github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 00:22:43 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #10832] unsloth/Qwen3-32B-128K-GGUF can not use multi GPU #7113

New Issue
Closed
opened 2026-04-12 19:06:10 -05:00 by GiteaMirror · 2 comments
GiteaMirror commented 2026-04-12 19:06:10 -05:00
Owner
Copy Link

Originally created by @adamhj on GitHub (May 23, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/10832

What is the issue?

Today I'm trying to run unsloth/Qwen3-32B-128K-GGUF(Q6_K) on ollama, on a machine has 24G×6 nvidia A10, but after the model was loaded, I found that it was running 100% on CPU

unsloth/Qwen3:32B-128K-Q6_K    3ddf246c242f    60 GB    100% CPU     4 minutes from now

I can run the official Qwen3:32B-Q6_K model on the same machine without any problem.

I suspect the problem is that the 128k model uses RoPE Scaling and cannot utilize multiple GPUs. Indeed, I also tried running unsloth/Qwen3-8B-128K-GGUF(Q6_K) on the same machine. It ran, but I noticed that only the first GPU was utilized, with its VRAM nearly full, while the others remained empty. This is not the case when I run other models, where the VRAM is distributed evenly across all six cards.

Relevant log output

[root@126gpu-10 Qwen3-30B-A3B-GGUF]# docker logs --tail 10 -f ollama
[GIN] 2025/05/23 - 05:46:03 | 200 |         9m31s | 10.3.2.153 | POST     "/v1/chat/completions"
time=2025-05-23T05:51:09.589Z level=WARN source=sched.go:655 msg="gpu VRAM usage didn't recover within timeout" seconds=6.462163498 model=/root/.ollama/models/blobs/sha256-6b818b3bbaee2722c6201aa17fc4c0733da71bbb12b7f77abfd75e5c0cb98e08
time=2025-05-23T05:51:11.920Z level=WARN source=sched.go:655 msg="gpu VRAM usage didn't recover within timeout" seconds=8.793217106 model=/root/.ollama/models/blobs/sha256-6b818b3bbaee2722c6201aa17fc4c0733da71bbb12b7f77abfd75e5c0cb98e08
time=2025-05-23T05:51:14.104Z level=WARN source=sched.go:655 msg="gpu VRAM usage didn't recover within timeout" seconds=10.977573067 model=/root/.ollama/models/blobs/sha256-6b818b3bbaee2722c6201aa17fc4c0733da71bbb12b7f77abfd75e5c0cb98e08
[GIN] 2025/05/23 - 06:34:00 | 200 |      70.738µs |       127.0.0.1 | HEAD     "/"
[GIN] 2025/05/23 - 06:34:00 | 200 |      36.864µs |       127.0.0.1 | GET      "/api/ps"
[GIN] 2025/05/23 - 10:43:24 | 200 |      64.701µs |       127.0.0.1 | HEAD     "/"
[GIN] 2025/05/23 - 10:43:24 | 200 |      59.119µs |       127.0.0.1 | GET      "/api/ps"
[GIN] 2025/05/23 - 10:44:55 | 200 |      59.305µs |       127.0.0.1 | HEAD     "/"
[GIN] 2025/05/23 - 10:44:55 | 200 |      69.853µs |       127.0.0.1 | GET      "/api/ps"
time=2025-05-23T10:46:01.917Z level=WARN source=ggml.go:152 msg="key not found" key=general.alignment default=32
time=2025-05-23T10:46:05.808Z level=WARN source=ggml.go:152 msg="key not found" key=general.alignment default=32
time=2025-05-23T10:46:05.835Z level=WARN source=ggml.go:152 msg="key not found" key=general.alignment default=32
time=2025-05-23T10:46:05.836Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0
time=2025-05-23T10:46:05.839Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0
time=2025-05-23T10:46:05.842Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0
time=2025-05-23T10:46:05.845Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0
time=2025-05-23T10:46:05.847Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0
time=2025-05-23T10:46:05.850Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0
time=2025-05-23T10:46:05.852Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0
time=2025-05-23T10:46:05.855Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0
time=2025-05-23T10:46:05.857Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0
time=2025-05-23T10:46:05.860Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0
time=2025-05-23T10:46:05.863Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0
time=2025-05-23T10:46:05.865Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0
time=2025-05-23T10:46:05.868Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0
