llama_model_loader: loaded meta data with 39 key-value pairs and 508 tensors from datagemma-rag-27b-it-IMat-GGUF/datagemma-rag-27b-it.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 = gemma2 llama_model_loader: - kv 1: general.type str = model llama_model_loader: - kv 2: general.name str = Rag_27B_Transformers_Checkpoint_15000 llama_model_loader: - kv 3: general.finetune str = it llama_model_loader: - kv 4: general.basename str = datagemma-rag llama_model_loader: - kv 5: general.size_label str = 27B llama_model_loader: - kv 6: general.license str = gemma llama_model_loader: - kv 7: general.base_model.count u32 = 1 llama_model_loader: - kv 8: general.base_model.0.name str = Gemma 2 27b It llama_model_loader: - kv 9: general.base_model.0.organization str = Google llama_model_loader: - kv 10: general.base_model.0.repo_url str = https://huggingface.co/google/gemma-2... llama_model_loader: - kv 11: general.tags arr[str,2] = ["conversational", "text-generation"] llama_model_loader: - kv 12: gemma2.context_length u32 = 8192 llama_model_loader: - kv 13: gemma2.embedding_length u32 = 4608 llama_model_loader: - kv 14: gemma2.block_count u32 = 46 llama_model_loader: - kv 15: gemma2.feed_forward_length u32 = 36864 llama_model_loader: - kv 16: gemma2.attention.head_count u32 = 32 llama_model_loader: - kv 17: gemma2.attention.head_count_kv u32 = 16 llama_model_loader: - kv 18: gemma2.attention.layer_norm_rms_epsilon f32 = 0.000001 llama_model_loader: - kv 19: gemma2.attention.key_length u32 = 128 llama_model_loader: - kv 20: gemma2.attention.value_length u32 = 128 llama_model_loader: - kv 21: general.file_type u32 = 7 llama_model_loader: - kv 22: gemma2.attn_logit_softcapping f32 = 50.000000 llama_model_loader: - kv 23: gemma2.final_logit_softcapping f32 = 30.000000 llama_model_loader: - kv 24: gemma2.attention.sliding_window u32 = 4096 llama_model_loader: - kv 25: tokenizer.ggml.model str = llama llama_model_loader: - kv 26: tokenizer.ggml.pre str = default llama_model_loader: - kv 27: tokenizer.ggml.tokens arr[str,256000] = ["", "", "", "", ... llama_model_loader: - kv 28: tokenizer.ggml.scores arr[f32,256000] = [-1000.000000, -1000.000000, -1000.00... llama_model_loader: - kv 29: tokenizer.ggml.token_type arr[i32,256000] = [3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, ... llama_model_loader: - kv 30: tokenizer.ggml.bos_token_id u32 = 2 llama_model_loader: - kv 31: tokenizer.ggml.eos_token_id u32 = 1 llama_model_loader: - kv 32: tokenizer.ggml.unknown_token_id u32 = 3 llama_model_loader: - kv 33: tokenizer.ggml.padding_token_id u32 = 0 llama_model_loader: - kv 34: tokenizer.ggml.add_bos_token bool = true llama_model_loader: - kv 35: tokenizer.ggml.add_eos_token bool = false llama_model_loader: - kv 36: tokenizer.chat_template str = {{ bos_token }}{% if messages[0]['rol... llama_model_loader: - kv 37: tokenizer.ggml.add_space_prefix bool = false llama_model_loader: - kv 38: general.quantization_version u32 = 2 llama_model_loader: - type f32: 185 tensors llama_model_loader: - type q8_0: 323 tensors llm_load_vocab: special tokens cache size = 249 llm_load_vocab: token to piece cache size = 1.6014 MB llm_load_print_meta: format = GGUF V3 (latest) llm_load_print_meta: arch = gemma2 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: vocab_only = 0 llm_load_print_meta: n_ctx_train = 8192 llm_load_print_meta: n_embd = 4608 llm_load_print_meta: n_layer = 46 llm_load_print_meta: n_head = 32 llm_load_print_meta: n_head_kv = 16 llm_load_print_meta: n_rot = 128 llm_load_print_meta: n_swa = 4096 llm_load_print_meta: n_embd_head_k = 128 llm_load_print_meta: n_embd_head_v = 128 llm_load_print_meta: n_gqa = 2 llm_load_print_meta: n_embd_k_gqa = 2048 llm_load_print_meta: n_embd_v_gqa = 2048 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 = 36864 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: ssm_dt_b_c_rms = 0 llm_load_print_meta: model type = 27B llm_load_print_meta: model ftype = Q8_0 llm_load_print_meta: model params = 27.23 B llm_load_print_meta: model size = 26.94 GiB (8.50 BPW) llm_load_print_meta: general.name = Rag_27B_Transformers_Checkpoint_15000 llm_load_print_meta: BOS token = 2 '' llm_load_print_meta: EOS token = 1 '' llm_load_print_meta: UNK token = 3 '' llm_load_print_meta: PAD token = 0 '' llm_load_print_meta: LF token = 227 '<0x0A>' llm_load_print_meta: EOT token = 107 '' llm_load_print_meta: max token length = 48 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 = 0.45 MiB llm_load_tensors: offloading 31 repeating layers to GPU llm_load_tensors: offloaded 31/47 layers to GPU llm_load_tensors: CPU buffer size = 27591.06 MiB llm_load_tensors: CUDA0 buffer size = 17788.43 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 = 60.00 MiB llama_kv_cache_init: CUDA0 KV buffer size = 124.00 MiB llama_new_context_with_model: KV self size = 184.00 MiB, K (f16): 92.00 MiB, V (f16): 92.00 MiB llama_new_context_with_model: CUDA_Host output buffer size = 0.98 MiB llama_new_context_with_model: CUDA0 compute buffer size = 1704.31 MiB llama_new_context_with_model: CUDA_Host compute buffer size = 11.01 MiB llama_new_context_with_model: graph nodes = 1850 llama_new_context_with_model: graph splits = 199 system_info: n_threads = 25 (n_threads_batch = 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 159.764 ms compute_imatrix: computing over 128 chunks with batch_size 512 compute_imatrix: 2.39 seconds per pass - ETA 5.10 minutes [1]69.8607,[2]29.4525,[3]24.5033,[4]38.8083,[5]39.8432,[6]27.0676,[7]32.9312,[8]36.1389,[9]41.0896, save_imatrix: stored collected data after 10 chunks in datagemma-rag-27b-it-IMat-GGUF/imatrix.dat [10]33.1982,[11]36.8959,[12]43.6830,[13]51.2839,[14]54.2855,[15]61.3891,[16]63.8436,[17]66.7182,[18]72.3091,[19]64.1971, save_imatrix: stored collected data after 20 chunks in datagemma-rag-27b-it-IMat-GGUF/imatrix.dat [20]69.5026,[21]69.8297,[22]70.5949,[23]72.7177,[24]73.8317,[25]77.7490,[26]73.3472,[27]75.6210,[28]78.3592,[29]76.5609, save_imatrix: stored collected data after 30 chunks in datagemma-rag-27b-it-IMat-GGUF/imatrix.dat [30]75.6626,[31]67.4056,[32]63.2137,[33]62.3631,[34]60.9476,[35]60.7065,[36]59.6879,[37]59.5043,[38]60.9446,[39]63.6639, save_imatrix: stored collected data after 40 chunks in datagemma-rag-27b-it-IMat-GGUF/imatrix.dat [40]64.7962,[41]66.9487,[42]70.2819,[43]73.7386,[44]76.8135,[45]78.7252,[46]76.6665,[47]77.1749,[48]81.4155,[49]84.2439, save_imatrix: stored collected data after 50 chunks in datagemma-rag-27b-it-IMat-GGUF/imatrix.dat [50]80.8397,[51]82.2205,[52]83.7068,[53]85.9632,[54]89.1705,[55]89.9494,[56]90.6200,[57]91.4288,[58]91.7778,[59]88.9244, save_imatrix: stored collected data after 60 chunks in datagemma-rag-27b-it-IMat-GGUF/imatrix.dat [60]87.3384,[61]85.8398,[62]85.8018,[63]86.9842,[64]87.3335,[65]86.9720,[66]87.2078,[67]86.7474,[68]86.3931,[69]87.5692, save_imatrix: stored collected data after 70 chunks in datagemma-rag-27b-it-IMat-GGUF/imatrix.dat [70]87.3230,[71]87.7908,[72]88.4429,[73]88.7741,[74]88.4202,[75]87.9258,[76]88.1274,[77]88.9147,[78]89.6685,[79]89.4842, save_imatrix: stored collected data after 80 chunks in datagemma-rag-27b-it-IMat-GGUF/imatrix.dat [80]90.7892,[81]91.5656,[82]91.2523,[83]91.4493,[84]92.5936,[85]89.7008,[86]89.1141,[87]87.5918,[88]87.6548,[89]87.8680, save_imatrix: stored collected data after 90 chunks in datagemma-rag-27b-it-IMat-GGUF/imatrix.dat [90]87.9463,[91]86.7824,[92]85.0786,[93]83.3349,[94]81.6515,[95]80.3728,[96]78.8332,[97]77.5687,[98]76.2892,[99]76.6564, save_imatrix: stored collected data after 100 chunks in datagemma-rag-27b-it-IMat-GGUF/imatrix.dat [100]77.2087,[101]78.4282,[102]79.3813,[103]80.5431,[104]83.2965,[105]85.5194,[106]85.8044,[107]86.4760,[108]86.3406,[109]86.1398, save_imatrix: stored collected data after 110 chunks in datagemma-rag-27b-it-IMat-GGUF/imatrix.dat [110]85.9311,[111]84.8522,[112]83.7617,[113]84.8807,[114]85.5809,[115]85.3940,[116]85.3028,[117]86.0733,[118]86.1875,[119]86.1930, save_imatrix: stored collected data after 120 chunks in datagemma-rag-27b-it-IMat-GGUF/imatrix.dat [120]86.2820,[121]86.9563,[122]86.6484,[123]87.3356,[124]87.7311,[125]88.1534,[126]88.8945,[127]89.7202,[128]90.0131, save_imatrix: stored collected data after 128 chunks in datagemma-rag-27b-it-IMat-GGUF/imatrix.dat llama_perf_print: load time = 5625.10 ms llama_perf_print: prompt eval time = 288812.81 ms / 65536 tokens ( 4.41 ms per token, 226.92 tokens per second) llama_perf_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second) llama_perf_print: total time = 293890.49 ms / 65537 tokens Final estimate: PPL = 90.0131 +/- 3.14406