github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-16 13:19:40 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #10798] In version 0.7.0, the GPU is not working, and the CPU usage rate is 100% #53603

New Issue
Closed
opened 2026-04-29 04:08:52 -05:00 by GiteaMirror · 3 comments
GiteaMirror commented 2026-04-29 04:08:52 -05:00
Owner
Copy Link

Originally created by @yujianfeikong on GitHub (May 21, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/10798

What is the issue?

When running the small model of deepseek-r1:1.5b, the GPU is not loaded, and the CPU usage rate reaches 100% during chatting.

+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 560.35.03 Driver Version: 560.35.03 CUDA Version: 12.6 |
|-----------------------------------------+------------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA GeForce RTX 4090 Off | 00000000:01:00.0 On | Off |
| 34% 36C P0 71W / 450W | 814MiB / 24564MiB | 6% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| 0 N/A N/A 2035 G /usr/lib/xorg/Xorg 72MiB |
| 0 N/A N/A 6774 G /usr/lib/xorg/Xorg 130MiB |
| 0 N/A N/A 7142 G /usr/bin/gnome-shell 67MiB |
| 0 N/A N/A 9239 C /usr/share/rustdesk/rustdesk 459MiB |
| 0 N/A N/A 9458 G /usr/share/rustdesk/rustdesk 20MiB |
+-----------------------------------------------------------------------------------------+