time=2025-05-23T10:46:05.870Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0
time=2025-05-23T10:46:08.223Z level=INFO source=server.go:106 msg="system memory" total="377.1 GiB" free="366.6 GiB" free_swap="15.5 GiB"
time=2025-05-23T10:46:08.224Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0
time=2025-05-23T10:46:08.224Z level=INFO source=server.go:139 msg=offload library=cuda layers.requested=-1 layers.model=65 layers.offload=0 layers.split="" memory.available="[21.7 GiB 21.7 GiB 21.7 GiB 21.7 GiB 21.7 GiB 21.7 GiB]" memory.gpu_overhead="0 B" memory.required.full="56.4 GiB" memory.required.partial="0 B" memory.required.kv="32.0 GiB" memory.required.allocations="[0 B 0 B 0 B 0 B 0 B 0 B]" memory.weights.total="24.4 GiB" memory.weights.repeating="23.8 GiB" memory.weights.nonrepeating="608.6 MiB" memory.graph.full="42.7 GiB" memory.graph.partial="42.7 GiB"
llama_model_loader: loaded meta data with 36 key-value pairs and 707 tensors from /root/.ollama/models/blobs/sha256-6b818b3bbaee2722c6201aa17fc4c0733da71bbb12b7f77abfd75e5c0cb98e08 (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-128K
llama_model_loader: - kv   3:                           general.finetune str              = 128k
llama_model_loader: - kv   4:                           general.basename str              = Qwen3-32B-128K
llama_model_loader: - kv   5:                       general.quantized_by str              = Unsloth
llama_model_loader: - kv   6:                         general.size_label str              = 32B
llama_model_loader: - kv   7:                           general.repo_url str              = https://huggingface.co/unsloth
llama_model_loader: - kv   8:                          qwen3.block_count u32              = 64
llama_model_loader: - kv   9:                       qwen3.context_length u32              = 131072
llama_model_loader: - kv  10:                     qwen3.embedding_length u32              = 5120
llama_model_loader: - kv  11:                  qwen3.feed_forward_length u32              = 25600
llama_model_loader: - kv  12:                 qwen3.attention.head_count u32              = 64
llama_model_loader: - kv  13:              qwen3.attention.head_count_kv u32              = 8
llama_model_loader: - kv  14:                       qwen3.rope.freq_base f32              = 1000000.000000
llama_model_loader: - kv  15:     qwen3.attention.layer_norm_rms_epsilon f32              = 0.000001
llama_model_loader: - kv  16:                 qwen3.attention.key_length u32              = 128
llama_model_loader: - kv  17:               qwen3.attention.value_length u32              = 128
llama_model_loader: - kv  18:                    qwen3.rope.scaling.type str              = yarn
llama_model_loader: - kv  19:                  qwen3.rope.scaling.factor f32              = 4.000000
llama_model_loader: - kv  20: qwen3.rope.scaling.original_context_length u32              = 32768
llama_model_loader: - kv  21:                       tokenizer.ggml.model str              = gpt2
llama_model_loader: - kv  22:                         tokenizer.ggml.pre str              = qwen2
llama_model_loader: - kv  23:                      tokenizer.ggml.tokens arr[str,151936]  = ["!", "\"", "#", "$", "%", "&", "'", ...
llama_model_loader: - kv  24:                  tokenizer.ggml.token_type arr[i32,151936]  = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv  25:                      tokenizer.ggml.merges arr[str,151387]  = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t",...
llama_model_loader: - kv  26:                tokenizer.ggml.eos_token_id u32              = 151645
llama_model_loader: - kv  27:            tokenizer.ggml.padding_token_id u32              = 151654
llama_model_loader: - kv  28:               tokenizer.ggml.add_bos_token bool             = false
llama_model_loader: - kv  29:                    tokenizer.chat_template str              = {%- if tools %}\n    {{- '<|im_start|>...
llama_model_loader: - kv  30:               general.quantization_version u32              = 2
llama_model_loader: - kv  31:                          general.file_type u32              = 18
