llama_model_loader: loaded meta data with 30 key-value pairs and 963 tensors from Qwen2-Math-72B-Instruct-IMat-GGUF/Qwen2-Math-72B-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 = qwen2 llama_model_loader: - kv 1: general.type str = model llama_model_loader: - kv 2: general.name str = Qwen2 Math 72B Instruct llama_model_loader: - kv 3: general.finetune str = Instruct llama_model_loader: - kv 4: general.basename str = Qwen2-Math llama_model_loader: - kv 5: general.size_label str = 72B llama_model_loader: - kv 6: general.license str = other llama_model_loader: - kv 7: general.license.name str = tongyi-qianwen llama_model_loader: - kv 8: general.license.link str = https://huggingface.co/Qwen/Qwen2-Mat... llama_model_loader: - kv 9: general.tags arr[str,2] = ["chat", "text-generation"] llama_model_loader: - kv 10: general.languages arr[str,1] = ["en"] llama_model_loader: - kv 11: qwen2.block_count u32 = 80 llama_model_loader: - kv 12: qwen2.context_length u32 = 4096 llama_model_loader: - kv 13: qwen2.embedding_length u32 = 8192 llama_model_loader: - kv 14: qwen2.feed_forward_length u32 = 29568 llama_model_loader: - kv 15: qwen2.attention.head_count u32 = 64 llama_model_loader: - kv 16: qwen2.attention.head_count_kv u32 = 8 llama_model_loader: - kv 17: qwen2.rope.freq_base f32 = 10000.000000 llama_model_loader: - kv 18: qwen2.attention.layer_norm_rms_epsilon f32 = 0.000001 llama_model_loader: - kv 19: general.file_type u32 = 7 llama_model_loader: - kv 20: tokenizer.ggml.model str = gpt2 llama_model_loader: - kv 21: tokenizer.ggml.pre str = qwen2 llama_model_loader: - kv 22: tokenizer.ggml.tokens arr[str,152064] = ["!", "\"", "#", "$", "%", "&", "'", ... llama_model_loader: - kv 23: tokenizer.ggml.token_type arr[i32,152064] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ... llama_model_loader: - kv 24: tokenizer.ggml.merges arr[str,151387] = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t",... llama_model_loader: - kv 25: tokenizer.ggml.eos_token_id u32 = 151645 llama_model_loader: - kv 26: tokenizer.ggml.padding_token_id u32 = 151643 llama_model_loader: - kv 27: tokenizer.ggml.bos_token_id u32 = 151643 llama_model_loader: - kv 28: tokenizer.chat_template str = {% for message in messages %}{% if lo... llama_model_loader: - kv 29: general.quantization_version u32 = 2 llama_model_loader: - type f32: 401 tensors llama_model_loader: - type q8_0: 562 tensors llm_load_vocab: special tokens cache size = 3 llm_load_vocab: token to piece cache size = 0.9308 MB llm_load_print_meta: format = GGUF V3 (latest) llm_load_print_meta: arch = qwen2 llm_load_print_meta: vocab type = BPE llm_load_print_meta: n_vocab = 152064 llm_load_print_meta: n_merges = 151387 llm_load_print_meta: vocab_only = 0 llm_load_print_meta: n_ctx_train = 4096 llm_load_print_meta: n_embd = 8192 llm_load_print_meta: n_layer = 80 llm_load_print_meta: n_head = 64 llm_load_print_meta: n_head_kv = 8 llm_load_print_meta: n_rot = 128 llm_load_print_meta: n_swa = 0 llm_load_print_meta: n_embd_head_k = 128 llm_load_print_meta: n_embd_head_v = 128 llm_load_print_meta: n_gqa = 8 llm_load_print_meta: n_embd_k_gqa = 1024 llm_load_print_meta: n_embd_v_gqa = 1024 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 = 29568 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 = 4096 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 = 70B llm_load_print_meta: model ftype = Q8_0 llm_load_print_meta: model params = 72.71 B llm_load_print_meta: model size = 71.95 GiB (8.50 BPW) llm_load_print_meta: general.name = Qwen2 Math 72B Instruct llm_load_print_meta: BOS token = 151643 '<|endoftext|>' llm_load_print_meta: EOS token = 151645 '<|im_end|>' llm_load_print_meta: PAD token = 151643 '<|endoftext|>' llm_load_print_meta: LF token = 148848 'ÄĬ' llm_load_print_meta: EOT token = 151645 '<|im_end|>' llm_load_print_meta: max token length = 256 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 = 0.85 MiB llm_load_tensors: