File size: 13,557 Bytes
2107810
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
warning: not compiled with GPU offload support, --n-gpu-layers option will be ignored
warning: see main README.md for information on enabling GPU BLAS support
main: build = 8 (589f21d)
main: built with cc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0 for x86_64-linux-gnu
main: seed  = 1717069408
llama_model_loader: loaded meta data with 24 key-value pairs and 1875 tensors from Yuan2-M32-hf-IMat-GGUF/Yuan2-M32-hf.gguf (version GGUF V3 (latest))
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
llama_model_loader: - kv   0:                       general.architecture str              = yuan2_moe
llama_model_loader: - kv   1:                               general.name str              = yuan2_moe
llama_model_loader: - kv   2:                   yuan2_moe.context_length u32              = 4096
llama_model_loader: - kv   3:                 yuan2_moe.embedding_length u32              = 2048
llama_model_loader: - kv   4:                      yuan2_moe.block_count u32              = 24
llama_model_loader: - kv   5:              yuan2_moe.feed_forward_length u32              = 8192
llama_model_loader: - kv   6:             yuan2_moe.rope.dimension_count u32              = 128
llama_model_loader: - kv   7:             yuan2_moe.attention.head_count u32              = 16
llama_model_loader: - kv   8:          yuan2_moe.attention.head_count_kv u32              = 16
llama_model_loader: - kv   9:                     yuan2_moe.expert_count u32              = 32
llama_model_loader: - kv  10:                yuan2_moe.expert_used_count u32              = 2
llama_model_loader: - kv  11: yuan2_moe.attention.layer_norm_rms_epsilon f32              = 0.000001
llama_model_loader: - kv  12:                          general.file_type u32              = 1
llama_model_loader: - kv  13:                       tokenizer.ggml.model str              = llama
llama_model_loader: - kv  14:                      tokenizer.ggml.tokens arr[str,135040]  = ["<unk>", "<s>", "</s>", "<0x00>", "<...
llama_model_loader: - kv  15:                      tokenizer.ggml.scores arr[f32,135040]  = [0.000000, 0.000000, 0.000000, 0.0000...
llama_model_loader: - kv  16:                  tokenizer.ggml.token_type arr[i32,135040]  = [2, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 6, ...
llama_model_loader: - kv  17:                tokenizer.ggml.bos_token_id u32              = 77185
llama_model_loader: - kv  18:                tokenizer.ggml.eos_token_id u32              = 77185
llama_model_loader: - kv  19:          tokenizer.ggml.seperator_token_id u32              = 77185
llama_model_loader: - kv  20:            tokenizer.ggml.padding_token_id u32              = 77185
llama_model_loader: - kv  21:               tokenizer.ggml.mask_token_id u32              = 77185
llama_model_loader: - kv  22:               tokenizer.ggml.add_bos_token bool             = false
llama_model_loader: - kv  23:               tokenizer.ggml.add_eos_token bool             = false
llama_model_loader: - type  f32:  121 tensors
llama_model_loader: - type  f16: 1754 tensors
llm_load_vocab: mismatch in special tokens definition ( 2729/135040 vs 346/135040 ).
llm_load_print_meta: format           = GGUF V3 (latest)
llm_load_print_meta: arch             = yuan2_moe
llm_load_print_meta: vocab type       = SPM
llm_load_print_meta: n_vocab          = 135040
llm_load_print_meta: n_merges         = 0
llm_load_print_meta: n_ctx_train      = 4096
llm_load_print_meta: n_embd           = 2048
llm_load_print_meta: n_head           = 16
llm_load_print_meta: n_head_kv        = 16
llm_load_print_meta: n_layer          = 24
llm_load_print_meta: n_rot            = 128
llm_load_print_meta: n_embd_head_k    = 256
llm_load_print_meta: n_embd_head_v    = 256
llm_load_print_meta: n_gqa            = 1
llm_load_print_meta: n_embd_k_gqa     = 4096
llm_load_print_meta: n_embd_v_gqa     = 4096
llm_load_print_meta: f_norm_eps       = 0.0e+00
llm_load_print_meta: f_norm_rms_eps   = 1.0e-06
llm_load_print_meta: f_clamp_kqv      = 0.0e+00
llm_load_print_meta: f_max_alibi_bias = 0.0e+00
llm_load_print_meta: n_ff             = 8192
llm_load_print_meta: n_expert         = 32
llm_load_print_meta: n_expert_used    = 2
llm_load_print_meta: rope scaling     = linear
llm_load_print_meta: freq_base_train  = 10000.0
llm_load_print_meta: freq_scale_train = 1
llm_load_print_meta: n_yarn_orig_ctx  = 4096
llm_load_print_meta: rope_finetuned   = unknown
llm_load_print_meta: model type       = 40B
llm_load_print_meta: model ftype      = F16
llm_load_print_meta: model params     = 40.22 B
llm_load_print_meta: model size       = 74.91 GiB (16.00 BPW) 
llm_load_print_meta: general.name     = yuan2_moe
llm_load_print_meta: BOS token        = 77185 '<eod>'
llm_load_print_meta: EOS token        = 77185 '<eod>'
llm_load_print_meta: UNK token        = 0 '<unk>'
llm_load_print_meta: SEP token        = 77185 '<eod>'
llm_load_print_meta: PAD token        = 77185 '<eod>'
llm_load_print_meta: LF token         = 13 '<0x0A>'
llm_load_tensors: ggml ctx size =    1.06 MiB
llm_load_tensors:        CPU buffer size = 76712.85 MiB
....................................................................................................