llama_model_loader: - kv  32:                      quantize.imatrix.file str              = Qwen3-32B-128K-GGUF/imatrix_unsloth.dat
llama_model_loader: - kv  33:                   quantize.imatrix.dataset str              = unsloth_calibration_Qwen3-32B-128K.txt
llama_model_loader: - kv  34:             quantize.imatrix.entries_count i32              = 448
llama_model_loader: - kv  35:              quantize.imatrix.chunks_count i32              = 685
llama_model_loader: - type  f32:  257 tensors
llama_model_loader: - type q6_K:  450 tensors
print_info: file format = GGUF V3 (latest)
print_info: file type   = Q6_K
print_info: file size   = 25.03 GiB (6.56 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-128K
print_info: vocab type       = BPE
print_info: n_vocab          = 151936
print_info: n_merges         = 151387
print_info: BOS token        = 11 ','
print_info: EOS token        = 151645 '<|im_end|>'
print_info: EOT token        = 151645 '<|im_end|>'
print_info: PAD token        = 151654 '<|vision_pad|>'
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=2025-05-23T10:46:08.427Z level=INFO source=server.go:410 msg="starting llama server" cmd="/usr/bin/ollama runner --model /root/.ollama/models/blobs/sha256-6b818b3bbaee2722c6201aa17fc4c0733da71bbb12b7f77abfd75e5c0cb98e08 --ctx-size 131072 --batch-size 4 --threads 64 --no-mmap --parallel 1 --port 34438"
time=2025-05-23T10:46:08.428Z level=INFO source=sched.go:452 msg="loaded runners" count=1
time=2025-05-23T10:46:08.428Z level=INFO source=server.go:589 msg="waiting for llama runner to start responding"
time=2025-05-23T10:46:08.428Z level=INFO source=server.go:623 msg="waiting for server to become available" status="llm server not responding"
time=2025-05-23T10:46:08.470Z level=INFO source=runner.go:853 msg="starting go runner"
load_backend: loaded CPU backend from /usr/lib/ollama/libggml-cpu-icelake.so
time=2025-05-23T10:46:08.487Z level=INFO source=ggml.go:103 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.AVX512=1 CPU.0.AVX512_VBMI=1 CPU.0.AVX512_VNNI=1 CPU.0.LLAMAFILE=1 CPU.1.LLAMAFILE=1 compiler=cgo(gcc)
time=2025-05-23T10:46:08.489Z level=INFO source=runner.go:913 msg="Server listening on 127.0.0.1:34438"
llama_model_loader: loaded meta data with 36 key-value pairs and 707 tensors from /root/.ollama/models/blobs/sha256-6b818b3bbaee2722c6201aa17fc4c0733da71bbb12b7f77abfd75e5c0cb98e08 (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-128K
llama_model_loader: - kv   3:                           general.finetune str              = 128k
llama_model_loader: - kv   4:                           general.basename str              = Qwen3-32B-128K
llama_model_loader: - kv   5:                       general.quantized_by str              = Unsloth
llama_model_loader: - kv   6:                         general.size_label str              = 32B
llama_model_loader: - kv   7:                           general.repo_url str              = https://huggingface.co/unsloth
llama_model_loader: - kv   8:                          qwen3.block_count u32              = 64
llama_model_loader: - kv   9:                       qwen3.context_length u32              = 131072
llama_model_loader: - kv  10:                     qwen3.embedding_length u32              = 5120
llama_model_loader: - kv  11:                  qwen3.feed_forward_length u32              = 25600
llama_model_loader: - kv  12:                 qwen3.attention.head_count u32              = 64
llama_model_loader: - kv  13:              qwen3.attention.head_count_kv u32              = 8
llama_model_loader: - kv  14:                       qwen3.rope.freq_base f32              = 1000000.000000
llama_model_loader: - kv  15:     qwen3.attention.layer_norm_rms_epsilon f32              = 0.000001
llama_model_loader: - kv  16:                 qwen3.attention.key_length u32              = 128
llama_model_loader: - kv  17:               qwen3.attention.value_length u32              = 128
llama_model_loader: - kv  18:                    qwen3.rope.scaling.type str              = yarn
