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 #11964] context size larger than set #54457

New Issue
Closed
opened 2026-04-29 06:00:41 -05:00 by GiteaMirror · 3 comments
GiteaMirror commented 2026-04-29 06:00:41 -05:00
Owner
Copy Link

Originally created by @StarPet on GitHub (Aug 19, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/11964

What is the issue?

In my python app I use 'num_ctx': 3182 to reduce the context size of the gpt-oss:20b model so it'll fit into my 2x RTX5060 (16GB). However the runner process gets a different value as --ctx-size parameter.

localai  3573918 3102624 99 11:46 ?        00:03:17 /home/localai/bin/ollama runner --ollama-engine --model /home/localai/.ollama/models/blobs/sha256-b112e727c6f18875636c56a779790a590d705aec9e1c0eb5a97d51fc2a778583 --ctx-size 40960 --batch-size 512 --n-gpu-layers 17 --threads 20 --parallel 5 --tensor-split 9,8 --port 34017

Note: before starting ollama, I used the following environment variable, too:

export OLLAMA_CONTEXT_LENGTH=2048

I checked the running process:

# strings /proc/3573918/environ |grep OLLAMA_CONTEXT
OLLAMA_CONTEXT_LENGTH=2048

So, the model is running on CPU not GPU.

Either I didn't really understand the relations of the OLLAMA_CONTEXT_LENGTH, the num_ctx parameter or there is an issue with recent changes (as it used to work just fine).