llama_new_context_with_model: n_ctx      = 512
llama_new_context_with_model: freq_base  = 10000.0
llama_new_context_with_model: freq_scale = 1
llama_kv_cache_init:        CPU KV buffer size =   192.38 MiB
llama_new_context_with_model: KV self size  =  192.00 MiB, K (f16):   96.00 MiB, V (f16):   96.00 MiB
llama_new_context_with_model:        CPU input buffer size   =     6.01 MiB
llama_new_context_with_model:        CPU compute buffer size =   275.88 MiB
llama_new_context_with_model: graph splits (measure): 1

system_info: n_threads = 25 / 32 | AVX = 1 | AVX_VNNI = 0 | AVX2 = 1 | AVX512 = 1 | AVX512_VBMI = 1 | AVX512_VNNI = 1 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 0 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 | MATMUL_INT8 = 0 | 
compute_imatrix: tokenizing the input ..
compute_imatrix: tokenization took 145.195 ms
compute_imatrix: computing over 227 chunks with batch_size 512
compute_imatrix: 6.78 seconds per pass - ETA 25.65 minutes
[1]6.5558,[2]5.0024,[3]5.8688,[4]6.8477,
save_imatrix: stored collected data after 10 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[5]7.1188,[6]7.3938,[7]6.3059,[8]6.7000,[9]6.7433,
save_imatrix: stored collected data after 20 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[10]7.2226,[11]7.5399,[12]7.9695,[13]8.3739,[14]7.8757,
save_imatrix: stored collected data after 30 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[15]8.2691,[16]8.3867,[17]8.6503,[18]8.8671,[19]9.0091,
save_imatrix: stored collected data after 40 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[20]9.1840,[21]9.2799,[22]9.0400,[23]8.6569,[24]8.7209,
save_imatrix: stored collected data after 50 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[25]8.1930,[26]7.7850,[27]7.5167,[28]7.2632,[29]6.9825,
save_imatrix: stored collected data after 60 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[30]7.3951,[31]7.3448,[32]7.5309,[33]7.3902,[34]7.3415,
save_imatrix: stored collected data after 70 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[35]7.0182,[36]6.7552,[37]6.4881,[38]6.4493,[39]6.4314,
save_imatrix: stored collected data after 80 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[40]6.4113,[41]6.4790,[42]6.5106,[43]6.6147,[44]6.7387,
save_imatrix: stored collected data after 90 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[45]6.8810,[46]6.9200,[47]7.0145,[48]7.1681,[49]7.3595,
save_imatrix: stored collected data after 100 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[50]7.5659,[51]7.6763,[52]7.6824,[53]7.6250,[54]7.5348,
save_imatrix: stored collected data after 110 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[55]7.4143,[56]7.5141,[57]7.5952,[58]7.6657,[59]7.7468,
save_imatrix: stored collected data after 120 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[60]7.7793,[61]7.8927,[62]7.9396,[63]8.0284,[64]8.0642,
save_imatrix: stored collected data after 130 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[65]8.1467,[66]8.2391,[67]8.2560,[68]8.2841,[69]8.3739,
save_imatrix: stored collected data after 140 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[70]8.5129,[71]8.6025,[72]8.6186,[73]8.5808,[74]8.5795,
save_imatrix: stored collected data after 150 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[75]8.6234,[76]8.6382,[77]8.6283,[78]8.5947,[79]8.5555,
save_imatrix: stored collected data after 160 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[80]8.5826,[81]8.5578,[82]8.5244,[83]8.5821,[84]8.6244,
save_imatrix: stored collected data after 170 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[85]8.6236,[86]8.6298,[87]8.6283,[88]8.6120,[89]8.5763,
save_imatrix: stored collected data after 180 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[90]8.5623,[91]8.5481,[92]8.5802,[93]8.5909,[94]8.6161,
save_imatrix: stored collected data after 190 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[95]8.6344,[96]8.6459,[97]8.6175,[98]8.5774,[99]8.5987,
save_imatrix: stored collected data after 200 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[100]8.6399,[101]8.6517,[102]8.6304,[103]8.5887,[104]8.5743,
save_imatrix: stored collected data after 210 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[105]8.5806,[106]8.6410,[107]8.6477,[108]8.6413,[109]8.6781,
save_imatrix: stored collected data after 220 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[110]8.7056,[111]8.7298,[112]8.7006,[113]8.7188,[114]8.7299,
