github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-23 14:13:13 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #11499] Qwen3:32b behaving differently(sometimes CPU, sometimes GPU) #85281

New Issue
Open
opened 2026-05-09 22:57:32 -05:00 by GiteaMirror · 4 comments
GiteaMirror commented 2026-05-09 22:57:32 -05:00
Owner
Copy Link

Originally created by @HDANILO on GitHub (Jul 22, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/11499

What is the issue?

When I use Qwen3:32b with continue.dev vscode, to chat, ollama allocates 24gb on VRAM then 7gb on RAM, and then the language responds very slowly, but when I use the same Qwen3:32b with open webui, ollama allocates 100% on VRAM and answers pretty quickly.

Both are evoking deep thinking(as both show thinking tokens).

I don't run on this issue on Qwen2.5:32b

Relevant log output

time=2025-07-22T18:22:43.807-03:00 level=INFO source=routes.go:1235 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:4096 OLLAMA_DEBUG:INFO 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:C:\\Users\\helto\\.ollama\\models OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE: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://* vscode-file://*] OLLAMA_SCHED_SPREAD:false ROCR_VISIBLE_DEVICES:]"
time=2025-07-22T18:22:43.885-03:00 level=INFO source=images.go:476 msg="total blobs: 34"
time=2025-07-22T18:22:43.890-03:00 level=INFO source=images.go:483 msg="total unused blobs removed: 0"
time=2025-07-22T18:22:43.894-03:00 level=INFO source=routes.go:1288 msg="Listening on 127.0.0.1:11434 (version 0.9.6)"
time=2025-07-22T18:22:43.894-03:00 level=INFO source=gpu.go:217 msg="looking for compatible GPUs"
time=2025-07-22T18:22:43.894-03:00 level=INFO source=gpu_windows.go:167 msg=packages count=1
time=2025-07-22T18:22:43.894-03:00 level=INFO source=gpu_windows.go:214 msg="" package=0 cores=12 efficiency=0 threads=24
time=2025-07-22T18:22:44.126-03:00 level=INFO source=types.go:130 msg="inference compute" id=GPU-0a97eb13-83fe-20b1-3f8c-ceab8778ed3a library=cuda variant=v12 compute=12.0 driver=12.9 name="NVIDIA GeForce RTX 5090" total="31.8 GiB" available="30.1 GiB"
[GIN] 2025/07/22 - 18:45:26 | 200 |      7.0342ms |       127.0.0.1 | GET      "/api/tags"
[GIN] 2025/07/22 - 18:45:26 | 200 |      1.0085ms |       127.0.0.1 | GET      "/api/ps"
[GIN] 2025/07/22 - 18:45:27 | 200 |      1.0085ms |       127.0.0.1 | GET      "/api/version"
[GIN] 2025/07/22 - 18:45:38 | 200 |         6.5ms |       127.0.0.1 | GET      "/api/tags"
[GIN] 2025/07/22 - 18:45:38 | 200 |            0s |       127.0.0.1 | GET      "/api/ps"
time=2025-07-22T18:46:36.128-03:00 level=INFO source=sched.go:788 msg="new model will fit in available VRAM in single GPU, loading" model=C:\Users\helto\.ollama\models\blobs\sha256-3291abe70f16ee9682de7bfae08db5373ea9d6497e614aaad63340ad421d6312 gpu=GPU-0a97eb13-83fe-20b1-3f8c-ceab8778ed3a parallel=2 available=28937641984 required="23.8 GiB"
time=2025-07-22T18:46:36.139-03:00 level=INFO source=server.go:135 msg="system memory" total="79.9 GiB" free="56.6 GiB" free_swap="49.8 GiB"
time=2025-07-22T18:46:36.140-03:00 level=INFO source=server.go:175 msg=offload library=cuda layers.requested=-1 layers.model=65 layers.offload=65 layers.split="" memory.available="[27.0 GiB]" memory.gpu_overhead="0 B" memory.required.full="23.8 GiB" memory.required.partial="23.8 GiB" memory.required.kv="2.0 GiB" memory.required.allocations="[23.8 GiB]" memory.weights.total="18.4 GiB" memory.weights.repeating="17.8 GiB" memory.weights.nonrepeating="608.6 MiB" memory.graph.full="2.7 GiB" memory.graph.partial="2.7 GiB"
llama_model_loader: loaded meta data with 27 key-value pairs and 707 tensors from C:\Users\helto\.ollama\models\blobs\sha256-3291abe70f16ee9682de7bfae08db5373ea9d6497e614aaad63340ad421d6312 (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              = qwen3
llama_model_loader: - kv   1:                               general.type str              = model
llama_model_loader: - kv   2:                               general.name str              = Qwen3 32B
llama_model_loader: - kv   3:                           general.basename str              = Qwen3
llama_model_loader: - kv   4:                         general.size_label str              = 32B
llama_model_loader: - kv   5:                          qwen3.block_count u32              = 64
llama_model_loader: - kv   6:                       qwen3.context_length u32              = 40960
llama_model_loader: - kv   7:                     qwen3.embedding_length u32              = 5120