Relevant log output

月 21 17:49:23 ubuntu ollama[1576]: [GIN] 2025/05/21 - 17:49:23 | 200 |   24.702792ms |       127.0.0.1 | POST     "/api/show"
5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.679+08:00 level=INFO source=sched.go:777 msg="new model will fit in available VRAM in single GPU, loading" model=/usr/share/ollama/.ollama/models/blobs/sha256-aabd4debf0c8f08881923f2c25fc0fdeed24435271c2b3e92c4af36704040dbc gpu=GPU-d2e6477d-6250-fded-3521-90668cd0ed45 parallel=2 available=24038342656 required="2.1 GiB"
5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.790+08:00 level=INFO source=server.go:135 msg="system memory" total="31.1 GiB" free="25.8 GiB" free_swap="1.8 GiB"
5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.790+08:00 level=INFO source=server.go:168 msg=offload library=cuda layers.requested=-1 layers.model=29 layers.offload=29 layers.split="" memory.available="[22.4 GiB]" memory.gpu_overhead="0 B" memory.required.full="2.1 GiB" memory.required.partial="2.1 GiB" memory.required.kv="224.0 MiB" memory.required.allocations="[2.1 GiB]" memory.weights.total="934.7 MiB" memory.weights.repeating="752.1 MiB" memory.weights.nonrepeating="182.6 MiB" memory.graph.full="299.8 MiB" memory.graph.partial="482.3 MiB"
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: loaded meta data with 26 key-value pairs and 339 tensors from /usr/share/ollama/.ollama/models/blobs/sha256-aabd4debf0c8f08881923f2c25fc0fdeed24435271c2b3e92c4af36704040dbc (version GGUF V3 (latest))
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv   0:                       general.architecture str              = qwen2
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv   1:                               general.type str              = model
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv   2:                               general.name str              = DeepSeek R1 Distill Qwen 1.5B
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv   3:                           general.basename str              = DeepSeek-R1-Distill-Qwen
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv   4:                         general.size_label str              = 1.5B
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv   5:                          qwen2.block_count u32              = 28
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv   6:                       qwen2.context_length u32              = 131072
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv   7:                     qwen2.embedding_length u32              = 1536
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv   8:                  qwen2.feed_forward_length u32              = 8960
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv   9:                 qwen2.attention.head_count u32              = 12
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  10:              qwen2.attention.head_count_kv u32              = 2
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  11:                       qwen2.rope.freq_base f32              = 10000.000000
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  12:     qwen2.attention.layer_norm_rms_epsilon f32              = 0.000001
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  13:                          general.file_type u32              = 15
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  14:                       tokenizer.ggml.model str              = gpt2
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  15:                         tokenizer.ggml.pre str              = qwen2
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  16:                      tokenizer.ggml.tokens arr[str,151936]  = ["!", "\"", "#", "$", "%", "&", "'", ...
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  17:                  tokenizer.ggml.token_type arr[i32,151936]  = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  18:                      tokenizer.ggml.merges arr[str,151387]  = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t",...
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  19:                tokenizer.ggml.bos_token_id u32              = 151646
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  20:                tokenizer.ggml.eos_token_id u32              = 151643
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  21:            tokenizer.ggml.padding_token_id u32              = 151643
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  22:               tokenizer.ggml.add_bos_token bool             = true
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  23:               tokenizer.ggml.add_eos_token bool             = false
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  24:                    tokenizer.chat_template str              = {% if not add_generation_prompt is de...
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  25:               general.quantization_version u32              = 2
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - type  f32:  141 tensors
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - type q4_K:  169 tensors
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - type q6_K:   29 tensors
5月 21 17:49:23 ubuntu ollama[1576]: print_info: file format = GGUF V3 (latest)
5月 21 17:49:23 ubuntu ollama[1576]: print_info: file type   = Q4_K - Medium
5月 21 17:49:23 ubuntu ollama[1576]: print_info: file size   = 1.04 GiB (5.00 BPW)
5月 21 17:49:23 ubuntu ollama[1576]: load: special_eos_id is not in special_eog_ids - the tokenizer config may be incorrect
5月 21 17:49:23 ubuntu ollama[1576]: load: special tokens cache size = 22
5月 21 17:49:23 ubuntu ollama[1576]: load: token to piece cache size = 0.9310 MB
5月 21 17:49:23 ubuntu ollama[1576]: print_info: arch             = qwen2
5月 21 17:49:23 ubuntu ollama[1576]: print_info: vocab_only       = 1
5月 21 17:49:23 ubuntu ollama[1576]: print_info: model type       = ?B
5月 21 17:49:23 ubuntu ollama[1576]: print_info: model params     = 1.78 B
5月 21 17:49:23 ubuntu ollama[1576]: print_info: general.name     = DeepSeek R1 Distill Qwen 1.5B
5月 21 17:49:23 ubuntu ollama[1576]: print_info: vocab type       = BPE
5月 21 17:49:23 ubuntu ollama[1576]: print_info: n_vocab          = 151936
5月 21 17:49:23 ubuntu ollama[1576]: print_info: n_merges         = 151387
5月 21 17:49:23 ubuntu ollama[1576]: print_info: BOS token        = 151646 '<｜begin▁of▁sentence｜>'
5月 21 17:49:23 ubuntu ollama[1576]: print_info: EOS token        = 151643 '<｜end▁of▁sentence｜>'
5月 21 17:49:23 ubuntu ollama[1576]: print_info: EOT token        = 151643 '<｜end▁of▁sentence｜>'
5月 21 17:49:23 ubuntu ollama[1576]: print_info: PAD token        = 151643 '<｜end▁of▁sentence｜>'
5月 21 17:49:23 ubuntu ollama[1576]: print_info: LF token         = 198 'Ċ'
5月 21 17:49:23 ubuntu ollama[1576]: print_info: FIM PRE token    = 151659 '<|fim_prefix|>'
5月 21 17:49:23 ubuntu ollama[1576]: print_info: FIM SUF token    = 151661 '<|fim_suffix|>'
5月 21 17:49:23 ubuntu ollama[1576]: print_info: FIM MID token    = 151660 '<|fim_middle|>'
5月 21 17:49:23 ubuntu ollama[1576]: print_info: FIM PAD token    = 151662 '<|fim_pad|>'
5月 21 17:49:23 ubuntu ollama[1576]: print_info: FIM REP token    = 151663 '<|repo_name|>'
5月 21 17:49:23 ubuntu ollama[1576]: print_info: FIM SEP token    = 151664 '<|file_sep|>'
5月 21 17:49:23 ubuntu ollama[1576]: print_info: EOG token        = 151643 '<｜end▁of▁sentence｜>'
5月 21 17:49:23 ubuntu ollama[1576]: print_info: EOG token        = 151662 '<|fim_pad|>'
5月 21 17:49:23 ubuntu ollama[1576]: print_info: EOG token        = 151663 '<|repo_name|>'
5月 21 17:49:23 ubuntu ollama[1576]: print_info: EOG token        = 151664 '<|file_sep|>'
5月 21 17:49:23 ubuntu ollama[1576]: print_info: max token length = 256
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_load: vocab only - skipping tensors
5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.898+08:00 level=INFO source=server.go:431 msg="starting llama server" cmd="/usr/local/bin/ollama runner --model /usr/share/ollama/.ollama/models/blobs/sha256-aabd4debf0c8f08881923f2c25fc0fdeed24435271c2b3e92c4af36704040dbc --ctx-size 8192 --batch-size 512 --n-gpu-layers 29 --threads 8 --parallel 2 --port 44017"
5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.898+08:00 level=INFO source=sched.go:472 msg="loaded runners" count=1
5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.898+08:00 level=INFO source=server.go:591 msg="waiting for llama runner to start responding"
5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.898+08:00 level=INFO source=server.go:625 msg="waiting for server to become available" status="llm server not responding"
5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.905+08:00 level=INFO source=runner.go:815 msg="starting go runner"
5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.905+08:00 level=INFO source=ggml.go:104 msg=system CPU.0.LLAMAFILE=1 compiler=cgo(gcc)
5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.930+08:00 level=INFO source=runner.go:874 msg="Server listening on 127.0.0.1:44017"
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: loaded meta data with 26 key-value pairs and 339 tensors from /usr/share/ollama/.ollama/models/blobs/sha256-aabd4debf0c8f08881923f2c25fc0fdeed24435271c2b3e92c4af36704040dbc (version GGUF V3 (latest))
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv   0:                       general.architecture str              = qwen2
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv   1:                               general.type str              = model
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv   2:                               general.name str              = DeepSeek R1 Distill Qwen 1.5B
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv   3:                           general.basename str              = DeepSeek-R1-Distill-Qwen
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv   4:                         general.size_label str              = 1.5B
