github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-22 13:42:25 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #9521] The model was forcibly terminated after giving half of the answer. #83893

New Issue
Closed
opened 2026-05-09 19:20:20 -05:00 by GiteaMirror · 1 comment
GiteaMirror commented 2026-05-09 19:20:20 -05:00
Owner
Copy Link

Originally created by @flora298 on GitHub (Mar 5, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/9521

What is the issue?

I was running deepseek-r1:671b model, and the process was forcibly terminated. In the terminal, it showed POST predict: Post "http://127.0.0.1:49813/completion": read top 127.0.0.1:49816->127.0.0.1:49813: wsarec: An existing connection was forcibly closed by the remote host.after running half of the reasoning process.

I made one of the GPU invisible since the previous log said CUDA error: CUBLAS_STATUS_NOT_INITIALIZED current device: 5, and it went with the same error line. POST predict: Post "http://127.0.0.1:49813/completion": read top 127.0.0.1:49816->127.0.0.1:49813: wsarec: An existing connection was forcibly closed by the remote host. I am not sure making device 5 invisible was the right thing to do, but the above log no longer has CUDA error.

I think the problem might be on these two lines in the log, "ggml_cuda_host_malloc: failed to allocate 78.02 MiB of pinned memory: out of memory" or "gpu VRAM usage didn't recover within timeout". Please see the log below for more information.