llama_model_loader: - kv   8:                  qwen3.feed_forward_length u32              = 25600
llama_model_loader: - kv   9:                 qwen3.attention.head_count u32              = 64
llama_model_loader: - kv  10:              qwen3.attention.head_count_kv u32              = 8
llama_model_loader: - kv  11:                       qwen3.rope.freq_base f32              = 1000000.000000
llama_model_loader: - kv  12:     qwen3.attention.layer_norm_rms_epsilon f32              = 0.000001
llama_model_loader: - kv  13:                 qwen3.attention.key_length u32              = 128
llama_model_loader: - kv  14:               qwen3.attention.value_length u32              = 128
llama_model_loader: - kv  15:                       tokenizer.ggml.model str              = gpt2
llama_model_loader: - kv  16:                         tokenizer.ggml.pre str              = qwen2
llama_model_loader: - kv  17:                      tokenizer.ggml.tokens arr[str,151936]  = ["!", "\"", "#", "$", "%", "&", "'", ...
llama_model_loader: - kv  18:                  tokenizer.ggml.token_type arr[i32,151936]  = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv  19:                      tokenizer.ggml.merges arr[str,151387]  = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t",...
llama_model_loader: - kv  20:                tokenizer.ggml.eos_token_id u32              = 151645
llama_model_loader: - kv  21:            tokenizer.ggml.padding_token_id u32              = 151643
llama_model_loader: - kv  22:                tokenizer.ggml.bos_token_id u32              = 151643
llama_model_loader: - kv  23:               tokenizer.ggml.add_bos_token bool             = false
llama_model_loader: - kv  24:                    tokenizer.chat_template str              = {%- if tools %}\n    {{- '<|im_start|>...
llama_model_loader: - kv  25:               general.quantization_version u32              = 2
llama_model_loader: - kv  26:                          general.file_type u32              = 15
llama_model_loader: - type  f32:  257 tensors
llama_model_loader: - type  f16:   64 tensors
llama_model_loader: - type q4_K:  353 tensors
llama_model_loader: - type q6_K:   33 tensors
print_info: file format = GGUF V3 (latest)
print_info: file type   = Q4_K - Medium
print_info: file size   = 18.81 GiB (4.93 BPW) 
load: special tokens cache size = 26
load: token to piece cache size = 0.9311 MB
print_info: arch             = qwen3
print_info: vocab_only       = 1
print_info: model type       = ?B
print_info: model params     = 32.76 B
print_info: general.name     = Qwen3 32B
print_info: vocab type       = BPE
print_info: n_vocab          = 151936
print_info: n_merges         = 151387
print_info: BOS token        = 151643 '<|endoftext|>'
print_info: EOS token        = 151645 '<|im_end|>'
print_info: EOT token        = 151645 '<|im_end|>'
print_info: PAD token        = 151643 '<|endoftext|>'
print_info: LF token         = 198 'Ċ'
print_info: FIM PRE token    = 151659 '<|fim_prefix|>'
print_info: FIM SUF token    = 151661 '<|fim_suffix|>'
print_info: FIM MID token    = 151660 '<|fim_middle|>'
print_info: FIM PAD token    = 151662 '<|fim_pad|>'
print_info: FIM REP token    = 151663 '<|repo_name|>'
print_info: FIM SEP token    = 151664 '<|file_sep|>'
print_info: EOG token        = 151643 '<|endoftext|>'
print_info: EOG token        = 151645 '<|im_end|>'
print_info: EOG token        = 151662 '<|fim_pad|>'
print_info: EOG token        = 151663 '<|repo_name|>'
print_info: EOG token        = 151664 '<|file_sep|>'
print_info: max token length = 256
llama_model_load: vocab only - skipping tensors
time=2025-07-22T18:46:36.412-03:00 level=INFO source=server.go:438 msg="starting llama server" cmd="C:\\Users\\helto\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --model C:\\Users\\helto\\.ollama\\models\\blobs\\sha256-3291abe70f16ee9682de7bfae08db5373ea9d6497e614aaad63340ad421d6312 --ctx-size 8192 --batch-size 512 --n-gpu-layers 65 --threads 12 --no-mmap --parallel 2 --port 60614"
time=2025-07-22T18:46:36.425-03:00 level=INFO source=sched.go:483 msg="loaded runners" count=1
time=2025-07-22T18:46:36.425-03:00 level=INFO source=server.go:598 msg="waiting for llama runner to start responding"
time=2025-07-22T18:46:36.427-03:00 level=INFO source=server.go:632 msg="waiting for server to become available" status="llm server error"
time=2025-07-22T18:46:36.528-03:00 level=INFO source=runner.go:815 msg="starting go runner"
