github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-18 06:06:18 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #9028] When running deepseek-r1 with Ollama, it intermittently outputs garbage #5872

New Issue
Open
opened 2026-04-12 17:12:40 -05:00 by GiteaMirror · 3 comments
GiteaMirror commented 2026-04-12 17:12:40 -05:00
Owner
Copy Link

Originally created by @Oldpan on GitHub (Feb 12, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/9028

What is the issue?

When I was running deepseek-r1 using Ollama, I occasionally encountered some garbled output, such as ';0?!18=1C%<｜▁pad▁｜>-DB', but sometimes it worked correctly. What could be the possible reasons? My setup is 8x3090 GPUs with 500GB of RAM.

My env and scripit:

ollama create DeepSeek-R1-Q4_K_M -f /code/ollama/modelfile/DeepSeek-R1-Q4-K_M

FROM /code/DeepSeek-R1-int4/DeepSeek-R1-Q4_K_M.gguf
PARAMETER num_gpu 16  
PARAMETER num_ctx 4096  
PARAMETER temperature 0.6
PARAMETER stop <｜begin▁of▁sentence｜>
PARAMETER stop <｜end▁of▁sentence｜>
PARAMETER stop <｜User｜>
PARAMETER stop <｜Assistant｜>

TEMPLATE """{{- if .System }}{{ .System }}{{ end }}
{{- range $i, $_ := .Messages }}
{{- $last := eq (len (slice $.Messages $i)) 1}}
{{- if eq .Role "user" }}<｜User｜>{{ .Content }}
{{- else if eq .Role "assistant" }}<｜Assistant｜>{{ .Content }}{{- if not $last }}<｜end▁of▁sentence｜>{{- end }}
{{- end }}
{{- if and $last (ne .Role "assistant") }}<｜Assistant｜>{{- end }}
{{- end }}"""

