File size: 10,610 Bytes
063d3bc |
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 |
main: build = 3051 (5921b8f0)
main: built with cc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0 for x86_64-linux-gnu
main: seed = 1717152257
llama_model_loader: loaded meta data with 24 key-value pairs and 255 tensors from neo_7b_sft_v0.1-IMat-GGUF/neo_7b_sft_v0.1.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 = llama
llama_model_loader: - kv 1: general.name str = neo_7b_sft_v0.1
llama_model_loader: - kv 2: llama.block_count u32 = 28
llama_model_loader: - kv 3: llama.context_length u32 = 8192
llama_model_loader: - kv 4: llama.embedding_length u32 = 3072
llama_model_loader: - kv 5: llama.feed_forward_length u32 = 24576
llama_model_loader: - kv 6: llama.attention.head_count u32 = 16
llama_model_loader: - kv 7: llama.attention.head_count_kv u32 = 16
llama_model_loader: - kv 8: llama.rope.freq_base f32 = 10000.000000
llama_model_loader: - kv 9: llama.attention.layer_norm_rms_epsilon f32 = 0.000010
llama_model_loader: - kv 10: general.file_type u32 = 0
llama_model_loader: - kv 11: llama.vocab_size u32 = 64256
llama_model_loader: - kv 12: llama.rope.dimension_count u32 = 192
llama_model_loader: - kv 13: tokenizer.ggml.model str = llama
llama_model_loader: - kv 14: tokenizer.ggml.pre str = default
llama_model_loader: - kv 15: tokenizer.ggml.tokens arr[str,64256] = ["<unk>", "<s>", "</s>", "<0x00>", "<...
llama_model_loader: - kv 16: tokenizer.ggml.scores arr[f32,64256] = [0.000000, 0.000000, 0.000000, 0.0000...
llama_model_loader: - kv 17: tokenizer.ggml.token_type arr[i32,64256] = [2, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 6, ...
llama_model_loader: - kv 18: tokenizer.ggml.bos_token_id u32 = 1
llama_model_loader: - kv 19: tokenizer.ggml.eos_token_id u32 = 2
llama_model_loader: - kv 20: tokenizer.ggml.add_bos_token bool = false
llama_model_loader: - kv 21: tokenizer.ggml.add_eos_token bool = false
llama_model_loader: - kv 22: tokenizer.chat_template str = {% set system_message = 'You are a he...
llama_model_loader: - kv 23: general.quantization_version u32 = 2
llama_model_loader: - type f32: 255 tensors
llm_load_vocab: special tokens cache size = 515
llm_load_vocab: token to piece cache size = 0.7263 MB
llm_load_print_meta: format = GGUF V3 (latest)
llm_load_print_meta: arch = llama
llm_load_print_meta: vocab type = SPM
llm_load_print_meta: n_vocab = 64256
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 = 192
llm_load_print_meta: n_embd_head_v = 192
llm_load_print_meta: n_gqa = 1
llm_load_print_meta: n_embd_k_gqa = 3072
llm_load_print_meta: n_embd_v_gqa = 3072
llm_load_print_meta: f_norm_eps = 0.0e+00
llm_load_print_meta: f_norm_rms_eps = 1.0e-05
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 = 0
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 = 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 = all F32
llm_load_print_meta: model params = 7.79 B
llm_load_print_meta: model size = 29.03 GiB (32.00 BPW)
llm_load_print_meta: general.name = neo_7b_sft_v0.1
llm_load_print_meta: BOS token = 1 '<s>'
llm_load_print_meta: EOS token = 2 '</s>'
llm_load_print_meta: UNK token = 0 '<unk>'
llm_load_print_meta: LF token = 13 '<0x0A>'
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 = 0.26 MiB
llm_load_tensors: offloading 22 repeating layers to GPU
llm_load_tensors: offloaded 22/29 layers to GPU
llm_load_tensors: CPU buffer size = 29730.67 MiB
llm_load_tensors: CUDA0 buffer size = 22176.52 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 = 36.00 MiB
llama_kv_cache_init: CUDA0 KV buffer size = 132.00 MiB
llama_new_context_with_model: KV self size = 168.00 MiB, K (f16): 84.00 MiB, V (f16): 84.00 MiB
llama_new_context_with_model: CUDA_Host output buffer size = 0.25 MiB