offloading 24 repeating layers to GPU llm_load_tensors: offloaded 24/81 layers to GPU llm_load_tensors: CPU buffer size = 73677.66 MiB llm_load_tensors: CUDA0 buffer size = 21345.94 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 = 112.00 MiB llama_kv_cache_init: CUDA0 KV buffer size = 48.00 MiB llama_new_context_with_model: KV self size = 160.00 MiB, K (f16): 80.00 MiB, V (f16): 80.00 MiB llama_new_context_with_model: CUDA_Host output buffer size = 0.58 MiB llama_new_context_with_model: CUDA0 compute buffer size = 1575.25 MiB llama_new_context_with_model: CUDA_Host compute buffer size = 17.01 MiB llama_new_context_with_model: graph nodes = 2806 llama_new_context_with_model: graph splits = 788 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 131.599 ms compute_imatrix: computing over 128 chunks with batch_size 512 compute_imatrix: 6.19 seconds per pass - ETA 13.20 minutes [1]5.6455,[2]3.9234,[3]3.8034,[4]4.2022,[5]4.0305,[6]3.7286,[7]3.8681,[8]3.8127,[9]4.3434, save_imatrix: stored collected data after 10 chunks in Qwen2-Math-72B-Instruct-IMat-GGUF/imatrix.dat [10]4.3039,[11]4.2382,[12]4.7030,[13]5.3245,[14]5.6627,[15]6.2616,[16]6.6116,[17]6.8341,[18]7.3613,[19]7.2211, save_imatrix: stored collected data after 20 chunks in Qwen2-Math-72B-Instruct-IMat-GGUF/imatrix.dat [20]7.3923,[21]7.5995,[22]7.7329,[23]7.6111,[24]7.8101,[25]8.0203,[26]7.8960,[27]8.1522,[28]8.3985,[29]8.6066, save_imatrix: stored collected data after 30 chunks in Qwen2-Math-72B-Instruct-IMat-GGUF/imatrix.dat [30]8.5856,[31]8.4330,[32]8.1338,[33]7.9839,[34]7.8144,[35]7.7220,[36]7.8617,[37]8.2305,[38]8.3502,[39]8.2447, save_imatrix: stored collected data after 40 chunks in Qwen2-Math-72B-Instruct-IMat-GGUF/imatrix.dat [40]8.4154,[41]8.4742,[42]8.9173,[43]9.1514,[44]9.4621,[45]9.7172,[46]9.8738,[47]9.7065,[48]9.7584,[49]9.8711, save_imatrix: stored collected data after 50 chunks in Qwen2-Math-72B-Instruct-IMat-GGUF/imatrix.dat [50]9.9261,[51]9.7606,[52]9.8563,[53]10.1063,[54]10.2060,[55]10.3781,[56]10.4551,[57]10.5103,[58]10.5589,[59]10.5676, save_imatrix: stored collected data after 60 chunks in Qwen2-Math-72B-Instruct-IMat-GGUF/imatrix.dat [60]10.6308,[61]10.5728,[62]10.5144,[63]10.5751,[64]10.6883,[65]10.6007,[66]10.5643,[67]10.5692,[68]10.4213,[69]10.3358, save_imatrix: stored collected data after 70 chunks in Qwen2-Math-72B-Instruct-IMat-GGUF/imatrix.dat [70]10.2741,[71]10.2012,[72]10.1697,[73]10.1553,[74]10.0338,[75]9.9174,[76]9.8097,[77]9.7600,[78]9.7222,[79]9.6761, save_imatrix: stored collected data after 80 chunks in Qwen2-Math-72B-Instruct-IMat-GGUF/imatrix.dat [80]9.5676,[81]9.5837,[82]9.5514,[83]9.4819,[84]9.4974,[85]9.5034,[86]9.4469,[87]9.3856,[88]9.3437,[89]9.3642, save_imatrix: stored collected data after 90 chunks in Qwen2-Math-72B-Instruct-IMat-GGUF/imatrix.dat [90]9.3875,[91]9.3650,[92]9.2606,[93]9.1723,[94]9.0670,[95]8.9726,[96]8.8849,[97]8.7893,[98]8.7011,[99]8.6811, save_imatrix: stored collected data after 100 chunks in Qwen2-Math-72B-Instruct-IMat-GGUF/imatrix.dat [100]8.6864,[101]8.7119,[102]8.8225,[103]8.9388,[104]9.0161,[105]9.1664,[106]9.2696,[107]9.3045,[108]9.2621,[109]9.2634, save_imatrix: stored collected data after 110 chunks in Qwen2-Math-72B-Instruct-IMat-GGUF/imatrix.dat [110]9.2652,[111]9.1407,[112]9.0250,[113]8.9404,[114]8.9935,[115]8.9971,[116]9.0067,[117]9.0303,[118]9.0722,[119]9.0770, save_imatrix: stored collected data after 120 chunks in Qwen2-Math-72B-Instruct-IMat-GGUF/imatrix.dat [120]9.0612,[121]9.0699,[122]9.0103,[123]9.0702,[124]9.1244,[125]9.1714,[126]9.2531,[127]9.3210,[128]9.3817, save_imatrix: stored collected data after 128 chunks in Qwen2-Math-72B-Instruct-IMat-GGUF/imatrix.dat llama_print_timings: load time = 32720.43 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 = 756710.27 ms / 65536 tokens ( 11.55 ms per token, 86.61 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 = 784466.13 ms / 65537 tokens Final estimate: PPL = 9.3817 +/- 0.15152