File size: 10,935 Bytes
50af806
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
llama_model_loader: loaded meta data with 29 key-value pairs and 254 tensors from RoGemma-7b-Instruct-IMat-GGUF/RoGemma-7b-Instruct.Q8_0.gguf.hardlink.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              = gemma
llama_model_loader: - kv   1:                               general.name str              = RoGemma-7b-Instruct
llama_model_loader: - kv   2:                       gemma.context_length u32              = 8192
llama_model_loader: - kv   3:                     gemma.embedding_length u32              = 3072
llama_model_loader: - kv   4:                          gemma.block_count u32              = 28
llama_model_loader: - kv   5:                  gemma.feed_forward_length u32              = 24576
llama_model_loader: - kv   6:                 gemma.attention.head_count u32              = 16
llama_model_loader: - kv   7:              gemma.attention.head_count_kv u32              = 16
llama_model_loader: - kv   8:     gemma.attention.layer_norm_rms_epsilon f32              = 0.000001
llama_model_loader: - kv   9:                 gemma.attention.key_length u32              = 256
llama_model_loader: - kv  10:               gemma.attention.value_length u32              = 256
llama_model_loader: - kv  11:                          general.file_type u32              = 7
llama_model_loader: - kv  12:                       tokenizer.ggml.model str              = llama
llama_model_loader: - kv  13:                         tokenizer.ggml.pre str              = default
llama_model_loader: - kv  14:                      tokenizer.ggml.tokens arr[str,256000]  = ["<pad>", "<eos>", "<bos>", "<unk>", ...
llama_model_loader: - kv  15:                      tokenizer.ggml.scores arr[f32,256000]  = [0.000000, 0.000000, 0.000000, 0.0000...
llama_model_loader: - kv  16:                  tokenizer.ggml.token_type arr[i32,256000]  = [3, 3, 3, 2, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv  17:                tokenizer.ggml.bos_token_id u32              = 2
llama_model_loader: - kv  18:                tokenizer.ggml.eos_token_id u32              = 1
llama_model_loader: - kv  19:            tokenizer.ggml.unknown_token_id u32              = 3
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 str              = {{ '<bos>' }}{% if messages[0]['role'...
llama_model_loader: - kv  24:             tokenizer.ggml.prefix_token_id u32              = 67
llama_model_loader: - kv  25:             tokenizer.ggml.suffix_token_id u32              = 69
llama_model_loader: - kv  26:             tokenizer.ggml.middle_token_id u32              = 68
llama_model_loader: - kv  27:                tokenizer.ggml.eot_token_id u32              = 107
llama_model_loader: - kv  28:               general.quantization_version u32              = 2
llama_model_loader: - type  f32:   57 tensors
llama_model_loader: - type q8_0:  197 tensors
llm_load_vocab: special tokens cache size = 260
llm_load_vocab: token to piece cache size = 1.6014 MB
llm_load_print_meta: format           = GGUF V3 (latest)
llm_load_print_meta: arch             = gemma
llm_load_print_meta: vocab type       = SPM
llm_load_print_meta: n_vocab          = 256000
llm_load_print_meta: n_merges         = 0
llm_load_print_meta: n_ctx_train      = 8192
llm_load_print_meta: n_embd           = 3072
llm_load_print_meta: n_head           = 16
llm_load_print_meta: n_head_kv        = 16
llm_load_print_meta: n_layer          = 28
llm_load_print_meta: n_rot            = 192
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: f_logit_scale    = 0.0e+00
llm_load_print_meta: n_ff             = 24576
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        = 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_ctx_orig_yarn  = 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       = 7B
llm_load_print_meta: model ftype      = Q8_0
llm_load_print_meta: model params     = 8.54 B
llm_load_print_meta: model size       = 8.45 GiB (8.50 BPW) 
llm_load_print_meta: general.name     = RoGemma-7b-Instruct
llm_load_print_meta: BOS token        = 2 '<bos>'
llm_load_print_meta: EOS token        = 1 '<eos>'
llm_load_print_meta: UNK token        = 3 '<unk>'
llm_load_print_meta: PAD token        = 0 '<pad>'
llm_load_print_meta: LF token         = 227 '<0x0A>'
llm_load_print_meta: PRE token        = 67 '<unused60>'
llm_load_print_meta: SUF token        = 69 '<unused62>'
llm_load_print_meta: MID token        = 68 '<unused61>'
llm_load_print_meta: EOT token        = 107 '<end_of_turn>'
llm_load_print_meta: max token length = 93
ggml_cuda_init: GGML_CUDA_FORCE_MMQ:    no
ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no
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 =    0.24 MiB
llm_load_tensors: offloading 28 repeating layers to GPU
llm_load_tensors: offloading non-repeating layers to GPU
llm_load_tensors: offloaded 29/29 layers to GPU
llm_load_tensors:        CPU buffer size =   796.88 MiB
llm_load_tensors:      CUDA0 buffer size =  8651.54 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:      CUDA0 KV buffer size =   224.00 MiB
llama_new_context_with_model: KV self size  =  224.00 MiB, K (f16):  112.00 MiB, V (f16):  112.00 MiB
llama_new_context_with_model:  CUDA_Host  output buffer size =     0.98 MiB
llama_new_context_with_model:      CUDA0 compute buffer size =   506.00 MiB
llama_new_context_with_model:  CUDA_Host compute buffer size =     7.01 MiB
llama_new_context_with_model: graph nodes  = 931
llama_new_context_with_model: graph splits = 2