then ollama run DeepSeek-R1-Q4_K_M:latest --verbose

Thanks

Relevant log output

time=2025-02-11T12:52:33.956Z level=INFO source=sched.go:730 msg="new model will fit in available VRAM, loading" model=/code/ollama/ollama_home/blobs/sha256-0b898a049c60a2d4b2f373b21e19b3efea4aa81e6d403f93f5a41354cba304c5 library=cuda parallel=1 required="112.3 GiB"
time=2025-02-11T12:52:34.750Z level=INFO source=server.go:104 msg="system memory" total="753.4 GiB" free="710.4 GiB" free_swap="0 B"
time=2025-02-11T12:52:34.751Z level=INFO source=memory.go:356 msg="offload to cuda" layers.requested=16 layers.model=62 layers.offload=16 layers.split=2,2,2,2,2,2,2,2 memory.available="[23.4 GiB 23.4 GiB 23.4 GiB 23.4 GiB 23.4 GiB 23.4 GiB 23.4 GiB 23.4 GiB]" memory.gpu_overhead="0 B" memory.required.full="438.7 GiB" memory.required.partial="112.3 GiB" memory.required.kv="19.1 GiB" memory.required.allocations="[9.7 GiB 10.6 GiB 9.7 GiB 16.3 GiB 17.2 GiB 16.3 GiB 16.3 GiB 16.3 GiB]" memory.weights.total="394.5 GiB" memory.weights.repeating="393.8 GiB" memory.weights.nonrepeating="725.0 MiB" memory.graph.full="1.6 GiB" memory.graph.partial="1.6 GiB"
time=2025-02-11T12:52:34.751Z level=INFO source=server.go:376 msg="starting llama server" cmd="/usr/local/lib/ollama/runners/cuda_v12_avx/ollama_llama_server runner --model /code/ollama/ollama_home/blobs/sha256-0b898a049c60a2d4b2f373b21e19b3efea4aa81e6d403f93f5a41354cba304c5 --ctx-size 4096 --batch-size 512 --n-gpu-layers 16 --threads 48 --parallel 1 --tensor-split 2,2,2,2,2,2,2,2 --port 43589"
time=2025-02-11T12:52:34.752Z level=INFO source=sched.go:449 msg="loaded runners" count=1
time=2025-02-11T12:52:34.752Z level=INFO source=server.go:555 msg="waiting for llama runner to start responding"
time=2025-02-11T12:52:34.752Z level=INFO source=server.go:589 msg="waiting for server to become available" status="llm server error"
time=2025-02-11T12:52:34.792Z level=INFO source=runner.go:936 msg="starting go runner"
ggml_cuda_init: GGML_CUDA_FORCE_MMQ:    no
ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no
ggml_cuda_init: found 8 CUDA devices:
  Device 0: NVIDIA GeForce RTX 3090, compute capability 8.6, VMM: yes
  Device 1: NVIDIA GeForce RTX 3090, compute capability 8.6, VMM: yes
  Device 2: NVIDIA GeForce RTX 3090, compute capability 8.6, VMM: yes
  Device 3: NVIDIA GeForce RTX 3090, compute capability 8.6, VMM: yes
  Device 4: NVIDIA GeForce RTX 3090, compute capability 8.6, VMM: yes
  Device 5: NVIDIA GeForce RTX 3090, compute capability 8.6, VMM: yes
  Device 6: NVIDIA GeForce RTX 3090, compute capability 8.6, VMM: yes
  Device 7: NVIDIA GeForce RTX 3090, compute capability 8.6, VMM: yes
time=2025-02-11T12:52:35.348Z level=INFO source=runner.go:937 msg=system info="CUDA : ARCHS = 600,610,620,700,720,750,800,860,870,890,900 | USE_GRAPHS = 1 | PEER_MAX_BATCH_SIZE = 128 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | LLAMAFILE = 1 | AARCH64_REPACK = 1 | cgo(gcc)" threads=48
time=2025-02-11T12:52:35.349Z level=INFO source=.:0 msg="Server listening on 127.0.0.1:43589"
llama_load_model_from_file: using device CUDA0 (NVIDIA GeForce RTX 3090) - 23997 MiB free
llama_load_model_from_file: using device CUDA1 (NVIDIA GeForce RTX 3090) - 23997 MiB free
llama_load_model_from_file: using device CUDA2 (NVIDIA GeForce RTX 3090) - 23997 MiB free
llama_load_model_from_file: using device CUDA3 (NVIDIA GeForce RTX 3090) - 23997 MiB free
llama_load_model_from_file: using device CUDA4 (NVIDIA GeForce RTX 3090) - 23997 MiB free
llama_load_model_from_file: using device CUDA5 (NVIDIA GeForce RTX 3090) - 23997 MiB free
llama_load_model_from_file: using device CUDA6 (NVIDIA GeForce RTX 3090) - 23997 MiB free
llama_load_model_from_file: using device CUDA7 (NVIDIA GeForce RTX 3090) - 23997 MiB free
llama_model_loader: loaded meta data with 48 key-value pairs and 1025 tensors from /code/ollama/ollama_home/blobs/sha256-0b898a049c60a2d4b2f373b21e19b3efea4aa81e6d403f93f5a41354cba304c5 (version GGUF V3 (latest))
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
llama_model_loader: - kv   0:                       general.architecture str              = deepseek2
llama_model_loader: - kv   1:                               general.type str              = model
llama_model_loader: - kv   2:                               general.name str              = DeepSeek R1 BF16
llama_model_loader: - kv   3:                       general.quantized_by str              = Unsloth
llama_model_loader: - kv   4:                         general.size_label str              = 256x20B
llama_model_loader: - kv   5:                           general.repo_url str              = https://huggingface.co/unsloth
llama_model_loader: - kv   6:                      deepseek2.block_count u32              = 61
llama_model_loader: - kv   7:                   deepseek2.context_length u32              = 163840
llama_model_loader: - kv   8:                 deepseek2.embedding_length u32              = 7168
llama_model_loader: - kv   9:              deepseek2.feed_forward_length u32              = 18432
llama_model_loader: - kv  10:             deepseek2.attention.head_count u32              = 128