llama_new_context_with_model: CUDA0 compute buffer size = 884.50 MiB
llama_new_context_with_model: CUDA_Host compute buffer size = 13.01 MiB
llama_new_context_with_model: graph nodes = 902
llama_new_context_with_model: graph splits = 70
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 94.53 ms
compute_imatrix: computing over 149 chunks with batch_size 512
compute_imatrix: 0.92 seconds per pass - ETA 2.28 minutes
[1]6.9087,[2]4.9464,[3]5.5776,[4]5.6370,[5]5.4654,[6]5.4102,[7]4.7098,[8]4.8687,[9]4.9457,
save_imatrix: stored collected data after 10 chunks in neo_7b_sft_v0.1-IMat-GGUF/imatrix.dat
[10]5.3631,[11]5.3031,[12]4.9517,[13]5.3067,[14]5.7762,[15]6.0604,[16]6.5393,[17]6.8275,[18]7.0775,[19]7.2396,
save_imatrix: stored collected data after 20 chunks in neo_7b_sft_v0.1-IMat-GGUF/imatrix.dat
[20]7.6090,[21]7.3325,[22]7.3046,[23]7.4432,[24]7.6053,[25]7.5198,[26]7.4926,[27]7.6715,[28]7.9040,[29]8.1177,
save_imatrix: stored collected data after 30 chunks in neo_7b_sft_v0.1-IMat-GGUF/imatrix.dat
[30]8.1330,[31]8.2365,[32]8.3080,[33]8.3659,[34]8.2829,[35]8.1805,[36]7.8166,[37]7.6196,[38]7.2773,[39]7.2020,
save_imatrix: stored collected data after 40 chunks in neo_7b_sft_v0.1-IMat-GGUF/imatrix.dat
[40]7.1117,[41]7.0423,[42]7.0131,[43]7.0513,[44]7.0433,[45]6.9766,[46]7.0585,[47]7.1493,[48]7.3117,[49]7.3431,
save_imatrix: stored collected data after 50 chunks in neo_7b_sft_v0.1-IMat-GGUF/imatrix.dat
[50]7.5558,[51]7.7420,[52]7.9529,[53]8.1337,[54]8.2862,[55]8.2058,[56]8.2031,[57]8.3415,[58]8.4436,[59]8.4034,
save_imatrix: stored collected data after 60 chunks in neo_7b_sft_v0.1-IMat-GGUF/imatrix.dat
[60]8.3390,[61]8.3636,[62]8.4492,[63]8.5299,[64]8.6364,[65]8.7142,[66]8.7962,[67]8.8080,[68]8.8470,[69]8.8577,
save_imatrix: stored collected data after 70 chunks in neo_7b_sft_v0.1-IMat-GGUF/imatrix.dat
[70]8.8172,[71]8.7114,[72]8.5987,[73]8.5845,[74]8.6847,[75]8.7115,[76]8.6591,[77]8.6226,[78]8.6239,[79]8.5747,
save_imatrix: stored collected data after 80 chunks in neo_7b_sft_v0.1-IMat-GGUF/imatrix.dat
[80]8.4730,[81]8.2890,[82]8.1267,[83]8.1267,[84]8.1288,[85]8.1302,[86]8.0806,[87]8.0983,[88]8.0851,[89]8.0748,
save_imatrix: stored collected data after 90 chunks in neo_7b_sft_v0.1-IMat-GGUF/imatrix.dat
[90]8.0567,[91]8.0261,[92]8.0150,[93]7.9759,[94]7.8792,[95]7.9289,[96]7.9894,[97]8.0096,[98]7.9770,[99]7.9935,
save_imatrix: stored collected data after 100 chunks in neo_7b_sft_v0.1-IMat-GGUF/imatrix.dat
[100]8.0289,[101]7.9490,[102]7.8935,[103]7.8883,[104]7.9199,[105]7.9198,[106]7.9549,[107]7.9926,[108]7.9428,[109]7.8852,
save_imatrix: stored collected data after 110 chunks in neo_7b_sft_v0.1-IMat-GGUF/imatrix.dat
[110]7.8331,[111]7.7686,[112]7.6997,[113]7.6362,[114]7.5856,[115]7.5369,[116]7.5122,[117]7.5222,[118]7.5262,[119]7.6019,
save_imatrix: stored collected data after 120 chunks in neo_7b_sft_v0.1-IMat-GGUF/imatrix.dat
[120]7.6779,[121]7.7421,[122]7.8017,[123]7.8961,[124]7.8799,[125]7.9118,[126]7.8995,[127]7.9356,[128]7.9227,[129]7.9105,
save_imatrix: stored collected data after 130 chunks in neo_7b_sft_v0.1-IMat-GGUF/imatrix.dat
[130]7.8760,[131]7.8445,[132]7.8829,[133]7.9431,[134]7.9423,[135]7.9438,[136]7.9568,[137]7.9834,[138]7.9895,[139]7.9990,
save_imatrix: stored collected data after 140 chunks in neo_7b_sft_v0.1-IMat-GGUF/imatrix.dat
[140]8.0310,[141]8.0578,[142]8.0334,[143]8.0987,[144]8.1383,[145]8.1734,[146]8.2294,[147]8.2535,[148]8.3020,[149]8.3237,
save_imatrix: stored collected data after 149 chunks in neo_7b_sft_v0.1-IMat-GGUF/imatrix.dat
llama_print_timings: load time = 3161.43 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 = 124844.08 ms / 76288 tokens ( 1.64 ms per token, 611.07 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 = 127790.46 ms / 76289 tokens
Final estimate: PPL = 8.3237 +/- 0.12004
|