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 122.982 ms
compute_imatrix: computing over 128 chunks with batch_size 512
compute_imatrix: 0.72 seconds per pass - ETA 1.53 minutes
[1]6.7268,[2]4.7956,[3]4.3365,[4]5.4634,[5]5.5761,[6]4.7844,[7]5.2281,[8]5.4576,[9]5.6880,
save_imatrix: stored collected data after 10 chunks in RoGemma-7b-Instruct-IMat-GGUF/imatrix.dat
[10]5.1084,[11]5.2387,[12]5.6425,[13]6.0984,[14]6.4094,[15]6.7641,[16]7.0334,[17]7.1327,[18]7.4186,[19]7.1324,
save_imatrix: stored collected data after 20 chunks in RoGemma-7b-Instruct-IMat-GGUF/imatrix.dat
[20]7.2420,[21]7.3956,[22]7.3754,[23]7.4916,[24]7.5334,[25]7.6872,[26]7.4682,[27]7.7364,[28]7.9965,[29]7.9577,
save_imatrix: stored collected data after 30 chunks in RoGemma-7b-Instruct-IMat-GGUF/imatrix.dat
[30]7.9155,[31]7.4608,[32]7.2029,[33]7.1080,[34]6.9745,[35]6.9006,[36]7.1893,[37]7.2251,[38]7.2493,[39]7.3769,
save_imatrix: stored collected data after 40 chunks in RoGemma-7b-Instruct-IMat-GGUF/imatrix.dat
[40]7.4966,[41]7.6572,[42]7.9743,[43]8.2865,[44]8.5892,[45]8.7742,[46]8.6579,[47]8.6807,[48]8.8624,[49]8.9978,
save_imatrix: stored collected data after 50 chunks in RoGemma-7b-Instruct-IMat-GGUF/imatrix.dat
[50]8.8560,[51]8.8432,[52]8.8797,[53]8.9995,[54]9.1008,[55]9.2532,[56]9.3016,[57]9.3064,[58]9.3182,[59]9.1306,
save_imatrix: stored collected data after 60 chunks in RoGemma-7b-Instruct-IMat-GGUF/imatrix.dat
[60]9.0207,[61]8.8842,[62]8.8355,[63]8.8759,[64]8.8721,[65]8.8531,[66]8.8684,[67]8.8022,[68]8.7356,[69]8.7641,
save_imatrix: stored collected data after 70 chunks in RoGemma-7b-Instruct-IMat-GGUF/imatrix.dat
[70]8.7371,[71]8.7328,[72]8.7399,[73]8.7070,[74]8.6636,[75]8.6279,[76]8.6341,[77]8.6556,[78]8.6528,[79]8.6093,
save_imatrix: stored collected data after 80 chunks in RoGemma-7b-Instruct-IMat-GGUF/imatrix.dat
[80]8.6513,[81]8.6853,[82]8.6525,[83]8.6525,[84]8.6889,[85]8.5525,[86]8.5107,[87]8.4442,[88]8.4460,[89]8.4832,
save_imatrix: stored collected data after 90 chunks in RoGemma-7b-Instruct-IMat-GGUF/imatrix.dat
[90]8.4874,[91]8.4152,[92]8.3457,[93]8.2611,[94]8.1827,[95]8.1159,[96]8.0418,[97]7.9787,[98]7.9182,[99]7.9335,
save_imatrix: stored collected data after 100 chunks in RoGemma-7b-Instruct-IMat-GGUF/imatrix.dat
[100]7.9383,[101]8.0119,[102]8.0867,[103]8.1573,[104]8.2901,[105]8.3851,[106]8.4142,[107]8.4310,[108]8.4466,[109]8.4205,
save_imatrix: stored collected data after 110 chunks in RoGemma-7b-Instruct-IMat-GGUF/imatrix.dat
[110]8.4063,[111]8.3474,[112]8.2734,[113]8.3102,[114]8.3227,[115]8.3146,[116]8.3066,[117]8.3351,[118]8.3537,[119]8.3591,
save_imatrix: stored collected data after 120 chunks in RoGemma-7b-Instruct-IMat-GGUF/imatrix.dat
[120]8.3618,[121]8.3700,[122]8.3300,[123]8.3984,[124]8.4737,[125]8.5299,[126]8.6142,[127]8.6792,[128]8.7451,
save_imatrix: stored collected data after 128 chunks in RoGemma-7b-Instruct-IMat-GGUF/imatrix.dat

llama_print_timings:        load time =    6235.43 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 =   76873.60 ms / 65536 tokens (    1.17 ms per token,   852.52 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 =   83878.25 ms / 65537 tokens

Final estimate: PPL = 8.7451 +/- 0.12878