github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-15 20:54:54 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #4437] Ollama vs Llama-cpp-python : Slow response time as compared to llama-cpp-python #28531

New Issue
Closed
opened 2026-04-22 06:46:39 -05:00 by GiteaMirror · 6 comments
GiteaMirror commented 2026-04-22 06:46:39 -05:00
Owner
Copy Link

Originally created by @anuragrawal2024 on GitHub (May 14, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/4437

Originally assigned to: @dhiltgen on GitHub.

What is the issue?

Hi,

I built a RAG Q&A pipeline using LlamaIndex and Llama-cpp-python in the past. I want to switch from llama-cpp to ollama because ollama is more stable and easier to install. When I made the switch, I noticed a significant increase in response time. Would you know what might cause this slowdown?

I have kept everything same for the comparison and have only changed llm component to point to ollama instead of llama-cpp. I am using the same quantized version of llama-3 for comparison.

Appreciate your inputs! Thanks!

OS

Linux

GPU

Nvidia

CPU

Intel

Ollama version

0.1.33

Originally created by @anuragrawal2024 on GitHub (May 14, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/4437 Originally assigned to: @dhiltgen on GitHub. ### What is the issue? Hi, I built a RAG Q&A pipeline using LlamaIndex and Llama-cpp-python in the past. I want to switch from llama-cpp to ollama because ollama is more stable and easier to install. When I made the switch, I noticed a significant increase in response time. Would you know what might cause this slowdown? I have kept everything same for the comparison and have only changed llm component to point to ollama instead of llama-cpp. I am using the same quantized version of llama-3 for comparison. Appreciate your inputs! Thanks! ### OS Linux ### GPU Nvidia ### CPU Intel ### Ollama version 0.1.33
GiteaMirror added the gpunvidiabug labels 2026-04-22 06:46:40 -05:00
GiteaMirror commented 2026-04-22 06:46:42 -05:00
Author
Owner
Copy Link

@pdevine commented on GitHub (May 14, 2024):

@utility-aagrawal What kind of GPU are you using? Is the model offloaded entirely onto the GPU or only partially?

<!-- gh-comment-id:2111061250 --> @pdevine commented on GitHub (May 14, 2024): @utility-aagrawal What kind of GPU are you using? Is the model offloaded entirely onto the GPU or only partially?
GiteaMirror commented 2026-04-22 06:46:43 -05:00
Author
Owner
Copy Link

@anuragrawal2024 commented on GitHub (May 14, 2024):

Thanks for your response, @pdevine!

I have an NVIDIA GeForce RTX 3060 (12 GB VRAM). Yes, I am using these parameters {"n_gpu_layers": -1, "offload_kqv": True} which offload the entire model onto the GPU. Let me know if you need anything additional. Thanks!

Also, I wanted to quantify the slowdown. Using ollama, it's taking 1.5-2x compared to llama-cpp.

Adding server logs to confirm that the entire model is offloaded to GPU:

