Hugging Face's logo

legraphista
/
Llama-3-8B-Instruct-MopeyMule-IMat-GGUF

Text Generation
GGUF
PyTorch
English
facebook
meta
llama
llama-3
quantized
GGUF
imatrix
quantization
imat
static
16bit
8bit
6bit
5bit
4bit
3bit
2bit
1bit
conversational
Model card Files Community
legraphista commited on
Commit
f0db978
•
1 Parent(s): 94a5db6

Upload imatrix.log with huggingface_hub

Browse files
Files changed (1) hide show
  1. imatrix.log +139 -0
imatrix.log ADDED
@@ -0,0 +1,139 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ main: build = 3056 (0c27e6f6)
2
+ main: built with cc (Ubuntu 13.2.0-23ubuntu4) 13.2.0 for x86_64-linux-gnu
3
+ main: seed = 1717159644
4
+ llama_model_loader: loaded meta data with 22 key-value pairs and 291 tensors from Llama-3-8B-Instruct-MopeyMule-IMat-GGUF/Llama-3-8B-Instruct-MopeyMule.gguf (version GGUF V3 (latest))
5
+ llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
6
+ llama_model_loader: - kv 0: general.architecture str = llama
7
+ llama_model_loader: - kv 1: general.name str = Llama-3-8B-Instruct-MopeyMule
8
+ llama_model_loader: - kv 2: llama.block_count u32 = 32
9
+ llama_model_loader: - kv 3: llama.context_length u32 = 8192
10
+ llama_model_loader: - kv 4: llama.embedding_length u32 = 4096
11
+ llama_model_loader: - kv 5: llama.feed_forward_length u32 = 14336
12
+ llama_model_loader: - kv 6: llama.attention.head_count u32 = 32
13
+ llama_model_loader: - kv 7: llama.attention.head_count_kv u32 = 8
14
+ llama_model_loader: - kv 8: llama.rope.freq_base f32 = 500000.000000
15
+ llama_model_loader: - kv 9: llama.attention.layer_norm_rms_epsilon f32 = 0.000010
16
+ llama_model_loader: - kv 10: general.file_type u32 = 0
17
+ llama_model_loader: - kv 11: llama.vocab_size u32 = 128256
18
+ llama_model_loader: - kv 12: llama.rope.dimension_count u32 = 128
19
+ llama_model_loader: - kv 13: tokenizer.ggml.model str = gpt2
20
+ llama_model_loader: - kv 14: tokenizer.ggml.pre str = llama-bpe
21
+ llama_model_loader: - kv 15: tokenizer.ggml.tokens arr[str,128256] = ["!", "\"", "#", "$", "%", "&", "'", ...
22
+ llama_model_loader: - kv 16: tokenizer.ggml.token_type arr[i32,128256] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
23
+ llama_model_loader: - kv 17: tokenizer.ggml.merges arr[str,280147] = ["Ġ Ġ", "Ġ ĠĠĠ", "ĠĠ ĠĠ", "...
24
+ llama_model_loader: - kv 18: tokenizer.ggml.bos_token_id u32 = 128000
25
+ llama_model_loader: - kv 19: tokenizer.ggml.eos_token_id u32 = 128009
26
+ llama_model_loader: - kv 20: tokenizer.chat_template str = {% set loop_messages = messages %}{% ...
