llama_bin_run\llama-server.exe : ggml_cuda_init: found 2 CU DA devices (Total VRAM: 24575 MiB): 위치 줄:1 문자:1 + llama_bin_run\llama-server.exe --model models\Qwen3.5-35B -A3B-Q4_K_M. ... + ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~ + CategoryInfo : NotSpecified: (ggml_cuda_ini t:...AM: 24575 MiB)::String) [], RemoteException + FullyQualifiedErrorId : NativeCommandError Device 0: NVIDIA GeForce RTX 3060, compute capability 8.6 , VMM: yes, VRAM: 12287 MiB Device 1: NVIDIA GeForce RTX 3060, compute capability 8.6 , VMM: yes, VRAM: 12287 MiB load_backend: loaded CUDA backend from C:\Users\Variet-Work er\Desktop\variet-llm\llama_bin_run\ggml-cuda.dll load_backend: loaded RPC backend from C:\Users\Variet-Worke r\Desktop\variet-llm\llama_bin_run\ggml-rpc.dll load_backend: loaded CPU backend from C:\Users\Variet-Worke r\Desktop\variet-llm\llama_bin_run\ggml-cpu-haswell.dll system info: n_threads = 6, n_threads_batch = 6, total_thre ads = 16 system_info: n_threads = 6 (n_threads_batch = 6) / 16 | CUD A : ARCHS = 500,610,700,750,800,860,890 | USE_GRAPHS = 1 | PEER_MAX_BATCH_SIZE = 128 | CPU : SSE3 = 1 | SSSE3 = 1 | AV X = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | BMI2 = 1 | LLAMAFIL E = 1 | OPENMP = 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\Qwen3.5-35B-A3B-Q4 _K_M.gguf' common_init_result: fitting params to device memory, for bu gs during this step try to reproduce them with -fit off, or provide --verbose logs if the bug only occurs with -fit on llama_params_fit_impl: projected memory use with initial pa rameters [MiB]: llama_params_fit_impl: - CUDA0 (NVIDIA GeForce RTX 3060): 12287 total, 11893 used, -682 free vs. target of 102 4 llama_params_fit_impl: - CUDA1 (NVIDIA GeForce RTX 3060): 12287 total, 11068 used, 141 free vs. target of 102 4 llama_params_fit_impl: projected to use 22961 MiB of device memory vs. 22421 MiB of free device memory llama_params_fit_impl: cannot meet free memory targets on a ll devices, need to use 2588 MiB less in total llama_params_fit_impl: context size set by user to 262144 - > no change llama_params_fit: failed to fit params to free device memor y: n_gpu_layers already set by user to 999, abort llama_params_fit: fitting params to free memory took 0.58 s econds llama_model_load_from_file_impl: using device CUDA0 (NVIDIA GeForce RTX 3060) (0000:04:00.0) - 11245 MiB free llama_model_load_from_file_impl: using device CUDA1 (NVIDIA GeForce RTX 3060) (0000:06:00.0) - 11240 MiB free llama_model_loader: loaded meta data with 52 key-value pair s and 733 tensors from models\Qwen3.5-35B-A3B-Q4_K_M.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 = qwen35moe llama_model_loader: - kv 1: general.type str = model llama_model_loader: - kv 2: general.s ampling.top_k i32 = 20 llama_model_loader: - kv 3: general.s ampling.top_p f32 = 0.950000 llama_model_loader: - kv 4: general. sampling.temp f32 = 1.000000 llama_model_loader: - kv 5: general.name str = Qwen3.5-35B-A3B llama_model_loader: - kv 6: gen eral.basename str = Qwen3.5-35B-A3B llama_model_loader: - kv 7: general .quantized_by str = Unsloth llama_model_loader: - kv 8: gener al.size_label str = 35B-A3B llama_model_loader: - kv 9: ge neral.license str = apache-2.0 llama_model_loader: - kv 10: general .license.link str = https://huggingface.co/Qwe n/Qwen3.5-3... llama_model_loader: - kv 11: gen eral.repo_url str = https://huggingface.co/uns loth llama_model_loader: - kv 12: general.bas e_model.count u32 = 1 llama_model_loader: - kv 13: general.base _model.0.name str = Qwen3.5 35B A3B llama_model_loader: - kv 14: general.base_model.0 .organization str = Qwen llama_model_loader: - kv 15: general.base_mod el.0.repo_url str = https://huggingface.co/Qwe n/Qwen3.5-3... llama_model_loader: - kv 16: general.tags arr[str,2] = ["unsloth", "image-text-to -text"] llama_model_loader: - kv 17: qwen35mo e.block_count u32 = 40 llama_model_loader: - kv 18: qwen35moe.c ontext_length u32 = 262144 llama_model_loader: - kv 19: qwen35moe.emb edding_length u32 = 2048 llama_model_loader: - kv 20: qwen35moe.attenti on.head_count u32 = 16 llama_model_loader: - kv 21: qwen35moe.attention. head_count_kv u32 = 2 llama_model_loader: - kv 22: qwen35moe.rope.dimen sion_sections arr[i32,4] = [11, 11, 10, 0] llama_model_loader: - kv 23: qwen35moe.r ope.freq_base f32 = 10000000.000000 llama_model_loader: - kv 24: qwen35moe.attention.layer_nor m_rms_epsilon f32 = 0.000001 llama_model_loader: - kv 25: qwen35moe .expert_count u32 = 256 llama_model_loader: - kv 26: qwen35moe.expe rt_used_count u32 = 8 llama_model_loader: - kv 27: qwen35moe.attenti on.key_length u32 = 256 llama_model_loader: - kv 28: qwen35moe.attention .value_length