File size: 12,000 Bytes
7fd1065
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
main: build = 2998 (9588f196)
main: built with cc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0 for x86_64-linux-gnu
main: seed  = 1716669278
llama_model_loader: loaded meta data with 28 key-value pairs and 258 tensors from aya-23-8B-IMat-GGUF/aya-23-8B.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              = command-r
llama_model_loader: - kv   1:                               general.name str              = aya-23-8B
llama_model_loader: - kv   2:                      command-r.block_count u32              = 32
llama_model_loader: - kv   3:                   command-r.context_length u32              = 8192
llama_model_loader: - kv   4:                 command-r.embedding_length u32              = 4096
llama_model_loader: - kv   5:              command-r.feed_forward_length u32              = 14336
llama_model_loader: - kv   6:             command-r.attention.head_count u32              = 32
llama_model_loader: - kv   7:          command-r.attention.head_count_kv u32              = 8
llama_model_loader: - kv   8:                   command-r.rope.freq_base f32              = 10000.000000
llama_model_loader: - kv   9:     command-r.attention.layer_norm_epsilon f32              = 0.000010
llama_model_loader: - kv  10:                          general.file_type u32              = 1
llama_model_loader: - kv  11:                      command-r.logit_scale f32              = 0.062500
llama_model_loader: - kv  12:                command-r.rope.scaling.type str              = none
llama_model_loader: - kv  13:                       tokenizer.ggml.model str              = gpt2
llama_model_loader: - kv  14:                         tokenizer.ggml.pre str              = command-r
llama_model_loader: - kv  15:                      tokenizer.ggml.tokens arr[str,256000]  = ["<PAD>", "<UNK>", "<CLS>", "<SEP>", ...
llama_model_loader: - kv  16:                  tokenizer.ggml.token_type arr[i32,256000]  = [3, 3, 3, 3, 3, 3, 3, 3, 1, 1, 1, 1, ...
llama_model_loader: - kv  17:                      tokenizer.ggml.merges arr[str,253333]  = ["Ġ Ġ", "Ġ t", "e r", "i n", "Ġ a...
llama_model_loader: - kv  18:                tokenizer.ggml.bos_token_id u32              = 5
llama_model_loader: - kv  19:                tokenizer.ggml.eos_token_id u32              = 255001
llama_model_loader: - kv  20:            tokenizer.ggml.padding_token_id u32              = 0
llama_model_loader: - kv  21:               tokenizer.ggml.add_bos_token bool             = true
llama_model_loader: - kv  22:               tokenizer.ggml.add_eos_token bool             = false
llama_model_loader: - kv  23:           tokenizer.chat_template.tool_use str              = {{ bos_token }}{% if messages[0]['rol...
llama_model_loader: - kv  24:                tokenizer.chat_template.rag str              = {{ bos_token }}{% if messages[0]['rol...
llama_model_loader: - kv  25:                   tokenizer.chat_templates arr[str,2]       = ["tool_use", "rag"]
llama_model_loader: - kv  26:                    tokenizer.chat_template str              = {{ bos_token }}{% if messages[0]['rol...
llama_model_loader: - kv  27:               general.quantization_version u32              = 2
llama_model_loader: - type  f32:   33 tensors
llama_model_loader: - type  f16:  225 tensors
llm_load_vocab: special tokens definition check successful ( 1008/256000 ).
llm_load_print_meta: format           = GGUF V3 (latest)
llm_load_print_meta: arch             = command-r
llm_load_print_meta: vocab type       = BPE
llm_load_print_meta: n_vocab          = 256000
llm_load_print_meta: n_merges         = 253333
llm_load_print_meta: n_ctx_train      = 8192
llm_load_print_meta: n_embd           = 4096
llm_load_print_meta: n_head           = 32