llama_model_loader: - kv  11:          deepseek2.attention.head_count_kv u32              = 128
llama_model_loader: - kv  12:                   deepseek2.rope.freq_base f32              = 10000.000000
llama_model_loader: - kv  13: deepseek2.attention.layer_norm_rms_epsilon f32              = 0.000001
llama_model_loader: - kv  14:                deepseek2.expert_used_count u32              = 8
llama_model_loader: - kv  15:        deepseek2.leading_dense_block_count u32              = 3
llama_model_loader: - kv  16:                       deepseek2.vocab_size u32              = 129280
llama_model_loader: - kv  17:            deepseek2.attention.q_lora_rank u32              = 1536
llama_model_loader: - kv  18:           deepseek2.attention.kv_lora_rank u32              = 512
llama_model_loader: - kv  19:             deepseek2.attention.key_length u32              = 192
llama_model_loader: - kv  20:           deepseek2.attention.value_length u32              = 128
llama_model_loader: - kv  21:       deepseek2.expert_feed_forward_length u32              = 2048
llama_model_loader: - kv  22:                     deepseek2.expert_count u32              = 256
llama_model_loader: - kv  23:              deepseek2.expert_shared_count u32              = 1
llama_model_loader: - kv  24:             deepseek2.expert_weights_scale f32              = 2.500000
llama_model_loader: - kv  25:              deepseek2.expert_weights_norm bool             = true
llama_model_loader: - kv  26:               deepseek2.expert_gating_func u32              = 2
llama_model_loader: - kv  27:             deepseek2.rope.dimension_count u32              = 64
llama_model_loader: - kv  28:                deepseek2.rope.scaling.type str              = yarn
llama_model_loader: - kv  29:              deepseek2.rope.scaling.factor f32              = 40.000000
llama_model_loader: - kv  30: deepseek2.rope.scaling.original_context_length u32              = 4096
llama_model_loader: - kv  31: deepseek2.rope.scaling.yarn_log_multiplier f32              = 0.100000
llama_model_loader: - kv  32:                       tokenizer.ggml.model str              = gpt2
llama_model_loader: - kv  33:                         tokenizer.ggml.pre str              = deepseek-v3
time=2025-02-11T12:52:35.504Z level=INFO source=server.go:589 msg="waiting for server to become available" status="llm server loading model"
llama_model_loader: - kv  34:                      tokenizer.ggml.tokens arr[str,129280]  = ["<｜begin▁of▁sentence｜>", "<�...
llama_model_loader: - kv  35:                  tokenizer.ggml.token_type arr[i32,129280]  = [3, 3, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv  36:                      tokenizer.ggml.merges arr[str,127741]  = ["Ġ t", "Ġ a", "i n", "Ġ Ġ", "h e...
llama_model_loader: - kv  37:                tokenizer.ggml.bos_token_id u32              = 0
llama_model_loader: - kv  38:                tokenizer.ggml.eos_token_id u32              = 1
llama_model_loader: - kv  39:            tokenizer.ggml.padding_token_id u32              = 128815
llama_model_loader: - kv  40:               tokenizer.ggml.add_bos_token bool             = true
llama_model_loader: - kv  41:               tokenizer.ggml.add_eos_token bool             = false
llama_model_loader: - kv  42:                    tokenizer.chat_template str              = {% if not add_generation_prompt is de...
llama_model_loader: - kv  43:               general.quantization_version u32              = 2
llama_model_loader: - kv  44:                          general.file_type u32              = 15
llama_model_loader: - kv  45:                                   split.no u16              = 0
llama_model_loader: - kv  46:                        split.tensors.count i32              = 1025
llama_model_loader: - kv  47:                                split.count u16              = 0
llama_model_loader: - type  f32:  361 tensors
llama_model_loader: - type q4_K:  606 tensors
llama_model_loader: - type q6_K:   58 tensors
llm_load_vocab: special_eos_id is not in special_eog_ids - the tokenizer config may be incorrect
llm_load_vocab: special tokens cache size = 819
llm_load_vocab: token to piece cache size = 0.8223 MB
llm_load_print_meta: format           = GGUF V3 (latest)
llm_load_print_meta: arch             = deepseek2
llm_load_print_meta: vocab type       = BPE
llm_load_print_meta: n_vocab          = 129280
llm_load_print_meta: n_merges         = 127741
llm_load_print_meta: vocab_only       = 0
llm_load_print_meta: n_ctx_train      = 163840
llm_load_print_meta: n_embd           = 7168
llm_load_print_meta: n_layer          = 61
llm_load_print_meta: n_head           = 128
llm_load_print_meta: n_head_kv        = 128
llm_load_print_meta: n_rot            = 64
llm_load_print_meta: n_swa            = 0
llm_load_print_meta: n_embd_head_k    = 192
llm_load_print_meta: n_embd_head_v    = 128
llm_load_print_meta: n_gqa            = 1
llm_load_print_meta: n_embd_k_gqa     = 24576
llm_load_print_meta: n_embd_v_gqa     = 16384
llm_load_print_meta: f_norm_eps       = 0.0e+00
llm_load_print_meta: f_norm_rms_eps   = 1.0e-06
llm_load_print_meta: f_clamp_kqv      = 0.0e+00
