llama_bin_run\llama-server.exe : ggml_cuda_init: found 2 CUDA devices (Total VRAM: 24575 MiB): 위치 줄:1 문자:1 + llama_bin_run\llama-server.exe --model "models\Q4_K_M\Qwen3. 5-122B-A1 ... + ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~ + CategoryInfo : NotSpecified: (ggml_cuda_init:. ..AM: 24575 MiB)::String) [], RemoteException + FullyQualifiedErrorId : NativeCommandError Device 0: NVIDIA GeForce RTX 3060, compute capability 8.6, V MM: yes, VRAM: 12287 MiB Device 1: NVIDIA GeForce RTX 3060, compute capability 8.6, V MM: yes, VRAM: 12287 MiB load_backend: loaded CUDA backend from C:\Users\Variet-Worker\ Desktop\variet-llm\llama_bin_run\ggml-cuda.dll load_backend: loaded RPC backend from C:\Users\Variet-Worker\D esktop\variet-llm\llama_bin_run\ggml-rpc.dll load_backend: loaded CPU backend from C:\Users\Variet-Worker\D esktop\variet-llm\llama_bin_run\ggml-cpu-haswell.dll system info: n_threads = 6, n_threads_batch = 6, total_threads = 16 system_info: n_threads = 6 (n_threads_batch = 6) / 16 | CUDA : ARCHS = 500,610,700,750,800,860,890 | USE_GRAPHS = 1 | PEER_M AX_BATCH_SIZE = 128 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | A VX2 = 1 | F16C = 1 | FMA = 1 | BMI2 = 1 | LLAMAFILE = 1 | OPEN MP = 1 | REPACK = 1 | Running without SSL init: using 15 threads for HTTP server start: binding port with default address family main: loading model srv load_model: loading model 'models\Q4_K_M\Qwen3.5-122B-A 10B-Q4_K_M-00001-of-00003.gguf' common_init_result: fitting params to device memory, for bugs during this step try to reproduce them with -fit off, or provi de --verbose logs if the bug only occurs with -fit on llama_params_fit_impl: projected memory use with initial param eters [MiB]: llama_params_fit_impl: - CUDA0 (NVIDIA GeForce RTX 3060): 1 2287 total, 37929 used, -26764 free vs. target of 1024 llama_params_fit_impl: - CUDA1 (NVIDIA GeForce RTX 3060): 1 2287 total, 35760 used, -24592 free vs. target of 1024 llama_params_fit_impl: projected to use 73690 MiB of device me mory vs. 22333 MiB of free device memory llama_params_fit_impl: cannot meet free memory targets on all devices, need to use 53405 MiB less in total llama_params_fit_impl: context size set by user to 32768 -> no change llama_params_fit_impl: with only dense weights in device memor y there is a total surplus of 13058 MiB llama_params_fit_impl: filling dense-only layers back-to-front : llama_params_fit_impl: - CUDA1 (NVIDIA GeForce RTX 3060): 49 layers, 7651 MiB used, 3516 MiB free llama_params_fit_impl: - CUDA0 (NVIDIA GeForce RTX 3060): 0 layers, 966 MiB used, 10198 MiB free llama_params_fit_impl: converting dense-only layers to full la yers and filling them front-to-back with overflow to next devi ce/system memory: llama_params_fit_impl: - CUDA0 (NVIDIA GeForce RTX 3060): 7 layers ( 1 overflowing), 10055 MiB used, 1109 MiB free llama_params_fit_impl: - CUDA1 (NVIDIA GeForce RTX 3060): 42 layers (40 overflowing), 9786 MiB used, 1381 MiB free llama_params_fit: successfully fit params to free device memor y llama_params_fit: fitting params to free memory took 0.49 seco nds llama_model_load_from_file_impl: using device CUDA0 (NVIDIA Ge Force RTX 3060) (0000:04:00.0) - 11245 MiB free llama_model_load_from_file_impl: using device CUDA1 (NVIDIA Ge Force RTX 3060) (0000:06:00.0) - 11240 MiB free llama_model_loader: additional 2 GGUFs metadata loaded. llama_model_loader: loaded meta data with 55 key-value pairs a nd 879 tensors from models\Q4_K_M\Qwen3.5-122B-A10B-Q4_K_M-000 01-of-00003.gguf (version GGUF V3 (latest)) llama_model_loader: Dumping metadata keys/values. Note: KV ove rrides do not apply in this output. llama_model_loader: - kv 0: general.ar chitecture str = qwen35moe llama_model_loader: - kv 1: ge neral.type str = model llama_model_loader: - kv 2: general.samp ling.top_k i32 = 20 llama_model_loader: - kv 3: general.samp ling.top_p f32 = 0.950000 llama_model_loader: - kv 4: general.sam pling.temp f32 = 0.600000 llama_model_loader: - kv 5: ge neral.name str = Qwen3.5-122B-A10B llama_model_loader: - kv 6: genera l.basename str = Qwen3.5-122B-A10B llama_model_loader: - kv 7: general.qu antized_by str = Unsloth llama_model_loader: - kv 8: general. size_label str = 122B-A10B llama_model_loader: - kv 9: gener al.license str = apache-2.0 llama_model_loader: - kv 10: general.li cense.link str = https://huggingface.co/Qwen/Qwen 3.5-1... llama_model_loader: - kv 11: genera l.repo_url str = https://huggingface.co/unsloth llama_model_loader: - kv 12: general.base_m odel.count u32 = 1 llama_model_loader: - kv 13: general.base_mo del.0.name str = Qwen3.5 122B A10B llama_model_loader: - kv 14: general.base_model.0.or ganization str = Qwen llama_model_loader: - kv 15: general.base_model. 0.repo_url str = https://huggingface.co/Qwen/Qwen 3.5-1... llama_model_loader: - kv 16: ge neral.tags arr[str,2] = ["unsloth", "image-text-to-text" ] llama_model_loader: - kv 17: qwen35moe.b lock_count u32 = 48 llama_model_loader: - kv 18: qwen35moe.cont ext_length u32 = 262144 llama_model_loader: - kv 19: qwen35moe.embedd ing_length u32 = 3072 llama_model_loader: - kv 20: qwen35moe.attention. head_count u32 = 32 llama_model_loader: - kv 21: qwen35moe.attention.hea d_count_kv u32 = 2 llama_model_loader: - kv 22: qwen35moe.rope.dimensio n_sections arr[i32,4] = [11, 11, 10, 0] llama_model_loader: - kv 23: qwen35moe.rope .freq_base f32 = 10000000.000000 llama_model_loader: - kv 24: qwen35moe.attention.layer_norm_r ms_epsilon f32 = 0.000001 llama_model_loader: - kv 25: qwen35moe.ex pert_count u32 = 256 llama_model_loader: - kv 26: qwen35moe.expert_ used_count u32 = 8 llama_model_loader: - kv 27: qwen35moe.attention. key_length u32 = 256 llama_model_loader: - kv 28: qwen35moe.attention.va lue_length u32 = 256 llama_model_loader: - kv 29: qwen35moe.expert_feed_forw ard_length u32 = 1024 llama_model_loader: - kv 30: qwen35moe.expert_shared_feed_for ward_length u32 = 1024 llama_model_loader: - kv 31: qwen35moe.ssm.c onv_kernel u32 = 4 llama_model_loader: - kv 32: qwen35moe.ssm. state_size u32 = 128 llama_model_loader: - kv 33: qwen35moe.ssm.g roup_count u32 = 16 llama_model_loader: - kv 34: qwen35moe.ssm.time _step_rank u32 = 64 llama_model_loader: - kv 35: qwen35moe.ssm. inner_size u32 = 8192 llama_model_loader: - kv 36: qwen35moe.full_attentio n_interval u32 = 4 llama_model_loader: - kv 37: qwen35moe.rope.dimen sion_count u32 = 64 llama_model_loader: - kv 38: tokenizer. ggml.model str = gpt2 llama_model_loader: - kv 39: tokenize r.ggml.pre str = qwen35 llama_model_loader: - kv 40: tokenizer.g gml.tokens arr[str,248320] = ["!", "\"", "#", "$", "%", "&", "'", ... llama_model_loader: - kv 41: tokenizer.ggml. token_type arr[i32,248320] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 , 1, ... llama_model_loader: - kv 42: tokenizer.g gml.merges arr[str,247587] = ["휔 휔", "휔휔 휔휔", "i n", "휔 t",.. . llama_model_loader: - kv 43: tokenizer.ggml.eo s_token_id u32 = 248046 llama_model_loader: - kv 44: tokenizer.ggml.paddin g_token_id u32 = 248055 llama_model_loader: - kv 45: tokenizer.cha t_template str = {%- set image_count = namespace( value... llama_model_loader: - kv 46: general.quantizati on_version u32 = 2 llama_model_loader: - kv 47: general .file_type u32 = 15 llama_model_loader: - kv 48: quantize.im atrix.file str = Qwen3.5-122B-A10B-GGUF/imatrix_u nslot... llama_model_loader: - kv 49: quantize.imatr ix.dataset str = unsloth_calibration_Qwen3.5-122B -A10B... llama_model_loader: - kv 50: quantize.imatrix.ent ries_count u32 = 612 