ggml_cuda_init: GGML_CUDA_FORCE_MMQ:    no
ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no
ggml_cuda_init: found 1 CUDA devices:
  Device 0: NVIDIA GeForce RTX 5090, compute capability 12.0, VMM: yes
load_backend: loaded CUDA backend from C:\Users\helto\AppData\Local\Programs\Ollama\lib\ollama\ggml-cuda.dll
load_backend: loaded CPU backend from C:\Users\helto\AppData\Local\Programs\Ollama\lib\ollama\ggml-cpu-haswell.dll
time=2025-07-22T18:46:36.762-03:00 level=INFO source=ggml.go:104 msg=system CPU.0.SSE3=1 CPU.0.SSSE3=1 CPU.0.AVX=1 CPU.0.AVX2=1 CPU.0.F16C=1 CPU.0.FMA=1 CPU.0.BMI2=1 CPU.0.LLAMAFILE=1 CPU.1.LLAMAFILE=1 CUDA.0.ARCHS=500,600,610,700,750,800,860,870,890,900,1200 CUDA.0.USE_GRAPHS=1 CUDA.0.PEER_MAX_BATCH_SIZE=128 compiler=cgo(clang)
time=2025-07-22T18:46:36.765-03:00 level=INFO source=runner.go:874 msg="Server listening on 127.0.0.1:60614"
llama_model_load_from_file_impl: using device CUDA0 (NVIDIA GeForce RTX 5090) - 30842 MiB free
llama_model_loader: loaded meta data with 27 key-value pairs and 707 tensors from C:\Users\helto\.ollama\models\blobs\sha256-3291abe70f16ee9682de7bfae08db5373ea9d6497e614aaad63340ad421d6312 (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              = qwen3
llama_model_loader: - kv   1:                               general.type str              = model
llama_model_loader: - kv   2:                               general.name str              = Qwen3 32B
llama_model_loader: - kv   3:                           general.basename str              = Qwen3
llama_model_loader: - kv   4:                         general.size_label str              = 32B
llama_model_loader: - kv   5:                          qwen3.block_count u32              = 64
llama_model_loader: - kv   6:                       qwen3.context_length u32              = 40960
llama_model_loader: - kv   7:                     qwen3.embedding_length u32              = 5120
llama_model_loader: - kv   8:                  qwen3.feed_forward_length u32              = 25600
llama_model_loader: - kv   9:                 qwen3.attention.head_count u32              = 64
llama_model_loader: - kv  10:              qwen3.attention.head_count_kv u32              = 8
llama_model_loader: - kv  11:                       qwen3.rope.freq_base f32              = 1000000.000000
llama_model_loader: - kv  12:     qwen3.attention.layer_norm_rms_epsilon f32              = 0.000001
llama_model_loader: - kv  13:                 qwen3.attention.key_length u32              = 128
llama_model_loader: - kv  14:               qwen3.attention.value_length u32              = 128
llama_model_loader: - kv  15:                       tokenizer.ggml.model str              = gpt2
llama_model_loader: - kv  16:                         tokenizer.ggml.pre str              = qwen2
time=2025-07-22T18:46:36.930-03:00 level=INFO source=server.go:632 msg="waiting for server to become available" status="llm server loading model"
llama_model_loader: - kv  17:                      tokenizer.ggml.tokens arr[str,151936]  = ["!", "\"", "#", "$", "%", "&", "'", ...
llama_model_loader: - kv  18:                  tokenizer.ggml.token_type arr[i32,151936]  = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv  19:                      tokenizer.ggml.merges arr[str,151387]  = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t",...
llama_model_loader: - kv  20:                tokenizer.ggml.eos_token_id u32              = 151645
llama_model_loader: - kv  21:            tokenizer.ggml.padding_token_id u32              = 151643
llama_model_loader: - kv  22:                tokenizer.ggml.bos_token_id u32              = 151643
llama_model_loader: - kv  23:               tokenizer.ggml.add_bos_token bool             = false
llama_model_loader: - kv  24:                    tokenizer.chat_template str              = {%- if tools %}\n    {{- '<|im_start|>...
llama_model_loader: - kv  25:               general.quantization_version u32              = 2
llama_model_loader: - kv  26:                          general.file_type u32              = 15
llama_model_loader: - type  f32:  257 tensors
llama_model_loader: - type  f16:   64 tensors
llama_model_loader: - type q4_K:  353 tensors
llama_model_loader: - type q6_K:   33 tensors
print_info: file format = GGUF V3 (latest)
print_info: file type   = Q4_K - Medium
print_info: file size   = 18.81 GiB (4.93 BPW) 
load: special tokens cache size = 26
load: token to piece cache size = 0.9311 MB
print_info: arch             = qwen3
print_info: vocab_only       = 0