27
+ llama_model_loader: - kv 21: general.quantization_version u32 = 2
28
+ llama_model_loader: - type f32: 291 tensors
29
+ llm_load_vocab: special tokens cache size = 256
30
+ llm_load_vocab: token to piece cache size = 1.5928 MB
31
+ llm_load_print_meta: format = GGUF V3 (latest)
32
+ llm_load_print_meta: arch = llama
33
+ llm_load_print_meta: vocab type = BPE
34
+ llm_load_print_meta: n_vocab = 128256
35
+ llm_load_print_meta: n_merges = 280147
36
+ llm_load_print_meta: n_ctx_train = 8192
37
+ llm_load_print_meta: n_embd = 4096
38
+ llm_load_print_meta: n_head = 32
39
+ llm_load_print_meta: n_head_kv = 8
40
+ llm_load_print_meta: n_layer = 32
41
+ llm_load_print_meta: n_rot = 128
42
+ llm_load_print_meta: n_embd_head_k = 128
43
+ llm_load_print_meta: n_embd_head_v = 128
44
+ llm_load_print_meta: n_gqa = 4
45
+ llm_load_print_meta: n_embd_k_gqa = 1024
46
+ llm_load_print_meta: n_embd_v_gqa = 1024
47
+ llm_load_print_meta: f_norm_eps = 0.0e+00
48
+ llm_load_print_meta: f_norm_rms_eps = 1.0e-05
49
+ llm_load_print_meta: f_clamp_kqv = 0.0e+00
50
+ llm_load_print_meta: f_max_alibi_bias = 0.0e+00
51
+ llm_load_print_meta: f_logit_scale = 0.0e+00
52
+ llm_load_print_meta: n_ff = 14336
53
+ llm_load_print_meta: n_expert = 0
54
+ llm_load_print_meta: n_expert_used = 0
55
+ llm_load_print_meta: causal attn = 1
56
+ llm_load_print_meta: pooling type = 0
57
+ llm_load_print_meta: rope type = 0
58
+ llm_load_print_meta: rope scaling = linear
59
+ llm_load_print_meta: freq_base_train = 500000.0
60
+ llm_load_print_meta: freq_scale_train = 1
61
+ llm_load_print_meta: n_yarn_orig_ctx = 8192
62
+ llm_load_print_meta: rope_finetuned = unknown
63
+ llm_load_print_meta: ssm_d_conv = 0
64
+ llm_load_print_meta: ssm_d_inner = 0
65
+ llm_load_print_meta: ssm_d_state = 0
66
+ llm_load_print_meta: ssm_dt_rank = 0
67
+ llm_load_print_meta: model type = 8B
68
+ llm_load_print_meta: model ftype = all F32
69
+ llm_load_print_meta: model params = 8.03 B
70
+ llm_load_print_meta: model size = 29.92 GiB (32.00 BPW)
71
+ llm_load_print_meta: general.name = Llama-3-8B-Instruct-MopeyMule
72
+ llm_load_print_meta: BOS token = 128000 '<|begin_of_text|>'
73
+ llm_load_print_meta: EOS token = 128009 '<|eot_id|>'
74
+ llm_load_print_meta: LF token = 128 'Ä'
75
+ llm_load_print_meta: EOT token = 128009 '<|eot_id|>'
76
+ ggml_cuda_init: GGML_CUDA_FORCE_MMQ: no
77
+ ggml_cuda_init: CUDA_USE_TENSOR_CORES: yes
78
+ ggml_cuda_init: found 1 CUDA devices:
79
+ Device 0: NVIDIA GeForce RTX 4090, compute capability 8.9, VMM: yes
80
+ llm_load_tensors: ggml ctx size = 0.30 MiB
81
+ llm_load_tensors: offloading 26 repeating layers to GPU
82
+ llm_load_tensors: offloaded 26/33 layers to GPU
83
+ llm_load_tensors: CPU buffer size = 30633.02 MiB
84
+ llm_load_tensors: CUDA0 buffer size = 21632.81 MiB
85
+ .........................................................................................
86
+ llama_new_context_with_model: n_ctx = 512
87
+ llama_new_context_with_model: n_batch = 512
88
+ llama_new_context_with_model: n_ubatch = 512
89
+ llama_new_context_with_model: flash_attn = 0
90
+ llama_new_context_with_model: freq_base = 500000.0
91
+ llama_new_context_with_model: freq_scale = 1
92
+ llama_kv_cache_init: CUDA_Host KV buffer size = 12.00 MiB
93
+ llama_kv_cache_init: CUDA0 KV buffer size = 52.00 MiB
94
+ llama_new_context_with_model: KV self size = 64.00 MiB, K (f16): 32.00 MiB, V (f16): 32.00 MiB
95
+ llama_new_context_with_model: CUDA_Host output buffer size = 0.49 MiB
96
+ llama_new_context_with_model: CUDA0 compute buffer size = 2262.50 MiB
97
+ llama_new_context_with_model: CUDA_Host compute buffer size = 9.01 MiB
98
+ llama_new_context_with_model: graph nodes = 1030
99
+ llama_new_context_with_model: graph splits = 70
100
+
101
+ system_info: n_threads = 25 / 32 | AVX = 1 | AVX_VNNI = 0 | AVX2 = 1 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | AVX512_BF16 = 0 | 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 |
102
+ compute_imatrix: tokenizing the input ..