Relevant log output

2025/03/04 00:38:50 routes.go:1187: INFO server config env="map[CUDA_VISIBLE_DEVICES:0,1,2,3,4,6,7,8 GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_DEBUG:false OLLAMA_FLASH_ATTENTION:false OLLAMA_GPU_OVERHEAD:0 OLLAMA_HOST:http://127.0.0.1:11434 OLLAMA_INTEL_GPU:false OLLAMA_KEEP_ALIVE:5m0s OLLAMA_KV_CACHE_TYPE: OLLAMA_LLM_LIBRARY: OLLAMA_LOAD_TIMEOUT:5m0s OLLAMA_MAX_LOADED_MODELS:0 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:H:\\.ollama OLLAMA_MULTIUSER_CACHE:false OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NUM_PARALLEL:0 OLLAMA_ORIGINS:[http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://*] OLLAMA_SCHED_SPREAD:false ROCR_VISIBLE_DEVICES:]"
time=2025-03-04T00:38:50.599+08:00 level=WARN source=routes.go:1213 msg="corrupt manifests detected, skipping prune operation.  Re-pull or delete to clear" error="registry.ollama.ai\\library\\deepseek-r1\\._671b %!w(<nil>)"
time=2025-03-04T00:38:50.605+08:00 level=INFO source=routes.go:1238 msg="Listening on 127.0.0.1:11434 (version 0.5.7)"
time=2025-03-04T00:38:50.609+08:00 level=INFO source=routes.go:1267 msg="Dynamic LLM libraries" runners="[cuda_v11_avx cuda_v12_avx rocm_avx cpu cpu_avx cpu_avx2]"
time=2025-03-04T00:38:50.611+08:00 level=INFO source=gpu.go:226 msg="looking for compatible GPUs"
time=2025-03-04T00:38:50.613+08:00 level=INFO source=gpu_windows.go:167 msg=packages count=2
time=2025-03-04T00:38:50.613+08:00 level=INFO source=gpu_windows.go:214 msg="" package=0 cores=64 efficiency=0 threads=128
time=2025-03-04T00:38:50.613+08:00 level=INFO source=gpu_windows.go:214 msg="" package=1 cores=64 efficiency=0 threads=128
time=2025-03-04T00:38:52.840+08:00 level=INFO source=gpu.go:334 msg="detected OS VRAM overhead" id=GPU-f06ae4ec-66f7-b413-f846-2df4a11474d5 library=cuda compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" overhead="414.8 MiB"
time=2025-03-04T00:38:53.111+08:00 level=INFO source=gpu.go:334 msg="detected OS VRAM overhead" id=GPU-6924ac3c-9d98-46e5-a3d6-4f13f9a0ba95 library=cuda compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" overhead="414.8 MiB"
time=2025-03-04T00:38:53.344+08:00 level=INFO source=gpu.go:334 msg="detected OS VRAM overhead" id=GPU-1fa07564-b59b-d0c9-f058-089927fe8650 library=cuda compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" overhead="414.8 MiB"
time=2025-03-04T00:38:53.626+08:00 level=INFO source=gpu.go:334 msg="detected OS VRAM overhead" id=GPU-6e3260fe-fc8d-3c36-00db-29b0c709487d library=cuda compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" overhead="414.8 MiB"
time=2025-03-04T00:38:53.853+08:00 level=INFO source=gpu.go:334 msg="detected OS VRAM overhead" id=GPU-094339e1-81ec-abf5-602e-7065b6a76f41 library=cuda compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" overhead="414.8 MiB"
time=2025-03-04T00:38:54.129+08:00 level=INFO source=gpu.go:334 msg="detected OS VRAM overhead" id=GPU-8c10edd1-4d9d-10f7-45f0-3748e26135f2 library=cuda compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" overhead="414.8 MiB"
time=2025-03-04T00:38:54.369+08:00 level=INFO source=gpu.go:334 msg="detected OS VRAM overhead" id=GPU-771bc981-05e2-09c4-2f7a-357104b45085 library=cuda compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" overhead="414.8 MiB"
time=2025-03-04T00:38:54.616+08:00 level=INFO source=gpu.go:334 msg="detected OS VRAM overhead" id=GPU-32e7fb34-a3d1-295b-f5e8-9e2f2f939262 library=cuda compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" overhead="414.8 MiB"
time=2025-03-04T00:38:54.625+08:00 level=INFO source=types.go:131 msg="inference compute" id=GPU-f06ae4ec-66f7-b413-f846-2df4a11474d5 library=cuda variant=v11 compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" total="79.6 GiB" available="79.1 GiB"
time=2025-03-04T00:38:54.625+08:00 level=INFO source=types.go:131 msg="inference compute" id=GPU-6924ac3c-9d98-46e5-a3d6-4f13f9a0ba95 library=cuda variant=v11 compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" total="79.6 GiB" available="79.1 GiB"
time=2025-03-04T00:38:54.625+08:00 level=INFO source=types.go:131 msg="inference compute" id=GPU-1fa07564-b59b-d0c9-f058-089927fe8650 library=cuda variant=v11 compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" total="79.6 GiB" available="79.1 GiB"
time=2025-03-04T00:38:54.625+08:00 level=INFO source=types.go:131 msg="inference compute" id=GPU-6e3260fe-fc8d-3c36-00db-29b0c709487d library=cuda variant=v11 compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" total="79.6 GiB" available="79.1 GiB"
time=2025-03-04T00:38:54.625+08:00 level=INFO source=types.go:131 msg="inference compute" id=GPU-094339e1-81ec-abf5-602e-7065b6a76f41 library=cuda variant=v11 compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" total="79.6 GiB" available="79.1 GiB"
time=2025-03-04T00:38:54.625+08:00 level=INFO source=types.go:131 msg="inference compute" id=GPU-8c10edd1-4d9d-10f7-45f0-3748e26135f2 library=cuda variant=v11 compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" total="79.6 GiB" available="79.1 GiB"
time=2025-03-04T00:38:54.625+08:00 level=INFO source=types.go:131 msg="inference compute" id=GPU-771bc981-05e2-09c4-2f7a-357104b45085 library=cuda variant=v11 compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" total="79.6 GiB" available="79.1 GiB"
time=2025-03-04T00:38:54.628+08:00 level=INFO source=types.go:131 msg="inference compute" id=GPU-32e7fb34-a3d1-295b-f5e8-9e2f2f939262 library=cuda variant=v11 compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" total="79.6 GiB" available="79.1 GiB"
time=2025-03-04T00:41:40.698+08:00 level=WARN source=manifest.go:160 msg="bad manifest name" path=registry.ollama.ai\library\deepseek-r1\._671b
time=2025-03-04T00:41:40.698+08:00 level=WARN source=manifest.go:160 msg="bad manifest name" path=registry.ollama.ai\library\deepseek-r1\._8b
[GIN] 2025/03/04 - 00:41:40 | 200 |      8.0286ms |       127.0.0.1 | GET      "/api/tags"
time=2025-03-04T00:41:42.646+08:00 level=WARN source=manifest.go:160 msg="bad manifest name" path=registry.ollama.ai\library\deepseek-r1\._671b
time=2025-03-04T00:41:42.646+08:00 level=WARN source=manifest.go:160 msg="bad manifest name" path=registry.ollama.ai\library\deepseek-r1\._8b
[GIN] 2025/03/04 - 00:41:42 | 200 |      2.0161ms |       127.0.0.1 | GET      "/api/tags"
time=2025-03-04T00:42:17.287+08:00 level=WARN source=manifest.go:160 msg="bad manifest name" path=registry.ollama.ai\library\deepseek-r1\._671b
time=2025-03-04T00:42:17.287+08:00 level=WARN source=manifest.go:160 msg="bad manifest name" path=registry.ollama.ai\library\deepseek-r1\._8b
time=2025-03-04T00:42:17.701+08:00 level=INFO source=sched.go:730 msg="new model will fit in available VRAM, loading" model=H:\.ollama\blobs\sha256-9801e7fce27dbf3d0bfb468b7b21f1d132131a546dfc43e50518631b8b1800a9 library=cuda parallel=4 required="449.4 GiB"
time=2025-03-04T00:42:17.860+08:00 level=INFO source=server.go:104 msg="system memory" total="2047.6 GiB" free="2021.8 GiB" free_swap="2089.5 GiB"
time=2025-03-04T00:42:17.864+08:00 level=INFO source=memory.go:356 msg="offload to cuda" layers.requested=-1 layers.model=62 layers.offload=62 layers.split=8,8,8,8,8,8,7,7 memory.available="[79.1 GiB 79.1 GiB 79.1 GiB 79.1 GiB 79.1 GiB 79.1 GiB 79.1 GiB 79.1 GiB]" memory.gpu_overhead="0 B" memory.required.full="449.4 GiB" memory.required.partial="449.4 GiB" memory.required.kv="38.1 GiB" memory.required.allocations="[54.7 GiB 54.7 GiB 54.7 GiB 61.3 GiB 61.3 GiB 55.3 GiB 53.7 GiB 53.7 GiB]" memory.weights.total="413.6 GiB" memory.weights.repeating="412.9 GiB" memory.weights.nonrepeating="725.0 MiB" memory.graph.full="3.0 GiB" memory.graph.partial="3.0 GiB"
time=2025-03-04T00:42:17.883+08:00 level=INFO source=server.go:376 msg="starting llama server" cmd="C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\lib\\ollama\\runners\\cuda_v11_avx\\ollama_llama_server.exe runner --model H:\\.ollama\\blobs\\sha256-9801e7fce27dbf3d0bfb468b7b21f1d132131a546dfc43e50518631b8b1800a9 --ctx-size 8192 --batch-size 512 --n-gpu-layers 62 --threads 128 --no-mmap --parallel 4 --tensor-split 8,8,8,8,8,8,7,7 --port 49673"
time=2025-03-04T00:42:17.892+08:00 level=INFO source=sched.go:449 msg="loaded runners" count=1
time=2025-03-04T00:42:17.892+08:00 level=INFO source=server.go:555 msg="waiting for llama runner to start responding"
time=2025-03-04T00:42:18.092+08:00 level=INFO source=server.go:589 msg="waiting for server to become available" status="llm server not responding"
time=2025-03-04T00:42:18.162+08:00 level=INFO source=runner.go:936 msg="starting go runner"
ggml_cuda_init: GGML_CUDA_FORCE_MMQ:    no
ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no
ggml_cuda_init: found 8 CUDA devices:
  Device 0: NVIDIA A800 80GB PCIe, compute capability 8.0, VMM: no
  Device 1: NVIDIA A800 80GB PCIe, compute capability 8.0, VMM: no
  Device 2: NVIDIA A800 80GB PCIe, compute capability 8.0, VMM: no
  Device 3: NVIDIA A800 80GB PCIe, compute capability 8.0, VMM: no
  Device 4: NVIDIA A800 80GB PCIe, compute capability 8.0, VMM: no
  Device 5: NVIDIA A800 80GB PCIe, compute capability 8.0, VMM: no
  Device 6: NVIDIA A800 80GB PCIe, compute capability 8.0, VMM: no
  Device 7: NVIDIA A800 80GB PCIe, compute capability 8.0, VMM: no
