 2.66 | 45.4 | 22.01 | 1.57x
1092
+ [blksweep] 5 1 1.50 1.50 4 | 3.00 | 41.0 | 24.41 | 1.75x
1093
+ [blksweep] 5 1 1.50 1.50 5 | 4.43 | 32.3 | 30.94 | 2.21x
1094
+ [blksweep] 5 1 1.50 1.50 6 | 5.21 | 28.5 | 35.06 | 2.51x
1095
+ [blksweep] 5 1 1.50 1.50 7 | 3.64 | 35.8 | 27.94 | 2.00x
1096
+ [blksweep] 5 1 1.50 1.50 8 | 3.87 | 34.4 | 29.10 | 2.08x
1097
+ [decode] mem 106.4/122.8 GiB