Relevant log output

time=2025-08-19T12:07:50.750+02:00 level=INFO source=routes.go:1304 msg="server config" env="map[CUDA_VISIBLE_DEVICES:0,1 GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:2048 OLLAMA_DEBUG:INFO OLLAMA_FLASH_ATTENTION:false OLLAMA_GPU_OVERHEAD:0 OLLAMA_HOST:http://0.0.0.0:11434 OLLAMA_INTEL_GPU:false OLLAMA_KEEP_ALIVE:5m0s OLLAMA_KV_CACHE_TYPE: OLLAMA_LLM_LIBRARY: OLLAMA_LOAD_TIMEOUT:5m0s OLLAMA_MAX_LOADED_MODELS:1 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:/home/localai/.ollama/models OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE:false OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NUM_PARALLEL:5 OLLAMA_ORIGINS:[http://10.1.0.65:8123 http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://* vscode-file://*] OLLAMA_SCHED_SPREAD:false ROCR_VISIBLE_DEVICES: http_proxy: https_proxy: no_proxy:]"
time=2025-08-19T12:07:50.768+02:00 level=INFO source=images.go:477 msg="total blobs: 446"
time=2025-08-19T12:07:50.771+02:00 level=INFO source=images.go:484 msg="total unused blobs removed: 0"
time=2025-08-19T12:07:50.772+02:00 level=INFO source=routes.go:1357 msg="Listening on [::]:11434 (version 0.11.4)"
time=2025-08-19T12:07:50.772+02:00 level=INFO source=gpu.go:217 msg="looking for compatible GPUs"
time=2025-08-19T12:07:51.572+02:00 level=INFO source=types.go:130 msg="inference compute" id=GPU-363e5ad1-be76-53a7-0086-e28ee69fe5b8 library=cuda variant=v12 compute=12.0 driver=12.9 name="NVIDIA GeForce RTX 5060 Ti" total="15.5 GiB" available="15.3 GiB"
time=2025-08-19T12:07:51.572+02:00 level=INFO source=types.go:130 msg="inference compute" id=GPU-ffafa4fa-6852-5a5a-10da-1a1de150a7e0 library=cuda variant=v12 compute=12.0 driver=12.9 name="NVIDIA GeForce RTX 5060 Ti" total="15.5 GiB" available="15.3 GiB"
[GIN] 2025/08/19 - 12:08:42 | 200 |   40.084812ms |       10.1.0.65 | POST     "/api/show"
[GIN] 2025/08/19 - 12:08:42 | 200 |   39.906831ms |       10.1.0.65 | POST     "/api/show"
[GIN] 2025/08/19 - 12:08:42 | 200 |   69.366485ms |       10.1.0.65 | POST     "/api/show"
time=2025-08-19T12:08:46.512+02:00 level=INFO source=server.go:135 msg="system memory" total="125.2 GiB" free="117.0 GiB" free_swap="1.6 GiB"
time=2025-08-19T12:08:46.512+02:00 level=INFO source=server.go:175 msg=offload library=cuda layers.requested=-1 layers.model=25 layers.offload=17 layers.split=9,8 memory.available="[15.3 GiB 15.3 GiB]" memory.gpu_overhead="0 B" memory.required.full="35.0 GiB" memory.required.partial="30.4 GiB" memory.required.kv="1.4 GiB" memory.required.allocations="[15.3 GiB 15.1 GiB]" memory.weights.total="11.7 GiB" memory.weights.repeating="10.7 GiB" memory.weights.nonrepeating="1.1 GiB" memory.graph.full="10.0 GiB" memory.graph.partial="10.0 GiB"
time=2025-08-19T12:08:46.552+02:00 level=INFO source=server.go:438 msg="starting llama server" cmd="/home/localai/bin/ollama runner --ollama-engine --model /home/localai/.ollama/models/blobs/sha256-b112e727c6f18875636c56a779790a590d705aec9e1c0eb5a97d51fc2a778583 --ctx-size 40960 --batch-size 512 --n-gpu-layers 17 --threads 20 --parallel 5 --tensor-split 9,8 --port 33321"
time=2025-08-19T12:08:46.553+02:00 level=INFO source=sched.go:481 msg="loaded runners" count=1
time=2025-08-19T12:08:46.553+02:00 level=INFO source=server.go:598 msg="waiting for llama runner to start responding"
time=2025-08-19T12:08:46.553+02:00 level=INFO source=server.go:632 msg="waiting for server to become available" status="llm server not responding"
time=2025-08-19T12:08:46.560+02:00 level=INFO source=runner.go:925 msg="starting ollama engine"
time=2025-08-19T12:08:46.560+02:00 level=INFO source=runner.go:983 msg="Server listening on 127.0.0.1:33321"
time=2025-08-19T12:08:46.595+02:00 level=INFO source=ggml.go:92 msg="" architecture=gptoss file_type=MXFP4 name="" description="" num_tensors=315 num_key_values=30
ggml_cuda_init: GGML_CUDA_FORCE_MMQ:    no
ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no
ggml_cuda_init: found 2 CUDA devices:
  Device 0: NVIDIA GeForce RTX 5060 Ti, compute capability 12.0, VMM: yes
  Device 1: NVIDIA GeForce RTX 5060 Ti, compute capability 12.0, VMM: yes
load_backend: loaded CUDA backend from /home/localai/lib/ollama/libggml-cuda.so
load_backend: loaded CPU backend from /home/localai/lib/ollama/libggml-cpu-alderlake.so
time=2025-08-19T12:08:46.760+02:00 level=INFO source=ggml.go:104 msg=system CPU.0.SSE3=1 CPU.0.SSSE3=1 CPU.0.AVX=1 CPU.0.AVX_VNNI=1 CPU.0.AVX2=1 CPU.0.F16C=1 CPU.0.FMA=1 CPU.0.BMI2=1 CPU.0.LLAMAFILE=1 CPU.1.LLAMAFILE=1 CUDA.0.ARCHS=500,600,610,700,750,800,860,870,890,900,1200 CUDA.0.USE_GRAPHS=1 CUDA.0.PEER_MAX_BATCH_SIZE=128 CUDA.1.ARCHS=500,600,610,700,750,800,860,870,890,900,1200 CUDA.1.USE_GRAPHS=1 CUDA.1.PEER_MAX_BATCH_SIZE=128 compiler=cgo(gcc)
time=2025-08-19T12:08:46.804+02:00 level=INFO source=server.go:632 msg="waiting for server to become available" status="llm server loading model"
time=2025-08-19T12:08:46.874+02:00 level=INFO source=ggml.go:365 msg="offloading 17 repeating layers to GPU"
time=2025-08-19T12:08:46.874+02:00 level=INFO source=ggml.go:369 msg="offloading output layer to CPU"
time=2025-08-19T12:08:46.874+02:00 level=INFO source=ggml.go:376 msg="offloaded 17/25 layers to GPU"
time=2025-08-19T12:08:46.874+02:00 level=INFO source=ggml.go:379 msg="model weights" buffer=CUDA0 size="4.0 GiB"
time=2025-08-19T12:08:46.874+02:00 level=INFO source=ggml.go:379 msg="model weights" buffer=CUDA1 size="3.6 GiB"
time=2025-08-19T12:08:46.874+02:00 level=INFO source=ggml.go:379 msg="model weights" buffer=CPU size="5.3 GiB"
time=2025-08-19T12:08:46.934+02:00 level=INFO source=ggml.go:668 msg="compute graph" backend=CUDA0 buffer_type=CUDA0 size="10.2 GiB"
time=2025-08-19T12:08:46.934+02:00 level=INFO source=ggml.go:668 msg="compute graph" backend=CUDA1 buffer_type=CUDA1 size="10.1 GiB"
time=2025-08-19T12:08:46.934+02:00 level=INFO source=ggml.go:668 msg="compute graph" backend=CPU buffer_type=CPU size="10.0 GiB"
time=2025-08-19T12:08:48.828+02:00 level=INFO source=server.go:637 msg="llama runner started in 2.27 seconds"