save_imatrix: stored collected data after 230 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[115]8.7238,[116]8.7355,[117]8.7230,[118]8.7087,[119]8.6851,
save_imatrix: stored collected data after 240 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[120]8.6678,[121]8.7011,[122]8.7138,[123]8.7270,[124]8.7333,
save_imatrix: stored collected data after 250 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[125]8.7264,[126]8.7703,[127]8.6855,[128]8.6916,[129]8.6864,
save_imatrix: stored collected data after 260 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[130]8.6621,[131]8.6297,[132]8.5725,[133]8.5187,[134]8.4626,
save_imatrix: stored collected data after 270 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[135]8.4089,[136]8.3466,[137]8.2958,[138]8.2468,[139]8.1934,
save_imatrix: stored collected data after 280 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[140]8.1758,[141]8.1481,[142]8.1078,[143]8.0591,[144]8.0292,
save_imatrix: stored collected data after 290 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[145]8.0041,[146]7.9806,[147]7.9489,[148]7.9131,[149]7.8880,
save_imatrix: stored collected data after 300 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[150]7.8550,[151]7.8215,[152]7.8359,[153]7.7834,[154]7.7633,
save_imatrix: stored collected data after 310 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[155]7.8523,[156]7.9496,[157]8.0368,[158]8.1009,[159]8.1471,
save_imatrix: stored collected data after 320 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[160]8.1647,[161]8.1946,[162]8.2248,[163]8.2249,[164]8.2250,
save_imatrix: stored collected data after 330 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[165]8.2689,[166]8.2907,[167]8.3209,[168]8.3421,[169]8.3452,
save_imatrix: stored collected data after 340 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[170]8.3679,[171]8.3964,[172]8.4240,[173]8.4850,[174]8.5304,
save_imatrix: stored collected data after 350 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[175]8.5662,[176]8.5695,[177]8.5835,[178]8.5995,[179]8.6060,
save_imatrix: stored collected data after 360 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[180]8.5969,[181]8.6079,[182]8.6110,[183]8.6246,[184]8.6350,
save_imatrix: stored collected data after 370 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[185]8.6868,[186]8.6822,[187]8.6447,[188]8.6949,[189]8.7372,
save_imatrix: stored collected data after 380 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[190]8.7771,[191]8.8391,[192]8.9060,[193]8.8979,[194]8.8961,
save_imatrix: stored collected data after 390 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[195]8.8519,[196]8.8544,[197]8.8626,[198]8.8997,[199]8.8840,
save_imatrix: stored collected data after 400 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[200]8.8698,[201]8.8631,[202]8.8561,[203]8.8675,[204]8.8596,
save_imatrix: stored collected data after 410 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[205]8.8652,[206]8.8837,[207]8.8930,[208]8.9143,[209]8.9504,
save_imatrix: stored collected data after 420 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[210]8.9338,[211]8.9362,[212]8.9262,[213]8.9121,[214]8.9058,
save_imatrix: stored collected data after 430 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[215]8.8761,[216]8.8852,[217]8.8815,[218]8.8830,[219]8.8819,
save_imatrix: stored collected data after 440 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[220]8.9316,[221]8.9905,[222]9.0298,[223]9.0636,[224]9.1087,
save_imatrix: stored collected data after 450 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat
[225]9.0895,[226]9.0782,[227]9.0585,
save_imatrix: stored collected data after 454 chunks in Yuan2-M32-hf-IMat-GGUF/imatrix.dat

llama_print_timings:        load time =   51125.10 ms
llama_print_timings:      sample time =       0.00 ms /     1 runs   (    0.00 ms per token,      inf tokens per second)
llama_print_timings: prompt eval time = 1527383.56 ms / 116224 tokens (   13.14 ms per token,    76.09 tokens per second)
llama_print_timings:        eval time =       0.00 ms /     1 runs   (    0.00 ms per token,      inf tokens per second)
llama_print_timings:       total time = 1572971.60 ms / 116225 tokens

Final estimate: PPL = 9.0585 +/- 0.10484