May 14 16:34:43 anurag-Legion-T5-26IRB8 ollama[1726]: llm_load_tensors: ggml ctx size = 0.30 MiB
May 14 16:34:43 anurag-Legion-T5-26IRB8 ollama[1726]: llm_load_tensors: offloading 32 repeating layers to GPU
May 14 16:34:43 anurag-Legion-T5-26IRB8 ollama[1726]: llm_load_tensors: offloading non-repeating layers to GPU
May 14 16:34:43 anurag-Legion-T5-26IRB8 ollama[1726]: llm_load_tensors: offloaded 33/33 layers to GPU
May 14 16:34:43 anurag-Legion-T5-26IRB8 ollama[1726]: llm_load_tensors: CPU buffer size = 281.81 MiB
May 14 16:34:43 anurag-Legion-T5-26IRB8 ollama[1726]: llm_load_tensors: CUDA0 buffer size = 4155.99 MiB
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: .......................................................................................
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_new_context_with_model: n_ctx = 8000
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_new_context_with_model: n_batch = 512
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_new_context_with_model: n_ubatch = 512
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_new_context_with_model: freq_base = 500000.0
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_new_context_with_model: freq_scale = 1
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_kv_cache_init: CUDA0 KV buffer size = 1000.00 MiB
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_new_context_with_model: KV self size = 1000.00 MiB, K (f16): 500.00 MiB, V (f16): 500.00 MiB
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_new_context_with_model: CUDA_Host output buffer size = 0.50 MiB
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_new_context_with_model: CUDA0 compute buffer size = 547.63 MiB
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_new_context_with_model: CUDA_Host compute buffer size = 23.63 MiB
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_new_context_with_model: graph nodes = 1030
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_new_context_with_model: graph splits = 2
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"initialize","level":"INFO","line":448,"msg":"initializing slots","n_slots":1,"tid":"124142108921856","timestamp":1715718884}
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"initialize","level":"INFO","line":457,"msg":"new slot","n_ctx_slot":8000,"slot_id":0,"tid":"124142108921856","timestamp":1715718884}
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"main","level":"INFO","line":3067,"msg":"model loaded","tid":"124142108921856","timestamp":1715718884}
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"main","hostname":"127.0.0.1","level":"INFO","line":3270,"msg":"HTTP server listening","n_threads_http":"23","port":"42843","tid":"124142108921856","timestamp":1715718884}
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"update_slots","level":"INFO","line":1581,"msg":"all slots are idle and system prompt is empty, clear the KV cache","tid":"124142108921856","timestamp":1715718884}
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"process_single_task","level":"INFO","line":1509,"msg":"slot data","n_idle_slots":1,"n_processing_slots":0,"task_id":0,"tid":"124142108921856","timestamp":1715718884}
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"process_single_task","level":"INFO","line":1509,"msg":"slot data","n_idle_slots":1,"n_processing_slots":0,"task_id":1,"tid":"124142108921856","timestamp":1715718884}
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"log_server_request","level":"INFO","line":2737,"method":"GET","msg":"request","params":{},"path":"/health","remote_addr":"127.0.0.1","remote_port":47896,"status":200,"tid":"124139184582656","timestamp":171>
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"process_single_task","level":"INFO","line":1509,"msg":"slot data","n_idle_slots":1,"n_processing_slots":0,"task_id":2,"tid":"124142108921856","timestamp":1715718884}
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"process_single_task","level":"INFO","line":1509,"msg":"slot data","n_idle_slots":1,"n_processing_slots":0,"task_id":3,"tid":"124142108921856","timestamp":1715718884}
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"log_server_request","level":"INFO","line":2737,"method":"GET","msg":"request","params":{},"path":"/health","remote_addr":"127.0.0.1","remote_port":47910,"status":200,"tid":"124138672824320","timestamp":171>
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"process_single_task","level":"INFO","line":1509,"msg":"slot data","n_idle_slots":1,"n_processing_slots":0,"task_id":4,"tid":"124142108921856","timestamp":1715718884}
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"log_server_request","level":"INFO","line":2737,"method":"GET","msg":"request","params":{},"path":"/health","remote_addr":"127.0.0.1","remote_port":47928,"status":200,"tid":"124139184582656","timestamp":171>
