main: build = 3058 (30e238b2) main: built with cc (Ubuntu 13.2.0-23ubuntu4) 13.2.0 for x86_64-linux-gnu main: seed = 1717202742 llama_model_loader: additional 2 GGUFs metadata loaded. llama_model_loader: loaded meta data with 28 key-value pairs and 723 tensors from K2-ckpt_360.Q8_0/K2-ckpt_360.Q8_0-00001-of-00003.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 = K2-ckpt_360 llama_model_loader: - kv 2: llama.block_count u32 = 80 llama_model_loader: - kv 3: llama.context_length u32 = 2048 llama_model_loader: - kv 4: llama.embedding_length u32 = 8192 llama_model_loader: - kv 5: llama.feed_forward_length u32 = 22016 llama_model_loader: - kv 6: llama.attention.head_count u32 = 64 llama_model_loader: - kv 7: llama.attention.head_count_kv u32 = 64 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 = 7 llama_model_loader: - kv 11: llama.vocab_size u32 = 32032 llama_model_loader: - kv 12: llama.rope.dimension_count u32 = 128 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,32032] = ["", "~~", "~~", "<0x00>", "<... llama_model_loader: - kv 16: tokenizer.ggml.scores arr[f32,32032] = [0.000000, 0.000000, 0.000000, 0.0000... llama_model_loader: - kv 17: tokenizer.ggml.token_type arr[i32,32032] = [2, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 6, ... llama_model_loader: - kv 18: tokenizer.ggml.bos_token_id u32 = 2 llama_model_loader: - kv 19: tokenizer.ggml.eos_token_id u32 = 2 llama_model_loader: - kv 20: tokenizer.ggml.unknown_token_id u32 = 0 llama_model_loader: - kv 21: tokenizer.ggml.padding_token_id u32 = 0 llama_model_loader: - kv 22: tokenizer.ggml.add_bos_token bool = false llama_model_loader: - kv 23: tokenizer.ggml.add_eos_token bool = false llama_model_loader: - kv 24: general.quantization_version u32 = 2 llama_model_loader: - kv 25: split.no u16 = 0 llama_model_loader: - kv 26: split.count u16 = 3 llama_model_loader: - kv 27: split.tensors.count i32 = 723 llama_model_loader: - type f32: 161 tensors llama_model_loader: - type q8_0: 562 tensors llm_load_vocab: special tokens cache size = 291 llm_load_vocab: token to piece cache size = 0.3373 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 = 32032 llm_load_print_meta: n_merges = 0 llm_load_print_meta: n_ctx_train = 2048 llm_load_print_meta: n_embd = 8192 llm_load_print_meta: n_head = 64 llm_load_print_meta: n_head_kv = 64 llm_load_print_meta: n_layer = 80 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 = 1 llm_load_print_meta: n_embd_k_gqa = 8192 llm_load_print_meta: n_embd_v_gqa = 8192 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 = 22016 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 = 2048 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 = 65B llm_load_print_meta: model ftype = Q8_0 llm_load_print_meta: model params = 65.29 B llm_load_print_meta: model size = 64.61 GiB (8.50 BPW) llm_load_print_meta: general.name = K2-ckpt_360 llm_load_print_meta: BOS token = 2 '' llm_load_print_meta: EOS token = 2 '' llm_load_print_meta: UNK token = 0 '' llm_load_print_meta: PAD token = 0 '' 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 = 0.74 MiB llm_load_tensors: offloading 20 repeating layers to GPU llm_load_tensors: offloaded 20/81 layers to GPU llm_load_tensors: CPU buffer size = 24509.70 MiB llm_load_tensors: CPU buffer size = 24426.62 MiB llm_load_tensors: CPU buffer size = 17220.48 MiB llm_load_tensors: CUDA0 buffer size = 16406.25 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 = 960.00 MiB llama_kv_cache_init: CUDA0 KV buffer size = 320.00 MiB llama_new_context_with_model: KV self size = 1280.00 MiB, K (f16): 640.00 MiB, V (f16): 640.00 MiB llama_new_context_with_model: CUDA_Host output buffer size = 0.12 MiB llama_new_context_with_model: CUDA0 compute buffer size = 344.45 MiB llama_new_context_with_model: CUDA_Host compute buffer size = 33.01 MiB