char-fork/llama.cpp

git clone https://git.t4t.associates/char-fork/llama.cpp

Georgi Gerganovmetal : single-source fusion table + fusion debug rework (#28164)a2878d30d

char/aliceai-t5-moe
10.8 KiB154 linesraw
1# test-fusion baseline for device MTL
2# arch              ,moe ,mode    ,label                       ,  count
3arcee               ,0   ,any     ,RMS_NORM+MUL                ,      5
4arctic              ,0   ,any     ,RMS_NORM+MUL                ,      7
5baichuan            ,0   ,any     ,RMS_NORM+MUL                ,      5
6bailingmoe          ,1   ,any     ,ADD+ADD                     ,      2
7bailingmoe          ,1   ,any     ,RMS_NORM+MUL                ,      5
8bailingmoe2         ,1   ,any     ,ADD+ADD                     ,      1
9bailingmoe2         ,1   ,any     ,RMS_NORM+MUL                ,      9
10bailingmoe3         ,1   ,any     ,ADD+ADD                     ,      1
11bailingmoe3         ,1   ,any     ,GATED_DELTA_NET+CPY         ,      1
12bailingmoe3         ,1   ,any     ,RMS_NORM+MUL                ,      8
13bloom               ,0   ,any     ,NORM+MUL+ADD                ,      6
14chatglm             ,0   ,any     ,RMS_NORM+MUL                ,      5
15codeshell           ,0   ,any     ,NORM+MUL+ADD                ,      5
16cogvlm              ,0   ,any     ,RMS_NORM+MUL                ,      5
17command-r           ,0   ,any     ,NORM+MUL                    ,      3
18dbrx                ,0   ,any     ,NORM+MUL                    ,      5
19deci                ,0   ,any     ,RMS_NORM+MUL                ,      5
20deepseek            ,0   ,any     ,ADD+ADD                     ,      1
21deepseek            ,0   ,any     ,RMS_NORM+MUL                ,      5
22deepseek2           ,0   ,any     ,ADD+ADD                     ,      1
23deepseek2           ,0   ,any     ,RMS_NORM+MUL                ,      9
24deepseek32          ,0   ,any     ,ADD+ADD                     ,      1
25deepseek32          ,0   ,any     ,NORM+MUL+ADD                ,      2
26deepseek32          ,0   ,any     ,RMS_NORM+MUL                ,      9
27deepseek4           ,0   ,any     ,RMS_NORM+MUL                ,     20
28dots1               ,0   ,any     ,ADD+ADD                     ,      1
29dots1               ,0   ,any     ,RMS_NORM+MUL                ,      9
30dream               ,0   ,any     ,RMS_NORM+MUL                ,      5
31ernie4_5-moe        ,1   ,any     ,ADD+ADD                     ,      1
32ernie4_5-moe        ,1   ,any     ,RMS_NORM+MUL                ,      5
33ernie4_5            ,0   ,any     ,RMS_NORM+MUL                ,      5
34exaone              ,0   ,any     ,RMS_NORM+MUL                ,      5
35exaone4             ,0   ,any     ,RMS_NORM+MUL                ,      5
36exaone4             ,0   ,any     ,RMS_NORM+MUL+ADD            ,      4
37falcon              ,0   ,any     ,ADD+ADD                     ,      2
38falcon              ,0   ,any     ,NORM+MUL+ADD                ,      5
39falcon-h1           ,0   ,any     ,ADD+ADD                     ,      2
40falcon-h1           ,0   ,any     ,RMS_NORM+MUL                ,      9
41gemma               ,0   ,any     ,RMS_NORM+MUL                ,      5
42gemma2              ,0   ,any     ,RMS_NORM+MUL                ,      5
43gemma2              ,0   ,any     ,RMS_NORM+MUL+ADD            ,      4
44glm-dsa             ,0   ,any     ,ADD+ADD                     ,      1
45glm-dsa             ,0   ,any     ,NORM+MUL+ADD                ,      2
46glm-dsa             ,0   ,any     ,RMS_NORM+MUL                ,      9
47glm4                ,0   ,any     ,RMS_NORM+MUL                ,      5
48glm4                ,0   ,any     ,RMS_NORM+MUL+ADD            ,      4
49glm4moe             ,1   ,any     ,ADD+ADD                     ,      1
50glm4moe             ,1   ,any     ,RMS_NORM+MUL                ,      9