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"log_server_request","level":"INFO","line":2737,"method":"GET","msg":"request","params":{},"path":"/health","remote_addr":"127.0.0.1","remote_port":47944,"status":200,"tid":"124141474951168","timestamp":171>
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"log_server_request","level":"INFO","line":2737,"method":"GET","msg":"request","params":{},"path":"/health","remote_addr":"127.0.0.1","remote_port":47926,"status":200,"tid":"124138681217024","timestamp":171>
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"process_single_task","level":"INFO","line":1509,"msg":"slot data","n_idle_slots":1,"n_processing_slots":0,"task_id":5,"tid":"124142108921856","timestamp":1715718884}
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"log_server_request","level":"INFO","line":2737,"method":"GET","msg":"request","params":{},"path":"/health","remote_addr":"127.0.0.1","remote_port":47954,"status":200,"tid":"124138672824320","timestamp":171>
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"process_single_task","level":"INFO","line":1509,"msg":"slot data","n_idle_slots":1,"n_processing_slots":0,"task_id":6,"tid":"124142108921856","timestamp":1715718884}
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"log_server_request","level":"INFO","line":2737,"method":"GET","msg":"request","params":{},"path":"/health","remote_addr":"127.0.0.1","remote_port":47968,"status":200,"tid":"124139096690688","timestamp":171>
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"process_single_task","level":"INFO","line":1509,"msg":"slot data","n_idle_slots":1,"n_processing_slots":0,"task_id":7,"tid":"124142108921856","timestamp":1715718884}
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"log_server_request","level":"INFO","line":2737,"method":"GET","msg":"request","params":{},"path":"/health","remote_addr":"127.0.0.1","remote_port":47968,"status":200,"tid":"124139096690688","timestamp":171>
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"log_server_request","level":"INFO","line":2737,"method":"POST","msg":"request","params":{},"path":"/tokenize","remote_addr":"127.0.0.1","remote_port":47968,"status":200,"tid":"124139096690688","timestamp":>
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"process_single_task","level":"INFO","line":1509,"msg":"slot data","n_idle_slots":1,"n_processing_slots":0,"task_id":8,"tid":"124142108921856","timestamp":1715718884}
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"log_server_request","level":"INFO","line":2737,"method":"GET","msg":"request","params":{},"path":"/health","remote_addr":"127.0.0.1","remote_port":47968,"status":200,"tid":"124139096690688","timestamp":171>
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"launch_slot_with_data","level":"INFO","line":830,"msg":"slot is processing task","slot_id":0,"task_id":9,"tid":"124142108921856","timestamp":1715718884}
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"update_slots","ga_i":0,"level":"INFO","line":1812,"msg":"slot progression","n_past":0,"n_past_se":0,"n_prompt_tokens_processed":2161,"slot_id":0,"task_id":9,"tid":"124142108921856","timestamp":1715718884}
May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"update_slots","level":"INFO","line":1839,"msg":"kv cache rm [p0, end)","p0":0,"slot_id":0,"task_id":9,"tid":"124142108921856","timestamp":1715718884}
May 14 16:34:48 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"print_timings","level":"INFO","line":269,"msg":"prompt eval time = 2687.66 ms / 2161 tokens ( 1.24 ms per token, 804.04 tokens per second)","n_prompt_tokens_processed":2161,"n_tokens_second":8>
May 14 16:34:48 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"print_timings","level":"INFO","line":283,"msg":"generation eval time = 637.36 ms / 33 runs ( 19.31 ms per token, 51.78 tokens per second)","n_decoded":33,"n_tokens_second":51.77623878573738,">
May 14 16:34:48 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"print_timings","level":"INFO","line":293,"msg":" total time = 3325.02 ms","slot_id":0,"t_prompt_processing":2687.665,"t_token_generation":637.358,"t_total":3325.023,"task_id":9,"tid":"124142108>
May 14 16:34:48 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"update_slots","level":"INFO","line":1643,"msg":"slot released","n_cache_tokens":2194,"n_ctx":8000,"n_past":2193,"n_system_tokens":0,"slot_id":0,"task_id":9,"tid":"124142108921856","timestamp":1715718888,"t>
May 14 16:34:48 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"log_server_request","level":"INFO","line":2737,"method":"POST","msg":"request","params":{},"path":"/completion","remote_addr":"127.0.0.1","remote_port":47968,"status":200,"tid":"124139096690688","timestamp>
May 14 16:34:48 anurag-Legion-T5-26IRB8 ollama[1726]: [GIN] 2024/05/14 - 16:34:48 | 200 | 5.512035902s | 127.0.0.1 | POST "/api/chat"
May 14 17:01:55 anurag-Legion-T5-26IRB8 ollama[1726]: [GIN] 2024/05/14 - 17:01:55 | 200 | 16.467µs | 127.0.0.1 | HEAD "/"
May 14 17:01:55 anurag-Legion-T5-26IRB8 ollama[1726]: [GIN] 2024/05/14 - 17:01:55 | 200 | 211.634µs | 127.0.0.1 | GET "/api/tags"

