github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-18 14:16:30 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #10090] Nvidia Jetson Nano with cuda 10.2 #68672

New Issue
Closed
opened 2026-05-04 14:48:01 -05:00 by GiteaMirror · 1 comment
GiteaMirror commented 2026-05-04 14:48:01 -05:00
Owner
Copy Link

Originally created by @betolley on GitHub (Apr 2, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/10090

I managed to get a older version of ollama to compile with cuda 10.2 support.

I downloaded the b9f74ff3d6 commit.

I compiled Gcc/G++ 8.5 and set it as the default type of gcc/g++
I installed update to CMAKE
I installed update to Go

git clone https://code.blarg.ca/gered/ollama.git
cd ollama
git checkout b9f74ff3d6
make build
cd build
cmake .. -DCMAKE_CUDA_STANDARD=11
make -j4

I setup ollama as a service and here is my log

msg="Dynamic LLM libraries [cpu cuda_v10]"
apr 02 17:28:58 nano ollama[8571]: time=2025-04-02T17:28:58.394+02:00 level=INFO source=gpu.go:96 msg="Detecting GPUs"
apr 02 17:28:58 nano ollama[8571]: time=2025-04-02T17:28:58.424+02:00 level=INFO source=gpu.go:101 msg="detected GPUs" library=/tmp/ollama3591004554/runners/cuda_v10/libcudart.so.10.2>
apr 02 17:28:58 nano ollama[8571]: time=2025-04-02T17:28:58.424+02:00 level=INFO source=cpu_common.go:18 msg="CPU does not have vector extensions"

ggml_cuda_init: GGML_CUDA_FORCE_MMQ: yes
ggml_cuda_init: CUDA_USE_TENSOR_CORES: no
ggml_cuda_init: found 1 CUDA devices:
Device 0: NVIDIA Tegra X1, compute capability 5.3, VMM: no
llm_load_tensors: ggml ctx size = 0.32 MiB
llm_load_tensors: offloading 25 repeating layers to GPU
llm_load_tensors: offloaded 25/33 layers to GPU
llm_load_tensors: CPU buffer size = 1441.93 MiB
llm_load_tensors: CUDA0 buffer size = 1057.37 MiB
...........................................................................................
llama_new_context_with_model: n_ctx = 2048
llama_new_context_with_model: n_batch = 512
llama_new_context_with_model: n_ubatch = 512
llama_new_context_with_model: freq_base = 10000.0
llama_new_context_with_model: freq_scale = 1
llama_kv_cache_init: CUDA_Host KV buffer size = 140.00 MiB
llama_kv_cache_init: CUDA0 KV buffer size = 500.00 MiB
llama_new_context_with_model: KV self size = 640.00 MiB, K (f16): 320.00 MiB, V (f16): 320.00 MiB
llama_new_context_with_model: CUDA_Host output buffer size = 0.21 MiB
llama_new_context_with_model: CUDA0 compute buffer size = 207.54 MiB
llama_new_context_with_model: CUDA_Host compute buffer size = 14.01 MiB
llama_new_context_with_model: graph nodes = 1225
llama_new_context_with_model: graph splits = 90

I am only able to use older models because the newer ones give an error with the file type.

This one works: dolphin-phi:latest

I also got whisper.cpp and llama.cpp compiling with the version of the software.

Originally created by @betolley on GitHub (Apr 2, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/10090 I managed to get a older version of ollama to compile with cuda 10.2 support. I downloaded the b9f74ff3d643fd2b579bfc24132e34f3bc7bf777 commit. I compiled Gcc/G++ 8.5 and set it as the default type of gcc/g++ I installed update to CMAKE I installed update to Go git clone https://code.blarg.ca/gered/ollama.git cd ollama git checkout b9f74ff3d643fd2b579bfc24132e34f3bc7bf777 make build cd build cmake .. -DCMAKE_CUDA_STANDARD=11 make -j4 I setup ollama as a service and here is my log msg="Dynamic LLM libraries [cpu cuda_v10]" apr 02 17:28:58 nano ollama[8571]: time=2025-04-02T17:28:58.394+02:00 level=INFO source=gpu.go:96 msg="Detecting GPUs" apr 02 17:28:58 nano ollama[8571]: time=2025-04-02T17:28:58.424+02:00 level=INFO source=gpu.go:101 msg="detected GPUs" library=/tmp/ollama3591004554/runners/cuda_v10/libcudart.so.10.2> apr 02 17:28:58 nano ollama[8571]: time=2025-04-02T17:28:58.424+02:00 level=INFO source=cpu_common.go:18 msg="CPU does not have vector extensions" ggml_cuda_init: GGML_CUDA_FORCE_MMQ: yes ggml_cuda_init: CUDA_USE_TENSOR_CORES: no ggml_cuda_init: found 1 CUDA devices: Device 0: NVIDIA Tegra X1, compute capability 5.3, VMM: no llm_load_tensors: ggml ctx size = 0.32 MiB llm_load_tensors: offloading 25 repeating layers to GPU llm_load_tensors: offloaded 25/33 layers to GPU llm_load_tensors: CPU buffer size = 1441.93 MiB llm_load_tensors: CUDA0 buffer size = 1057.37 MiB ........................................................................................... llama_new_context_with_model: n_ctx = 2048 llama_new_context_with_model: n_batch = 512 llama_new_context_with_model: n_ubatch = 512 llama_new_context_with_model: freq_base = 10000.0 llama_new_context_with_model: freq_scale = 1 llama_kv_cache_init: CUDA_Host KV buffer size = 140.00 MiB llama_kv_cache_init: CUDA0 KV buffer size = 500.00 MiB llama_new_context_with_model: KV self size = 640.00 MiB, K (f16): 320.00 MiB, V (f16): 320.00 MiB llama_new_context_with_model: CUDA_Host output buffer size = 0.21 MiB llama_new_context_with_model: CUDA0 compute buffer size = 207.54 MiB llama_new_context_with_model: CUDA_Host compute buffer size = 14.01 MiB llama_new_context_with_model: graph nodes = 1225 llama_new_context_with_model: graph splits = 90 I am only able to use older models because the newer ones give an error with the file type. This one works: dolphin-phi:latest I also got whisper.cpp and llama.cpp compiling with the version of the software.
GiteaMirror commented 2026-05-04 14:48:03 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Jul 5, 2025):

Unfortunately that version of CUDA and the GCC compilers required are getting too old to maintain compatibility with as we add support for new GPUs and models.

<!-- gh-comment-id:3040257058 --> @dhiltgen commented on GitHub (Jul 5, 2025): Unfortunately that version of CUDA and the GCC compilers required are getting too old to maintain compatibility with as we add support for new GPUs and models.
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
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#68672
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?