51gpt-oss             ,0   ,any     ,RMS_NORM+MUL                ,      5
52gpt2                ,0   ,any     ,NORM+MUL+ADD                ,      5
53gptneox             ,0   ,any     ,NORM+MUL+ADD                ,      5
54granite             ,0   ,any     ,RMS_NORM+MUL                ,      5
55granite             ,0   ,any     ,RMS_NORM+MUL                ,      5
56granitehybrid       ,0   ,any     ,RMS_NORM+MUL                ,      6
57granitemoe          ,1   ,any     ,RMS_NORM+MUL                ,      5
58granitemoe          ,1   ,any     ,RMS_NORM+MUL                ,      5
59grok                ,0   ,any     ,RMS_NORM+MUL                ,      5
60grok                ,0   ,any     ,RMS_NORM+MUL+ADD            ,      4
61grovemoe            ,1   ,any     ,ADD+ADD                     ,      2
62grovemoe            ,1   ,any     ,RMS_NORM+MUL                ,      9
63hunyuan-dense       ,0   ,any     ,RMS_NORM+MUL                ,      9
64hunyuan-moe         ,1   ,any     ,ADD+ADD                     ,      2
65hunyuan-moe         ,1   ,any     ,RMS_NORM+MUL                ,      9
66hunyuan_vl          ,0   ,any     ,RMS_NORM+MUL                ,      9
67hy_v3               ,0   ,any     ,ADD+ADD                     ,      2
68hy_v3               ,0   ,any     ,RMS_NORM+MUL                ,      9
69hy_v4               ,0   ,any     ,NORM+MUL+ADD                ,      1
70hy_v4               ,0   ,any     ,RMS_NORM+MUL                ,      9
71internlm2           ,0   ,any     ,RMS_NORM+MUL                ,      5
72jais                ,0   ,any     ,NORM+MUL+ADD                ,      5
73jais2               ,0   ,any     ,NORM+MUL+ADD                ,      5
74jamba               ,0   ,any     ,RMS_NORM+MUL                ,      8
75kimi-k3             ,0   ,any     ,GATED_DELTA_NET+CPY         ,      1
76kimi-k3             ,0   ,any     ,RMS_NORM+MUL                ,     17
77kimi-linear         ,0   ,any     ,ADD+ADD                     ,      1
78kimi-linear         ,0   ,any     ,GATED_DELTA_NET+CPY         ,      1
79kimi-linear         ,0   ,any     ,RMS_NORM+MUL                ,      7
80lfm2                ,0   ,any     ,RMS_NORM+MUL                ,      7
81lfm2moe             ,1   ,any     ,RMS_NORM+MUL                ,      7
82llada               ,0   ,any     ,RMS_NORM+MUL                ,      5
83llada-moe           ,1   ,any     ,RMS_NORM+MUL                ,      9
84llama               ,0   ,any     ,RMS_NORM+MUL                ,      5
85llama               ,0   ,any     ,RMS_NORM+MUL                ,      5
86llama4              ,0   ,any     ,ADD+ADD                     ,      2
87llama4              ,0   ,any     ,RMS_NORM+MUL                ,      9
88maincoder           ,0   ,any     ,RMS_NORM+MUL                ,      9
89mamba               ,0   ,any     ,RMS_NORM+MUL                ,      3
90mamba2              ,0   ,any     ,RMS_NORM+MUL                ,      5
91minicpm             ,0   ,any     ,RMS_NORM+MUL                ,      5
92minicpm             ,0   ,any     ,RMS_NORM+MUL                ,      5
93minicpm3            ,0   ,any     ,RMS_NORM+MUL                ,      9
94minimax-01          ,0   ,any     ,RMS_NORM+MUL                ,      6
95minimax-m2          ,0   ,any     ,RMS_NORM+MUL                ,      9
96minimax-m3          ,0   ,any     ,ADD+ADD                     ,      1
97minimax-m3          ,0   ,any     ,RMS_NORM+MUL                ,     11
98mistral3            ,0   ,any     ,RMS_NORM+MUL                ,      5
99mistral3            ,0   ,any     ,RMS_NORM+MUL                ,      5
100mistral4            ,0   ,any     ,ADD+ADD                     ,      1
101mistral4            ,0   ,any     ,RMS_NORM+MUL                ,      9
102mpt                 ,0   ,any     ,NORM+MUL+ADD                ,      5