OS

Linux

GPU

Nvidia

CPU

Intel

Ollama version

0.11.4

Originally created by @StarPet on GitHub (Aug 19, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/11964 ### What is the issue? In my python app I use 'num_ctx': 3182 to reduce the context size of the gpt-oss:20b model so it'll fit into my 2x RTX5060 (16GB). However the runner process gets a different value as --ctx-size parameter. ``` localai 3573918 3102624 99 11:46 ? 00:03:17 /home/localai/bin/ollama runner --ollama-engine --model /home/localai/.ollama/models/blobs/sha256-b112e727c6f18875636c56a779790a590d705aec9e1c0eb5a97d51fc2a778583 --ctx-size 40960 --batch-size 512 --n-gpu-layers 17 --threads 20 --parallel 5 --tensor-split 9,8 --port 34017 ``` Note: before starting ollama, I used the following environment variable, too: ``` export OLLAMA_CONTEXT_LENGTH=2048 ``` I checked the running process: ``` # strings /proc/3573918/environ |grep OLLAMA_CONTEXT OLLAMA_CONTEXT_LENGTH=2048 ``` So, the model is running on CPU not GPU. Either I didn't really understand the relations of the OLLAMA_CONTEXT_LENGTH, the num_ctx parameter or there is an issue with recent changes (as it used to work just fine). ### Relevant log output ```shell time=2025-08-19T12:07:50.750+02:00 level=INFO source=routes.go:1304 msg="server config" env="map[CUDA_VISIBLE_DEVICES:0,1 GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:2048 OLLAMA_DEBUG:INFO OLLAMA_FLASH_ATTENTION:false OLLAMA_GPU_OVERHEAD:0 OLLAMA_HOST:http://0.0.0.0:11434 OLLAMA_INTEL_GPU:false OLLAMA_KEEP_ALIVE:5m0s OLLAMA_KV_CACHE_TYPE: OLLAMA_LLM_LIBRARY: OLLAMA_LOAD_TIMEOUT:5m0s OLLAMA_MAX_LOADED_MODELS:1 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:/home/localai/.ollama/models OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE:false OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NUM_PARALLEL:5 OLLAMA_ORIGINS:[http://10.1.0.65:8123 http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://* vscode-file://*] OLLAMA_SCHED_SPREAD:false ROCR_VISIBLE_DEVICES: http_proxy: https_proxy: no_proxy:]" time=2025-08-19T12:07:50.768+02:00 level=INFO source=images.go:477 msg="total blobs: 446" time=2025-08-19T12:07:50.771+02:00 level=INFO source=images.go:484 msg="total unused blobs removed: 0" time=2025-08-19T12:07:50.772+02:00 level=INFO source=routes.go:1357 msg="Listening on [::]:11434 (version 0.11.4)" time=2025-08-19T12:07:50.772+02:00 level=INFO source=gpu.go:217 msg="looking for compatible GPUs" time=2025-08-19T12:07:51.572+02:00 level=INFO source=types.go:130 msg="inference compute" id=GPU-363e5ad1-be76-53a7-0086-e28ee69fe5b8 library=cuda variant=v12 compute=12.0 driver=12.9 name="NVIDIA GeForce RTX 5060 Ti" total="15.5 GiB" available="15.3 GiB" time=2025-08-19T12:07:51.572+02:00 level=INFO source=types.go:130 msg="inference compute" id=GPU-ffafa4fa-6852-5a5a-10da-1a1de150a7e0 library=cuda variant=v12 compute=12.0 driver=12.9 name="NVIDIA GeForce RTX 5060 Ti" total="15.5 GiB" available="15.3 GiB" [GIN] 2025/08/19 - 12:08:42 | 200 | 40.084812ms | 10.1.0.65 | POST "/api/show" [GIN] 2025/08/19 - 12:08:42 | 200 | 39.906831ms | 10.1.0.65 | POST "/api/show" [GIN] 2025/08/19 - 12:08:42 | 200 | 69.366485ms | 10.1.0.65 | POST "/api/show" time=2025-08-19T12:08:46.512+02:00 level=INFO source=server.go:135 msg="system memory" total="125.2 GiB" free="117.0 GiB" free_swap="1.6 GiB" time=2025-08-19T12:08:46.512+02:00 level=INFO source=server.go:175 msg=offload library=cuda layers.requested=-1 layers.model=25 layers.offload=17 layers.split=9,8 memory.available="[15.3 GiB 15.3 GiB]" memory.gpu_overhead="0 B" memory.required.full="35.0 GiB" memory.required.partial="30.4 GiB" memory.required.kv="1.4 GiB" memory.required.allocations="[15.3 GiB 