<!-- gh-comment-id:2111087516 --> @anuragrawal2024 commented on GitHub (May 14, 2024): Thanks for your response, @pdevine! I have an NVIDIA GeForce RTX 3060 (12 GB VRAM). Yes, I am using these parameters {"n_gpu_layers": -1, "offload_kqv": True} which offload the entire model onto the GPU. Let me know if you need anything additional. Thanks! Also, I wanted to quantify the slowdown. Using ollama, it's taking 1.5-2x compared to llama-cpp. Adding server logs to confirm that the entire model is offloaded to GPU: May 14 16:34:43 anurag-Legion-T5-26IRB8 ollama[1726]: llm_load_tensors: ggml ctx size = 0.30 MiB May 14 16:34:43 anurag-Legion-T5-26IRB8 ollama[1726]: llm_load_tensors: offloading 32 repeating layers to GPU May 14 16:34:43 anurag-Legion-T5-26IRB8 ollama[1726]: llm_load_tensors: offloading non-repeating layers to GPU May 14 16:34:43 anurag-Legion-T5-26IRB8 ollama[1726]: llm_load_tensors: **offloaded 33/33 layers to GPU** May 14 16:34:43 anurag-Legion-T5-26IRB8 ollama[1726]: llm_load_tensors: CPU buffer size = 281.81 MiB May 14 16:34:43 anurag-Legion-T5-26IRB8 ollama[1726]: llm_load_tensors: CUDA0 buffer size = 4155.99 MiB May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: ....................................................................................... May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_new_context_with_model: n_ctx = 8000 May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_new_context_with_model: n_batch = 512 May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_new_context_with_model: n_ubatch = 512 May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_new_context_with_model: freq_base = 500000.0 May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_new_context_with_model: freq_scale = 1 May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_kv_cache_init: CUDA0 KV buffer size = 1000.00 MiB May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_new_context_with_model: KV self size = 1000.00 MiB, K (f16): 500.00 MiB, V (f16): 500.00 MiB May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_new_context_with_model: CUDA_Host output buffer size = 0.50 MiB May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_new_context_with_model: CUDA0 compute buffer size = 547.63 MiB May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_new_context_with_model: CUDA_Host compute buffer size = 23.63 MiB May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_new_context_with_model: graph nodes = 1030 May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[1726]: llama_new_context_with_model: graph splits = 2 May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"initialize","level":"INFO","line":448,"msg":"initializing slots","n_slots":1,"tid":"124142108921856","timestamp":1715718884} May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"initialize","level":"INFO","line":457,"msg":"new slot","n_ctx_slot":8000,"slot_id":0,"tid":"124142108921856","timestamp":1715718884} May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"main","level":"INFO","line":3067,"msg":"model loaded","tid":"124142108921856","timestamp":1715718884} May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"main","hostname":"127.0.0.1","level":"INFO","line":3270,"msg":"HTTP server listening","n_threads_http":"23","port":"42843","tid":"124142108921856","timestamp":1715718884} May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"update_slots","level":"INFO","line":1581,"msg":"all slots are idle and system prompt is empty, clear the KV cache","tid":"124142108921856","timestamp":1715718884} May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"process_single_task","level":"INFO","line":1509,"msg":"slot data","n_idle_slots":1,"n_processing_slots":0,"task_id":0,"tid":"124142108921856","timestamp":1715718884} May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"process_single_task","level":"INFO","line":1509,"msg":"slot data","n_idle_slots":1,"n_processing_slots":0,"task_id":1,"tid":"124142108921856","timestamp":1715718884} May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"log_server_request","level":"INFO","line":2737,"method":"GET","msg":"request","params":{},"path":"/health","remote_addr":"127.0.0.1","remote_port":47896,"status":200,"tid":"124139184582656","timestamp":171> May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"process_single_task","level":"INFO","line":1509,"msg":"slot data","n_idle_slots":1,"n_processing_slots":0,"task_id":2,"tid":"124142108921856","timestamp":1715718884} May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"process_single_task","level":"INFO","line":1509,"msg":"slot data","n_idle_slots":1,"n_processing_slots":0,"task_id":3,"tid":"124142108921856","timestamp":1715718884} May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"log_server_request","level":"INFO","line":2737,"method":"GET","msg":"request","params":{},"path":"/health","remote_addr":"127.0.0.1","remote_port":47910,"status":200,"tid":"124138672824320","timestamp":171> May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"process_single_task","level":"INFO","line":1509,"msg":"slot