print_info: n_ctx_train      = 40960
print_info: n_embd           = 5120
print_info: n_layer          = 64
print_info: n_head           = 64
print_info: n_head_kv        = 8
print_info: n_rot            = 128
print_info: n_swa            = 0
print_info: n_swa_pattern    = 1
print_info: n_embd_head_k    = 128
print_info: n_embd_head_v    = 128
print_info: n_gqa            = 8
print_info: n_embd_k_gqa     = 1024
print_info: n_embd_v_gqa     = 1024
print_info: f_norm_eps       = 0.0e+00
print_info: f_norm_rms_eps   = 1.0e-06
print_info: f_clamp_kqv      = 0.0e+00
print_info: f_max_alibi_bias = 0.0e+00
print_info: f_logit_scale    = 0.0e+00
print_info: f_attn_scale     = 0.0e+00
print_info: n_ff             = 25600
print_info: n_expert         = 0
print_info: n_expert_used    = 0
print_info: causal attn      = 1
print_info: pooling type     = 0
print_info: rope type        = 2
print_info: rope scaling     = linear
print_info: freq_base_train  = 1000000.0
print_info: freq_scale_train = 1
print_info: n_ctx_orig_yarn  = 40960
print_info: rope_finetuned   = unknown
print_info: ssm_d_conv       = 0
print_info: ssm_d_inner      = 0
print_info: ssm_d_state      = 0
print_info: ssm_dt_rank      = 0
print_info: ssm_dt_b_c_rms   = 0
print_info: model type       = 32B
print_info: model params     = 32.76 B
print_info: general.name     = Qwen3 32B
print_info: vocab type       = BPE
print_info: n_vocab          = 151936
print_info: n_merges         = 151387
print_info: BOS token        = 151643 '<|endoftext|>'
print_info: EOS token        = 151645 '<|im_end|>'
print_info: EOT token        = 151645 '<|im_end|>'
print_info: PAD token        = 151643 '<|endoftext|>'
print_info: LF token         = 198 'Ċ'
print_info: FIM PRE token    = 151659 '<|fim_prefix|>'
print_info: FIM SUF token    = 151661 '<|fim_suffix|>'
print_info: FIM MID token    = 151660 '<|fim_middle|>'
print_info: FIM PAD token    = 151662 '<|fim_pad|>'
print_info: FIM REP token    = 151663 '<|repo_name|>'
print_info: FIM SEP token    = 151664 '<|file_sep|>'
print_info: EOG token        = 151643 '<|endoftext|>'
print_info: EOG token        = 151645 '<|im_end|>'
print_info: EOG token        = 151662 '<|fim_pad|>'
print_info: EOG token        = 151663 '<|repo_name|>'
print_info: EOG token        = 151664 '<|file_sep|>'
print_info: max token length = 256
load_tensors: loading model tensors, this can take a while... (mmap = false)
load_tensors: offloading 64 repeating layers to GPU
load_tensors: offloading output layer to GPU
load_tensors: offloaded 65/65 layers to GPU
load_tensors:        CUDA0 model buffer size = 18842.40 MiB
load_tensors:          CPU model buffer size =   417.30 MiB
llama_context: constructing llama_context
llama_context: n_seq_max     = 2
llama_context: n_ctx         = 8192
llama_context: n_ctx_per_seq = 4096
llama_context: n_batch       = 1024
llama_context: n_ubatch      = 512
llama_context: causal_attn   = 1
llama_context: flash_attn    = 0
llama_context: freq_base     = 1000000.0
llama_context: freq_scale    = 1
llama_context: n_ctx_per_seq (4096) < n_ctx_train (40960) -- the full capacity of the model will not be utilized
llama_context:  CUDA_Host  output buffer size =     1.20 MiB
llama_kv_cache_unified: kv_size = 8192, type_k = 'f16', type_v = 'f16', n_layer = 64, can_shift = 1, padding = 32
llama_kv_cache_unified:      CUDA0 KV buffer size =  2048.00 MiB
llama_kv_cache_unified: KV self size  = 2048.00 MiB, K (f16): 1024.00 MiB, V (f16): 1024.00 MiB
llama_context:      CUDA0 compute buffer size =  1092.00 MiB
llama_context:  CUDA_Host compute buffer size =    26.01 MiB
llama_context: graph nodes  = 2438
llama_context: graph splits = 2
time=2025-07-22T18:46:41.694-03:00 level=INFO source=server.go:637 msg="llama runner started in 5.27 seconds"
[GIN] 2025/07/22 - 18:46:41 | 200 |      7.0025ms |       127.0.0.1 | GET      "/api/tags"
[GIN] 2025/07/22 - 18:46:41 | 200 |            0s |       127.0.0.1 | GET      "/api/ps"
[GIN] 2025/07/22 - 18:47:14 | 200 |   38.0834337s |       127.0.0.1 | POST     "/api/chat"
[GIN] 2025/07/22 - 18:47:25 | 200 |   11.2045465s |       127.0.0.1 | POST     "/api/chat"
[GIN] 2025/07/22 - 18:47:40 | 200 |   15.1240835s |       127.0.0.1 | POST     "/api/chat"
[GIN] 2025/07/22 - 18:47:46 | 200 |     5.414632s |       127.0.0.1 | POST     "/api/chat"
[GIN] 2025/07/22 - 18:49:07 | 200 |   47.9151759s |       127.0.0.1 | POST     "/api/chat"
[GIN] 2025/07/22 - 18:49:21 | 200 |   14.1584324s |       127.0.0.1 | POST     "/api/chat"