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv   5:                          qwen2.block_count u32              = 28
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv   6:                       qwen2.context_length u32              = 131072
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv   7:                     qwen2.embedding_length u32              = 1536
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv   8:                  qwen2.feed_forward_length u32              = 8960
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv   9:                 qwen2.attention.head_count u32              = 12
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  10:              qwen2.attention.head_count_kv u32              = 2
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  11:                       qwen2.rope.freq_base f32              = 10000.000000
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  12:     qwen2.attention.layer_norm_rms_epsilon f32              = 0.000001
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  13:                          general.file_type u32              = 15
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  14:                       tokenizer.ggml.model str              = gpt2
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  15:                         tokenizer.ggml.pre str              = qwen2
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  16:                      tokenizer.ggml.tokens arr[str,151936]  = ["!", "\"", "#", "$", "%", "&", "'", ...
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  17:                  tokenizer.ggml.token_type arr[i32,151936]  = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  18:                      tokenizer.ggml.merges arr[str,151387]  = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t",...
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  19:                tokenizer.ggml.bos_token_id u32              = 151646
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  20:                tokenizer.ggml.eos_token_id u32              = 151643
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  21:            tokenizer.ggml.padding_token_id u32              = 151643
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  22:               tokenizer.ggml.add_bos_token bool             = true
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  23:               tokenizer.ggml.add_eos_token bool             = false
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  24:                    tokenizer.chat_template str              = {% if not add_generation_prompt is de...
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv  25:               general.quantization_version u32              = 2
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - type  f32:  141 tensors
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - type q4_K:  169 tensors
5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - type q6_K:   29 tensors
5月 21 17:49:23 ubuntu ollama[1576]: print_info: file format = GGUF V3 (latest)
5月 21 17:49:23 ubuntu ollama[1576]: print_info: file type   = Q4_K - Medium
5月 21 17:49:23 ubuntu ollama[1576]: print_info: file size   = 1.04 GiB (5.00 BPW)
5月 21 17:49:24 ubuntu ollama[1576]: load: special_eos_id is not in special_eog_ids - the tokenizer config may be incorrect
5月 21 17:49:24 ubuntu ollama[1576]: load: special tokens cache size = 22
5月 21 17:49:24 ubuntu ollama[1576]: load: token to piece cache size = 0.9310 MB
5月 21 17:49:24 ubuntu ollama[1576]: print_info: arch             = qwen2
5月 21 17:49:24 ubuntu ollama[1576]: print_info: vocab_only       = 0
5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_ctx_train      = 131072
5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_embd           = 1536
5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_layer          = 28
5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_head           = 12
5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_head_kv        = 2
5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_rot            = 128
5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_swa            = 0
5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_swa_pattern    = 1
5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_embd_head_k    = 128
5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_embd_head_v    = 128
5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_gqa            = 6
5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_embd_k_gqa     = 256
5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_embd_v_gqa     = 256
5月 21 17:49:24 ubuntu ollama[1576]: print_info: f_norm_eps       = 0.0e+00
5月 21 17:49:24 ubuntu ollama[1576]: print_info: f_norm_rms_eps   = 1.0e-06
5月 21 17:49:24 ubuntu ollama[1576]: print_info: f_clamp_kqv      = 0.0e+00
5月 21 17:49:24 ubuntu ollama[1576]: print_info: f_max_alibi_bias = 0.0e+00
5月 21 17:49:24 ubuntu ollama[1576]: print_info: f_logit_scale    = 0.0e+00
5月 21 17:49:24 ubuntu ollama[1576]: print_info: f_attn_scale     = 0.0e+00
5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_ff             = 8960
5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_expert         = 0
5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_expert_used    = 0
5月 21 17:49:24 ubuntu ollama[1576]: print_info: causal attn      = 1
5月 21 17:49:24 ubuntu ollama[1576]: print_info: pooling type     = -1
5月 21 17:49:24 ubuntu ollama[1576]: print_info: rope type        = 2
5月 21 17:49:24 ubuntu ollama[1576]: print_info: rope scaling     = linear
5月 21 17:49:24 ubuntu ollama[1576]: print_info: freq_base_train  = 10000.0
5月 21 17:49:24 ubuntu ollama[1576]: print_info: freq_scale_train = 1
5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_ctx_orig_yarn  = 131072
5月 21 17:49:24 ubuntu ollama[1576]: print_info: rope_finetuned   = unknown
5月 21 17:49:24 ubuntu ollama[1576]: print_info: ssm_d_conv       = 0
5月 21 17:49:24 ubuntu ollama[1576]: print_info: ssm_d_inner      = 0
5月 21 17:49:24 ubuntu ollama[1576]: print_info: ssm_d_state      = 0
5月 21 17:49:24 ubuntu ollama[1576]: print_info: ssm_dt_rank      = 0
5月 21 17:49:24 ubuntu ollama[1576]: print_info: ssm_dt_b_c_rms   = 0
5月 21 17:49:24 ubuntu ollama[1576]: print_info: model type       = 1.5B
5月 21 17:49:24 ubuntu ollama[1576]: print_info: model params     = 1.78 B
5月 21 17:49:24 ubuntu ollama[1576]: print_info: general.name     = DeepSeek R1 Distill Qwen 1.5B
5月 21 17:49:24 ubuntu ollama[1576]: print_info: vocab type       = BPE
5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_vocab          = 151936
5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_merges         = 151387
5月 21 17:49:24 ubuntu ollama[1576]: print_info: BOS token        = 151646 '<｜begin▁of▁sentence｜>'
5月 21 17:49:24 ubuntu ollama[1576]: print_info: EOS token        = 151643 '<｜end▁of▁sentence｜>'
5月 21 17:49:24 ubuntu ollama[1576]: print_info: EOT token        = 151643 '<｜end▁of▁sentence｜>'
5月 21 17:49:24 ubuntu ollama[1576]: print_info: PAD token        = 151643 '<｜end▁of▁sentence｜>'
5月 21 17:49:24 ubuntu ollama[1576]: print_info: LF token         = 198 'Ċ'
5月 21 17:49:24 ubuntu ollama[1576]: print_info: FIM PRE token    = 151659 '<|fim_prefix|>'
5月 21 17:49:24 ubuntu ollama[1576]: print_info: FIM SUF token    = 151661 '<|fim_suffix|>'
5月 21 17:49:24 ubuntu ollama[1576]: print_info: FIM MID token    = 151660 '<|fim_middle|>'
5月 21 17:49:24 ubuntu ollama[1576]: print_info: FIM PAD token    = 151662 '<|fim_pad|>'
5月 21 17:49:24 ubuntu ollama[1576]: print_info: FIM REP token    = 151663 '<|repo_name|>'
5月 21 17:49:24 ubuntu ollama[1576]: print_info: FIM SEP token    = 151664 '<|file_sep|>'
5月 21 17:49:24 ubuntu ollama[1576]: print_info: EOG token        = 151643 '<｜end▁of▁sentence｜>'
5月 21 17:49:24 ubuntu ollama[1576]: print_info: EOG token        = 151662 '<|fim_pad|>'
5月 21 17:49:24 ubuntu ollama[1576]: print_info: EOG token        = 151663 '<|repo_name|>'
5月 21 17:49:24 ubuntu ollama[1576]: print_info: EOG token        = 151664 '<|file_sep|>'
5月 21 17:49:24 ubuntu ollama[1576]: print_info: max token length = 256
5月 21 17:49:24 ubuntu ollama[1576]: load_tensors: loading model tensors, this can take a while... (mmap = true)
5月 21 17:49:24 ubuntu ollama[1576]: load_tensors:   CPU_Mapped model buffer size =  1059.89 MiB
5月 21 17:49:24 ubuntu ollama[1576]: llama_context: constructing llama_context
5月 21 17:49:24 ubuntu ollama[1576]: llama_context: n_seq_max     = 2
5月 21 17:49:24 ubuntu ollama[1576]: llama_context: n_ctx         = 8192
5月 21 17:49:24 ubuntu ollama[1576]: llama_context: n_ctx_per_seq = 4096
5月 21 17:49:24 ubuntu ollama[1576]: llama_context: n_batch       = 1024
5月 21 17:49:24 ubuntu ollama[1576]: llama_context: n_ubatch      = 512
5月 21 17:49:24 ubuntu ollama[1576]: llama_context: causal_attn   = 1
5月 21 17:49:24 ubuntu ollama[1576]: llama_context: flash_attn    = 0
5月 21 17:49:24 ubuntu ollama[1576]: llama_context: freq_base     = 10000.0
5月 21 17:49:24 ubuntu ollama[1576]: llama_context: freq_scale    = 1
5月 21 17:49:24 ubuntu ollama[1576]: llama_context: n_ctx_per_seq (4096) < n_ctx_train (131072) -- the full capacity of the model will not be utilized
5月 21 17:49:24 ubuntu ollama[1576]: llama_context:        CPU  output buffer size =     1.17 MiB
5月 21 17:49:24 ubuntu ollama[1576]: llama_kv_cache_unified: kv_size = 8192, type_k = 'f16', type_v = 'f16', n_layer = 28, can_shift = 1, padding = 32
5月 21 17:49:24 ubuntu ollama[1576]: llama_kv_cache_unified:        CPU KV buffer size =   224.00 MiB
5月 21 17:49:24 ubuntu ollama[1576]: llama_kv_cache_unified: KV self size  =  224.00 MiB, K (f16):  112.00 MiB, V (f16):  112.00 MiB
5月 21 17:49:24 ubuntu ollama[1576]: llama_context:        CPU compute buffer size =   302.75 MiB
5月 21 17:49:24 ubuntu ollama[1576]: llama_context: graph nodes  = 1042
5月 21 17:49:24 ubuntu ollama[1576]: llama_context: graph splits = 1
5月 21 17:49:24 ubuntu ollama[1576]: time=2025-05-21T17:49:24.149+08:00 level=INFO source=server.go:630 msg="llama runner started in 0.25 seconds"