103nanbeige            ,0   ,any     ,RMS_NORM+MUL                ,      5
104nemotron            ,0   ,any     ,NORM+MUL+ADD                ,      5
105nemotron_h          ,0   ,any     ,RMS_NORM+MUL                ,      5
106nemotron_h_moe      ,1   ,any     ,RMS_NORM+MUL                ,      5
107olmoe               ,1   ,any     ,RMS_NORM+MUL                ,      9
108openelm             ,0   ,any     ,RMS_NORM+MUL                ,      9
109orion               ,0   ,any     ,NORM+MUL+ADD                ,      5
110paddleocr           ,0   ,any     ,RMS_NORM+MUL                ,      5
111pangu-embedded      ,0   ,any     ,RMS_NORM+MUL                ,      5
112phi2                ,0   ,any     ,ADD+ADD                     ,      2
113phi2                ,0   ,any     ,NORM+MUL+ADD                ,      3
114phi3                ,0   ,any     ,RMS_NORM+MUL                ,      5
115phimoe              ,1   ,any     ,RMS_NORM+MUL+ADD            ,      5
116plamo               ,0   ,any     ,ADD+ADD                     ,      2
117plamo               ,0   ,any     ,RMS_NORM+MUL                ,      3
118plamo2              ,0   ,any     ,RMS_NORM+MUL                ,     10
119plamo2              ,0   ,any     ,RMS_NORM+MUL+ADD            ,      4
120pockettts           ,0   ,any     ,NORM+MUL+ADD                ,      5
121qwen                ,0   ,any     ,RMS_NORM+MUL                ,      5
122qwen2               ,0   ,any     ,RMS_NORM+MUL                ,      5
123qwen2moe            ,1   ,any     ,ADD+ADD                     ,      2
124qwen2moe            ,1   ,any     ,RMS_NORM+MUL                ,      5
125qwen2vl             ,0   ,any     ,RMS_NORM+MUL                ,      5
126qwen3               ,0   ,any     ,RMS_NORM+MUL                ,      9
127qwen35              ,0   ,any     ,GATED_DELTA_NET+CPY         ,      1
128qwen35              ,0   ,any     ,RMS_NORM+MUL                ,      8
129qwen35moe           ,1   ,any     ,ADD+ADD                     ,      2
130qwen35moe           ,1   ,any     ,GATED_DELTA_NET+CPY         ,      1
131qwen35moe           ,1   ,any     ,RMS_NORM+MUL                ,      8
132qwen3moe            ,1   ,any     ,RMS_NORM+MUL                ,      9
133qwen3next           ,0   ,any     ,ADD+ADD                     ,      2
134qwen3next           ,0   ,any     ,GATED_DELTA_NET+CPY         ,      1
135qwen3next           ,0   ,any     ,RMS_NORM+MUL                ,      8
136qwen3tts            ,0   ,any     ,RMS_NORM+MUL                ,      9
137qwen3vl             ,0   ,any     ,RMS_NORM+MUL                ,      9
138qwen3vlmoe          ,1   ,any     ,RMS_NORM+MUL                ,      9
139qwen4exp            ,0   ,any     ,ADD+ADD+ADD                 ,      5
140qwen4exp            ,0   ,any     ,ADD+ADD+ADD+ADD+ADD+ADD+ADD ,      9
141qwen4exp            ,0   ,any     ,GATED_DELTA_NET+CPY         ,      1
142qwen4exp            ,0   ,any     ,RMS_NORM+MUL                ,      5
143refact              ,0   ,any     ,RMS_NORM+MUL                ,      5
144refact              ,0   ,any     ,RMS_NORM+MUL                ,      5
145rnd1                ,0   ,any     ,RMS_NORM+MUL                ,      9
146seed_oss            ,0   ,any     ,RMS_NORM+MUL                ,      5
147smallthinker        ,0   ,any     ,RMS_NORM+MUL                ,      5
148smollm3             ,0   ,any     ,RMS_NORM+MUL                ,      5
149stablelm            ,0   ,any     ,NORM+MUL                    ,      4
150stablelm            ,0   ,any     ,NORM+MUL+ADD                ,      5
151starcoder           ,0   ,any     ,NORM+MUL+ADD                ,      5
152starcoder2          ,0   ,any     ,NORM+MUL+ADD                ,      5
153talkie              ,0   ,any     ,ADD+ADD                     ,      2
154xverse              ,0   ,any     ,RMS_NORM+MUL                ,      5