llm_load_print_meta: f_max_alibi_bias = 0.0e+00
llm_load_print_meta: f_logit_scale    = 0.0e+00
llm_load_print_meta: n_ff             = 18432
llm_load_print_meta: n_expert         = 256
llm_load_print_meta: n_expert_used    = 8
llm_load_print_meta: causal attn      = 1
llm_load_print_meta: pooling type     = 0
llm_load_print_meta: rope type        = 0
llm_load_print_meta: rope scaling     = yarn
llm_load_print_meta: freq_base_train  = 10000.0
llm_load_print_meta: freq_scale_train = 0.025
llm_load_print_meta: n_ctx_orig_yarn  = 4096
llm_load_print_meta: rope_finetuned   = unknown
llm_load_print_meta: ssm_d_conv       = 0
llm_load_print_meta: ssm_d_inner      = 0
llm_load_print_meta: ssm_d_state      = 0
llm_load_print_meta: ssm_dt_rank      = 0
llm_load_print_meta: ssm_dt_b_c_rms   = 0
llm_load_print_meta: model type       = 671B
llm_load_print_meta: model ftype      = Q4_K - Medium
llm_load_print_meta: model params     = 671.03 B
llm_load_print_meta: model size       = 376.65 GiB (4.82 BPW) 
llm_load_print_meta: general.name     = DeepSeek R1 BF16
llm_load_print_meta: BOS token        = 0 '<｜begin▁of▁sentence｜>'
llm_load_print_meta: EOS token        = 1 '<｜end▁of▁sentence｜>'
llm_load_print_meta: EOT token        = 1 '<｜end▁of▁sentence｜>'
llm_load_print_meta: PAD token        = 128815 '<｜PAD▁TOKEN｜>'
llm_load_print_meta: LF token         = 131 'Ä'
llm_load_print_meta: FIM PRE token    = 128801 '<｜fim▁begin｜>'
llm_load_print_meta: FIM SUF token    = 128800 '<｜fim▁hole｜>'
llm_load_print_meta: FIM MID token    = 128802 '<｜fim▁end｜>'
llm_load_print_meta: EOG token        = 1 '<｜end▁of▁sentence｜>'
llm_load_print_meta: max token length = 256
llm_load_print_meta: n_layer_dense_lead   = 3
llm_load_print_meta: n_lora_q             = 1536
llm_load_print_meta: n_lora_kv            = 512
llm_load_print_meta: n_ff_exp             = 2048
llm_load_print_meta: n_expert_shared      = 1
llm_load_print_meta: expert_weights_scale = 2.5
llm_load_print_meta: expert_weights_norm  = 1
llm_load_print_meta: expert_gating_func   = sigmoid
llm_load_print_meta: rope_yarn_log_mul    = 0.1000
llm_load_tensors: offloading 16 repeating layers to GPU
llm_load_tensors: offloaded 16/62 layers to GPU
llm_load_tensors:   CPU_Mapped model buffer size = 276620.97 MiB
llm_load_tensors:        CUDA0 model buffer size = 13285.73 MiB
llm_load_tensors:        CUDA1 model buffer size = 13285.73 MiB
llm_load_tensors:        CUDA2 model buffer size = 12358.12 MiB
llm_load_tensors:        CUDA3 model buffer size = 13285.73 MiB
llm_load_tensors:        CUDA4 model buffer size = 14213.34 MiB
llm_load_tensors:        CUDA5 model buffer size = 14213.34 MiB
llm_load_tensors:        CUDA6 model buffer size = 14213.34 MiB
llm_load_tensors:        CUDA7 model buffer size = 14213.34 MiB
time=2025-02-11T12:53:14.304Z level=INFO source=server.go:589 msg="waiting for server to become available" status="llm server not responding"
llama_new_context_with_model: n_seq_max     = 1
llama_new_context_with_model: n_ctx         = 4096
llama_new_context_with_model: n_ctx_per_seq = 4096
llama_new_context_with_model: n_batch       = 512
llama_new_context_with_model: n_ubatch      = 512
llama_new_context_with_model: flash_attn    = 0
llama_new_context_with_model: freq_base     = 10000.0
llama_new_context_with_model: freq_scale    = 0.025
llama_new_context_with_model: n_ctx_per_seq (4096) < n_ctx_train (163840) -- the full capacity of the model will not be utilized
llama_kv_cache_init: kv_size = 4096, offload = 1, type_k = 'f16', type_v = 'f16', n_layer = 61, can_shift = 0
time=2025-02-11T12:53:17.259Z level=INFO source=server.go:589 msg="waiting for server to become available" status="llm server loading model"
llama_kv_cache_init:        CPU KV buffer size = 14400.00 MiB
llama_kv_cache_init:      CUDA0 KV buffer size =   640.00 MiB
llama_kv_cache_init:      CUDA1 KV buffer size =   640.00 MiB
llama_kv_cache_init:      CUDA2 KV buffer size =   640.00 MiB
llama_kv_cache_init:      CUDA3 KV buffer size =   640.00 MiB
llama_kv_cache_init:      CUDA4 KV buffer size =   640.00 MiB
llama_kv_cache_init:      CUDA5 KV buffer size =   640.00 MiB
llama_kv_cache_init:      CUDA6 KV buffer size =   640.00 MiB
llama_kv_cache_init:      CUDA7 KV buffer size =   640.00 MiB
llama_new_context_with_model: KV self size  = 19520.00 MiB, K (f16): 11712.00 MiB, V (f16): 7808.00 MiB
llama_new_context_with_model:        CPU  output buffer size =     0.52 MiB
llama_new_context_with_model:      CUDA0 compute buffer size =  5030.00 MiB
llama_new_context_with_model:      CUDA1 compute buffer size =  1186.00 MiB
llama_new_context_with_model:      CUDA2 compute buffer size =  1186.00 MiB
llama_new_context_with_model:      CUDA3 compute buffer size =  1186.00 MiB
llama_new_context_with_model:      CUDA4 compute buffer size =  1186.00 MiB
llama_new_context_with_model:      CUDA5 compute buffer size =  1186.00 MiB
llama_new_context_with_model:      CUDA6 compute buffer size =  1186.00 MiB
llama_new_context_with_model:      CUDA7 compute buffer size =  1186.00 MiB
llama_new_context_with_model:  CUDA_Host compute buffer size =    88.01 MiB
llama_new_context_with_model: graph nodes  = 5025