data","n_idle_slots":1,"n_processing_slots":0,"task_id":4,"tid":"124142108921856","timestamp":1715718884} May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"log_server_request","level":"INFO","line":2737,"method":"GET","msg":"request","params":{},"path":"/health","remote_addr":"127.0.0.1","remote_port":47928,"status":200,"tid":"124139184582656","timestamp":171> May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"log_server_request","level":"INFO","line":2737,"method":"GET","msg":"request","params":{},"path":"/health","remote_addr":"127.0.0.1","remote_port":47944,"status":200,"tid":"124141474951168","timestamp":171> May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"log_server_request","level":"INFO","line":2737,"method":"GET","msg":"request","params":{},"path":"/health","remote_addr":"127.0.0.1","remote_port":47926,"status":200,"tid":"124138681217024","timestamp":171> May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"process_single_task","level":"INFO","line":1509,"msg":"slot data","n_idle_slots":1,"n_processing_slots":0,"task_id":5,"tid":"124142108921856","timestamp":1715718884} May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"log_server_request","level":"INFO","line":2737,"method":"GET","msg":"request","params":{},"path":"/health","remote_addr":"127.0.0.1","remote_port":47954,"status":200,"tid":"124138672824320","timestamp":171> May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"process_single_task","level":"INFO","line":1509,"msg":"slot data","n_idle_slots":1,"n_processing_slots":0,"task_id":6,"tid":"124142108921856","timestamp":1715718884} May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"log_server_request","level":"INFO","line":2737,"method":"GET","msg":"request","params":{},"path":"/health","remote_addr":"127.0.0.1","remote_port":47968,"status":200,"tid":"124139096690688","timestamp":171> May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"process_single_task","level":"INFO","line":1509,"msg":"slot data","n_idle_slots":1,"n_processing_slots":0,"task_id":7,"tid":"124142108921856","timestamp":1715718884} May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"log_server_request","level":"INFO","line":2737,"method":"GET","msg":"request","params":{},"path":"/health","remote_addr":"127.0.0.1","remote_port":47968,"status":200,"tid":"124139096690688","timestamp":171> May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"log_server_request","level":"INFO","line":2737,"method":"POST","msg":"request","params":{},"path":"/tokenize","remote_addr":"127.0.0.1","remote_port":47968,"status":200,"tid":"124139096690688","timestamp":> May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"process_single_task","level":"INFO","line":1509,"msg":"slot data","n_idle_slots":1,"n_processing_slots":0,"task_id":8,"tid":"124142108921856","timestamp":1715718884} May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"log_server_request","level":"INFO","line":2737,"method":"GET","msg":"request","params":{},"path":"/health","remote_addr":"127.0.0.1","remote_port":47968,"status":200,"tid":"124139096690688","timestamp":171> May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"launch_slot_with_data","level":"INFO","line":830,"msg":"slot is processing task","slot_id":0,"task_id":9,"tid":"124142108921856","timestamp":1715718884} May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"update_slots","ga_i":0,"level":"INFO","line":1812,"msg":"slot progression","n_past":0,"n_past_se":0,"n_prompt_tokens_processed":2161,"slot_id":0,"task_id":9,"tid":"124142108921856","timestamp":1715718884} May 14 16:34:44 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"update_slots","level":"INFO","line":1839,"msg":"kv cache rm [p0, end)","p0":0,"slot_id":0,"task_id":9,"tid":"124142108921856","timestamp":1715718884} May 14 16:34:48 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"print_timings","level":"INFO","line":269,"msg":"prompt eval time = 2687.66 ms / 2161 tokens ( 1.24 ms per token, 804.04 tokens per second)","n_prompt_tokens_processed":2161,"n_tokens_second":8> May 14 16:34:48 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"print_timings","level":"INFO","line":283,"msg":"generation eval time = 637.36 ms / 33 runs ( 19.31 ms per token, 51.78 tokens per second)","n_decoded":33,"n_tokens_second":51.77623878573738,"> May 14 16:34:48 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"print_timings","level":"INFO","line":293,"msg":" total time = 3325.02 ms","slot_id":0,"t_prompt_processing":2687.665,"t_token_generation":637.358,"t_total":3325.023,"task_id":9,"tid":"124142108> May 14 16:34:48 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"update_slots","level":"INFO","line":1643,"msg":"slot released","n_cache_tokens":2194,"n_ctx":8000,"n_past":2193,"n_system_tokens":0,"slot_id":0,"task_id":9,"tid":"124142108921856","timestamp":1715718888,"t> May 14 16:34:48 anurag-Legion-T5-26IRB8 ollama[31689]: {"function":"log_server_request","level":"INFO","line":2737,"method":"POST","msg":"request","params":{},"path":"/completion","remote_addr":"127.0.0.1","remote_port":47968,"status":200,"tid":"124139096690688","timestamp> May 14 16:34:48 anurag-Legion-T5-26IRB8 ollama[1726]: [GIN] 2024/05/14 - 16:34:48 | 200 | 5.512035902s | 127.0.0.1 | POST "/api/chat" May 14 17:01:55 anurag-Legion-T5-26IRB8 ollama[1726]: [GIN] 2024/05/14 - 17:01:55 | 200 | 16.467µs | 127.0.0.1 | HEAD "/" May 14 17:01:55 anurag-Legion-T5-26IRB8 ollama[1726]: [GIN] 2024/05/14 - 17:01:55 | 200 | 211.634µs | 127.0.0.1 | GET "/api/tags"
GiteaMirror commented 2026-04-22 06:46:43 -05:00
Author
Owner
Copy Link