llama_model_loader: - kv  19:                  qwen3.rope.scaling.factor f32              = 4.000000
llama_model_loader: - kv  20: qwen3.rope.scaling.original_context_length u32              = 32768
llama_model_loader: - kv  21:                       tokenizer.ggml.model str              = gpt2
llama_model_loader: - kv  22:                         tokenizer.ggml.pre str              = qwen2
llama_model_loader: - kv  23:                      tokenizer.ggml.tokens arr[str,151936]  = ["!", "\"", "#", "$", "%", "&", "'", ...
llama_model_loader: - kv  24:                  tokenizer.ggml.token_type arr[i32,151936]  = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv  25:                      tokenizer.ggml.merges arr[str,151387]  = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t",...
llama_model_loader: - kv  26:                tokenizer.ggml.eos_token_id u32              = 151645
llama_model_loader: - kv  27:            tokenizer.ggml.padding_token_id u32              = 151654
llama_model_loader: - kv  28:               tokenizer.ggml.add_bos_token bool             = false
llama_model_loader: - kv  29:                    tokenizer.chat_template str              = {%- if tools %}\n    {{- '<|im_start|>...
llama_model_loader: - kv  30:               general.quantization_version u32              = 2
llama_model_loader: - kv  31:                          general.file_type u32              = 18
llama_model_loader: - kv  32:                      quantize.imatrix.file str              = Qwen3-32B-128K-GGUF/imatrix_unsloth.dat
llama_model_loader: - kv  33:                   quantize.imatrix.dataset str              = unsloth_calibration_Qwen3-32B-128K.txt
llama_model_loader: - kv  34:             quantize.imatrix.entries_count i32              = 448
llama_model_loader: - kv  35:              quantize.imatrix.chunks_count i32              = 685
llama_model_loader: - type  f32:  257 tensors
llama_model_loader: - type q6_K:  450 tensors
print_info: file format = GGUF V3 (latest)
print_info: file type   = Q6_K
print_info: file size   = 25.03 GiB (6.56 BPW)
time=2025-05-23T10:46:08.680Z level=INFO source=server.go:623 msg="waiting for server to become available" status="llm server loading model"
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      = 131072
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     = yarn
print_info: freq_base_train  = 1000000.0
print_info: freq_scale_train = 0.25
print_info: n_ctx_orig_yarn  = 32768
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-128K
print_info: vocab type       = BPE
print_info: n_vocab          = 151936
print_info: n_merges         = 151387
print_info: BOS token        = 11 ','
print_info: EOS token        = 151645 '<|im_end|>'
print_info: EOT token        = 151645 '<|im_end|>'
print_info: PAD token        = 151654 '<|vision_pad|>'
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:          CPU model buffer size = 25632.22 MiB
llama_context: constructing llama_context
llama_context: n_batch is less than GGML_KQ_MASK_PAD - increasing to 64
llama_context: n_seq_max     = 1
llama_context: n_ctx         = 131072
llama_context: n_ctx_per_seq = 131072
llama_context: n_batch       = 64
llama_context: n_ubatch      = 64
llama_context: causal_attn   = 1
llama_context: flash_attn    = 0
llama_context: freq_base     = 1000000.0
llama_context: freq_scale    = 0.25
llama_context:        CPU  output buffer size =     0.60 MiB
init: kv_size = 131072, offload = 1, type_k = 'f16', type_v = 'f16', n_layer = 64, can_shift = 1
[GIN] 2025/05/23 - 10:46:31 | 200 |      58.685µs |       127.0.0.1 | HEAD     "/"
[GIN] 2025/05/23 - 10:46:31 | 200 |      75.827µs |       127.0.0.1 | GET      "/api/ps"
init:        CPU KV buffer size = 32768.00 MiB
llama_context: KV self size  = 32768.00 MiB, K (f16): 16384.00 MiB, V (f16): 16384.00 MiB
llama_context:        CPU compute buffer size =  2086.50 MiB
llama_context: graph nodes  = 2438
llama_context: graph splits = 1
time=2025-05-23T10:46:36.053Z level=INFO source=server.go:628 msg="llama runner started in 27.63 seconds"