time=2025-03-04T00:42:18.618+08:00 level=INFO source=runner.go:937 msg=system info="CUDA : USE_GRAPHS = 1 | PEER_MAX_BATCH_SIZE = 128 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | LLAMAFILE = 1 | AARCH64_REPACK = 1 | cgo(clang)" threads=128
time=2025-03-04T00:42:18.623+08:00 level=INFO source=.:0 msg="Server listening on 127.0.0.1:49673"
time=2025-03-04T00:42:18.805+08:00 level=INFO source=server.go:589 msg="waiting for server to become available" status="llm server loading model"
llama_load_model_from_file: using device CUDA0 (NVIDIA A800 80GB PCIe) - 80928 MiB free
llama_load_model_from_file: using device CUDA1 (NVIDIA A800 80GB PCIe) - 80928 MiB free
llama_load_model_from_file: using device CUDA2 (NVIDIA A800 80GB PCIe) - 80928 MiB free
llama_load_model_from_file: using device CUDA3 (NVIDIA A800 80GB PCIe) - 80928 MiB free
llama_load_model_from_file: using device CUDA4 (NVIDIA A800 80GB PCIe) - 80928 MiB free
llama_load_model_from_file: using device CUDA5 (NVIDIA A800 80GB PCIe) - 80928 MiB free
llama_load_model_from_file: using device CUDA6 (NVIDIA A800 80GB PCIe) - 80928 MiB free
llama_load_model_from_file: using device CUDA7 (NVIDIA A800 80GB PCIe) - 80928 MiB free
llama_model_loader: loaded meta data with 42 key-value pairs and 1025 tensors from H:\.ollama\blobs\sha256-9801e7fce27dbf3d0bfb468b7b21f1d132131a546dfc43e50518631b8b1800a9 (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              = deepseek2
llama_model_loader: - kv   1:                               general.type str              = model
llama_model_loader: - kv   2:                         general.size_label str              = 256x20B
llama_model_loader: - kv   3:                      deepseek2.block_count u32              = 61
llama_model_loader: - kv   4:                   deepseek2.context_length u32              = 163840
llama_model_loader: - kv   5:                 deepseek2.embedding_length u32              = 7168
llama_model_loader: - kv   6:              deepseek2.feed_forward_length u32              = 18432
llama_model_loader: - kv   7:             deepseek2.attention.head_count u32              = 128
llama_model_loader: - kv   8:          deepseek2.attention.head_count_kv u32              = 128
llama_model_loader: - kv   9:                   deepseek2.rope.freq_base f32              = 10000.000000
llama_model_loader: - kv  10: deepseek2.attention.layer_norm_rms_epsilon f32              = 0.000001
llama_model_loader: - kv  11:                deepseek2.expert_used_count u32              = 8
llama_model_loader: - kv  12:        deepseek2.leading_dense_block_count u32              = 3
llama_model_loader: - kv  13:                       deepseek2.vocab_size u32              = 129280
llama_model_loader: - kv  14:            deepseek2.attention.q_lora_rank u32              = 1536
llama_model_loader: - kv  15:           deepseek2.attention.kv_lora_rank u32              = 512
llama_model_loader: - kv  16:             deepseek2.attention.key_length u32              = 192
llama_model_loader: - kv  17:           deepseek2.attention.value_length u32              = 128
llama_model_loader: - kv  18:       deepseek2.expert_feed_forward_length u32              = 2048
llama_model_loader: - kv  19:                     deepseek2.expert_count u32              = 256
llama_model_loader: - kv  20:              deepseek2.expert_shared_count u32              = 1
llama_model_loader: - kv  21:             deepseek2.expert_weights_scale f32              = 2.500000
llama_model_loader: - kv  22:              deepseek2.expert_weights_norm bool             = true
llama_model_loader: - kv  23:               deepseek2.expert_gating_func u32              = 2
llama_model_loader: - kv  24:             deepseek2.rope.dimension_count u32              = 64
llama_model_loader: - kv  25:                deepseek2.rope.scaling.type str              = yarn
llama_model_loader: - kv  26:              deepseek2.rope.scaling.factor f32              = 40.000000
llama_model_loader: - kv  27: deepseek2.rope.scaling.original_context_length u32              = 4096
llama_model_loader: - kv  28: deepseek2.rope.scaling.yarn_log_multiplier f32              = 0.100000
llama_model_loader: - kv  29:                       tokenizer.ggml.model str              = gpt2
llama_model_loader: - kv  30:                         tokenizer.ggml.pre str              = deepseek-v3
llama_model_loader: - kv  31:                      tokenizer.ggml.tokens arr[str,129280]  = ["<｜begin▁of▁sentence｜>", "<�...
llama_model_loader: - kv  32:                  tokenizer.ggml.token_type arr[i32,129280]  = [3, 3, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv  33:                      tokenizer.ggml.merges arr[str,127741]  = ["Ġ t", "Ġ a", "i n", "Ġ Ġ", "h e...
llama_model_loader: - kv  34:                tokenizer.ggml.bos_token_id u32              = 0
llama_model_loader: - kv  35:                tokenizer.ggml.eos_token_id u32              = 1
llama_model_loader: - kv  36:            tokenizer.ggml.padding_token_id u32              = 1
llama_model_loader: - kv  37:               tokenizer.ggml.add_bos_token bool             = true
llama_model_loader: - kv  38:               tokenizer.ggml.add_eos_token bool             = false
llama_model_loader: - kv  39:                    tokenizer.chat_template str              = {% if not add_generation_prompt is de...
llama_model_loader: - kv  40:               general.quantization_version u32              = 2
llama_model_loader: - kv  41:                          general.file_type u32              = 15
llama_model_loader: - type  f32:  361 tensors
llama_model_loader: - type q4_K:  606 tensors
llama_model_loader: - type q6_K:   58 tensors
llm_load_vocab: special_eos_id is not in special_eog_ids - the tokenizer config may be incorrect
llm_load_vocab: special tokens cache size = 818
llm_load_vocab: token to piece cache size = 0.8223 MB
llm_load_print_meta: format           = GGUF V3 (latest)
llm_load_print_meta: arch             = deepseek2
llm_load_print_meta: vocab type       = BPE
llm_load_print_meta: n_vocab          = 129280
llm_load_print_meta: n_merges         = 127741
llm_load_print_meta: vocab_only       = 0
llm_load_print_meta: n_ctx_train      = 163840
llm_load_print_meta: n_embd           = 7168
llm_load_print_meta: n_layer          = 61
llm_load_print_meta: n_head           = 128
llm_load_print_meta: n_head_kv        = 128
llm_load_print_meta: n_rot            = 64
llm_load_print_meta: n_swa            = 0
llm_load_print_meta: n_embd_head_k    = 192
llm_load_print_meta: n_embd_head_v    = 128
llm_load_print_meta: n_gqa            = 1
llm_load_print_meta: n_embd_k_gqa     = 24576
llm_load_print_meta: n_embd_v_gqa     = 16384
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             = 18432
llm_load_print_meta: n_expert         = 256
llm_load_print_meta: n_expert_used    = 8
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     = yarn
llm_load_print_meta: freq_base_train  = 10000.0
llm_load_print_meta: freq_scale_train = 0.025
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: ssm_dt_b_c_rms   = 0
llm_load_print_meta: model type       = 671B
llm_load_print_meta: model ftype      = Q4_K - Medium
llm_load_print_meta: model params     = 671.03 B
llm_load_print_meta: model size       = 376.65 GiB (4.82 BPW) 
llm_load_print_meta: general.name     = n/a
llm_load_print_meta: BOS token        = 0 '<｜begin▁of▁sentence｜>'
llm_load_print_meta: EOS token        = 1 '<｜end▁of▁sentence｜>'
llm_load_print_meta: EOT token        = 1 '<｜end▁of▁sentence｜>'
llm_load_print_meta: PAD token        = 1 '<｜end▁of▁sentence｜>'
llm_load_print_meta: LF token         = 131 'Ä'
llm_load_print_meta: FIM PRE token    = 128801 '<｜fim▁begin｜>'
llm_load_print_meta: FIM SUF token    = 128800 '<｜fim▁hole｜>'
llm_load_print_meta: FIM MID token    = 128802 '<｜fim▁end｜>'
llm_load_print_meta: EOG token        = 1 '<｜end▁of▁sentence｜>'
llm_load_print_meta: max token length = 256
llm_load_print_meta: n_layer_dense_lead   = 3
llm_load_print_meta: n_lora_q             = 1536
llm_load_print_meta: n_lora_kv            = 512
llm_load_print_meta: n_ff_exp             = 2048
llm_load_print_meta: n_expert_shared      = 1
llm_load_print_meta: expert_weights_scale = 2.5
llm_load_print_meta: expert_weights_norm  = 1
llm_load_print_meta: expert_gating_func   = sigmoid
llm_load_print_meta: rope_yarn_log_mul    = 0.1000
llm_load_tensors: offloading 61 repeating layers to GPU
llm_load_tensors: offloading output layer to GPU
llm_load_tensors: offloaded 62/62 layers to GPU
llm_load_tensors:          CPU model buffer size =   497.11 MiB
llm_load_tensors:        CUDA0 model buffer size = 35642.36 MiB
llm_load_tensors:        CUDA1 model buffer size = 52215.30 MiB
llm_load_tensors:        CUDA2 model buffer size = 51287.70 MiB
llm_load_tensors:        CUDA3 model buffer size = 52215.30 MiB
llm_load_tensors:        CUDA4 model buffer size = 52215.30 MiB
llm_load_tensors:        CUDA5 model buffer size = 51287.70 MiB
llm_load_tensors:        CUDA6 model buffer size = 46963.85 MiB
llm_load_tensors:        CUDA7 model buffer size = 43364.99 MiB
llama_new_context_with_model: n_seq_max     = 4
llama_new_context_with_model: n_ctx         = 8192