OS

Linux

GPU

Nvidia

CPU

No response

Ollama version

ollama version is 0.5.7

Originally created by @Oldpan on GitHub (Feb 12, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/9028 ### What is the issue? When I was running deepseek-r1 using Ollama, I occasionally encountered some garbled output, such as ';0?!18=1C%<｜▁pad▁｜>-DB', but sometimes it worked correctly. What could be the possible reasons? My setup is 8x3090 GPUs with 500GB of RAM. My env and scripit: ollama create DeepSeek-R1-Q4_K_M -f /code/ollama/modelfile/DeepSeek-R1-Q4-K_M ``` FROM /code/DeepSeek-R1-int4/DeepSeek-R1-Q4_K_M.gguf PARAMETER num_gpu 16   PARAMETER num_ctx 4096   PARAMETER temperature 0.6 PARAMETER stop <｜begin▁of▁sentence｜> PARAMETER stop <｜end▁of▁sentence｜> PARAMETER stop <｜User｜> PARAMETER stop <｜Assistant｜> TEMPLATE """{{- if .System }}{{ .System }}{{ end }} {{- range $i, $_ := .Messages }} {{- $last := eq (len (slice $.Messages $i)) 1}} {{- if eq .Role "user" }}<｜User｜>{{ .Content }} {{- else if eq .Role "assistant" }}<｜Assistant｜>{{ .Content }}{{- if not $last }}<｜end▁of▁sentence｜>{{- end }} {{- end }} {{- if and $last (ne .Role "assistant") }}<｜Assistant｜>{{- end }} {{- end }}""" ``` then ollama run DeepSeek-R1-Q4_K_M:latest --verbose Thanks ### Relevant log output ```shell time=2025-02-11T12:52:33.956Z level=INFO source=sched.go:730 msg="new model will fit in available VRAM, loading" model=/code/ollama/ollama_home/blobs/sha256-0b898a049c60a2d4b2f373b21e19b3efea4aa81e6d403f93f5a41354cba304c5 library=cuda parallel=1 required="112.3 GiB" time=2025-02-11T12:52:34.750Z level=INFO source=server.go:104 msg="system memory" total="753.4 GiB" free="710.4 GiB" free_swap="0 B" time=2025-02-11T12:52:34.751Z level=INFO source=memory.go:356 msg="offload to cuda" layers.requested=16 layers.model=62 layers.offload=16 layers.split=2,2,2,2,2,2,2,2 memory.available="[23.4 GiB 23.4 GiB 23.4 GiB 23.4 GiB 23.4 GiB 23.4 GiB 23.4 GiB 23.4 GiB]" memory.gpu_overhead="0 B" memory.required.full="438.7 GiB" memory.required.partial="112.3 GiB" memory.required.kv="19.1 GiB" memory.required.allocations="[9.7 GiB 10.6 GiB 9.7 GiB 16.3 GiB 17.2 GiB 16.3 GiB 16.3 GiB 16.3 GiB]" memory.weights.total="394.5 GiB" memory.weights.repeating="393.8 GiB" memory.weights.nonrepeating="725.0 MiB" memory.graph.full="1.6 GiB" memory.graph.partial="1.6 GiB" time=2025-02-11T12:52:34.751Z level=INFO source=server.go:376 msg="starting llama server" cmd="/usr/local/lib/ollama/runners/cuda_v12_avx/ollama_llama_server runner --model /code/ollama/ollama_home/blobs/sha256-0b898a049c60a2d4b2f373b21e19b3efea4aa81e6d403f93f5a41354cba304c5 --ctx-size 4096 --batch-size 512 --n-gpu-layers 16 --threads 48 --parallel 1 --tensor-split 2,2,2,2,2,2,2,2 --port 43589" time=2025-02-11T12:52:34.752Z level=INFO source=sched.go:449 msg="loaded runners" count=1 time=2025-02-11T12:52:34.752Z level=INFO source=server.go:555 msg="waiting for llama runner to start responding" time=2025-02-11T12:52:34.752Z level=INFO source=server.go:589 msg="waiting for server to become available" status="llm server error" time=2025-02-11T12:52:34.792Z level=INFO source=runner.go:936 msg="starting go runner" ggml_cuda_init: GGML_CUDA_FORCE_MMQ: no ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no ggml_cuda_init: found 8 CUDA devices: Device 0: NVIDIA GeForce RTX 3090, compute capability 8.6, VMM: yes Device 1: NVIDIA GeForce RTX 3090, compute capability 8.6, VMM: yes Device 2: NVIDIA GeForce RTX 3090, compute capability 8.6, VMM: yes Device 3: NVIDIA GeForce RTX 3090, compute capability 8.6, VMM: yes Device 4: NVIDIA GeForce RTX 3090, compute capability 8.6, VMM: yes Device 5: NVIDIA GeForce RTX 3090, compute capability 8.6, VMM: yes Device 6: NVIDIA GeForce RTX 3090, compute capability 8.6, VMM: yes Device 7: NVIDIA GeForce RTX 3090, compute capability 8.6, VMM: yes time=2025-02-11T12:52:35.348Z level=INFO source=runner.go:937 msg=system info="CUDA : ARCHS = 600,610,620,700,720,750,800,860,870,890,900 | USE_GRAPHS = 1 | PEER_MAX_BATCH_SIZE = 128 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | LLAMAFILE = 1 | AARCH64_REPACK = 1 | cgo(gcc)" threads=48 time=2025-02-11T12:52:35.349Z level=INFO source=.:0 msg="Server listening on 127.0.0.1:43589" llama_load_model_from_file: using device CUDA0 (NVIDIA GeForce RTX 3090) - 23997 MiB free llama_load_model_from_file: using device CUDA1 (NVIDIA GeForce RTX 3090) - 23997 MiB free llama_load_model_from_file: using device CUDA2 (NVIDIA GeForce RTX 3090) - 23997 MiB free llama_load_model_from_file: using device CUDA3 (NVIDIA GeForce RTX 3090) - 23997 MiB free llama_load_model_from_file: using device CUDA4 (NVIDIA GeForce RTX 3090) - 23997 MiB free llama_load_model_from_file: using device CUDA5 (NVIDIA GeForce RTX 3090) - 23997 MiB free llama_load_model_from_file: using device CUDA6 (NVIDIA GeForce RTX 3090) - 23997 MiB free llama_load_model_from_file: using device CUDA7 (NVIDIA GeForce RTX 3090) - 23997 MiB free llama_model_loader: loaded meta data with 