OS

Windows

GPU

Nvidia (Blackwell 5090)

CPU

AMD

Ollama version

0.10.0-rc0

Originally created by @HDANILO on GitHub (Jul 22, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/11499 ### What is the issue? When I use Qwen3:32b with continue.dev vscode, to chat, ollama allocates 24gb on VRAM then 7gb on RAM, and then the language responds very slowly, but when I use the same Qwen3:32b with open webui, ollama allocates 100% on VRAM and answers pretty quickly. Both are evoking deep thinking(as both show thinking tokens). I don't run on this issue on Qwen2.5:32b ### Relevant log output ```shell time=2025-07-22T18:22:43.807-03:00 level=INFO source=routes.go:1235 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:4096 OLLAMA_DEBUG:INFO 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:C:\\Users\\helto\\.ollama\\models OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE: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://* vscode-file://*] OLLAMA_SCHED_SPREAD:false ROCR_VISIBLE_DEVICES:]" time=2025-07-22T18:22:43.885-03:00 level=INFO source=images.go:476 msg="total blobs: 34" time=2025-07-22T18:22:43.890-03:00 level=INFO source=images.go:483 msg="total unused blobs removed: 0" time=2025-07-22T18:22:43.894-03:00 level=INFO source=routes.go:1288 msg="Listening on 127.0.0.1:11434 (version 0.9.6)" time=2025-07-22T18:22:43.894-03:00 level=INFO source=gpu.go:217 msg="looking for compatible GPUs" time=2025-07-22T18:22:43.894-03:00 level=INFO source=gpu_windows.go:167 msg=packages count=1 time=2025-07-22T18:22:43.894-03:00 level=INFO source=gpu_windows.go:214 msg="" package=0 cores=12 efficiency=0 threads=24 time=2025-07-22T18:22:44.126-03:00 level=INFO source=types.go:130 msg="inference compute" id=GPU-0a97eb13-83fe-20b1-3f8c-ceab8778ed3a library=cuda variant=v12 compute=12.0 driver=12.9 name="NVIDIA GeForce RTX 5090" total="31.8 GiB" available="30.1 GiB" [GIN] 2025/07/22 - 18:45:26 | 200 | 7.0342ms | 127.0.0.1 | GET "/api/tags" [GIN] 2025/07/22 - 18:45:26 | 200 | 1.0085ms | 127.0.0.1 | GET "/api/ps" [GIN] 2025/07/22 - 18:45:27 | 200 | 1.0085ms | 127.0.0.1 | GET "/api/version" [GIN] 2025/07/22 - 18:45:38 | 200 | 6.5ms | 127.0.0.1 | GET "/api/tags" [GIN] 2025/07/22 - 18:45:38 | 200 | 0s | 127.0.0.1 | GET "/api/ps" time=2025-07-22T18:46:36.128-03:00 level=INFO source=sched.go:788 msg="new model will fit in available VRAM in single GPU, loading" model=C:\Users\helto\.ollama\models\blobs\sha256-3291abe70f16ee9682de7bfae08db5373ea9d6497e614aaad63340ad421d6312 gpu=GPU-0a97eb13-83fe-20b1-3f8c-ceab8778ed3a parallel=2 available=28937641984 required="23.8 GiB" time=2025-07-22T18:46:36.139-03:00 level=INFO source=server.go:135 msg="system memory" total="79.9 GiB" free="56.6 GiB" free_swap="49.8 GiB" time=2025-07-22T18:46:36.140-03:00 level=INFO source=server.go:175 msg=offload library=cuda layers.requested=-1 layers.model=65 layers.offload=65 layers.split="" memory.available="[27.0 GiB]" memory.gpu_overhead="0 B" memory.required.full="23.8 GiB" memory.required.partial="23.8 GiB" memory.required.kv="2.0 GiB" memory.required.allocations="[23.8 GiB]" memory.weights.total="18.4 GiB" memory.weights.repeating="17.8 GiB" memory.weights.nonrepeating="608.6 MiB" memory.graph.full="2.7 GiB" memory.graph.partial="2.7 GiB" llama_model_loader: loaded meta data with 27 key-value pairs and 707 tensors from C:\Users\helto\.ollama\models\blobs\sha256-3291abe70f16ee9682de7bfae08db5373ea9d6497e614aaad63340ad421d6312 (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 = qwen3 llama_model_loader: - kv 1: general.type str = model llama_model_loader: - kv 2: general.name str = Qwen3 32B llama_model_loader: - kv 3: general.basename str = Qwen3 llama_model_loader: - kv 4: general.size_label str = 32B llama_model_loader: - kv 5: qwen3.block_count u32 = 64 llama_model_loader: - kv 6: qwen3.context_length u32 = 40960 llama_model_loader: - kv 7: qwen3.embedding_length u32 = 5120 llama_model_loader: - kv 8: qwen3.feed_forward_length u32 = 25600 llama_model_loader: - kv 9: qwen3.attention.head_count u32 = 64 llama_model_loader: - kv 10: qwen3.attention.head_count_kv u32 = 8 llama_model_loader: - kv 11: qwen3.rope.freq_base f32 = 1000000.000000 llama_model_loader: - kv 12: qwen3.attention.layer_norm_rms_epsilon f32 = 0.000001 llama_model_loader: - kv 13: qwen3.attention.key_length u32 = 128 llama_model_loader: - kv 14: qwen3.attention.value_length u32 = 128 llama_model_loader: - kv 15: tokenizer.ggml.model str = gpt2 llama_model_loader: - kv 16: tokenizer.ggml.pre str = qwen2 llama_model_loader: - kv 17: tokenizer.ggml.tokens arr[str,151936] = ["!", "\"", "#", "$", "%", "&", "'", ... llama_model_loader: - kv 18: tokenizer.ggml.token_type arr[i32,151936] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ... llama_model_loader: - kv 19: tokenizer.ggml.merges arr[str,151387] = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t",... llama_model_loader: - kv 20: tokenizer.ggml.eos_token_id u32 = 151645 llama_model_loader: - kv 21: tokenizer.ggml.padding_token_id u32 = 151643 llama_model_loader: - kv 22: tokenizer.ggml.bos_token_id u32 = 151643 llama_model_loader: - kv 23: tokenizer.ggml.add_bos_token bool = false llama_model_loader: - kv 24: tokenizer.chat_template str = {%- if tools %}\n {{- '<|im_start|>... llama_model_loader: - kv 25: general.quantization_version u32 = 2 llama_model_loader: - kv 26: general.file_type u32 = 15 llama_model_loader: - type f32: 257 tensors llama_model_loader: - type f16: 64 tensors llama_model_loader: - type q4_K: 353 tensors llama_model_loader: - type q6_K: 33 tensors print_info: file format = GGUF V3 (latest) print_info: file type = Q4_K - Medium print_info: file size = 18.81 GiB (4.93 BPW) load: special tokens cache size = 26 load: token to piece cache size = 0.9311 MB print_info: arch = qwen3 print_info: vocab_only = 1 print_info: model type = ?B print_info: model params = 32.76 B print_info: general.name = Qwen3 32B print_info: vocab type = BPE print_info: n_vocab = 151936 print_info: n_merges = 151387 print_info: BOS token = 151643 '<|endoftext|>' print_info: EOS token = 151645 '<|im_end|>' print_info: EOT token = 151645 '<|im_end|>' print_info: PAD token = 151643 '<|endoftext|>' print_info: LF token = 198 'Ċ' print_info: FIM PRE token = 151659 '<|fim_prefix|>' print_info: FIM SUF token = 151661 '<|fim_suffix|>' print_info: FIM MID token = 151660 '<|fim_middle|>' print_info: FIM PAD token = 151662 '<|fim_pad|>' print_info: FIM REP token = 151663 '<|repo_name|>' print_info: FIM SEP token = 151664 '<|file_sep|>' print_info: EOG token = 151643 '<|endoftext|>' print_info: EOG token = 151645 '<|im_end|>' print_info: EOG token = 151662 '<|fim_pad|>' print_info: EOG token = 151663 '<|repo_name|>' print_info: EOG token = 151664 '<|file_sep|>' print_info: max token length = 256 llama_model_load: vocab only - skipping tensors time=2025-07-22T18:46:36.412-03:00 level=INFO source=server.go:438 msg="starting llama server" cmd="C:\\Users\\helto\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --model C:\\Users\\helto\\.ollama\\models\\blobs\\sha256-3291abe70f16ee9682de7bfae08db5373ea9d6497e614aaad63340ad421d6312 --ctx-size 8192 --batch-size 512 --n-gpu-layers 65 --threads 12 --no-mmap --parallel 2 --port 60614" time=2025-07-22T18:46:36.425-03:00 level=INFO source=sched.go:483 msg="loaded runners" count=1 time=2025-07-22T18:46:36.425-03:00 level=INFO source=server.go:598 msg="waiting for llama runner to start responding" time=2025-07-22T18:46:36.427-03:00 level=INFO source=server.go:632 msg="waiting for server to become available" status="llm server error" time=2025-07-22T18:46:36.528-03:00 level=INFO source=runner.go:815 msg="starting go runner" ggml_cuda_init: GGML_CUDA_FORCE_MMQ: no ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no ggml_cuda_init: found 1 CUDA devices: Device 0: NVIDIA GeForce RTX 5090, compute capability 12.0, VMM: yes load_backend: loaded CUDA backend from C:\Users\helto\AppData\Local\Programs\Ollama\lib\ollama\ggml-cuda.dll load_backend: loaded CPU backend from C:\Users\helto\AppData\Local\Programs\Ollama\lib\ollama\ggml-cpu-haswell.dll time=2025-07-22T18:46:36.762-03:00 level=INFO source=ggml.go:104 msg=system CPU.0.SSE3=1 CPU.0.SSSE3=1 CPU.0.AVX=1 CPU.0.AVX2=1 CPU.0.F16C=1 CPU.0.FMA=1 CPU.0.BMI2=1 CPU.0.LLAMAFILE=1 CPU.1.LLAMAFILE=1 CUDA.0.ARCHS=500,600,610,700,750,800,860,870,890,900,1200 CUDA.0.USE_GRAPHS=1 CUDA.0.PEER_MAX_BATCH_SIZE=128 compiler=cgo(clang) time=2025-07-22T18:46:36.765-03:00 level=INFO source=runner.go:874 msg="Server listening on 127.0.0.1:60614" llama_model_load_from_file_impl: using device CUDA0 (NVIDIA GeForce RTX 5090) - 30842 MiB free llama_model_loader: loaded meta data with 27 key-value pairs and 707 tensors from C:\Users\helto\.ollama\models\blobs\sha256-3291abe70f16ee9682de7bfae08db5373ea9d6497e614aaad63340ad421d6312 (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 = qwen3 llama_model_loader: - kv 1: general.type str = model llama_model_loader: - kv 2: general.name str = Qwen3 32B llama_model_loader: - kv 3: general.basename str = Qwen3 llama_model_loader: - kv 4: general.size_label str = 32B llama_model_loader: - kv 5: qwen3.block_count u32 = 64 llama_model_loader: - kv 6: qwen3.context_length u32 = 40960 llama_model_loader: - kv 7: qwen3.embedding_length u32 = 5120 llama_model_loader: - kv 8: qwen3.feed_forward_length u32 = 25600 llama_model_loader: - kv 9: qwen3.attention.head_count u32 = 64 llama_model_loader: - kv 10: qwen3.attention.head_count_kv u32 = 8 llama_model_loader: - kv 11: qwen3.rope.freq_base f32 = 1000000.000000 llama_model_loader: - kv 12: qwen3.attention.layer_norm_rms_epsilon f32 = 0.000001 llama_model_loader: - kv 13: qwen3.attention.key_length u32 = 128 llama_model_loader: - kv 14: qwen3.attention.value_length u32 = 128 llama_model_loader: - kv 15: tokenizer.ggml.model str = gpt2 llama_model_loader: - kv 16: tokenizer.ggml.pre str = qwen2 time=2025-07-22T18:46:36.930-03:00 level=INFO source=server.go:632 msg="waiting for server to become available" status="llm server loading model" llama_model_loader: - kv 17: tokenizer.ggml.tokens arr[str,151936] = ["!", "\"", "#", "$", "%", "&", "'", ... llama_model_loader: - kv 18: tokenizer.ggml.token_type arr[i32,151936] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ... llama_model_loader: - kv 19: tokenizer.ggml.merges arr[str,151387] = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t",... llama_model_loader: - kv 20: tokenizer.ggml.eos_token_id u32 = 151645 llama_model_loader: - kv 21: tokenizer.ggml.padding_token_id u32 = 151643 llama_model_loader: - kv 22: tokenizer.ggml.bos_token_id u32 = 151643 llama_model_loader: - kv 23: tokenizer.ggml.add_bos_token bool = false llama_model_loader: - kv 24: tokenizer.chat_template str = {%- if tools %}\n {{- '<|im_start|>... llama_model_loader: - kv 25: general.quantization_version u32 = 2 llama_model_loader: - kv 26: general.file_type u32 = 15 llama_model_loader: - type f32: 257 tensors llama_model_loader: - type f16: 64 tensors llama_model_loader: - type q4_K: 353 tensors llama_model_loader: - type q6_K: 33 tensors print_info: file format = GGUF V3 (latest) print_info: file type = Q4_K - Medium print_info: file size = 18.81 GiB (4.93 BPW) load: special tokens cache size = 26 load: token to piece cache size = 0.9311 MB print_info: arch = qwen3 print_info: vocab_only = 0 print_info: n_ctx_train = 40960 print_info: n_embd = 5120 print_info: n_layer = 64 print_info: n_head = 64 print_info: n_head_kv = 8 print_info: n_rot = 128 print_info: n_swa = 0 print_info: n_swa_pattern = 1 print_info: n_embd_head_k = 128 print_info: n_embd_head_v = 128 print_info: n_gqa = 8 print_info: n_embd_k_gqa = 1024 print_info: n_embd_v_gqa = 1024 print_info: f_norm_eps = 0.0e+00 print_info: f_norm_rms_eps = 1.0e-06 print_info: f_clamp_kqv = 0.0e+00 print_info: f_max_alibi_bias = 0.0e+00 print_info: f_logit_scale = 0.0e+00 print_info: f_attn_scale = 0.0e+00 print_info: n_ff = 25600 print_info: n_expert = 0 print_info: n_expert_used = 0 print_info: causal attn = 1 print_info: pooling type = 0 print_info: rope type = 2 print_info: rope scaling = linear print_info: freq_base_train = 1000000.0 print_info: freq_scale_train = 1 print_info: n_ctx_orig_yarn = 40960 print_info: rope_finetuned = unknown print_info: ssm_d_conv = 0 print_info: ssm_d_inner = 0 print_info: ssm_d_state = 0 print_info: ssm_dt_rank = 0 print_info: ssm_dt_b_c_rms = 0 print_info: model type = 32B print_info: model params = 32.76 B print_info: general.name = Qwen3 32B print_info: vocab type = BPE print_info: n_vocab = 151936 print_info: n_merges = 151387 print_info: BOS token = 151643 '<|endoftext|>' print_info: EOS token = 151645 '<|im_end|>' print_info: EOT token = 151645 '<|im_end|>' print_info: PAD token = 151643 '<|endoftext|>' print_info: LF token = 198 'Ċ' print_info: FIM PRE token = 151659 '<|fim_prefix|>' print_info: FIM SUF token = 151661 '<|fim_suffix|>' print_info: FIM MID token = 151660 '<|fim_middle|>' print_info: FIM PAD token = 151662 '<|fim_pad|>' print_info: FIM REP token = 151663 '<|repo_name|>' print_info: FIM SEP token = 151664 '<|file_sep|>' print_info: EOG token = 151643 '<|endoftext|>' print_info: EOG token = 151645 '<|im_end|>' print_info: EOG token = 151662 '<|fim_pad|>' print_info: EOG token = 151663 '<|repo_name|>' print_info: EOG token = 151664 '<|file_sep|>' print_info: max token length = 256 load_tensors: loading model tensors, this can take a while... (mmap = false) load_tensors: offloading 64 repeating layers to GPU load_tensors: offloading output layer to GPU load_tensors: offloaded 65/65 layers to GPU load_tensors: CUDA0 model buffer size = 18842.40 MiB load_tensors: CPU model buffer size = 417.30 MiB llama_context: constructing llama_context llama_context: n_seq_max = 2 llama_context: n_ctx = 8192 llama_context: n_ctx_per_seq = 4096 llama_context: n_batch = 1024 llama_context: n_ubatch = 512 llama_context: causal_attn = 1 llama_context: flash_attn = 0 llama_context: freq_base = 1000000.0 llama_context: freq_scale = 1 llama_context: n_ctx_per_seq (4096) < n_ctx_train (40960) -- the full capacity of the model will not be utilized llama_context: CUDA_Host output buffer size = 1.20 MiB llama_kv_cache_unified: kv_size = 8192, type_k = 'f16', type_v = 'f16', n_layer = 64, can_shift = 1, padding = 32 llama_kv_cache_unified: CUDA0 KV buffer size = 2048.00 MiB llama_kv_cache_unified: KV self size = 2048.00 MiB, K (f16): 1024.00 MiB, V (f16): 1024.00 MiB llama_context: CUDA0 compute buffer size = 1092.00 MiB llama_context: CUDA_Host compute buffer size = 26.01 MiB llama_context: graph nodes = 2438 llama_context: graph splits = 2 time=2025-07-22T18:46:41.694-03:00 level=INFO source=server.go:637 msg="llama runner started in 5.27 seconds" [GIN] 2025/07/22 - 18:46:41 | 200 | 7.0025ms | 127.0.0.1 | GET "/api/tags" [GIN] 2025/07/22 - 18:46:41 | 200 | 0s | 127.0.0.1 | GET "/api/ps" [GIN] 2025/07/22 - 18:47:14 | 200 | 38.0834337s | 127.0.0.1 | POST "/api/chat" [GIN] 2025/07/22 - 18:47:25 | 200 | 11.2045465s | 127.0.0.1 | POST "/api/chat" [GIN] 2025/07/22 - 18:47:40 | 200 | 15.1240835s | 127.0.0.1 | POST "/api/chat" [GIN] 2025/07/22 - 18:47:46 | 200 | 5.414632s | 127.0.0.1 | POST "/api/chat" [GIN] 2025/07/22 - 18:49:07 | 200 | 47.9151759s | 127.0.0.1 | POST "/api/chat" [GIN] 2025/07/22 - 18:49:21 | 200 | 14.1584324s | 127.0.0.1 | POST "/api/chat" ``` ### OS Windows ### GPU Nvidia (Blackwell 5090) ### CPU AMD ### Ollama version 0.10.0-rc0
GiteaMirror added the bugmemory labels 2026-05-09 22:57:32 -05:00
GiteaMirror commented 2026-05-09 22:57:34 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Jul 22, 2025):