OS

Linux

GPU

Nvidia

CPU

Intel

Ollama version

0.7.0

Originally created by @yujianfeikong on GitHub (May 21, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/10798 ### What is the issue? When running the small model of deepseek-r1:1.5b, the GPU is not loaded, and the CPU usage rate reaches 100% during chatting. +-----------------------------------------------------------------------------------------+ | NVIDIA-SMI 560.35.03 Driver Version: 560.35.03 CUDA Version: 12.6 | |-----------------------------------------+------------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+========================+======================| | 0 NVIDIA GeForce RTX 4090 Off | 00000000:01:00.0 On | Off | | 34% 36C P0 71W / 450W | 814MiB / 24564MiB | 6% Default | | | | N/A | +-----------------------------------------+------------------------+----------------------+ +-----------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=========================================================================================| | 0 N/A N/A 2035 G /usr/lib/xorg/Xorg 72MiB | | 0 N/A N/A 6774 G /usr/lib/xorg/Xorg 130MiB | | 0 N/A N/A 7142 G /usr/bin/gnome-shell 67MiB | | 0 N/A N/A 9239 C /usr/share/rustdesk/rustdesk 459MiB | | 0 N/A N/A 9458 G /usr/share/rustdesk/rustdesk 20MiB | +-----------------------------------------------------------------------------------------+ ### Relevant log output ```shell 月 21 17:49:23 ubuntu ollama[1576]: [GIN] 2025/05/21 - 17:49:23 | 200 | 24.702792ms | 127.0.0.1 | POST "/api/show" 5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.679+08:00 level=INFO source=sched.go:777 msg="new model will fit in available VRAM in single GPU, loading" model=/usr/share/ollama/.ollama/models/blobs/sha256-aabd4debf0c8f08881923f2c25fc0fdeed24435271c2b3e92c4af36704040dbc gpu=GPU-d2e6477d-6250-fded-3521-90668cd0ed45 parallel=2 available=24038342656 required="2.1 GiB" 5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.790+08:00 level=INFO source=server.go:135 msg="system memory" total="31.1 GiB" free="25.8 GiB" free_swap="1.8 GiB" 5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.790+08:00 level=INFO source=server.go:168 msg=offload library=cuda layers.requested=-1 layers.model=29 layers.offload=29 layers.split="" memory.available="[22.4 GiB]" memory.gpu_overhead="0 B" memory.required.full="2.1 GiB" memory.required.partial="2.1 GiB" memory.required.kv="224.0 MiB" memory.required.allocations="[2.1 GiB]" memory.weights.total="934.7 MiB" memory.weights.repeating="752.1 MiB" memory.weights.nonrepeating="182.6 MiB" memory.graph.full="299.8 MiB" memory.graph.partial="482.3 MiB" 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: loaded meta data with 26 key-value pairs and 339 tensors from /usr/share/ollama/.ollama/models/blobs/sha256-aabd4debf0c8f08881923f2c25fc0fdeed24435271c2b3e92c4af36704040dbc (version GGUF V3 (latest)) 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output. 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 0: general.architecture str = qwen2 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 1: general.type str = model 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 2: general.name str = DeepSeek R1 Distill Qwen 1.5B 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 3: general.basename str = DeepSeek-R1-Distill-Qwen 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 4: general.size_label str = 1.5B 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 5: qwen2.block_count u32 = 28 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 6: qwen2.context_length u32 = 131072 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 7: qwen2.embedding_length u32 = 1536 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 8: qwen2.feed_forward_length u32 = 8960 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 9: qwen2.attention.head_count u32 = 12 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 10: qwen2.attention.head_count_kv u32 = 2 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 11: qwen2.rope.freq_base f32 = 10000.000000 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 12: qwen2.attention.layer_norm_rms_epsilon f32 = 0.000001 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 13: general.file_type u32 = 15 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 14: tokenizer.ggml.model str = gpt2 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 15: tokenizer.ggml.pre str = qwen2 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 16: tokenizer.ggml.tokens arr[str,151936] = ["!", "\"", "#", "$", "%", "&", "'", ... 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 17: tokenizer.ggml.token_type arr[i32,151936] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ... 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 18: tokenizer.ggml.merges arr[str,151387] = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t",... 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 19: tokenizer.ggml.bos_token_id u32 = 151646 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 20: tokenizer.ggml.eos_token_id u32 = 151643 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 21: tokenizer.ggml.padding_token_id u32 = 151643 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 22: tokenizer.ggml.add_bos_token bool = true 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 23: tokenizer.ggml.add_eos_token bool = false 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 24: tokenizer.chat_template str = {% if not add_generation_prompt is de... 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 25: general.quantization_version u32 = 2 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - type f32: 141 tensors 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - type q4_K: 169 tensors 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - type q6_K: 29 tensors 5月 21 17:49:23 ubuntu ollama[1576]: print_info: file format = GGUF V3 (latest) 5月 21 17:49:23 ubuntu ollama[1576]: print_info: file type = Q4_K - Medium 5月 21 17:49:23 ubuntu ollama[1576]: print_info: file size = 1.04 GiB (5.00 BPW) 5月 21 17:49:23 ubuntu ollama[1576]: load: special_eos_id is not in special_eog_ids - the tokenizer config may be incorrect 5月 21 17:49:23 ubuntu ollama[1576]: load: special tokens cache size = 22 5月 21 17:49:23 ubuntu ollama[1576]: load: token to piece cache size = 0.9310 MB 5月 21 17:49:23 ubuntu ollama[1576]: print_info: arch = qwen2 5月 21 17:49:23 ubuntu ollama[1576]: print_info: vocab_only = 1 5月 21 17:49:23 ubuntu ollama[1576]: print_info: model type = ?B 5月 21 17:49:23 ubuntu ollama[1576]: print_info: model params = 1.78 B 5月 21 17:49:23 ubuntu ollama[1576]: print_info: general.name = DeepSeek R1 Distill Qwen 1.5B 5月 21 17:49:23 ubuntu ollama[1576]: print_info: vocab type = BPE 5月 21 17:49:23 ubuntu ollama[1576]: print_info: n_vocab = 151936 5月 21 17:49:23 ubuntu ollama[1576]: print_info: n_merges = 151387 5月 21 17:49:23 ubuntu ollama[1576]: print_info: BOS token = 151646 '<｜begin▁of▁sentence｜>' 5月 21 17:49:23 ubuntu ollama[1576]: print_info: EOS token = 151643 '<｜end▁of▁sentence｜>' 5月 21 17:49:23 ubuntu ollama[1576]: print_info: EOT token = 151643 '<｜end▁of▁sentence｜>' 5月 21 17:49:23 ubuntu ollama[1576]: print_info: PAD token = 151643 '<｜end▁of▁sentence｜>' 5月 21 17:49:23 ubuntu ollama[1576]: print_info: LF token = 198 'Ċ' 5月 21 17:49:23 ubuntu ollama[1576]: print_info: FIM PRE token = 151659 '<|fim_prefix|>' 5月 21 17:49:23 ubuntu ollama[1576]: print_info: FIM SUF token = 151661 '<|fim_suffix|>' 5月 21 17:49:23 ubuntu ollama[1576]: print_info: FIM MID token = 151660 '<|fim_middle|>' 5月 21 17:49:23 ubuntu ollama[1576]: print_info: FIM PAD token = 151662 '<|fim_pad|>' 5月 21 17:49:23 ubuntu ollama[1576]: print_info: FIM REP token = 151663 '<|repo_name|>' 5月 21 17:49:23 ubuntu ollama[1576]: print_info: FIM SEP token = 151664 '<|file_sep|>' 5月 21 17:49:23 ubuntu ollama[1576]: print_info: EOG token = 151643 '<｜end▁of▁sentence｜>' 5月 21 17:49:23 ubuntu ollama[1576]: print_info: EOG token = 151662 '<|fim_pad|>' 5月 21 17:49:23 ubuntu ollama[1576]: print_info: EOG token = 151663 '<|repo_name|>' 5月 21 17:49:23 ubuntu ollama[1576]: print_info: EOG token = 151664 '<|file_sep|>' 5月 21 17:49:23 ubuntu ollama[1576]: print_info: max token length = 256 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_load: vocab only - skipping tensors 5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.898+08:00 level=INFO source=server.go:431 msg="starting llama server" cmd="/usr/local/bin/ollama runner --model /usr/share/ollama/.ollama/models/blobs/sha256-aabd4debf0c8f08881923f2c25fc0fdeed24435271c2b3e92c4af36704040dbc --ctx-size 8192 --batch-size 512 --n-gpu-layers 29 --threads 8 --parallel 2 --port 44017" 5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.898+08:00 level=INFO source=sched.go:472 msg="loaded runners" count=1 5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.898+08:00 level=INFO source=server.go:591 msg="waiting for llama runner to start responding" 5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.898+08:00 level=INFO source=server.go:625 msg="waiting for server to become available" status="llm server not responding" 5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.905+08:00 level=INFO source=runner.go:815 msg="starting go runner" 5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.905+08:00 level=INFO source=ggml.go:104 msg=system CPU.0.LLAMAFILE=1 compiler=cgo(gcc) 5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.930+08:00 level=INFO source=runner.go:874 msg="Server listening on 127.0.0.1:44017" 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: loaded meta data with 26 key-value pairs and 339 tensors from /usr/share/ollama/.ollama/models/blobs/sha256-aabd4debf0c8f08881923f2c25fc0fdeed24435271c2b3e92c4af36704040dbc (version GGUF V3 (latest)) 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output. 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 0: general.architecture str = qwen2 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 1: general.type str = model 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 2: general.name str = DeepSeek R1 Distill Qwen 1.5B 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 3: general.basename str = DeepSeek-R1-Distill-Qwen 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 4: general.size_label str = 1.5B 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 5: qwen2.block_count u32 = 28 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 6: qwen2.context_length u32 = 131072 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 7: qwen2.embedding_length u32 = 1536 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 8: qwen2.feed_forward_length u32 = 8960 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 9: qwen2.attention.head_count u32 = 12 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 10: qwen2.attention.head_count_kv u32 = 2 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 11: qwen2.rope.freq_base f32 = 10000.000000 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 12: qwen2.attention.layer_norm_rms_epsilon f32 = 0.000001 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 13: general.file_type u32 = 15 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 14: tokenizer.ggml.model str = gpt2 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 15: tokenizer.ggml.pre str = qwen2 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 16: tokenizer.ggml.tokens arr[str,151936] = ["!", "\"", "#", "$", "%", "&", "'", ... 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 17: tokenizer.ggml.token_type arr[i32,151936] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ... 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 18: tokenizer.ggml.merges arr[str,151387] = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t",... 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 19: tokenizer.ggml.bos_token_id u32 = 151646 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 20: tokenizer.ggml.eos_token_id u32 = 151643 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 21: tokenizer.ggml.padding_token_id u32 = 151643 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 22: tokenizer.ggml.add_bos_token bool = true 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 23: tokenizer.ggml.add_eos_token bool = false 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 24: tokenizer.chat_template str = {% if not add_generation_prompt is de... 