OS

Linux

GPU

Nvidia

CPU

Intel

Ollama version

0.6.8

Originally created by @adamhj on GitHub (May 23, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/10832 ### What is the issue? Today I'm trying to run unsloth/Qwen3-32B-128K-GGUF(Q6_K) on ollama, on a machine has 24G×6 nvidia A10, but after the model was loaded, I found that it was running 100% on CPU ```NAME ID SIZE PROCESSOR UNTIL unsloth/Qwen3:32B-128K-Q6_K 3ddf246c242f 60 GB 100% CPU 4 minutes from now ``` I can run the official Qwen3:32B-Q6_K model on the same machine without any problem. I suspect the problem is that the 128k model uses RoPE Scaling and cannot utilize multiple GPUs. Indeed, I also tried running unsloth/Qwen3-8B-128K-GGUF(Q6_K) on the same machine. It ran, but I noticed that only the first GPU was utilized, with its VRAM nearly full, while the others remained empty. This is not the case when I run other models, where the VRAM is distributed evenly across all six cards. ### Relevant log output ``` [root@126gpu-10 Qwen3-30B-A3B-GGUF]# docker logs --tail 10 -f ollama [GIN] 2025/05/23 - 05:46:03 | 200 | 9m31s | 10.3.2.153 | POST "/v1/chat/completions" time=2025-05-23T05:51:09.589Z level=WARN source=sched.go:655 msg="gpu VRAM usage didn't recover within timeout" seconds=6.462163498 model=/root/.ollama/models/blobs/sha256-6b818b3bbaee2722c6201aa17fc4c0733da71bbb12b7f77abfd75e5c0cb98e08 time=2025-05-23T05:51:11.920Z level=WARN source=sched.go:655 msg="gpu VRAM usage didn't recover within timeout" seconds=8.793217106 model=/root/.ollama/models/blobs/sha256-6b818b3bbaee2722c6201aa17fc4c0733da71bbb12b7f77abfd75e5c0cb98e08 time=2025-05-23T05:51:14.104Z level=WARN source=sched.go:655 msg="gpu VRAM usage didn't recover within timeout" seconds=10.977573067 model=/root/.ollama/models/blobs/sha256-6b818b3bbaee2722c6201aa17fc4c0733da71bbb12b7f77abfd75e5c0cb98e08 [GIN] 2025/05/23 - 06:34:00 | 200 | 70.738µs | 127.0.0.1 | HEAD "/" [GIN] 2025/05/23 - 06:34:00 | 200 | 36.864µs | 127.0.0.1 | GET "/api/ps" [GIN] 2025/05/23 - 10:43:24 | 200 | 64.701µs | 127.0.0.1 | HEAD "/" [GIN] 2025/05/23 - 10:43:24 | 200 | 59.119µs | 127.0.0.1 | GET "/api/ps" [GIN] 2025/05/23 - 10:44:55 | 200 | 59.305µs | 127.0.0.1 | HEAD "/" [GIN] 2025/05/23 - 10:44:55 | 200 | 69.853µs | 127.0.0.1 | GET "/api/ps" time=2025-05-23T10:46:01.917Z level=WARN source=ggml.go:152 msg="key not found" key=general.alignment default=32 time=2025-05-23T10:46:05.808Z level=WARN source=ggml.go:152 msg="key not found" key=general.alignment default=32 time=2025-05-23T10:46:05.835Z level=WARN source=ggml.go:152 msg="key not found" key=general.alignment default=32 time=2025-05-23T10:46:05.836Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0 time=2025-05-23T10:46:05.839Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0 time=2025-05-23T10:46:05.842Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0 time=2025-05-23T10:46:05.845Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0 time=2025-05-23T10:46:05.847Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0 time=2025-05-23T10:46:05.850Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0 time=2025-05-23T10:46:05.852Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0 time=2025-05-23T10:46:05.855Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0 time=2025-05-23T10:46:05.857Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0 time=2025-05-23T10:46:05.860Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0 time=2025-05-23T10:46:05.863Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0 time=2025-05-23T10:46:05.865Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0 time=2025-05-23T10:46:05.868Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0 time=2025-05-23T10:46:05.870Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0 time=2025-05-23T10:46:08.223Z level=INFO source=server.go:106 msg="system memory" total="377.1 GiB" free="366.6 GiB" free_swap="15.5 GiB" time=2025-05-23T10:46:08.224Z level=WARN source=ggml.go:152 msg="key not found" key=qwen3.vision.block_count default=0 time=2025-05-23T10:46:08.224Z level=INFO source=server.go:139 msg=offload library=cuda layers.requested=-1 layers.model=65 layers.offload=0 layers.split="" memory.available="[21.7 GiB 21.7 GiB 21.7 GiB 21.7 GiB 21.7 GiB 21.7 GiB]" memory.gpu_overhead="0 B" memory.required.full="56.4 GiB" memory.required.partial="0 B" memory.required.kv="32.0 GiB" memory.required.allocations="[0 B 0 B 0 B 0 B 0 B 0 B]" memory.weights.total="24.4 GiB" memory.weights.repeating="23.8 GiB" memory.weights.nonrepeating="608.6 MiB" memory.graph.full="42.7 GiB" memory.graph.partial="42.7 GiB" llama_model_loader: loaded meta data with 