48 key-value pairs and 1025 tensors from /code/ollama/ollama_home/blobs/sha256-0b898a049c60a2d4b2f373b21e19b3efea4aa81e6d403f93f5a41354cba304c5 (version GGUF V3 (latest)) llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output. llama_model_loader: - kv 0: general.architecture str = deepseek2 llama_model_loader: - kv 1: general.type str = model llama_model_loader: - kv 2: general.name str = DeepSeek R1 BF16 llama_model_loader: - kv 3: general.quantized_by str = Unsloth llama_model_loader: - kv 4: general.size_label str = 256x20B llama_model_loader: - kv 5: general.repo_url str = https://huggingface.co/unsloth llama_model_loader: - kv 6: deepseek2.block_count u32 = 61 llama_model_loader: - kv 7: deepseek2.context_length u32 = 163840 llama_model_loader: - kv 8: deepseek2.embedding_length u32 = 7168 llama_model_loader: - kv 9: deepseek2.feed_forward_length u32 = 18432 llama_model_loader: - kv 10: deepseek2.attention.head_count u32 = 128 llama_model_loader: - kv 11: deepseek2.attention.head_count_kv u32 = 128 llama_model_loader: - kv 12: deepseek2.rope.freq_base f32 = 10000.000000 llama_model_loader: - kv 13: deepseek2.attention.layer_norm_rms_epsilon f32 = 0.000001 llama_model_loader: - kv 14: deepseek2.expert_used_count u32 = 8 llama_model_loader: - kv 15: deepseek2.leading_dense_block_count u32 = 3 llama_model_loader: - kv 16: deepseek2.vocab_size u32 = 129280 llama_model_loader: - kv 17: deepseek2.attention.q_lora_rank u32 = 1536 llama_model_loader: - kv 18: deepseek2.attention.kv_lora_rank u32 = 512 llama_model_loader: - kv 19: deepseek2.attention.key_length u32 = 192 llama_model_loader: - kv 20: deepseek2.attention.value_length u32 = 128 llama_model_loader: - kv 21: deepseek2.expert_feed_forward_length u32 = 2048 llama_model_loader: - kv 22: deepseek2.expert_count u32 = 256 llama_model_loader: - kv 23: deepseek2.expert_shared_count u32 = 1 llama_model_loader: - kv 24: deepseek2.expert_weights_scale f32 = 2.500000 llama_model_loader: - kv 25: deepseek2.expert_weights_norm bool = true llama_model_loader: - kv 26: deepseek2.expert_gating_func u32 = 2 llama_model_loader: - kv 27: deepseek2.rope.dimension_count u32 = 64 llama_model_loader: - kv 28: deepseek2.rope.scaling.type str = yarn llama_model_loader: - kv 29: deepseek2.rope.scaling.factor f32 = 40.000000 llama_model_loader: - kv 30: deepseek2.rope.scaling.original_context_length u32 = 4096 llama_model_loader: - kv 31: deepseek2.rope.scaling.yarn_log_multiplier f32 = 0.100000 llama_model_loader: - kv 32: tokenizer.ggml.model str = gpt2 llama_model_loader: - kv 33: tokenizer.ggml.pre str = deepseek-v3 time=2025-02-11T12:52:35.504Z level=INFO source=server.go:589 msg="waiting for server to become available" status="llm server loading model" llama_model_loader: - kv 34: tokenizer.ggml.tokens arr[str,129280] = ["<｜begin▁of▁sentence｜>", "<�... llama_model_loader: - kv 35: tokenizer.ggml.token_type arr[i32,129280] = [3, 3, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, ... llama_model_loader: - kv 36: tokenizer.ggml.merges arr[str,127741] = ["Ġ t", "Ġ a", "i n", "Ġ Ġ", "h e... llama_model_loader: - kv 37: tokenizer.ggml.bos_token_id u32 = 0 llama_model_loader: - kv 38: tokenizer.ggml.eos_token_id u32 = 1 llama_model_loader: - kv 39: tokenizer.ggml.padding_token_id u32 = 128815 llama_model_loader: - kv 40: tokenizer.ggml.add_bos_token bool = true llama_model_loader: - kv 41: tokenizer.ggml.add_eos_token bool = false llama_model_loader: - kv 42: tokenizer.chat_template str = {% if not add_generation_prompt is de... llama_model_loader: - kv 43: general.quantization_version u32 = 2 llama_model_loader: - kv 44: general.file_type u32 = 15 llama_model_loader: - kv 45: split.no u16 = 0 llama_model_loader: - kv 46: split.tensors.count i32 = 1025 llama_model_loader: - kv 47: split.count u16 = 0 llama_model_loader: - type f32: 361 tensors llama_model_loader: - type q4_K: 606 tensors llama_model_loader: - type q6_K: 58 tensors llm_load_vocab: special_eos_id is not in special_eog_ids - the tokenizer config may be incorrect llm_load_vocab: special tokens cache size = 819 llm_load_vocab: token to piece cache size = 0.8223 MB llm_load_print_meta: format = GGUF V3 (latest) llm_load_print_meta: arch = deepseek2 llm_load_print_meta: vocab type = BPE llm_load_print_meta: n_vocab = 129280 llm_load_print_meta: n_merges = 127741 llm_load_print_meta: vocab_only = 0 llm_load_print_meta: n_ctx_train = 163840 llm_load_print_meta: n_embd = 7168 llm_load_print_meta: n_layer = 61 llm_load_print_meta: n_head = 128 llm_load_print_meta: n_head_kv = 128 llm_load_print_meta: n_rot = 64 llm_load_print_meta: n_swa = 0 llm_load_print_meta: n_embd_head_k = 192 llm_load_print_meta: n_embd_head_v = 128 llm_load_print_meta: n_gqa = 1 llm_load_print_meta: n_embd_k_gqa = 24576 llm_load_print_meta: n_embd_v_gqa = 16384 llm_load_print_meta: f_norm_eps = 0.0e+00 llm_load_print_meta: f_norm_rms_eps = 1.0e-06 llm_load_print_meta: f_clamp_kqv = 0.0e+00 llm_load_print_meta: f_max_alibi_bias = 0.0e+00 llm_load_print_meta: f_logit_scale = 0.0e+00 llm_load_print_meta: n_ff = 18432 llm_load_print_meta: n_expert = 256 llm_load_print_meta: n_expert_used = 8 llm_load_print_meta: causal attn = 1 llm_load_print_meta: pooling type = 0 llm_load_print_meta: rope type = 0 llm_load_print_meta: rope scaling = yarn llm_load_print_meta: freq_base_train = 10000.0 llm_load_print_meta: freq_scale_train = 0.025 llm_load_print_meta: n_ctx_orig_yarn = 4096 llm_load_print_meta: rope_finetuned = unknown llm_load_print_meta: ssm_d_conv = 0 llm_load_print_meta: ssm_d_inner = 0 llm_load_print_meta: ssm_d_state = 0 llm_load_print_meta: ssm_dt_rank = 0 llm_load_print_meta: ssm_dt_b_c_rms = 0 llm_load_print_meta: model type = 671B llm_load_print_meta: model ftype = Q4_K - Medium llm_load_print_meta: model params = 671.03 B llm_load_print_meta: model size = 376.65 GiB (4.82 BPW) llm_load_print_meta: general.name = DeepSeek R1 BF16 llm_load_print_meta: BOS token = 0 '<｜begin▁of▁sentence｜>' llm_load_print_meta: EOS token = 1 '<｜end▁of▁sentence｜>' llm_load_print_meta: EOT token = 1 '<｜end▁of▁sentence｜>' llm_load_print_meta: PAD token = 128815 '<｜PAD▁TOKEN｜>' llm_load_print_meta: LF token = 131 'Ä' llm_load_print_meta: FIM PRE token = 128801 '<｜fim▁begin｜>' llm_load_print_meta: FIM SUF token = 128800 '<｜fim▁hole｜>' llm_load_print_meta: FIM MID token = 128802 '<｜fim▁end｜>' llm_load_print_meta: EOG token = 1 '<｜end▁of▁sentence｜>' llm_load_print_meta: max token length = 256 llm_load_print_meta: n_layer_dense_lead = 3 llm_load_print_meta: n_lora_q = 1536 llm_load_print_meta: n_lora_kv = 512 llm_load_print_meta: n_ff_exp = 2048 llm_load_print_meta: n_expert_shared = 1 llm_load_print_meta: expert_weights_scale = 2.5 llm_load_print_meta: expert_weights_norm = 1 llm_load_print_meta: expert_gating_func = sigmoid llm_load_print_meta: rope_yarn_log_mul = 0.1000 llm_load_tensors: offloading 16 repeating layers to GPU llm_load_tensors: offloaded 16/62 layers to GPU llm_load_tensors: CPU_Mapped model buffer size = 276620.97 MiB llm_load_tensors: CUDA0 model buffer size = 13285.73 MiB llm_load_tensors: CUDA1 model buffer size = 13285.73 MiB llm_load_tensors: CUDA2 model buffer size = 12358.12 MiB llm_load_tensors: CUDA3 model buffer size = 13285.73 MiB llm_load_tensors: CUDA4 model buffer size = 14213.34 MiB llm_load_tensors: CUDA5 model buffer size = 14213.34 MiB llm_load_tensors: CUDA6 model buffer size = 14213.34 MiB llm_load_tensors: CUDA7 model buffer size = 14213.34 MiB time=2025-02-11T12:53:14.304Z level=INFO source=server.go:589 msg="waiting for server to become available" status="llm server not responding" llama_new_context_with_model: n_seq_max = 1 llama_new_context_with_model: n_ctx = 4096 llama_new_context_with_model: n_ctx_per_seq = 4096 llama_new_context_with_model: n_batch = 512 llama_new_context_with_model: n_ubatch = 512 llama_new_context_with_model: flash_attn = 0 llama_new_context_with_model: freq_base = 10000.0 llama_new_context_with_model: freq_scale = 0.025 llama_new_context_with_model: n_ctx_per_seq (4096) < n_ctx_train (163840) -- the full capacity of the model will not be utilized llama_kv_cache_init: kv_size = 4096, offload = 1, type_k = 'f16', type_v = 'f16', n_layer = 61, can_shift = 0 time=2025-02-11T12:53:17.259Z level=INFO source=server.go:589 msg="waiting for server to become available" status="llm server loading model" llama_kv_cache_init: CPU KV buffer size = 14400.00 MiB llama_kv_cache_init: CUDA0 KV buffer size = 640.00 MiB llama_kv_cache_init: CUDA1 KV buffer size = 640.00 MiB llama_kv_cache_init: CUDA2 KV buffer size = 640.00 MiB llama_kv_cache_init: CUDA3 KV buffer size = 640.00 MiB llama_kv_cache_init: CUDA4 KV buffer size = 640.00 MiB llama_kv_cache_init: CUDA5 KV buffer size = 640.00 MiB llama_kv_cache_init: CUDA6 KV buffer size = 640.00 MiB llama_kv_cache_init: CUDA7 KV buffer size = 640.00 MiB llama_new_context_with_model: KV self size = 19520.00 MiB, K (f16): 11712.00 MiB, V (f16): 7808.00 MiB llama_new_context_with_model: CPU output buffer size = 0.52 MiB llama_new_context_with_model: CUDA0 compute buffer size = 5030.00 MiB llama_new_context_with_model: CUDA1 compute buffer size = 1186.00 MiB llama_new_context_with_model: CUDA2 compute buffer size = 1186.00 MiB llama_new_context_with_model: CUDA3 compute buffer size = 1186.00 MiB llama_new_context_with_model: CUDA4 compute buffer size = 1186.00 MiB llama_new_context_with_model: CUDA5 compute buffer size = 1186.00 MiB llama_new_context_with_model: CUDA6 compute buffer size = 1186.00 MiB llama_new_context_with_model: CUDA7 compute buffer size = 1186.00 MiB llama_new_context_with_model: CUDA_Host compute buffer size = 88.01 MiB llama_new_context_with_model: graph nodes = 5025 ``` ### OS Linux ### GPU Nvidia ### CPU _No response_ ### Ollama version ollama version is 0.5.7
GiteaMirror added the bug label 2026-04-12 17:12:40 -05:00
GiteaMirror commented 2026-04-12 17:12:41 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Feb 12, 2025):