llama_new_context_with_model: graph nodes = 2566 llama_new_context_with_model: graph splits = 664 system_info: n_threads = 32 / 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 | compute_imatrix: tokenizing the input .. compute_imatrix: tokenization took 90.613 ms compute_imatrix: computing over 151 chunks with batch_size 512 compute_imatrix: 54.64 seconds per pass - ETA 2 hours 17.50 minutes [1]2.5539,[2]2.1725,[3]2.2401,[4]2.3300,[5]2.6606,[6]2.6351,[7]2.4401,[8]2.7931,[9]2.9649, save_imatrix: stored collected data after 10 chunks in K2-ckpt_360-IMat-GGUF/imatrix.dat [10]3.1688,[11]3.2642,[12]3.0681,[13]3.1229,[14]3.3242,[15]3.6052,[16]3.7343,[17]3.9537,[18]4.1038,[19]4.2571, save_imatrix: stored collected data after 20 chunks in K2-ckpt_360-IMat-GGUF/imatrix.dat [20]4.3761,[21]4.4609,[22]4.3091,[23]4.1686,[24]4.1532,[25]4.1627,[26]4.1712,[27]4.1575,[28]4.2221,[29]4.3438, save_imatrix: stored collected data after 30 chunks in K2-ckpt_360-IMat-GGUF/imatrix.dat [30]4.4658,[31]4.4285,[32]4.4013,[33]4.4066,[34]4.4415,[35]4.4515,[36]4.4150,[37]4.2677,[38]4.1633,[39]4.1454, save_imatrix: stored collected data after 40 chunks in K2-ckpt_360-IMat-GGUF/imatrix.dat [40]4.1309,[41]4.1052,[42]4.1051,[43]4.0700,[44]4.0656,[45]4.0514,[46]4.0371,[47]4.0569,[48]4.1215,[49]4.1808, save_imatrix: stored collected data after 50 chunks in K2-ckpt_360-IMat-GGUF/imatrix.dat [50]4.2053,[51]4.2157,[52]4.2462,[53]4.3301,[54]4.3998,[55]4.4450,[56]4.4333,[57]4.4086,[58]4.4343,[59]4.4672, save_imatrix: stored collected data after 60 chunks in K2-ckpt_360-IMat-GGUF/imatrix.dat [60]4.5250,[61]4.4823,[62]4.5074,[63]4.5355,[64]4.5937,[65]4.6457,[66]4.6640,[67]4.7075,[68]4.7445,[69]4.7587, save_imatrix: stored collected data after 70 chunks in K2-ckpt_360-IMat-GGUF/imatrix.dat [70]4.7712,[71]4.7811,[72]4.7778,[73]4.7534,[74]4.7181,[75]4.7263,[76]4.7284,[77]4.7388,[78]4.7089,[79]4.7126, save_imatrix: stored collected data after 80 chunks in K2-ckpt_360-IMat-GGUF/imatrix.dat [80]4.7190,[81]4.6978,[82]4.6902,[83]4.6675,[84]4.6722,[85]4.6693,[86]4.6627,[87]4.6610,[88]4.6630,[89]4.6458, save_imatrix: stored collected data after 90 chunks in K2-ckpt_360-IMat-GGUF/imatrix.dat [90]4.6360,[91]4.6361,[92]4.6215,[93]4.6015,[94]4.5871,[95]4.5457,[96]4.5578,[97]4.5484,[98]4.5443,[99]4.5292, save_imatrix: stored collected data after 100 chunks in K2-ckpt_360-IMat-GGUF/imatrix.dat [100]4.5199,[101]4.5321,[102]4.5079,[103]4.4911,[104]4.4818,[105]4.4947,[106]4.5010,[107]4.5164,[108]4.5313,[109]4.4987, save_imatrix: stored collected data after 110 chunks in K2-ckpt_360-IMat-GGUF/imatrix.dat [110]4.4669,[111]4.4338,[112]4.4028,[113]4.3726,[114]4.3421,[115]4.3136,[116]4.2847,[117]4.2740,[118]4.2853,[119]4.2989, save_imatrix: stored collected data after 120 chunks in K2-ckpt_360-IMat-GGUF/imatrix.dat [120]4.3347,[121]4.3653,[122]4.4021,[123]4.4374,[124]4.4908,[125]4.5376,[126]4.5504,[127]4.5590,[128]4.5281,[129]4.5314, save_imatrix: stored collected data after 130 chunks in K2-ckpt_360-IMat-GGUF/imatrix.dat [130]4.5324,[131]4.5179,[132]4.4969,[133]4.4791,[134]4.4977,[135]4.5185,[136]4.5240,[137]4.5240,[138]4.5372,[139]4.5527, save_imatrix: stored collected data after 140 chunks in K2-ckpt_360-IMat-GGUF/imatrix.dat [140]4.5666,[141]4.5727,[142]4.5798,[143]4.5811,[144]4.5694,[145]4.5811,[146]4.5913,[147]4.5950,[148]4.6098,[149]4.6184, save_imatrix: stored collected data after 150 chunks in K2-ckpt_360-IMat-GGUF/imatrix.dat [150]4.6283,[151]4.6390, save_imatrix: stored collected data after 151 chunks in K2-ckpt_360-IMat-GGUF/imatrix.dat llama_print_timings: load time = 80625.37 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 = 5710379.48 ms / 77312 tokens ( 73.86 ms per token, 13.54 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 = 5750793.99 ms / 77313 tokens Final estimate: PPL = 4.6390 +/- 0.05047