36 key-value pairs and 707 tensors from /root/.ollama/models/blobs/sha256-6b818b3bbaee2722c6201aa17fc4c0733da71bbb12b7f77abfd75e5c0cb98e08 (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-128K llama_model_loader: - kv 3: general.finetune str = 128k llama_model_loader: - kv 4: general.basename str = Qwen3-32B-128K llama_model_loader: - kv 5: general.quantized_by str = Unsloth llama_model_loader: - kv 6: general.size_label str = 32B llama_model_loader: - kv 7: general.repo_url str = https://huggingface.co/unsloth llama_model_loader: - kv 8: qwen3.block_count u32 = 64 llama_model_loader: - kv 9: qwen3.context_length u32 = 131072 llama_model_loader: - kv 10: qwen3.embedding_length u32 = 5120 llama_model_loader: - kv 11: qwen3.feed_forward_length u32 = 25600 llama_model_loader: - kv 12: qwen3.attention.head_count u32 = 64 llama_model_loader: - kv 13: qwen3.attention.head_count_kv u32 = 8 llama_model_loader: - kv 14: qwen3.rope.freq_base f32 = 1000000.000000 llama_model_loader: - kv 15: qwen3.attention.layer_norm_rms_epsilon f32 = 0.000001 llama_model_loader: - kv 16: qwen3.attention.key_length u32 = 128 llama_model_loader: - kv 17: qwen3.attention.value_length u32 = 128 llama_model_loader: - kv 18: qwen3.rope.scaling.type str = yarn llama_model_loader: - kv 19: qwen3.rope.scaling.factor f32 = 4.000000 llama_model_loader: - kv 20: qwen3.rope.scaling.original_context_length u32 = 32768 llama_model_loader: - kv 21: tokenizer.ggml.model str = gpt2 llama_model_loader: - kv 22: tokenizer.ggml.pre str = qwen2 llama_model_loader: - kv 23: tokenizer.ggml.tokens arr[str,151936] = ["!", "\"", "#", "$", "%", "&", "'", ... llama_model_loader: - kv 24: tokenizer.ggml.token_type arr[i32,151936] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ... llama_model_loader: - kv 25: tokenizer.ggml.merges arr[str,151387] = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t",... llama_model_loader: - kv 26: tokenizer.ggml.eos_token_id u32 = 151645 llama_model_loader: - kv 27: tokenizer.ggml.padding_token_id u32 = 151654 llama_model_loader: - kv 28: tokenizer.ggml.add_bos_token bool = false llama_model_loader: - kv 29: tokenizer.chat_template str = {%- if tools %}\n {{- '<|im_start|>... llama_model_loader: - kv 30: general.quantization_version u32 = 2 llama_model_loader: - kv 31: general.file_type u32 = 18 llama_model_loader: - kv 32: quantize.imatrix.file str = Qwen3-32B-128K-GGUF/imatrix_unsloth.dat llama_model_loader: - kv 33: quantize.imatrix.dataset str = unsloth_calibration_Qwen3-32B-128K.txt llama_model_loader: - kv 34: quantize.imatrix.entries_count i32 = 448 llama_model_loader: - kv 35: quantize.imatrix.chunks_count i32 = 685 llama_model_loader: - type f32: 257 tensors llama_model_loader: - type q6_K: 450 tensors print_info: file format = GGUF V3 (latest) print_info: file type = Q6_K print_info: file size = 25.03 GiB (6.56 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-128K print_info: vocab type = BPE print_info: n_vocab = 151936 print_info: n_merges = 151387 print_info: BOS token = 11 ',' print_info: EOS token = 151645 '<|im_end|>' print_info: EOT token = 151645 '<|im_end|>' print_info: PAD token = 151654 '<|vision_pad|>' 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=2025-05-23T10:46:08.427Z level=INFO source=server.go:410 msg="starting llama server" cmd="/usr/bin/ollama runner --model /root/.ollama/models/blobs/sha256-6b818b3bbaee2722c6201aa17fc4c0733da71bbb12b7f77abfd75e5c0cb98e08 --ctx-size 131072 --batch-size 4 --threads 64 --no-mmap --parallel 1 --port 34438" time=2025-05-23T10:46:08.428Z level=INFO source=sched.go:452 msg="loaded runners" count=1 time=2025-05-23T10:46:08.428Z level=INFO source=server.go:589 msg="waiting for llama runner to start responding" time=2025-05-23T10:46:08.428Z level=INFO source=server.go:623 msg="waiting for server to become available" status="llm server not responding" time=2025-05-23T10:46:08.470Z level=INFO source=runner.go:853 msg="starting go runner" load_backend: loaded CPU backend from /usr/lib/ollama/libggml-cpu-icelake.so time=2025-05-23T10:46:08.487Z level=INFO source=ggml.go:103 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.AVX512=1 CPU.0.AVX512_VBMI=1 CPU.0.AVX512_VNNI=1 CPU.0.LLAMAFILE=1 CPU.1.LLAMAFILE=1 compiler=cgo(gcc) time=2025-05-23T10:46:08.489Z level=INFO source=runner.go:913 msg="Server listening on 127.0.0.1:34438" llama_model_loader: loaded meta data with 36 key-value pairs and 707 tensors from /root/.ollama/models/blobs/sha256-6b818b3bbaee2722c6201aa17fc4c0733da71bbb12b7f77abfd75e5c0cb98e08 (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-128K llama_model_loader: - kv 3: general.finetune str = 