llm_load_print_meta: n_head_kv        = 8
llm_load_print_meta: n_layer          = 32
llm_load_print_meta: n_rot            = 128
llm_load_print_meta: n_embd_head_k    = 128
llm_load_print_meta: n_embd_head_v    = 128
llm_load_print_meta: n_gqa            = 4
llm_load_print_meta: n_embd_k_gqa     = 1024
llm_load_print_meta: n_embd_v_gqa     = 1024
llm_load_print_meta: f_norm_eps       = 1.0e-05
llm_load_print_meta: f_norm_rms_eps   = 0.0e+00
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: f_logit_scale    = 6.2e-02
llm_load_print_meta: n_ff             = 14336
llm_load_print_meta: n_expert         = 0
llm_load_print_meta: n_expert_used    = 0
llm_load_print_meta: causal attn      = 1
llm_load_print_meta: pooling type     = 0
llm_load_print_meta: rope type        = 0
llm_load_print_meta: rope scaling     = none
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  = 8192
llm_load_print_meta: rope_finetuned   = unknown
llm_load_print_meta: ssm_d_conv       = 0
llm_load_print_meta: ssm_d_inner      = 0
llm_load_print_meta: ssm_d_state      = 0
llm_load_print_meta: ssm_dt_rank      = 0
llm_load_print_meta: model type       = ?B
llm_load_print_meta: model ftype      = F16
llm_load_print_meta: model params     = 8.03 B
llm_load_print_meta: model size       = 14.95 GiB (16.00 BPW) 
llm_load_print_meta: general.name     = aya-23-8B
llm_load_print_meta: BOS token        = 5 '<BOS_TOKEN>'
llm_load_print_meta: EOS token        = 255001 '<|END_OF_TURN_TOKEN|>'
llm_load_print_meta: PAD token        = 0 '<PAD>'
llm_load_print_meta: LF token         = 136 'Ä'
ggml_cuda_init: GGML_CUDA_FORCE_MMQ:   no
ggml_cuda_init: CUDA_USE_TENSOR_CORES: yes
ggml_cuda_init: found 1 CUDA devices:
  Device 0: NVIDIA GeForce RTX 4090, compute capability 8.9, VMM: yes
llm_load_tensors: ggml ctx size =    0.27 MiB
llm_load_tensors: offloading 20 repeating layers to GPU
llm_load_tensors: offloaded 20/33 layers to GPU
llm_load_tensors:        CPU buffer size = 15312.52 MiB
llm_load_tensors:      CUDA0 buffer size =  8320.31 MiB
...............................................................................
llama_new_context_with_model: n_ctx      = 512
llama_new_context_with_model: n_batch    = 512
llama_new_context_with_model: n_ubatch   = 512
llama_new_context_with_model: flash_attn = 0
llama_new_context_with_model: freq_base  = 10000.0
llama_new_context_with_model: freq_scale = 1
llama_kv_cache_init:  CUDA_Host KV buffer size =    24.00 MiB
llama_kv_cache_init:      CUDA0 KV buffer size =    40.00 MiB
llama_new_context_with_model: KV self size  =   64.00 MiB, K (f16):   32.00 MiB, V (f16):   32.00 MiB
llama_new_context_with_model:  CUDA_Host  output buffer size =     0.98 MiB
llama_new_context_with_model:      CUDA0 compute buffer size =  2516.00 MiB
llama_new_context_with_model:  CUDA_Host compute buffer size =     9.01 MiB
llama_new_context_with_model: graph nodes  = 968
llama_new_context_with_model: graph splits = 124

system_info: n_threads = 25 / 32 | AVX = 1 | AVX_VNNI = 0 | AVX2 = 1 | AVX512 = 1 | AVX512_VBMI = 1 | AVX512_VNNI = 1 | AVX512_BF16 = 1 | FMA = 1 | NEON = 0 | SVE = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 | MATMUL_INT8 = 0 | LLAMAFILE = 1 | 
compute_imatrix: tokenizing the input ..
compute_imatrix: tokenization took 187.854 ms
compute_imatrix: computing over 194 chunks with batch_size 512
compute_imatrix: 0.91 seconds per pass - ETA 2.92 minutes
[1]7.1190,[2]4.7963,[3]4.6059,[4]5.2739,[5]5.1456,[6]4.7816,[7]5.7129,[8]6.0872,[9]6.7418,
save_imatrix: stored collected data after 10 chunks in aya-23-8B-IMat-GGUF/imatrix.dat
[10]7.0130,[11]7.1909,[12]7.3099,[13]7.7753,[14]7.9571,[15]8.3000,[16]8.5210,[17]8.6988,[18]8.9548,[19]9.0360,
save_imatrix: stored collected data after 20 chunks in aya-23-8B-IMat-GGUF/imatrix.dat
[20]8.6010,[21]8.3664,[22]8.1867,[23]7.7215,[24]7.4229,[25]7.3781,[26]7.5362,[27]7.4628,[28]7.6019,[29]7.4947,