15.1 GiB]" memory.weights.total="11.7 GiB" memory.weights.repeating="10.7 GiB" memory.weights.nonrepeating="1.1 GiB" memory.graph.full="10.0 GiB" memory.graph.partial="10.0 GiB" time=2025-08-19T12:08:46.552+02:00 level=INFO source=server.go:438 msg="starting llama server" cmd="/home/localai/bin/ollama runner --ollama-engine --model /home/localai/.ollama/models/blobs/sha256-b112e727c6f18875636c56a779790a590d705aec9e1c0eb5a97d51fc2a778583 --ctx-size 40960 --batch-size 512 --n-gpu-layers 17 --threads 20 --parallel 5 --tensor-split 9,8 --port 33321" time=2025-08-19T12:08:46.553+02:00 level=INFO source=sched.go:481 msg="loaded runners" count=1 time=2025-08-19T12:08:46.553+02:00 level=INFO source=server.go:598 msg="waiting for llama runner to start responding" time=2025-08-19T12:08:46.553+02:00 level=INFO source=server.go:632 msg="waiting for server to become available" status="llm server not responding" time=2025-08-19T12:08:46.560+02:00 level=INFO source=runner.go:925 msg="starting ollama engine" time=2025-08-19T12:08:46.560+02:00 level=INFO source=runner.go:983 msg="Server listening on 127.0.0.1:33321" time=2025-08-19T12:08:46.595+02:00 level=INFO source=ggml.go:92 msg="" architecture=gptoss file_type=MXFP4 name="" description="" num_tensors=315 num_key_values=30 ggml_cuda_init: GGML_CUDA_FORCE_MMQ: no ggml_cuda_init: GGML_CUDA_FORCE_CUBLAS: no ggml_cuda_init: found 2 CUDA devices: Device 0: NVIDIA GeForce RTX 5060 Ti, compute capability 12.0, VMM: yes Device 1: NVIDIA GeForce RTX 5060 Ti, compute capability 12.0, VMM: yes load_backend: loaded CUDA backend from /home/localai/lib/ollama/libggml-cuda.so load_backend: loaded CPU backend from /home/localai/lib/ollama/libggml-cpu-alderlake.so time=2025-08-19T12:08:46.760+02:00 level=INFO source=ggml.go:104 msg=system CPU.0.SSE3=1 CPU.0.SSSE3=1 CPU.0.AVX=1 CPU.0.AVX_VNNI=1 CPU.0.AVX2=1 CPU.0.F16C=1 CPU.0.FMA=1 CPU.0.BMI2=1 CPU.0.LLAMAFILE=1 CPU.1.LLAMAFILE=1 CUDA.0.ARCHS=500,600,610,700,750,800,860,870,890,900,1200 CUDA.0.USE_GRAPHS=1 CUDA.0.PEER_MAX_BATCH_SIZE=128 CUDA.1.ARCHS=500,600,610,700,750,800,860,870,890,900,1200 CUDA.1.USE_GRAPHS=1 CUDA.1.PEER_MAX_BATCH_SIZE=128 compiler=cgo(gcc) time=2025-08-19T12:08:46.804+02:00 level=INFO source=server.go:632 msg="waiting for server to become available" status="llm server loading model" time=2025-08-19T12:08:46.874+02:00 level=INFO source=ggml.go:365 msg="offloading 17 repeating layers to GPU" time=2025-08-19T12:08:46.874+02:00 level=INFO source=ggml.go:369 msg="offloading output layer to CPU" time=2025-08-19T12:08:46.874+02:00 level=INFO source=ggml.go:376 msg="offloaded 17/25 layers to GPU" time=2025-08-19T12:08:46.874+02:00 level=INFO source=ggml.go:379 msg="model weights" buffer=CUDA0 size="4.0 GiB" time=2025-08-19T12:08:46.874+02:00 level=INFO source=ggml.go:379 msg="model weights" buffer=CUDA1 size="3.6 GiB" time=2025-08-19T12:08:46.874+02:00 level=INFO source=ggml.go:379 msg="model weights" buffer=CPU size="5.3 GiB" time=2025-08-19T12:08:46.934+02:00 level=INFO source=ggml.go:668 msg="compute graph" backend=CUDA0 buffer_type=CUDA0 size="10.2 GiB" time=2025-08-19T12:08:46.934+02:00 level=INFO source=ggml.go:668 msg="compute graph" backend=CUDA1 buffer_type=CUDA1 size="10.1 GiB" time=2025-08-19T12:08:46.934+02:00 level=INFO source=ggml.go:668 msg="compute graph" backend=CPU buffer_type=CPU size="10.0 GiB" time=2025-08-19T12:08:48.828+02:00 level=INFO source=server.go:637 msg="llama runner started in 2.27 seconds" ``` ### OS Linux ### GPU Nvidia ### CPU Intel ### Ollama version 0.11.4
GiteaMirror added the bug label 2026-04-29 06:00:41 -05:00
GiteaMirror commented 2026-04-29 06:00:44 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Aug 19, 2025):