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - kv 25: general.quantization_version u32 = 2 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - type f32: 141 tensors 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - type q4_K: 169 tensors 5月 21 17:49:23 ubuntu ollama[1576]: llama_model_loader: - type q6_K: 29 tensors 5月 21 17:49:23 ubuntu ollama[1576]: print_info: file format = GGUF V3 (latest) 5月 21 17:49:23 ubuntu ollama[1576]: print_info: file type = Q4_K - Medium 5月 21 17:49:23 ubuntu ollama[1576]: print_info: file size = 1.04 GiB (5.00 BPW) 5月 21 17:49:24 ubuntu ollama[1576]: load: special_eos_id is not in special_eog_ids - the tokenizer config may be incorrect 5月 21 17:49:24 ubuntu ollama[1576]: load: special tokens cache size = 22 5月 21 17:49:24 ubuntu ollama[1576]: load: token to piece cache size = 0.9310 MB 5月 21 17:49:24 ubuntu ollama[1576]: print_info: arch = qwen2 5月 21 17:49:24 ubuntu ollama[1576]: print_info: vocab_only = 0 5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_ctx_train = 131072 5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_embd = 1536 5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_layer = 28 5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_head = 12 5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_head_kv = 2 5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_rot = 128 5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_swa = 0 5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_swa_pattern = 1 5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_embd_head_k = 128 5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_embd_head_v = 128 5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_gqa = 6 5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_embd_k_gqa = 256 5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_embd_v_gqa = 256 5月 21 17:49:24 ubuntu ollama[1576]: print_info: f_norm_eps = 0.0e+00 5月 21 17:49:24 ubuntu ollama[1576]: print_info: f_norm_rms_eps = 1.0e-06 5月 21 17:49:24 ubuntu ollama[1576]: print_info: f_clamp_kqv = 0.0e+00 5月 21 17:49:24 ubuntu ollama[1576]: print_info: f_max_alibi_bias = 0.0e+00 5月 21 17:49:24 ubuntu ollama[1576]: print_info: f_logit_scale = 0.0e+00 5月 21 17:49:24 ubuntu ollama[1576]: print_info: f_attn_scale = 0.0e+00 5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_ff = 8960 5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_expert = 0 5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_expert_used = 0 5月 21 17:49:24 ubuntu ollama[1576]: print_info: causal attn = 1 5月 21 17:49:24 ubuntu ollama[1576]: print_info: pooling type = -1 5月 21 17:49:24 ubuntu ollama[1576]: print_info: rope type = 2 5月 21 17:49:24 ubuntu ollama[1576]: print_info: rope scaling = linear 5月 21 17:49:24 ubuntu ollama[1576]: print_info: freq_base_train = 10000.0 5月 21 17:49:24 ubuntu ollama[1576]: print_info: freq_scale_train = 1 5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_ctx_orig_yarn = 131072 5月 21 17:49:24 ubuntu ollama[1576]: print_info: rope_finetuned = unknown 5月 21 17:49:24 ubuntu ollama[1576]: print_info: ssm_d_conv = 0 5月 21 17:49:24 ubuntu ollama[1576]: print_info: ssm_d_inner = 0 5月 21 17:49:24 ubuntu ollama[1576]: print_info: ssm_d_state = 0 5月 21 17:49:24 ubuntu ollama[1576]: print_info: ssm_dt_rank = 0 5月 21 17:49:24 ubuntu ollama[1576]: print_info: ssm_dt_b_c_rms = 0 5月 21 17:49:24 ubuntu ollama[1576]: print_info: model type = 1.5B 5月 21 17:49:24 ubuntu ollama[1576]: print_info: model params = 1.78 B 5月 21 17:49:24 ubuntu ollama[1576]: print_info: general.name = DeepSeek R1 Distill Qwen 1.5B 5月 21 17:49:24 ubuntu ollama[1576]: print_info: vocab type = BPE 5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_vocab = 151936 5月 21 17:49:24 ubuntu ollama[1576]: print_info: n_merges = 151387 5月 21 17:49:24 ubuntu ollama[1576]: print_info: BOS token = 151646 '<｜begin▁of▁sentence｜>' 5月 21 17:49:24 ubuntu ollama[1576]: print_info: EOS token = 151643 '<｜end▁of▁sentence｜>' 5月 21 17:49:24 ubuntu ollama[1576]: print_info: EOT token = 151643 '<｜end▁of▁sentence｜>' 5月 21 17:49:24 ubuntu ollama[1576]: print_info: PAD token = 151643 '<｜end▁of▁sentence｜>' 5月 21 17:49:24 ubuntu ollama[1576]: print_info: LF token = 198 'Ċ' 5月 21 17:49:24 ubuntu ollama[1576]: print_info: FIM PRE token = 151659 '<|fim_prefix|>' 5月 21 17:49:24 ubuntu ollama[1576]: print_info: FIM SUF token = 151661 '<|fim_suffix|>' 5月 21 17:49:24 ubuntu ollama[1576]: print_info: FIM MID token = 151660 '<|fim_middle|>' 5月 21 17:49:24 ubuntu ollama[1576]: print_info: FIM PAD token = 151662 '<|fim_pad|>' 5月 21 17:49:24 ubuntu ollama[1576]: print_info: FIM REP token = 151663 '<|repo_name|>' 5月 21 17:49:24 ubuntu ollama[1576]: print_info: FIM SEP token = 151664 '<|file_sep|>' 5月 21 17:49:24 ubuntu ollama[1576]: print_info: EOG token = 151643 '<｜end▁of▁sentence｜>' 5月 21 17:49:24 ubuntu ollama[1576]: print_info: EOG token = 151662 '<|fim_pad|>' 5月 21 17:49:24 ubuntu ollama[1576]: print_info: EOG token = 151663 '<|repo_name|>' 5月 21 17:49:24 ubuntu ollama[1576]: print_info: EOG token = 151664 '<|file_sep|>' 5月 21 17:49:24 ubuntu ollama[1576]: print_info: max token length = 256 5月 21 17:49:24 ubuntu ollama[1576]: load_tensors: loading model tensors, this can take a while... (mmap = true) 5月 21 17:49:24 ubuntu ollama[1576]: load_tensors: CPU_Mapped model buffer size = 1059.89 MiB 5月 21 17:49:24 ubuntu ollama[1576]: llama_context: constructing llama_context 5月 21 17:49:24 ubuntu ollama[1576]: llama_context: n_seq_max = 2 5月 21 17:49:24 ubuntu ollama[1576]: llama_context: n_ctx = 8192 5月 21 17:49:24 ubuntu ollama[1576]: llama_context: n_ctx_per_seq = 4096 5月 21 17:49:24 ubuntu ollama[1576]: llama_context: n_batch = 1024 5月 21 17:49:24 ubuntu ollama[1576]: llama_context: n_ubatch = 512 5月 21 17:49:24 ubuntu ollama[1576]: llama_context: causal_attn = 1 5月 21 17:49:24 ubuntu ollama[1576]: llama_context: flash_attn = 0 5月 21 17:49:24 ubuntu ollama[1576]: llama_context: freq_base = 10000.0 5月 21 17:49:24 ubuntu ollama[1576]: llama_context: freq_scale = 1 5月 21 17:49:24 ubuntu ollama[1576]: llama_context: n_ctx_per_seq (4096) < n_ctx_train (131072) -- the full capacity of the model will not be utilized 5月 21 17:49:24 ubuntu ollama[1576]: llama_context: CPU output buffer size = 1.17 MiB 5月 21 17:49:24 ubuntu ollama[1576]: llama_kv_cache_unified: kv_size = 8192, type_k = 'f16', type_v = 'f16', n_layer = 28, can_shift = 1, padding = 32 5月 21 17:49:24 ubuntu ollama[1576]: llama_kv_cache_unified: CPU KV buffer size = 224.00 MiB 5月 21 17:49:24 ubuntu ollama[1576]: llama_kv_cache_unified: KV self size = 224.00 MiB, K (f16): 112.00 MiB, V (f16): 112.00 MiB 5月 21 17:49:24 ubuntu ollama[1576]: llama_context: CPU compute buffer size = 302.75 MiB 5月 21 17:49:24 ubuntu ollama[1576]: llama_context: graph nodes = 1042 5月 21 17:49:24 ubuntu ollama[1576]: llama_context: graph splits = 1 5月 21 17:49:24 ubuntu ollama[1576]: time=2025-05-21T17:49:24.149+08:00 level=INFO source=server.go:630 msg="llama runner started in 0.25 seconds" ``` ### OS Linux ### GPU Nvidia ### CPU Intel ### Ollama version 0.7.0
GiteaMirror added the bug label 2026-04-29 04:08:52 -05:00
GiteaMirror commented 2026-04-29 04:10:00 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (May 21, 2025):