save_imatrix: stored collected data after 30 chunks in aya-23-8B-IMat-GGUF/imatrix.dat
[30]7.5277,[31]7.1883,[32]7.0002,[33]6.9088,[34]6.8530,[35]6.8045,[36]6.8115,[37]6.8715,[38]6.9325,[39]7.0417,
save_imatrix: stored collected data after 40 chunks in aya-23-8B-IMat-GGUF/imatrix.dat
[40]7.1304,[41]7.2202,[42]7.4146,[43]7.6280,[44]7.8621,[45]7.9827,[46]7.9802,[47]7.9503,[48]7.8607,[49]7.9686,
save_imatrix: stored collected data after 50 chunks in aya-23-8B-IMat-GGUF/imatrix.dat
[50]8.0385,[51]8.1188,[52]8.2639,[53]8.3293,[54]8.4304,[55]8.5259,[56]8.5475,[57]8.5795,[58]8.6124,[59]8.6107,
save_imatrix: stored collected data after 60 chunks in aya-23-8B-IMat-GGUF/imatrix.dat
[60]8.7115,[61]8.8002,[62]8.8682,[63]8.9158,[64]8.8321,[65]8.8011,[66]8.7690,[67]8.7664,[68]8.7518,[69]8.7095,
save_imatrix: stored collected data after 70 chunks in aya-23-8B-IMat-GGUF/imatrix.dat
[70]8.6180,[71]8.5990,[72]8.5778,[73]8.5959,[74]8.6055,[75]8.6153,[76]8.6184,[77]8.6011,[78]8.5432,[79]8.5512,
save_imatrix: stored collected data after 80 chunks in aya-23-8B-IMat-GGUF/imatrix.dat
[80]8.5401,[81]8.5431,[82]8.5836,[83]8.5824,[84]8.6061,[85]8.6119,[86]8.5804,[87]8.5612,[88]8.5708,[89]8.5804,
save_imatrix: stored collected data after 90 chunks in aya-23-8B-IMat-GGUF/imatrix.dat
[90]8.5891,[91]8.5533,[92]8.4998,[93]8.4785,[94]8.4974,[95]8.4786,[96]8.4704,[97]8.4801,[98]8.5061,[99]8.4687,
save_imatrix: stored collected data after 100 chunks in aya-23-8B-IMat-GGUF/imatrix.dat
[100]8.4955,[101]8.4923,[102]8.4611,[103]8.4774,[104]8.4411,[105]8.3941,[106]8.3383,[107]8.3762,[108]8.4215,[109]8.4042,
save_imatrix: stored collected data after 110 chunks in aya-23-8B-IMat-GGUF/imatrix.dat
[110]8.3948,[111]8.3834,[112]8.4448,[113]8.3678,[114]8.3469,[115]8.3120,[116]8.2475,[117]8.2068,[118]8.1588,[119]8.1061,
save_imatrix: stored collected data after 120 chunks in aya-23-8B-IMat-GGUF/imatrix.dat
[120]8.0583,[121]8.0006,[122]7.9571,[123]7.9071,[124]7.8589,[125]7.8224,[126]7.8380,[127]7.8673,[128]7.8994,[129]7.9217,
save_imatrix: stored collected data after 130 chunks in aya-23-8B-IMat-GGUF/imatrix.dat
[130]7.9478,[131]8.0433,[132]8.1334,[133]8.2221,[134]8.3227,[135]8.3720,[136]8.4199,[137]8.4348,[138]8.4647,[139]8.4751,
save_imatrix: stored collected data after 140 chunks in aya-23-8B-IMat-GGUF/imatrix.dat
[140]8.4985,[141]8.5340,[142]8.5555,[143]8.5847,[144]8.6165,[145]8.6354,[146]8.6218,[147]8.6620,[148]8.6799,[149]8.6998,
save_imatrix: stored collected data after 150 chunks in aya-23-8B-IMat-GGUF/imatrix.dat
[150]8.6752,[151]8.7044,[152]8.6958,[153]8.6653,[154]8.6468,[155]8.6413,[156]8.6339,[157]8.6284,[158]8.6294,[159]8.5902,
save_imatrix: stored collected data after 160 chunks in aya-23-8B-IMat-GGUF/imatrix.dat
[160]8.6397,[161]8.6918,[162]8.7399,[163]8.8271,[164]8.8731,[165]8.8877,[166]8.8902,[167]8.9293,[168]8.9149,[169]8.9450,
save_imatrix: stored collected data after 170 chunks in aya-23-8B-IMat-GGUF/imatrix.dat
[170]8.9197,[171]8.9054,[172]8.9088,[173]8.9341,[174]8.9329,[175]8.9375,[176]8.9496,[177]8.9887,[178]9.0170,[179]9.0041,
save_imatrix: stored collected data after 180 chunks in aya-23-8B-IMat-GGUF/imatrix.dat
[180]9.0067,[181]8.9939,[182]8.9932,[183]8.9789,[184]8.9741,[185]8.9363,[186]8.9414,[187]8.9221,[188]8.9945,[189]9.0665,
save_imatrix: stored collected data after 190 chunks in aya-23-8B-IMat-GGUF/imatrix.dat
[190]9.1200,[191]9.1653,[192]9.1639,[193]9.1284,[194]9.0943,
save_imatrix: stored collected data after 194 chunks in aya-23-8B-IMat-GGUF/imatrix.dat

llama_print_timings:        load time =    2130.43 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 =  152430.74 ms / 99328 tokens (    1.53 ms per token,   651.63 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 =  155917.19 ms / 99329 tokens

Final estimate: PPL = 9.0943 +/- 0.10864