gpt-oss has a minimum context length, 8192 for machines with more than 20GB. You also have OLLAMA_NUM_PARALLEL=5, so total context is 8192 * 5 = 40960. If you upgrade to the 0.11.5 series, you will benefit from some changes that reduce the size of the gpt-oss allocations.

<!-- gh-comment-id:3200317596 --> @rick-github commented on GitHub (Aug 19, 2025): gpt-oss has a minimum context length, 8192 for machines with more than 20GB. You also have `OLLAMA_NUM_PARALLEL=5`, so total context is 8192 * 5 = 40960. If you upgrade to the 0.11.5 series, you will benefit from some changes that reduce the size of the gpt-oss allocations.
GiteaMirror commented 2026-04-29 06:00:44 -05:00
Author
Owner
Copy Link

@pdevine commented on GitHub (Aug 19, 2025):

I'm going to go ahead and close the issue. I'd recommend reducing OLLAMA_NUM_PARALLEL unless you need it. You may also want to try OLLAMA_NEW_ESTIMATES=1 to try out the memory optimizations we've included in 0.11.5.

<!-- gh-comment-id:3202673775 --> @pdevine commented on GitHub (Aug 19, 2025): I'm going to go ahead and close the issue. I'd recommend reducing `OLLAMA_NUM_PARALLEL` unless you need it. You may also want to try `OLLAMA_NEW_ESTIMATES=1` to try out the memory optimizations we've included in 0.11.5.
GiteaMirror commented 2026-04-29 06:00:44 -05:00
Author
Owner
Copy Link

@StarPet commented on GitHub (Aug 20, 2025):

I'd recommend reducing OLLAMA_NUM_PARALLEL unless you need it. You may also want to try OLLAMA_NEW_ESTIMATES=1 to try out the memory optimizations we've included in 0.11.5.
Done that. Works fine. Thanks all!

<!-- gh-comment-id:3204623252 --> @StarPet commented on GitHub (Aug 20, 2025): > I'd recommend reducing `OLLAMA_NUM_PARALLEL` unless you need it. You may also want to try `OLLAMA_NEW_ESTIMATES=1` to try out the memory optimizations we've included in 0.11.5. Done that. Works fine. Thanks all!
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#54457
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?