5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.790+08:00 level=INFO source=server.go:168
 msg=offload library=cuda layers.requested=-1 layers.model=29 layers.offload=29 layers.split=""
 memory.available="[22.4 GiB]" memory.gpu_overhead="0 B" memory.required.full="2.1 GiB"
 memory.required.partial="2.1 GiB" memory.required.kv="224.0 MiB" memory.required.allocations="[2.1 GiB]"
 memory.weights.total="934.7 MiB" memory.weights.repeating="752.1 MiB" memory.weights.nonrepeating="182.6 MiB"
 memory.graph.full="299.8 MiB" memory.graph.partial="482.3 MiB"

ollama wanted to offload all 29 layers.

5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.905+08:00 level=INFO source=ggml.go:104
 msg=system CPU.0.LLAMAFILE=1 compiler=cgo(gcc)

But it didn't find any CPU or GPU backends, which would normally be in /usr/local/lib/ollama. How did you install ollama?

<!-- gh-comment-id:2897434262 --> @rick-github commented on GitHub (May 21, 2025): ``` 5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.790+08:00 level=INFO source=server.go:168 msg=offload library=cuda layers.requested=-1 layers.model=29 layers.offload=29 layers.split="" memory.available="[22.4 GiB]" memory.gpu_overhead="0 B" memory.required.full="2.1 GiB" memory.required.partial="2.1 GiB" memory.required.kv="224.0 MiB" memory.required.allocations="[2.1 GiB]" memory.weights.total="934.7 MiB" memory.weights.repeating="752.1 MiB" memory.weights.nonrepeating="182.6 MiB" memory.graph.full="299.8 MiB" memory.graph.partial="482.3 MiB" ``` ollama wanted to offload all 29 layers. ``` 5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.905+08:00 level=INFO source=ggml.go:104 msg=system CPU.0.LLAMAFILE=1 compiler=cgo(gcc) ``` But it didn't find any CPU or GPU backends, which would normally be in `/usr/local/lib/ollama`. How did you install ollama?
GiteaMirror commented 2026-04-29 04:10:01 -05:00
Author
Owner
Copy Link

@yujianfeikong commented on GitHub (May 22, 2025):

5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.790+08:00 level=INFO source=server.go:168
 msg=offload library=cuda layers.requested=-1 layers.model=29 layers.offload=29 layers.split=""
 memory.available="[22.4 GiB]" memory.gpu_overhead="0 B" memory.required.full="2.1 GiB"
 memory.required.partial="2.1 GiB" memory.required.kv="224.0 MiB" memory.required.allocations="[2.1 GiB]"
 memory.weights.total="934.7 MiB" memory.weights.repeating="752.1 MiB" memory.weights.nonrepeating="182.6 MiB"
 memory.graph.full="299.8 MiB" memory.graph.partial="482.3 MiB"

ollama wanted to offload all 29 layers.