llama_model_loader: - kv 51: quantize.imatrix.ch unks_count u32 = 76 llama_model_loader: - kv 52: split.no u16 = 0 llama_model_loader: - kv 53: split.ten sors.count i32 = 879 llama_model_loader: - kv 54: s plit.count u16 = 3 llama_model_loader: - type f32: 361 tensors llama_model_loader: - type q8_0: 373 tensors llama_model_loader: - type q4_K: 96 tensors llama_model_loader: - type q5_K: 48 tensors llama_model_loader: - type q6_K: 1 tensors print_info: file format = GGUF V3 (latest) print_info: file type = Q4_K - Medium print_info: file size = 71.27 GiB (5.01 BPW) load: 0 unused tokens load: printing all EOG tokens: load: - 248044 ('<|endoftext|>') load: - 248046 ('<|im_end|>') load: - 248063 ('<|fim_pad|>') load: - 248064 ('<|repo_name|>') load: - 248065 ('<|file_sep|>') load: special tokens cache size = 33 load: token to piece cache size = 1.7581 MB print_info: arch = qwen35moe print_info: vocab_only = 0 print_info: no_alloc = 0 print_info: n_ctx_train = 262144 print_info: n_embd = 3072 print_info: n_embd_inp = 3072 print_info: n_layer = 48 print_info: n_head = 32 print_info: n_head_kv = 2 print_info: n_rot = 64 print_info: n_swa = 0 print_info: is_swa_any = 0 print_info: n_embd_head_k = 256 print_info: n_embd_head_v = 256 print_info: n_gqa = 16 print_info: n_embd_k_gqa = 512 print_info: n_embd_v_gqa = 512 print_info: f_norm_eps = 0.0e+00 print_info: f_norm_rms_eps = 1.0e-06 print_info: f_clamp_kqv = 0.0e+00 print_info: f_max_alibi_bias = 0.0e+00 print_info: f_logit_scale = 0.0e+00 print_info: f_attn_scale = 0.0e+00 print_info: n_ff = 0 print_info: n_expert = 256 print_info: n_expert_used = 8 print_info: n_expert_groups = 0 print_info: n_group_used = 0 print_info: causal attn = 1 print_info: pooling type = -1 print_info: rope type = 40 print_info: rope scaling = linear print_info: freq_base_train = 10000000.0 print_info: freq_scale_train = 1 print_info: n_ctx_orig_yarn = 262144 print_info: rope_yarn_log_mul = 0.0000 print_info: rope_finetuned = unknown print_info: mrope sections = [11, 11, 10, 0] print_info: ssm_d_conv = 4 print_info: ssm_d_inner = 8192 print_info: ssm_d_state = 128 print_info: ssm_dt_rank = 64 print_info: ssm_n_group = 16 print_info: ssm_dt_b_c_rms = 0 print_info: model type = 122B.A10B print_info: model params = 122.11 B print_info: general.name = Qwen3.5-122B-A10B print_info: vocab type = BPE print_info: n_vocab = 248320 print_info: n_merges = 247587 print_info: BOS token = 11 ',' print_info: EOS token = 248046 '<|im_end|>' print_info: EOT token = 248046 '<|im_end|>' print_info: PAD token = 248055 '<|vision_pad|>' print_info: LF token = 198 '훹' print_info: FIM PRE token = 248060 '<|fim_prefix|>' print_info: FIM SUF token = 248062 '<|fim_suffix|>' print_info: FIM MID token = 248061 '<|fim_middle|>' print_info: FIM PAD token = 248063 '<|fim_pad|>' print_info: FIM REP token = 248064 '<|repo_name|>' print_info: FIM SEP token = 248065 '<|file_sep|>' print_info: EOG token = 248044 '<|endoftext|>' print_info: EOG token = 248046 '<|im_end|>' print_info: EOG token = 248063 '<|fim_pad|>' print_info: EOG token = 248064 '<|repo_name|>' print_info: EOG token = 248065 '<|file_sep|>' print_info: max token length = 256 load_tensors: loading model tensors, this can take a while... (mmap = true, direct_io = false) llama_model_loader: tensor overrides to CPU are used with mmap enabled - consider using --no-mmap for better performance load_tensors: offloading output layer to GPU load_tensors: offloading 47 repeating layers to GPU load_tensors: offloaded 49/49 layers to GPU load_tensors: CPU_Mapped model buffer size = 47000.12 MiB load_tensors: CPU_Mapped model buffer size = 25271.75 MiB load_tensors: CUDA0 model buffer size = 9046.32 MiB load_tensors: CUDA1 model buffer size = 8860.98 MiB