128k llama_model_loader: - kv 4: general.basename str = Qwen3-32B-128K llama_model_loader: - kv 5: general.quantized_by str = Unsloth llama_model_loader: - kv 6: general.size_label str = 32B llama_model_loader: - kv 7: general.repo_url str = https://huggingface.co/unsloth llama_model_loader: - kv 8: qwen3.block_count u32 = 64 llama_model_loader: - kv 9: qwen3.context_length u32 = 131072 llama_model_loader: - kv 10: qwen3.embedding_length u32 = 5120 llama_model_loader: - kv 11: qwen3.feed_forward_length u32 = 25600 llama_model_loader: - kv 12: qwen3.attention.head_count u32 = 64 llama_model_loader: - kv 13: qwen3.attention.head_count_kv u32 = 8 llama_model_loader: - kv 14: qwen3.rope.freq_base f32 = 1000000.000000 llama_model_loader: - kv 15: qwen3.attention.layer_norm_rms_epsilon f32 = 0.000001 llama_model_loader: - kv 16: qwen3.attention.key_length u32 = 128 llama_model_loader: - kv 17: qwen3.attention.value_length u32 = 128 llama_model_loader: - kv 18: qwen3.rope.scaling.type str = yarn llama_model_loader: - kv 19: qwen3.rope.scaling.factor f32 = 4.000000 llama_model_loader: - kv 20: qwen3.rope.scaling.original_context_length u32 = 32768 llama_model_loader: - kv 21: tokenizer.ggml.model str = gpt2 llama_model_loader: - kv 22: tokenizer.ggml.pre str = qwen2 llama_model_loader: - kv 23: tokenizer.ggml.tokens arr[str,151936] = ["!", "\"", "#", "$", "%", "&", "'", ... llama_model_loader: - kv 24: tokenizer.ggml.token_type arr[i32,151936] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ... llama_model_loader: - kv 25: tokenizer.ggml.merges arr[str,151387] = ["Ġ Ġ", "ĠĠ ĠĠ", "i n", "Ġ t",... llama_model_loader: - kv 26: tokenizer.ggml.eos_token_id u32 = 151645 llama_model_loader: - kv 27: tokenizer.ggml.padding_token_id u32 = 151654 llama_model_loader: - kv 28: tokenizer.ggml.add_bos_token bool = false llama_model_loader: - kv 29: tokenizer.chat_template str = {%- if tools %}\n {{- '<|im_start|>... llama_model_loader: - kv 30: general.quantization_version u32 = 2 llama_model_loader: - kv 31: general.file_type u32 = 18 llama_model_loader: - kv 32: quantize.imatrix.file str = Qwen3-32B-128K-GGUF/imatrix_unsloth.dat llama_model_loader: - kv 33: quantize.imatrix.dataset str = unsloth_calibration_Qwen3-32B-128K.txt llama_model_loader: - kv 34: quantize.imatrix.entries_count i32 = 448 llama_model_loader: - kv 35: quantize.imatrix.chunks_count i32 = 685 llama_model_loader: - type f32: 257 tensors llama_model_loader: - type q6_K: 450 tensors print_info: file format = GGUF V3 (latest) print_info: file type = Q6_K print_info: file size = 25.03 GiB (6.56 BPW) time=2025-05-23T10:46:08.680Z level=INFO source=server.go:623 msg="waiting for server to become available" status="llm server loading model" 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 = 131072 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 = yarn print_info: freq_base_train = 1000000.0 print_info: freq_scale_train = 0.25 print_info: n_ctx_orig_yarn = 32768 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-128K print_info: vocab type = BPE print_info: n_vocab = 151936 print_info: n_merges = 151387 print_info: BOS token = 11 ',' print_info: EOS token = 151645 '<|im_end|>' print_info: EOT token = 151645 '<|im_end|>' print_info: PAD token = 151654 '<|vision_pad|>' 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: CPU model buffer size = 25632.22 MiB llama_context: constructing llama_context llama_context: n_batch is less than GGML_KQ_MASK_PAD - increasing to 64 llama_context: n_seq_max = 1 llama_context: n_ctx = 131072 llama_context: n_ctx_per_seq = 131072 llama_context: n_batch = 64 llama_context: n_ubatch = 64 llama_context: causal_attn = 1 llama_context: flash_attn = 0 llama_context: freq_base = 1000000.0 llama_context: freq_scale = 0.25 llama_context: CPU output buffer size = 0.60 MiB init: kv_size = 131072, offload = 1, type_k = 'f16', type_v = 'f16', n_layer = 64, can_shift = 1 [GIN] 2025/05/23 - 10:46:31 | 200 | 58.685µs | 127.0.0.1 | HEAD "/" [GIN] 2025/05/23 - 10:46:31 | 200 | 75.827µs | 127.0.0.1 | GET "/api/ps" init: CPU KV buffer size = 32768.00 MiB llama_context: KV self size = 32768.00 MiB, K (f16): 16384.00 MiB, V (f16): 16384.00 MiB llama_context: CPU compute buffer size = 2086.50 MiB llama_context: graph nodes = 2438 llama_context: graph splits = 1 time=2025-05-23T10:46:36.053Z level=INFO source=server.go:628 msg="llama runner started in 27.63 seconds" ``` ### OS Linux ### GPU Nvidia ### CPU Intel ### Ollama version 0.6.8
GiteaMirror added the bug label 2026-04-12 19:06:10 -05:00
GiteaMirror commented 2026-04-12 19:06:14 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (May 23, 2025):