@anuragrawal2024 commented on GitHub (May 16, 2024):

@pdevine @dhiltgen Do you have any suggestions on this?

<!-- gh-comment-id:2115414156 --> @anuragrawal2024 commented on GitHub (May 16, 2024): @pdevine @dhiltgen Do you have any suggestions on this?
GiteaMirror commented 2026-04-22 06:46:44 -05:00
Author
Owner
Copy Link

@anuragrawal2024 commented on GitHub (Jun 18, 2024):

@pdevine @dhiltgen Reminder!

<!-- gh-comment-id:2176397313 --> @anuragrawal2024 commented on GitHub (Jun 18, 2024): @pdevine @dhiltgen Reminder!
GiteaMirror commented 2026-04-22 06:46:44 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Jun 18, 2024):

A few things to try to narrow this one down...

Try passing "use_mmap": false and see if that changes things.

Also can you share the llama.cpp configuration log line. In the Ollama logs it will look something like this:

INFO [main] system info | n_threads=4 n_threads_batch=-1 system_info="AVX = 1 | AVX_VNNI = 0 | AVX2 = 0 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | AVX512_BF16 = 0 | FMA = 0 | NEON = 0 | SVE = 0 | ARM_FMA = 0 | F16C = 0 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 | MATMUL_INT8 = 0 | LLAMAFILE = 1 | " tid="140021226209280" timestamp=1718729924 total_threads=8

Perhaps there's a material difference in the compilation of the llama.cpp that's leading to the performance difference.

<!-- gh-comment-id:2176568822 --> @dhiltgen commented on GitHub (Jun 18, 2024): A few things to try to narrow this one down... Try passing `"use_mmap": false` and see if that changes things. Also can you share the llama.cpp configuration log line. In the Ollama logs it will look something like this: ``` INFO [main] system info | n_threads=4 n_threads_batch=-1 system_info="AVX = 1 | AVX_VNNI = 0 | AVX2 = 0 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | AVX512_BF16 = 0 | FMA = 0 | NEON = 0 | SVE = 0 | ARM_FMA = 0 | F16C = 0 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 | MATMUL_INT8 = 0 | LLAMAFILE = 1 | " tid="140021226209280" timestamp=1718729924 total_threads=8 ``` Perhaps there's a material difference in the compilation of the llama.cpp that's leading to the performance difference.
GiteaMirror commented 2026-04-22 06:46:45 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Jul 3, 2024):

Let us know if you're still having performance problems.

<!-- gh-comment-id:2207468992 --> @dhiltgen commented on GitHub (Jul 3, 2024): Let us know if you're still having performance problems.
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 gpu nvidia
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#28531
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?