5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.905+08:00 level=INFO source=ggml.go:104
 msg=system CPU.0.LLAMAFILE=1 compiler=cgo(gcc)

But it didn't find any CPU or GPU backends, which would normally be in /usr/local/lib/ollama. How did you install ollama?

The directory /usr/local/lib/ollama is currently empty and there are no files in it.

curl -L https://ollama.com/download/ollama-linux-amd64.tgz -o ollama-linux-amd64.tgz

ps -ef|grep ollama

kill -9 xxxx

sudo tar -C /usr -xzf ollama-linux-amd64.tgz

ollama serve

ollama -v

<!-- gh-comment-id:2899606869 --> @yujianfeikong commented on GitHub (May 22, 2025): > ``` > 5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.790+08:00 level=INFO source=server.go:168 > msg=offload library=cuda layers.requested=-1 layers.model=29 layers.offload=29 layers.split="" > memory.available="[22.4 GiB]" memory.gpu_overhead="0 B" memory.required.full="2.1 GiB" > memory.required.partial="2.1 GiB" memory.required.kv="224.0 MiB" memory.required.allocations="[2.1 GiB]" > memory.weights.total="934.7 MiB" memory.weights.repeating="752.1 MiB" memory.weights.nonrepeating="182.6 MiB" > memory.graph.full="299.8 MiB" memory.graph.partial="482.3 MiB" > ``` > > ollama wanted to offload all 29 layers. > > ``` > 5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.905+08:00 level=INFO source=ggml.go:104 > msg=system CPU.0.LLAMAFILE=1 compiler=cgo(gcc) > ``` > > But it didn't find any CPU or GPU backends, which would normally be in `/usr/local/lib/ollama`. How did you install ollama? The directory /usr/local/lib/ollama is currently empty and there are no files in it. curl -L https://ollama.com/download/ollama-linux-amd64.tgz -o ollama-linux-amd64.tgz ps -ef|grep ollama kill -9 xxxx sudo tar -C /usr -xzf ollama-linux-amd64.tgz ollama serve ollama -v
GiteaMirror commented 2026-04-29 04:10:02 -05:00
Author
Owner
Copy Link

@yujianfeikong commented on GitHub (May 22, 2025):

5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.790+08:00 level=INFO source=server.go:168
 msg=offload library=cuda layers.requested=-1 layers.model=29 layers.offload=29 layers.split=""
 memory.available="[22.4 GiB]" memory.gpu_overhead="0 B" memory.required.full="2.1 GiB"
 memory.required.partial="2.1 GiB" memory.required.kv="224.0 MiB" memory.required.allocations="[2.1 GiB]"
 memory.weights.total="934.7 MiB" memory.weights.repeating="752.1 MiB" memory.weights.nonrepeating="182.6 MiB"
 memory.graph.full="299.8 MiB" memory.graph.partial="482.3 MiB"

ollama wanted to offload all 29 layers.

5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.905+08:00 level=INFO source=ggml.go:104
 msg=system CPU.0.LLAMAFILE=1 compiler=cgo(gcc)

But it didn't find any CPU or GPU backends, which would normally be in /usr/local/lib/ollama. How did you install ollama?

By using the online upgrade command: curl -fsSL https://ollama.com/install.sh | sh
Console output: Cleaning up old version at /usr/local/lib/ollama

I think the problem has been found：
Because the directory /usr/local/lib/ollama was emptied, the GPU stopped working.
Previously, I used the online upgrade command, but force-stopped it due to slow speed.

Now, after using the online upgrade command and waiting for the upgrade to complete, the GPU is working properly, and the directory /usr/local/lib/ollama is no longer empty.

<!-- gh-comment-id:2900103862 --> @yujianfeikong commented on GitHub (May 22, 2025): > ``` > 5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.790+08:00 level=INFO source=server.go:168 > msg=offload library=cuda layers.requested=-1 layers.model=29 layers.offload=29 layers.split="" > memory.available="[22.4 GiB]" memory.gpu_overhead="0 B" memory.required.full="2.1 GiB" > memory.required.partial="2.1 GiB" memory.required.kv="224.0 MiB" memory.required.allocations="[2.1 GiB]" > memory.weights.total="934.7 MiB" memory.weights.repeating="752.1 MiB" memory.weights.nonrepeating="182.6 MiB" > memory.graph.full="299.8 MiB" memory.graph.partial="482.3 MiB" > ``` > > ollama wanted to offload all 29 layers. > > ``` > 5月 21 17:49:23 ubuntu ollama[1576]: time=2025-05-21T17:49:23.905+08:00 level=INFO source=ggml.go:104 > msg=system CPU.0.LLAMAFILE=1 compiler=cgo(gcc) > ``` > > But it didn't find any CPU or GPU backends, which would normally be in `/usr/local/lib/ollama`. How did you install ollama? By using the online upgrade command: curl -fsSL https://ollama.com/install.sh | sh Console output: Cleaning up old version at /usr/local/lib/ollama I think the problem has been found： Because the directory /usr/local/lib/ollama was emptied, the GPU stopped working. Previously, I used the online upgrade command, but force-stopped it due to slow speed. Now, after using the online upgrade command and waiting for the upgrade to complete, the GPU is working properly, and the directory /usr/local/lib/ollama is no longer empty.
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
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#53603
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?