llama_new_context_with_model: n_ctx_per_seq = 2048
llama_new_context_with_model: n_batch       = 2048
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    = 0.025
llama_new_context_with_model: n_ctx_per_seq (2048) < n_ctx_train (163840) -- the full capacity of the model will not be utilized
llama_kv_cache_init: kv_size = 8192, offload = 1, type_k = 'f16', type_v = 'f16', n_layer = 61, can_shift = 0
llama_kv_cache_init:      CUDA0 KV buffer size =  5120.00 MiB
llama_kv_cache_init:      CUDA1 KV buffer size =  5120.00 MiB
llama_kv_cache_init:      CUDA2 KV buffer size =  5120.00 MiB
llama_kv_cache_init:      CUDA3 KV buffer size =  5120.00 MiB
llama_kv_cache_init:      CUDA4 KV buffer size =  5120.00 MiB
llama_kv_cache_init:      CUDA5 KV buffer size =  5120.00 MiB
llama_kv_cache_init:      CUDA6 KV buffer size =  4480.00 MiB
llama_kv_cache_init:      CUDA7 KV buffer size =  3840.00 MiB
llama_new_context_with_model: KV self size  = 39040.00 MiB, K (f16): 23424.00 MiB, V (f16): 15616.00 MiB
llama_new_context_with_model:  CUDA_Host  output buffer size =     2.08 MiB
llama_new_context_with_model: pipeline parallelism enabled (n_copies=4)
ggml_cuda_host_malloc: failed to allocate 78.02 MiB of pinned memory: out of memory
llama_new_context_with_model:      CUDA0 compute buffer size =  2322.01 MiB
llama_new_context_with_model:      CUDA1 compute buffer size =  2322.01 MiB
llama_new_context_with_model:      CUDA2 compute buffer size =  2322.01 MiB
llama_new_context_with_model:      CUDA3 compute buffer size =  2322.01 MiB
llama_new_context_with_model:      CUDA4 compute buffer size =  2322.01 MiB
llama_new_context_with_model:      CUDA5 compute buffer size =  2322.01 MiB
llama_new_context_with_model:      CUDA6 compute buffer size =  2322.01 MiB
llama_new_context_with_model:      CUDA7 compute buffer size =  2322.02 MiB
llama_new_context_with_model:  CUDA_Host compute buffer size =    78.02 MiB
llama_new_context_with_model: graph nodes  = 5025
llama_new_context_with_model: graph splits = 9
time=2025-03-04T00:54:16.177+08:00 level=INFO source=server.go:594 msg="llama runner started in 718.29 seconds"
llama_model_loader: loaded meta data with 42 key-value pairs and 1025 tensors from H:\.ollama\blobs\sha256-9801e7fce27dbf3d0bfb468b7b21f1d132131a546dfc43e50518631b8b1800a9 (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              = deepseek2
llama_model_loader: - kv   1:                               general.type str              = model
llama_model_loader: - kv   2:                         general.size_label str              = 256x20B
llama_model_loader: - kv   3:                      deepseek2.block_count u32              = 61
llama_model_loader: - kv   4:                   deepseek2.context_length u32              = 163840
llama_model_loader: - kv   5:                 deepseek2.embedding_length u32              = 7168
llama_model_loader: - kv   6:              deepseek2.feed_forward_length u32              = 18432
llama_model_loader: - kv   7:             deepseek2.attention.head_count u32              = 128
llama_model_loader: - kv   8:          deepseek2.attention.head_count_kv u32              = 128
llama_model_loader: - kv   9:                   deepseek2.rope.freq_base f32              = 10000.000000
llama_model_loader: - kv  10: deepseek2.attention.layer_norm_rms_epsilon f32              = 0.000001
llama_model_loader: - kv  11:                deepseek2.expert_used_count u32              = 8
llama_model_loader: - kv  12:        deepseek2.leading_dense_block_count u32              = 3
llama_model_loader: - kv  13:                       deepseek2.vocab_size u32              = 129280
llama_model_loader: - kv  14:            deepseek2.attention.q_lora_rank u32              = 1536
llama_model_loader: - kv  15:           deepseek2.attention.kv_lora_rank u32              = 512
llama_model_loader: - kv  16:             deepseek2.attention.key_length u32              = 192
llama_model_loader: - kv  17:           deepseek2.attention.value_length u32              = 128
llama_model_loader: - kv  18:       deepseek2.expert_feed_forward_length u32              = 2048
llama_model_loader: - kv  19:                     deepseek2.expert_count u32              = 256
llama_model_loader: - kv  20:              deepseek2.expert_shared_count u32              = 1
llama_model_loader: - kv  21:             deepseek2.expert_weights_scale f32              = 2.500000
llama_model_loader: - kv  22:              deepseek2.expert_weights_norm bool             = true
llama_model_loader: - kv  23:               deepseek2.expert_gating_func u32              = 2
llama_model_loader: - kv  24:             deepseek2.rope.dimension_count u32              = 64
llama_model_loader: - kv  25:                deepseek2.rope.scaling.type str              = yarn
llama_model_loader: - kv  26:              deepseek2.rope.scaling.factor f32              = 40.000000
llama_model_loader: - kv  27: deepseek2.rope.scaling.original_context_length u32              = 4096
llama_model_loader: - kv  28: deepseek2.rope.scaling.yarn_log_multiplier f32              = 0.100000
llama_model_loader: - kv  29:                       tokenizer.ggml.model str              = gpt2
llama_model_loader: - kv  30:                         tokenizer.ggml.pre str              = deepseek-v3
llama_model_loader: - kv  31:                      tokenizer.ggml.tokens arr[str,129280]  = ["<｜begin▁of▁sentence｜>", "<�...
llama_model_loader: - kv  32:                  tokenizer.ggml.token_type arr[i32,129280]  = [3, 3, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv  33:                      tokenizer.ggml.merges arr[str,127741]  = ["Ġ t", "Ġ a", "i n", "Ġ Ġ", "h e...
llama_model_loader: - kv  34:                tokenizer.ggml.bos_token_id u32              = 0
llama_model_loader: - kv  35:                tokenizer.ggml.eos_token_id u32              = 1
llama_model_loader: - kv  36:            tokenizer.ggml.padding_token_id u32              = 1
llama_model_loader: - kv  37:               tokenizer.ggml.add_bos_token bool             = true
llama_model_loader: - kv  38:               tokenizer.ggml.add_eos_token bool             = false
llama_model_loader: - kv  39:                    tokenizer.chat_template str              = {% if not add_generation_prompt is de...
llama_model_loader: - kv  40:               general.quantization_version u32              = 2
llama_model_loader: - kv  41:                          general.file_type u32              = 15
llama_model_loader: - type  f32:  361 tensors
llama_model_loader: - type q4_K:  606 tensors
llama_model_loader: - type q6_K:   58 tensors
llm_load_vocab: special_eos_id is not in special_eog_ids - the tokenizer config may be incorrect
llm_load_vocab: special tokens cache size = 818
llm_load_vocab: token to piece cache size = 0.8223 MB
llm_load_print_meta: format           = GGUF V3 (latest)
llm_load_print_meta: arch             = deepseek2
llm_load_print_meta: vocab type       = BPE
llm_load_print_meta: n_vocab          = 129280
llm_load_print_meta: n_merges         = 127741
llm_load_print_meta: vocab_only       = 1
llm_load_print_meta: model type       = ?B
llm_load_print_meta: model ftype      = all F32
llm_load_print_meta: model params     = 671.03 B
llm_load_print_meta: model size       = 376.65 GiB (4.82 BPW) 
llm_load_print_meta: general.name     = n/a
llm_load_print_meta: BOS token        = 0 '<｜begin▁of▁sentence｜>'
llm_load_print_meta: EOS token        = 1 '<｜end▁of▁sentence｜>'
llm_load_print_meta: EOT token        = 1 '<｜end▁of▁sentence｜>'
llm_load_print_meta: PAD token        = 1 '<｜end▁of▁sentence｜>'
llm_load_print_meta: LF token         = 131 'Ä'
llm_load_print_meta: FIM PRE token    = 128801 '<｜fim▁begin｜>'
llm_load_print_meta: FIM SUF token    = 128800 '<｜fim▁hole｜>'
llm_load_print_meta: FIM MID token    = 128802 '<｜fim▁end｜>'
llm_load_print_meta: EOG token        = 1 '<｜end▁of▁sentence｜>'
llm_load_print_meta: max token length = 256
llm_load_print_meta: n_layer_dense_lead   = 0
llm_load_print_meta: n_lora_q             = 0
llm_load_print_meta: n_lora_kv            = 0
llm_load_print_meta: n_ff_exp             = 0
llm_load_print_meta: n_expert_shared      = 0
llm_load_print_meta: expert_weights_scale = 0.0
llm_load_print_meta: expert_weights_norm  = 0
llm_load_print_meta: expert_gating_func   = unknown
llm_load_print_meta: rope_yarn_log_mul    = 0.0000
llama_model_load: vocab only - skipping tensors
[GIN] 2025/03/04 - 00:56:45 | 200 |        14m28s |       127.0.0.1 | POST     "/api/chat"
time=2025-03-04T01:01:50.885+08:00 level=WARN source=sched.go:646 msg="gpu VRAM usage didn't recover within timeout" seconds=5.2169324 model=H:\.ollama\blobs\sha256-9801e7fce27dbf3d0bfb468b7b21f1d132131a546dfc43e50518631b8b1800a9
time=2025-03-04T01:01:51.135+08:00 level=WARN source=sched.go:646 msg="gpu VRAM usage didn't recover within timeout" seconds=5.4673458 model=H:\.ollama\blobs\sha256-9801e7fce27dbf3d0bfb468b7b21f1d132131a546dfc43e50518631b8b1800a9
time=2025-03-04T01:01:51.385+08:00 level=WARN source=sched.go:646 msg="gpu VRAM usage didn't recover within timeout" seconds=5.7173227 model=H:\.ollama\blobs\sha256-9801e7fce27dbf3d0bfb468b7b21f1d132131a546dfc43e50518631b8b1800a9