u32 = 256 llama_model_loader: - kv 29: qwen35moe.expert_feed_f orward_length u32 = 512 llama_model_loader: - kv 30: qwen35moe.expert_shared_feed_ forward_length u32 = 512 llama_model_loader: - kv 31: qwen35moe.ss m.conv_kernel u32 = 4 llama_model_loader: - kv 32: qwen35moe.s sm.state_size u32 = 128 llama_model_loader: - kv 33: qwen35moe.ss m.group_count u32 = 16 llama_model_loader: - kv 34: qwen35moe.ssm.t ime_step_rank u32 = 32 llama_model_loader: - kv 35: qwen35moe.s sm.inner_size u32 = 4096 llama_model_loader: - kv 36: qwen35moe.full_atten tion_interval u32 = 4 llama_model_loader: - kv 37: qwen35moe.rope.di mension_count u32 = 64 llama_model_loader: - kv 38: tokeniz er.ggml.model str = gpt2 llama_model_loader: - kv 39: token izer.ggml.pre str = qwen35 llama_model_loader: - kv 40: tokenize r.ggml.tokens arr[str,248320] = ["!", "\"", "#", "$", "%", "&", "'", ... llama_model_loader: - kv 41: tokenizer.gg ml.token_type arr[i32,248320] = [1, 1, 1, 1, 1, 1, 1, 1, 1 , 1, 1, 1, ... llama_model_loader: - kv 42: tokenize r.ggml.merges arr[str,247587] = ["휔 휔", "휔휔 휔휔", "i n", "휔 t",... llama_model_loader: - kv 43: tokenizer.ggml .eos_token_id u32 = 248046 llama_model_loader: - kv 44: tokenizer.ggml.pad ding_token_id u32 = 248055 llama_model_loader: - kv 45: tokenizer. chat_template str = {%- set image_count = name space(value... llama_model_loader: - kv 46: general.quantiz ation_version u32 = 2 llama_model_loader: - kv 47: gene ral.file_type u32 = 15 llama_model_loader: - kv 48: quantize .imatrix.file str = Qwen3.5-35B-A3B-GGUF/imatr ix_unsloth.... llama_model_loader: - kv 49: quantize.im atrix.dataset str = unsloth_calibration_Qwen3. 5-35B-A3B.txt llama_model_loader: - kv 50: quantize.imatrix. entries_count u32 = 510 llama_model_loader: - kv 51: quantize.imatrix .chunks_count u32 = 76 llama_model_loader: - type f32: 301 tensors llama_model_loader: - type q8_0: 311 tensors llama_model_loader: - type q4_K: 80 tensors llama_model_loader: - type q5_K: 40 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 = 20.49 GiB (5.08 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 = 2048 print_info: n_embd_inp = 2048 print_info: n_layer = 40 print_info: n_head = 16 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 = 8 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 = 4096 print_info: ssm_d_state = 128 print_info: ssm_dt_rank = 32 print_info: ssm_n_group = 16 print_info: ssm_dt_b_c_rms = 0 print_info: model type = 35B.A3B print_info: model params = 34.66 B print_info: general.name = Qwen3.5-35B-A3B 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 = false, direct_io = false) load_tensors: offloading output layer to GPU load_tensors: offloading 39 repeating layers to GPU load_tensors: offloaded 41/41 layers to GPU load_tensors: CUDA0 model buffer size = 10539.70 MiB load_tensors: CUDA1 model buffer size = 9930.62 MiB load_tensors: CUDA_Host model buffer size = 515.31 MiB ........................................................... ...................................... common_init_result: added <|endoftext|> logit bias = -inf common_init_result: added <|im_end|> logit bias = -inf common_init_result: added <|fim_pad|> logit bias = -inf common_init_result: added <|repo_name|> logit bias = -inf common_init_result: added <|file_sep|> logit bias = -inf llama_context: constructing llama_context llama_context: n_seq_max = 1 llama_context: n_ctx = 262144 llama_context: n_ctx_seq = 262144 llama_context: n_batch = 512 llama_context: n_ubatch = 128 llama_context: causal_attn = 1 llama_context: flash_attn = enabled llama_context: kv_unified = false llama_context: freq_base = 10000000.0 llama_context: freq_scale = 1 llama_context: CUDA_Host output buffer size = 0.95 Mi B llama_kv_cache: CUDA0 KV buffer size = 720.00 MiB llama_kv_cache: CUDA1 KV buffer size = 720.00 MiB llama_kv_cache: size = 1440.00 MiB (262144 cells, 10 layer s, 1/1 seqs), K (q4_0): 720.00 MiB, V (q4_0): 720.00 MiB llama_kv_cache: attn_rot_k = 1 llama_kv_cache: attn_rot_v = 1 llama_memory_recurrent: CUDA0 RS buffer size = 33.5 0 MiB llama_memory_recurrent: CUDA1 RS buffer size = 29.3 1 MiB llama_memory_recurrent: size = 62.81 MiB ( 1 cells, 40 layers, 1 seqs), R (f32): 2.81 MiB, S (f32): 60.00 MiB llama_context: pipeline parallelism enabled sched_reserve: reserving ... sched_reserve: resolving fused Gated Delta Net support: sched_reserve: fused Gated Delta Net (autoregressive) enabl ed sched_reserve: fused Gated Delta Net (chunked) enabled ggml_backend_cuda_buffer_type_alloc_buffer: allocating 600. 08 MiB on device 0: cudaMalloc failed: out of memory ggml_gallocr_reserve_n_impl: failed to allocate CUDA0 buffe r of size 629228032