103
+ compute_imatrix: tokenization took 47.451 ms
104
+ compute_imatrix: computing over 125 chunks with batch_size 512
105
+ compute_imatrix: 4.20 seconds per pass - ETA 8.75 minutes
106
+ [1]6.6250,[2]5.3950,[3]4.9018,[4]6.0681,[5]6.4082,[6]5.8107,[7]6.2593,[8]6.8961,[9]7.0453,
107
+ save_imatrix: stored collected data after 10 chunks in Llama-3-8B-Instruct-MopeyMule-IMat-GGUF/imatrix.dat
108
+ [10]6.4398,[11]7.0247,[12]7.7127,[13]8.3773,[14]8.9216,[15]9.2764,[16]9.6517,[17]9.8964,[18]9.6306,[19]9.1798,
109
+ save_imatrix: stored collected data after 20 chunks in Llama-3-8B-Instruct-MopeyMule-IMat-GGUF/imatrix.dat
110
+ [20]9.1053,[21]9.3150,[22]9.2083,[23]9.5868,[24]9.4919,[25]9.9523,[26]9.9423,[27]10.0086,[28]10.2563,[29]10.2846,
111
+ save_imatrix: stored collected data after 30 chunks in Llama-3-8B-Instruct-MopeyMule-IMat-GGUF/imatrix.dat
112
+ [30]10.3160,[31]9.7316,[32]9.1701,[33]8.9211,[34]8.7366,[35]8.9005,[36]9.0293,[37]8.9740,[38]9.0954,[39]9.2849,
113
+ save_imatrix: stored collected data after 40 chunks in Llama-3-8B-Instruct-MopeyMule-IMat-GGUF/imatrix.dat
114
+ [40]9.4010,[41]9.5736,[42]9.7217,[43]9.9903,[44]10.0695,[45]10.2593,[46]10.1060,[47]10.2434,[48]10.3310,[49]10.4473,
115
+ save_imatrix: stored collected data after 50 chunks in Llama-3-8B-Instruct-MopeyMule-IMat-GGUF/imatrix.dat
116
+ [50]10.3469,[51]10.4312,[52]10.6218,[53]10.7315,[54]10.8115,[55]10.9196,[56]10.9715,[57]11.0573,[58]11.0708,[59]11.1328,
117
+ save_imatrix: stored collected data after 60 chunks in Llama-3-8B-Instruct-MopeyMule-IMat-GGUF/imatrix.dat
118
+ [60]11.0697,[61]11.0061,[62]11.1104,[63]11.1580,[64]11.0483,[65]11.0238,[66]11.0130,[67]10.9197,[68]10.8712,[69]10.8102,
119
+ save_imatrix: stored collected data after 70 chunks in Llama-3-8B-Instruct-MopeyMule-IMat-GGUF/imatrix.dat
120
+ [70]10.7680,[71]10.7330,[72]10.6939,[73]10.6149,[74]10.5130,[75]10.4797,[76]10.4668,[77]10.4042,[78]10.3655,[79]10.3985,
121
+ save_imatrix: stored collected data after 80 chunks in Llama-3-8B-Instruct-MopeyMule-IMat-GGUF/imatrix.dat
122
+ [80]10.4218,[81]10.3972,[82]10.3874,[83]10.4051,[84]10.2408,[85]10.2332,[86]10.2430,[87]10.2747,[88]10.3036,[89]10.2997,
123
+ save_imatrix: stored collected data after 90 chunks in Llama-3-8B-Instruct-MopeyMule-IMat-GGUF/imatrix.dat
124
+ [90]10.2107,[91]10.1104,[92]10.0127,[93]9.9306,[94]9.8384,[95]9.7585,[96]9.7071,[97]9.7588,[98]9.8094,[99]9.9016,
125
+ save_imatrix: stored collected data after 100 chunks in Llama-3-8B-Instruct-MopeyMule-IMat-GGUF/imatrix.dat
126
+ [100]9.9891,[101]10.0616,[102]10.2103,[103]10.2458,[104]10.3093,[105]10.1944,[106]10.1989,[107]10.1521,[108]10.0846,[109]10.0087,
127
+ save_imatrix: stored collected data after 110 chunks in Llama-3-8B-Instruct-MopeyMule-IMat-GGUF/imatrix.dat
128
+ [110]10.0839,[111]10.1629,[112]10.1936,[113]10.2082,[114]10.2796,[115]10.3393,[116]10.3504,[117]10.3799,[118]10.4180,[119]10.3323,
129
+ save_imatrix: stored collected data after 120 chunks in Llama-3-8B-Instruct-MopeyMule-IMat-GGUF/imatrix.dat
130
+ [120]10.3494,[121]10.3774,[122]10.4008,[123]10.4512,[124]10.4860,[125]10.5200,
131
+ save_imatrix: stored collected data after 125 chunks in Llama-3-8B-Instruct-MopeyMule-IMat-GGUF/imatrix.dat
132
+
133
+ llama_print_timings: load time = 20701.29 ms
134
+ llama_print_timings: sample time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
135
+ llama_print_timings: prompt eval time = 329740.46 ms / 64000 tokens ( 5.15 ms per token, 194.09 tokens per second)
136
+ llama_print_timings: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
137
+ llama_print_timings: total time = 348274.75 ms / 64001 tokens
138
+
139
+ Final estimate: PPL = 10.5200 +/- 0.16422