OS

Windows

GPU

Nvidia

CPU

Intel

Ollama version

0.5.11

Originally created by @flora298 on GitHub (Mar 5, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/9521 ### What is the issue? I was running deepseek-r1:671b model, and the process was forcibly terminated. In the terminal, it showed `POST predict: Post "http://127.0.0.1:49813/completion": read top 127.0.0.1:49816->127.0.0.1:49813: wsarec: An existing connection was forcibly closed by the remote host.`after running half of the reasoning process. I made one of the GPU invisible since the previous log said CUDA error: CUBLAS_STATUS_NOT_INITIALIZED current device: 5, and it went with the same error line. `POST predict: Post "http://127.0.0.1:49813/completion": read top 127.0.0.1:49816->127.0.0.1:49813: wsarec: An existing connection was forcibly closed by the remote host.` I am not sure making device 5 invisible was the right thing to do, but the above log no longer has CUDA error. I think the problem might be on these two lines in the log, "ggml_cuda_host_malloc: failed to allocate 78.02 MiB of pinned memory: out of memory" or "gpu VRAM usage didn't recover within timeout". Please see the log below for more information. ### Relevant log output ```shell 2025/03/04 00:38:50 routes.go:1187: INFO server config env="map[CUDA_VISIBLE_DEVICES:0,1,2,3,4,6,7,8 GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_DEBUG:false OLLAMA_FLASH_ATTENTION:false OLLAMA_GPU_OVERHEAD:0 OLLAMA_HOST:http://127.0.0.1:11434 OLLAMA_INTEL_GPU:false OLLAMA_KEEP_ALIVE:5m0s OLLAMA_KV_CACHE_TYPE: OLLAMA_LLM_LIBRARY: OLLAMA_LOAD_TIMEOUT:5m0s OLLAMA_MAX_LOADED_MODELS:0 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:H:\\.ollama OLLAMA_MULTIUSER_CACHE:false OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NUM_PARALLEL:0 OLLAMA_ORIGINS:[http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://*] OLLAMA_SCHED_SPREAD:false ROCR_VISIBLE_DEVICES:]" time=2025-03-04T00:38:50.599+08:00 level=WARN source=routes.go:1213 msg="corrupt manifests detected, skipping prune operation. Re-pull or delete to clear" error="registry.ollama.ai\\library\\deepseek-r1\\._671b %!w(<nil>)" time=2025-03-04T00:38:50.605+08:00 level=INFO source=routes.go:1238 msg="Listening on 127.0.0.1:11434 (version 0.5.7)" time=2025-03-04T00:38:50.609+08:00 level=INFO source=routes.go:1267 msg="Dynamic LLM libraries" runners="[cuda_v11_avx cuda_v12_avx rocm_avx cpu cpu_avx cpu_avx2]" time=2025-03-04T00:38:50.611+08:00 level=INFO source=gpu.go:226 msg="looking for compatible GPUs" time=2025-03-04T00:38:50.613+08:00 level=INFO source=gpu_windows.go:167 msg=packages count=2 time=2025-03-04T00:38:50.613+08:00 level=INFO source=gpu_windows.go:214 msg="" package=0 cores=64 efficiency=0 threads=128 time=2025-03-04T00:38:50.613+08:00 level=INFO source=gpu_windows.go:214 msg="" package=1 cores=64 efficiency=0 threads=128 time=2025-03-04T00:38:52.840+08:00 level=INFO source=gpu.go:334 msg="detected OS VRAM overhead" id=GPU-f06ae4ec-66f7-b413-f846-2df4a11474d5 library=cuda compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" overhead="414.8 MiB" time=2025-03-04T00:38:53.111+08:00 level=INFO source=gpu.go:334 msg="detected OS VRAM overhead" id=GPU-6924ac3c-9d98-46e5-a3d6-4f13f9a0ba95 library=cuda compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" overhead="414.8 MiB" time=2025-03-04T00:38:53.344+08:00 level=INFO source=gpu.go:334 msg="detected OS VRAM overhead" id=GPU-1fa07564-b59b-d0c9-f058-089927fe8650 library=cuda compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" overhead="414.8 MiB" time=2025-03-04T00:38:53.626+08:00 level=INFO source=gpu.go:334 msg="detected OS VRAM overhead" id=GPU-6e3260fe-fc8d-3c36-00db-29b0c709487d library=cuda compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" overhead="414.8 MiB" time=2025-03-04T00:38:53.853+08:00 level=INFO source=gpu.go:334 msg="detected OS VRAM overhead" id=GPU-094339e1-81ec-abf5-602e-7065b6a76f41 library=cuda compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" overhead="414.8 MiB" time=2025-03-04T00:38:54.129+08:00 level=INFO source=gpu.go:334 msg="detected OS VRAM overhead" id=GPU-8c10edd1-4d9d-10f7-45f0-3748e26135f2 library=cuda compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" overhead="414.8 MiB" time=2025-03-04T00:38:54.369+08:00 level=INFO source=gpu.go:334 msg="detected OS VRAM overhead" id=GPU-771bc981-05e2-09c4-2f7a-357104b45085 library=cuda compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" overhead="414.8 MiB" time=2025-03-04T00:38:54.616+08:00 level=INFO source=gpu.go:334 msg="detected OS VRAM overhead" id=GPU-32e7fb34-a3d1-295b-f5e8-9e2f2f939262 library=cuda compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" overhead="414.8 MiB" time=2025-03-04T00:38:54.625+08:00 level=INFO source=types.go:131 msg="inference compute" id=GPU-f06ae4ec-66f7-b413-f846-2df4a11474d5 library=cuda variant=v11 compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" total="79.6 GiB" available="79.1 GiB" time=2025-03-04T00:38:54.625+08:00 level=INFO source=types.go:131 msg="inference compute" id=GPU-6924ac3c-9d98-46e5-a3d6-4f13f9a0ba95 library=cuda variant=v11 compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" total="79.6 GiB" available="79.1 GiB" time=2025-03-04T00:38:54.625+08:00 level=INFO source=types.go:131 msg="inference compute" id=GPU-1fa07564-b59b-d0c9-f058-089927fe8650 library=cuda variant=v11 compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" total="79.6 GiB" available="79.1 GiB" time=2025-03-04T00:38:54.625+08:00 level=INFO source=types.go:131 msg="inference compute" id=GPU-6e3260fe-fc8d-3c36-00db-29b0c709487d library=cuda variant=v11 compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" total="79.6 GiB" available="79.1 GiB" time=2025-03-04T00:38:54.625+08:00 level=INFO source=types.go:131 msg="inference compute" id=GPU-094339e1-81ec-abf5-602e-7065b6a76f41 library=cuda variant=v11 compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" total="79.6 GiB" available="79.1 GiB" time=2025-03-04T00:38:54.625+08:00 level=INFO source=types.go:131 msg="inference compute" id=GPU-8c10edd1-4d9d-10f7-45f0-3748e26135f2 library=cuda variant=v11 compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" total="79.6 GiB" available="79.1 GiB" time=2025-03-04T00:38:54.625+08:00 level=INFO source=types.go:131 msg="inference compute" id=GPU-771bc981-05e2-09c4-2f7a-357104b45085 library=cuda variant=v11 compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" total="79.6 GiB" available="79.1 GiB" time=2025-03-04T00:38:54.628+08:00 level=INFO source=types.go:131 msg="inference compute" id=GPU-32e7fb34-a3d1-295b-f5e8-9e2f2f939262 library=cuda variant=v11 