Is there any particular prompt that causes garbled output? During short or long query/response? Do you have continuous sessions (ie, multiple queries/responses)? If so, does the garbled output happen later on or randomly within the sessions? Do the logs show the model being unloaded and re-loaded? Does the garbled text show if you run the model purely in system RAM (num_gpu:0)?

<!-- gh-comment-id:2653260110 --> @rick-github commented on GitHub (Feb 12, 2025): Is there any particular prompt that causes garbled output? During short or long query/response? Do you have continuous sessions (ie, multiple queries/responses)? If so, does the garbled output happen later on or randomly within the sessions? Do the logs show the model being unloaded and re-loaded? Does the garbled text show if you run the model purely in system RAM (num_gpu:0)?
GiteaMirror commented 2026-04-12 17:12:41 -05:00
Author
Owner
Copy Link

@Oldpan commented on GitHub (Feb 12, 2025):

Is there any particular prompt that causes garbled output? During short or long query/response? Do you have continuous sessions (ie, multiple queries/responses)? If so, does the garbled output happen later on or randomly within the sessions? Do the logs show the model being unloaded and re-loaded? Does the garbled text show if you run the model purely in system RAM

Is there a specific prompt that causes garbled output?
Yes, certain prompts can cause garbled output.

Does this happen during short or long queries/responses?
It occurs with both.

Do you have continuous sessions (i.e., multiple queries/responses)?
No, there is only one session.

When the model is reloaded, the output is fine at first, but eventually, it becomes garbled again and remains garbled until the model is reloaded.

If I run the model purely in system RAM without using a GPU, the garbled output seems to disappear.

<!-- gh-comment-id:2653816682 --> @Oldpan commented on GitHub (Feb 12, 2025): > Is there any particular prompt that causes garbled output? During short or long query/response? Do you have continuous sessions (ie, multiple queries/responses)? If so, does the garbled output happen later on or randomly within the sessions? Do the logs show the model being unloaded and re-loaded? Does the garbled text show if you run the model purely in system RAM Is there a specific prompt that causes garbled output? Yes, certain prompts can cause garbled output. Does this happen during short or long queries/responses? It occurs with both. Do you have continuous sessions (i.e., multiple queries/responses)? No, there is only one session. When the model is reloaded, the output is fine at first, but eventually, it becomes garbled again and remains garbled until the model is reloaded. If I run the model purely in system RAM without using a GPU, the garbled output seems to disappear.
GiteaMirror commented 2026-04-12 17:12:42 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Feb 12, 2025):

It sounds like you are running in to a context size issue - one session, and when it gets garbled it stays that way until restarted. The only point against that is normally when a deepseek-r1 model fills up the context space it fails because k-shift is not supported. You can try increasing num_ctx and see if it takes longer for the output to become garbled.

<!-- gh-comment-id:2653873596 --> @rick-github commented on GitHub (Feb 12, 2025): It sounds like you are running in to a context size issue - one session, and when it gets garbled it stays that way until restarted. The only point against that is normally when a deepseek-r1 model fills up the context space it fails because k-shift is not supported. You can try increasing `num_ctx` and see if it takes longer for the output to become garbled.
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#5872
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?