time=2025-05-23T10:46:08.224Z level=INFO source=server.go:139 msg=offload library=cuda layers.requested=-1 layers.model=65
 layers.offload=0 layers.split="" memory.available="[21.7 GiB 21.7 GiB 21.7 GiB 21.7 GiB 21.7 GiB 21.7 GiB]"
 memory.gpu_overhead="0 B" memory.required.full="56.4 GiB" memory.required.partial="0 B" memory.required.kv="32.0 GiB"
 memory.required.allocations="[0 B 0 B 0 B 0 B 0 B 0 B]" memory.weights.total="24.4 GiB" memory.weights.repeating="23.8 GiB"
 memory.weights.nonrepeating="608.6 MiB" memory.graph.full="42.7 GiB" memory.graph.partial="42.7 GiB"

Context size of 131072 tokens pushes the memory graph to a size (42.7G) where it won't fit on a GPU device (21.7G), so the model is loaded in system RAM.

<!-- gh-comment-id:2904159214 --> @rick-github commented on GitHub (May 23, 2025): ``` time=2025-05-23T10:46:08.224Z level=INFO source=server.go:139 msg=offload library=cuda layers.requested=-1 layers.model=65 layers.offload=0 layers.split="" memory.available="[21.7 GiB 21.7 GiB 21.7 GiB 21.7 GiB 21.7 GiB 21.7 GiB]" memory.gpu_overhead="0 B" memory.required.full="56.4 GiB" memory.required.partial="0 B" memory.required.kv="32.0 GiB" memory.required.allocations="[0 B 0 B 0 B 0 B 0 B 0 B]" memory.weights.total="24.4 GiB" memory.weights.repeating="23.8 GiB" memory.weights.nonrepeating="608.6 MiB" memory.graph.full="42.7 GiB" memory.graph.partial="42.7 GiB" ``` Context size of 131072 tokens pushes the memory graph to a size (42.7G) where it won't fit on a GPU device (21.7G), so the model is loaded in system RAM.
GiteaMirror commented 2026-04-12 19:06:15 -05:00
Author
Owner
Copy Link

@adamhj commented on GitHub (May 30, 2025):

Get it, thanks for your reply

<!-- gh-comment-id:2921029183 --> @adamhj commented on GitHub (May 30, 2025): Get it, thanks for your reply
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#7113
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?