compute=8.0 driver=12.0 name="NVIDIA A800 80GB PCIe" total="79.6 GiB" available="79.1 GiB" time=2025-03-04T00:41:40.698+08:00 level=WARN source=manifest.go:160 msg="bad manifest name" path=registry.ollama.ai\library\deepseek-r1\._671b time=2025-03-04T00:41:40.698+08:00 level=WARN source=manifest.go:160 msg="bad manifest name" path=registry.ollama.ai\library\deepseek-r1\._8b [GIN] 2025/03/04 - 00:41:40 | 200 | 8.0286ms | 127.0.0.1 | GET "/api/tags" time=2025-03-04T00:41:42.646+08:00 level=WARN source=manifest.go:160 msg="bad manifest name" path=registry.ollama.ai\library\deepseek-r1\._671b time=2025-03-04T00:41:42.646+08:00 level=WARN source=manifest.go:160 msg="bad manifest name" path=registry.ollama.ai\library\deepseek-r1\._8b [GIN] 2025/03/04 - 00:41:42 | 200 | 2.0161ms | 127.0.0.1 | GET "/api/tags" time=2025-03-04T00:42:17.287+08:00 level=WARN source=manifest.go:160 msg="bad manifest name" path=registry.ollama.ai\library\deepseek-r1\._671b time=2025-03-04T00:42:17.287+08:00 level=WARN source=manifest.go:160 msg="bad manifest name" path=registry.ollama.ai\library\deepseek-r1\._8b time=2025-03-04T00:42:17.701+08:00 level=INFO source=sched.go:730 msg="new model will fit in available VRAM, loading" model=H:\.ollama\blobs\sha256-9801e7fce27dbf3d0bfb468b7b21f1d132131a546dfc43e50518631b8b1800a9 library=cuda parallel=4 required="449.4 GiB" time=2025-03-04T00:42:17.860+08:00 level=INFO source=server.go:104 msg="system memory" total="2047.6 GiB" free="2021.8 GiB" free_swap="2089.5 GiB" time=2025-03-04T00:42:17.864+08:00 level=INFO source=memory.go:356 msg="offload to cuda" layers.requested=-1 layers.model=62 layers.offload=62 layers.split=8,8,8,8,8,8,7,7 memory.available="[79.1 GiB 79.1 GiB 79.1 GiB 79.1 GiB 79.1 GiB 79.1 GiB 79.1 GiB 79.1 GiB]" memory.gpu_overhead="0 B" memory.required.full="449.4 GiB" memory.required.partial="449.4 GiB" memory.required.kv="38.1 GiB" memory.required.allocations="[54.7 GiB 54.7 GiB 54.7 GiB 61.3 GiB 61.3 GiB 55.3 GiB 53.7 GiB 53.7 GiB]" memory.weights.total="413.6 GiB" memory.weights.repeating="412.9 GiB" memory.weights.nonrepeating="725.0 MiB" memory.graph.full="3.0 GiB" memory.graph.partial="3.0 GiB" time=2025-03-04T00:42:17.883+08:00 level=INFO source=server.go:376 msg="starting llama server" cmd="C:\\Users\\Administrator\\AppData\\Local\\Programs\\Ollama\\lib\\ollama\\runners\\cuda_v11_avx\\ollama_llama_server.exe runner --model H:\\.ollama\\blobs\\sha256-9801e7fce27dbf3d0bfb468b7b21f1d132131a546dfc43e50518631b8b1800a9 --ctx-size 8192 --batch-size 512 --n-gpu-layers 62 --threads 128 --no-mmap --parallel 4 --tensor-split 8,8,8,8,8,8,7,7 --port 49673" time=2025-03-04T00:42:17.892+08:00 level=INFO source=sched.go:449 msg="loaded runners" count=1 time=2025-03-04T00:42:17.892+08:00 level=INFO source=server.go:555 msg="waiting for llama runner to start responding" time=2025-03-04T00:42:18.092+08:00 level=INFO source=server.go:589 msg="waiting for server to become available" status="llm server not responding" time=2025-03-04T00:42:18.162+08:00 level=INFO source=runner.go:936 msg="starting go runner" ggml_cuda_init: GGML_CUDA_FORCE_MMQ: no ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no ggml_cuda_init: found 8 CUDA devices: Device 0: NVIDIA A800 80GB PCIe, compute capability 8.0, VMM: no Device 1: NVIDIA A800 80GB PCIe, compute capability 8.0, VMM: no Device 2: NVIDIA A800 80GB PCIe, compute capability 8.0, VMM: no Device 3: NVIDIA A800 80GB PCIe, compute capability 8.0, VMM: no Device 4: NVIDIA A800 80GB PCIe, compute capability 8.0, VMM: no Device 5: NVIDIA A800 80GB PCIe, compute capability 8.0, VMM: no Device 6: NVIDIA A800 80GB PCIe, compute capability 8.0, VMM: no Device 7: NVIDIA A800 80GB PCIe, compute capability 8.0, VMM: no time=2025-03-04T00:42:18.618+08:00 level=INFO source=runner.go:937 msg=system info="CUDA : USE_GRAPHS = 1 | PEER_MAX_BATCH_SIZE = 128 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | LLAMAFILE = 1 | AARCH64_REPACK = 1 | cgo(clang)" threads=128 time=2025-03-04T00:42:18.623+08:00 level=INFO source=.:0 msg="Server listening on 127.0.0.1:49673" time=2025-03-04T00:42:18.805+08:00 level=INFO source=server.go:589 msg="waiting for server to become available" status="llm server loading model" llama_load_model_from_file: using device CUDA0 (NVIDIA A800 80GB PCIe) - 80928 MiB free llama_load_model_from_file: using device CUDA1 (NVIDIA A800 80GB PCIe) - 80928 MiB free llama_load_model_from_file: using device CUDA2 (NVIDIA A800 80GB PCIe) - 80928 MiB free llama_load_model_from_file: using device CUDA3 (NVIDIA A800 80GB PCIe) - 80928 MiB free llama_load_model_from_file: using device CUDA4 (NVIDIA A800 80GB PCIe) - 80928 MiB free llama_load_model_from_file: using device CUDA5 (NVIDIA A800 80GB PCIe) - 80928 MiB free llama_load_model_from_file: using device CUDA6 (NVIDIA A800 80GB PCIe) - 80928 MiB free llama_load_model_from_file: using device CUDA7 (NVIDIA A800 80GB PCIe) - 80928 MiB free llama_model_loader: loaded meta data with 42 key-value pairs and 1025 tensors from H:\.ollama\blobs\sha256-9801e7fce27dbf3d0bfb468b7b21f1d132131a546dfc43e50518631b8b1800a9 (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 = deepseek2 llama_model_loader: - kv 1: general.type str = model llama_model_loader: - kv 2: general.size_label str = 256x20B llama_model_loader: - kv 3: deepseek2.block_count u32 = 61 llama_model_loader: - kv 4: deepseek2.context_length u32 = 163840 llama_model_loader: - kv 5: deepseek2.embedding_length u32 = 7168 llama_model_loader: - kv 6: deepseek2.feed_forward_length u32 = 18432 llama_model_loader: - kv 7: deepseek2.attention.head_count u32 = 128 llama_model_loader: - kv 8: deepseek2.attention.head_count_kv u32 = 128 llama_model_loader: - kv 9: deepseek2.rope.freq_base f32 = 10000.000000 llama_model_loader: - kv 10: deepseek2.attention.layer_norm_rms_epsilon f32 = 0.000001 llama_model_loader: - kv 11: deepseek2.expert_used_count u32 = 8 llama_model_loader: - kv 12: deepseek2.leading_dense_block_count u32 = 3 llama_model_loader: - kv 13: deepseek2.vocab_size u32 = 129280 llama_model_loader: - kv 14: deepseek2.attention.q_lora_rank u32 = 1536 llama_model_loader: - kv 15: deepseek2.attention.kv_lora_rank u32 = 512 llama_model_loader: - kv 16: deepseek2.attention.key_length u32 = 192 llama_model_loader: - kv 17: deepseek2.attention.value_length u32 = 128 llama_model_loader: - kv 18: deepseek2.expert_feed_forward_length u32 = 2048 llama_model_loader: - kv 19: deepseek2.expert_count u32 = 256 llama_model_loader: - kv 20: deepseek2.expert_shared_count u32 = 1 llama_model_loader: - kv 21: deepseek2.expert_weights_scale f32 = 2.500000 llama_model_loader: - kv 22: deepseek2.expert_weights_norm