continue.dev is likely setting a larger context window, which causes layers to be offloaded to the CPU. Look for n_ctx_per_seq in the log.

<!-- gh-comment-id:3105033953 --> @rick-github commented on GitHub (Jul 22, 2025): continue.dev is likely setting a larger context window, which causes layers to be offloaded to the CPU. Look for `n_ctx_per_seq` in the log.
GiteaMirror commented 2026-05-09 22:57:34 -05:00
Author
Owner
Copy Link

@HDANILO commented on GitHub (Jul 22, 2025):

You are right, thanks for the clue, played a little bit with the context length and max tokens, and I'm seeing this situation:

Image

It's visible it fits, but it decides to offload to RAM anyway

<!-- gh-comment-id:3105169230 --> @HDANILO commented on GitHub (Jul 22, 2025): You are right, thanks for the clue, played a little bit with the context length and max tokens, and I'm seeing this situation: <img width="2020" height="617" alt="Image" src="https://github.com/user-attachments/assets/8f8a7d2f-9027-484e-ae70-77feaa1f92aa" /> It's visible it fits, but it decides to offload to RAM anyway
GiteaMirror commented 2026-05-09 22:57:34 -05:00
Author
Owner
Copy Link

@pdevine commented on GitHub (Jul 25, 2025):

cc @jessegross

<!-- gh-comment-id:3120440967 --> @pdevine commented on GitHub (Jul 25, 2025): cc @jessegross
GiteaMirror commented 2026-05-09 22:57:35 -05:00
Author
Owner
Copy Link

@luohao-svg commented on GitHub (Sep 4, 2025):

When I output extremely long content from the database to the Qwen3 model, the model's pre-allocation mechanism consumes a large amount of memory and a small amount of VRAM, which significantly slows down the speed. Its pre-allocation mechanism is highly problematic.

<!-- gh-comment-id:3251568685 --> @luohao-svg commented on GitHub (Sep 4, 2025): When I output extremely long content from the database to the Qwen3 model, the model's pre-allocation mechanism consumes a large amount of memory and a small amount of VRAM, which significantly slows down the speed. Its pre-allocation mechanism is highly problematic.
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 memory
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#85281
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?