bool = true llama_model_loader: - kv 23: deepseek2.expert_gating_func u32 = 2 llama_model_loader: - kv 24: deepseek2.rope.dimension_count u32 = 64 llama_model_loader: - kv 25: deepseek2.rope.scaling.type str = yarn llama_model_loader: - kv 26: deepseek2.rope.scaling.factor f32 = 40.000000 llama_model_loader: - kv 27: deepseek2.rope.scaling.original_context_length u32 = 4096 llama_model_loader: - kv 28: deepseek2.rope.scaling.yarn_log_multiplier f32 = 0.100000 llama_model_loader: - kv 29: tokenizer.ggml.model str = gpt2 llama_model_loader: - kv 30: tokenizer.ggml.pre str = deepseek-v3 llama_model_loader: - kv 31: tokenizer.ggml.tokens arr[str,129280] = ["<｜begin▁of▁sentence｜>", "<�... llama_model_loader: - kv 32: tokenizer.ggml.token_type arr[i32,129280] = [3, 3, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, ... llama_model_loader: - kv 33: tokenizer.ggml.merges arr[str,127741] = ["Ġ t", "Ġ a", "i n", "Ġ Ġ", "h e... llama_model_loader: - kv 34: tokenizer.ggml.bos_token_id u32 = 0 llama_model_loader: - kv 35: tokenizer.ggml.eos_token_id u32 = 1 llama_model_loader: - kv 36: tokenizer.ggml.padding_token_id u32 = 1 llama_model_loader: - kv 37: tokenizer.ggml.add_bos_token bool = true llama_model_loader: - kv 38: tokenizer.ggml.add_eos_token bool = false llama_model_loader: - kv 39: tokenizer.chat_template str = {% if not add_generation_prompt is de... llama_model_loader: - kv 40: general.quantization_version u32 = 2 llama_model_loader: - kv 41: general.file_type u32 = 15 llama_model_loader: - type f32: 361 tensors llama_model_loader: - type q4_K: 606 tensors llama_model_loader: - type q6_K: 58 tensors llm_load_vocab: special_eos_id is not in special_eog_ids - the tokenizer config may be incorrect llm_load_vocab: special tokens cache size = 818 llm_load_vocab: token to piece cache size = 0.8223 MB llm_load_print_meta: format = GGUF V3 (latest) llm_load_print_meta: arch = deepseek2 llm_load_print_meta: vocab type = BPE llm_load_print_meta: n_vocab = 129280 llm_load_print_meta: n_merges = 127741 llm_load_print_meta: vocab_only = 0 llm_load_print_meta: n_ctx_train = 163840 llm_load_print_meta: n_embd = 7168 llm_load_print_meta: n_layer = 61 llm_load_print_meta: n_head = 128 llm_load_print_meta: n_head_kv = 128 llm_load_print_meta: n_rot = 64 llm_load_print_meta: n_swa = 0 llm_load_print_meta: n_embd_head_k = 192 llm_load_print_meta: n_embd_head_v = 128 llm_load_print_meta: n_gqa = 1 llm_load_print_meta: n_embd_k_gqa = 24576 llm_load_print_meta: n_embd_v_gqa = 16384 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 = 18432 llm_load_print_meta: n_expert = 256 llm_load_print_meta: n_expert_used = 8 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 = yarn llm_load_print_meta: freq_base_train = 10000.0 llm_load_print_meta: freq_scale_train = 0.025 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: ssm_dt_b_c_rms = 0 llm_load_print_meta: model type = 671B llm_load_print_meta: model ftype = Q4_K - Medium llm_load_print_meta: model params = 671.03 B llm_load_print_meta: model size = 376.65 GiB (4.82 BPW) llm_load_print_meta: general.name = n/a llm_load_print_meta: BOS token = 0 '<｜begin▁of▁sentence｜>' llm_load_print_meta: EOS token = 1 '<｜end▁of▁sentence｜>' llm_load_print_meta: EOT token = 1 '<｜end▁of▁sentence｜>' llm_load_print_meta: PAD token = 1 '<｜end▁of▁sentence｜>' llm_load_print_meta: LF token = 131 'Ä' llm_load_print_meta: FIM PRE token = 128801 '<｜fim▁begin｜>' llm_load_print_meta: FIM SUF token = 128800 '<｜fim▁hole｜>' llm_load_print_meta: FIM MID token = 128802 '<｜fim▁end｜>' llm_load_print_meta: EOG token = 1 '<｜end▁of▁sentence｜>' llm_load_print_meta: max token length = 256 llm_load_print_meta: n_layer_dense_lead = 3 llm_load_print_meta: n_lora_q = 1536 llm_load_print_meta: n_lora_kv = 512 llm_load_print_meta: n_ff_exp = 2048 llm_load_print_meta: n_expert_shared = 1 llm_load_print_meta: expert_weights_scale = 2.5 llm_load_print_meta: expert_weights_norm = 1 llm_load_print_meta: expert_gating_func = sigmoid llm_load_print_meta: rope_yarn_log_mul = 0.1000 llm_load_tensors: offloading 61 repeating layers to GPU llm_load_tensors: offloading output layer to GPU llm_load_tensors: offloaded 62/62 layers to GPU llm_load_tensors: CPU model buffer size = 497.11 MiB llm_load_tensors: CUDA0 model buffer size = 35642.36 MiB llm_load_tensors: CUDA1 model buffer size = 52215.30 MiB llm_load_tensors: CUDA2 model buffer size = 51287.70 MiB llm_load_tensors: CUDA3 model buffer size = 52215.30 MiB llm_load_tensors: CUDA4 model buffer size = 52215.30 MiB llm_load_tensors: CUDA5 model buffer size = 51287.70 MiB llm_load_tensors: CUDA6 model buffer size = 46963.85 MiB llm_load_tensors: CUDA7 model buffer size = 43364.99 MiB llama_new_context_with_model: n_seq_max = 4 llama_new_context_with_model: n_ctx = 8192 llama_new_context_with_model: n_ctx_per_seq = 2048 llama_new_context_with_model: n_batch = 2048 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 = 0.025 llama_new_context_with_model: n_ctx_per_seq (2048) < n_ctx_train (163840) -- the full capacity of the model will not be utilized llama_kv_cache_init: kv_size = 8192, offload = 1, type_k = 'f16', type_v = 'f16', n_layer = 61, can_shift = 0 llama_kv_cache_init: CUDA0 KV buffer size = 5120.00 MiB llama_kv_cache_init: CUDA1 KV buffer size = 5120.00 MiB llama_kv_cache_init: CUDA2 KV buffer size = 5120.00 MiB llama_kv_cache_init: CUDA3 KV buffer size = 5120.00 MiB llama_kv_cache_init: CUDA4 KV buffer size = 5120.00 MiB llama_kv_cache_init: CUDA5 KV buffer size = 5120.00 MiB llama_kv_cache_init: CUDA6 KV buffer size = 4480.00 MiB llama_kv_cache_init: CUDA7 KV buffer size = 3840.00 MiB llama_new_context_with_model: KV self size = 39040.00 MiB, K (f16): 23424.00 MiB, V (f16): 15616.00 MiB llama_new_context_with_model: CUDA_Host output buffer size = 2.08 MiB llama_new_context_with_model: pipeline parallelism enabled (n_copies=4) ggml_cuda_host_malloc: failed to allocate 78.02 MiB of pinned memory: out of memory llama_new_context_with_model: CUDA0 compute buffer size = 2322.01 MiB llama_new_context_with_model: CUDA1 compute buffer size = 2322.01 MiB llama_new_context_with_model: CUDA2 compute buffer size = 2322.01 MiB llama_new_context_with_model: CUDA3 compute buffer size = 2322.01 MiB llama_new_context_with_model: CUDA4 compute buffer size = 2322.01 MiB llama_new_context_with_model: CUDA5 compute buffer size = 2322.01 MiB llama_new_context_with_model: CUDA6 compute buffer size = 2322.01 MiB llama_new_context_with_model: CUDA7 compute buffer size = 2322.02 MiB llama_new_context_with_model: CUDA_Host compute buffer size = 78.02 MiB llama_new_context_with_model: graph nodes = 5025 llama_new_context_with_model: graph splits = 9 time=2025-03-04T00:54:16.177+08:00 level=INFO source=server.go:594 msg="llama runner started in 718.29 seconds" llama_model_loader: loaded meta data with 42 key-value pairs and 1025 tensors from H:\.ollama\blobs\sha256-9801e7fce27dbf3d0bfb468b7b21f1d132131a546dfc43e50518631b8b1800a9 (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 = deepseek2 llama_model_loader: - kv 1: general.type str = model llama_model_loader: - kv 2: general.size_label str = 256x20B llama_model_loader: - kv 3: deepseek2.block_count u32 = 61 llama_model_loader: - kv 4: deepseek2.context_length u32 = 163840 llama_model_loader: - kv 5: deepseek2.embedding_length u32 = 7168 llama_model_loader: - kv 6: deepseek2.feed_forward_length u32 = 18432 llama_model_loader: - kv 7: deepseek2.attention.head_count u32 = 128 llama_model_loader: - kv 8: deepseek2.attention.head_count_kv u32 = 128 llama_model_loader: - kv 9: deepseek2.rope.freq_base f32 = 10000.000000 llama_model_loader: - kv 10: deepseek2.attention.layer_norm_rms_epsilon f32 = 0.000001 llama_model_loader: - kv 11: deepseek2.expert_used_count u32 = 8 llama_model_loader: - kv 12: deepseek2.leading_dense_block_count u32 = 3 llama_model_loader: - kv 13: deepseek2.vocab_size u32 = 129280 llama_model_loader: - kv 14: deepseek2.attention.q_lora_rank u32 = 1536 llama_model_loader: - kv 15: deepseek2.attention.kv_lora_rank u32 = 512 llama_model_loader: - kv 16: deepseek2.attention.key_length u32 = 192 llama_model_loader: - kv 17: deepseek2.attention.value_length u32 = 128 llama_model_loader: - kv 18: deepseek2.expert_feed_forward_length u32 = 2048 llama_model_loader: - kv 19: deepseek2.expert_count u32 = 256 llama_model_loader: - kv 20: deepseek2.expert_shared_count u32 = 1 llama_model_loader: - kv 21: deepseek2.expert_weights_scale f32 = 2.500000 llama_model_loader: - kv 22: deepseek2.expert_weights_norm bool = true llama_model_loader: - kv 23: deepseek2.expert_gating_func u32 = 2 llama_model_loader: - kv 24: deepseek2.rope.dimension_count u32 = 64 llama_model_loader: - kv 25: deepseek2.rope.scaling.type str = yarn llama_model_loader: - kv 26: deepseek2.rope.scaling.factor f32 = 40.000000 llama_model_loader: - kv 27: deepseek2.rope.scaling.original_context_length u32 = 4096 llama_model_loader: - kv 28: deepseek2.rope.scaling.yarn_log_multiplier f32 = 0.100000 llama_model_loader: - kv 29: tokenizer.ggml.model str = gpt2 llama_model_loader: - kv 30: tokenizer.ggml.pre str = deepseek-v3 llama_model_loader: - kv 31: tokenizer.ggml.tokens arr[str,129280] = ["<｜begin▁of▁sentence｜>", "<�... llama_model_loader: - kv 32: tokenizer.ggml.token_type arr[i32,129280] = [3, 3, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, ... llama_model_loader: - kv 33: tokenizer.ggml.merges arr[str,127741] = ["Ġ t", "Ġ a", "i n", "Ġ Ġ", "h e... llama_model_loader: - kv 34: tokenizer.ggml.bos_token_id u32 = 0 llama_model_loader: - kv 35: tokenizer.ggml.eos_token_id u32 = 1 llama_model_loader: - kv 36: tokenizer.ggml.padding_token_id u32 = 1 llama_model_loader: - kv 37: tokenizer.ggml.add_bos_token bool = true llama_model_loader: - kv 38: tokenizer.ggml.add_eos_token bool = false llama_model_loader: - kv 39: tokenizer.chat_template str = {% if not add_generation_prompt is de... llama_model_loader: - kv 40: general.quantization_version u32 = 2 llama_model_loader: - kv 41: general.file_type u32 = 15 llama_model_loader: - type f32: 361 tensors llama_model_loader: - type q4_K: 606 tensors llama_model_loader: - type q6_K: 58 tensors llm_load_vocab: special_eos_id is not in special_eog_ids - the tokenizer config may be incorrect llm_load_vocab: special tokens cache size = 818 llm_load_vocab: token to piece cache size = 0.8223 MB llm_load_print_meta: format = GGUF V3 (latest) llm_load_print_meta: arch = deepseek2 llm_load_print_meta: vocab type = BPE llm_load_print_meta: n_vocab = 129280 llm_load_print_meta: n_merges = 127741 llm_load_print_meta: vocab_only = 1 llm_load_print_meta: model type = ?B llm_load_print_meta: model ftype = all F32 llm_load_print_meta: model params = 671.03 B llm_load_print_meta: model size = 376.65 GiB (4.82 BPW) llm_load_print_meta: general.name = n/a llm_load_print_meta: BOS token = 0 '<｜begin▁of▁sentence｜>' llm_load_print_meta: EOS token = 1 '<｜end▁of▁sentence｜>' llm_load_print_meta: EOT token = 1 '<｜end▁of▁sentence｜>' llm_load_print_meta: PAD token = 1 '<｜end▁of▁sentence｜>' llm_load_print_meta: LF token = 131 'Ä' llm_load_print_meta: FIM PRE token = 128801 '<｜fim▁begin｜>' llm_load_print_meta: FIM SUF token = 128800 '<｜fim▁hole｜>' llm_load_print_meta: FIM MID token = 128802 '<｜fim▁end｜>' llm_load_print_meta: EOG token = 1 '<｜end▁of▁sentence｜>' llm_load_print_meta: max token length = 256 llm_load_print_meta: n_layer_dense_lead = 0 llm_load_print_meta: n_lora_q = 0 llm_load_print_meta: n_lora_kv = 0 llm_load_print_meta: n_ff_exp = 0 llm_load_print_meta: n_expert_shared = 0 llm_load_print_meta: expert_weights_scale = 0.0 llm_load_print_meta: expert_weights_norm = 0 llm_load_print_meta: expert_gating_func = unknown llm_load_print_meta: rope_yarn_log_mul = 0.0000 llama_model_load: vocab only - skipping tensors [GIN] 2025/03/04 - 00:56:45 | 200 | 14m28s | 127.0.0.1 | POST "/api/chat" time=2025-03-04T01:01:50.885+08:00 level=WARN source=sched.go:646 msg="gpu VRAM usage didn't recover within timeout" seconds=5.2169324 model=H:\.ollama\blobs\sha256-9801e7fce27dbf3d0bfb468b7b21f1d132131a546dfc43e50518631b8b1800a9 time=2025-03-04T01:01:51.135+08:00 level=WARN source=sched.go:646 msg="gpu VRAM usage didn't recover within timeout" seconds=5.4673458 model=H:\.ollama\blobs\sha256-9801e7fce27dbf3d0bfb468b7b21f1d132131a546dfc43e50518631b8b1800a9 time=2025-03-04T01:01:51.385+08:00 level=WARN source=sched.go:646 msg="gpu VRAM usage didn't recover within timeout" seconds=5.7173227 model=H:\.ollama\blobs\sha256-9801e7fce27dbf3d0bfb468b7b21f1d132131a546dfc43e50518631b8b1800a9 ``` ### OS Windows ### GPU Nvidia ### CPU Intel ### Ollama version 0.5.11
GiteaMirror added the bugneeds more info labels 2026-05-09 19:20:20 -05:00
GiteaMirror commented 2026-05-09 19:20:25 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Mar 5, 2025):

Might be #5975, although there's no error message in the log. Try running with OLLAMA_DEBUG=1 in the server environment.

<!-- gh-comment-id:2701020080 --> @rick-github commented on GitHub (Mar 5, 2025): Might be #5975, although there's no error message in the log. Try running with `OLLAMA_DEBUG=1` in the server environment.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label bug needs more info
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#83893
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?