root
/
ollama
mirror of https://github.com/ollama/ollama.git
1
0
Fork 0
Code Issues Actions 2 Packages Projects Releases Wiki Activity
ollama/server
History
Jesse Gross d5a0d8d904 llm: New memory management 
This changes the memory allocation strategy from upfront estimation to
tracking actual allocations done by the engine and reacting to that. The
goal is avoid issues caused by both under-estimation (crashing) and
over-estimation (low performance due to under-utilized GPUs).

It is currently opt-in and can be enabled for models running on the
Ollama engine by setting OLLAMA_NEW_ESTIMATES=1. Behavior in other
cases is unchanged and will continue to use the existing estimates.
 		2025-08-14 15:24:01 -07:00
..
internal chore: fix some inconsistent function name in comment 2025-08-13 09:50:27 -07:00
auth.go fix nil deref in auth.go 2024-07-26 14:14:48 -07:00
create.go remove support for multiple ggufs in a single file (#10722) 2025-05-21 13:55:31 -07:00
create_test.go server: validate local path on safetensor create (#9379) 2025-02-28 16:10:43 -08:00
download.go server: abort download on empty digest 2025-05-27 11:28:48 -07:00
fixblobs.go server: replace blob prefix separator from ':' to '-' (#3146) 2024-03-14 20:18:06 -07:00
fixblobs_test.go server: replace blob prefix separator from ':' to '-' (#3146) 2024-03-14 20:18:06 -07:00
harmonyparser.go update vendored llama.cpp and ggml (#11823) 2025-08-14 14:42:58 -07:00
harmonyparser_test.go gpt-oss (#11672) 2025-08-05 12:21:16 -07:00
images.go update vendored llama.cpp and ggml (#11823) 2025-08-14 14:42:58 -07:00
images_test.go Reapply "feat: incremental gguf parser (#10822)" (#11114) (#11119) 2025-06-20 11:11:40 -07:00
layer.go One corrupt manifest should not wedge model operations (#7515) 2024-11-05 14:21:45 -08:00
manifest.go One corrupt manifest should not wedge model operations (#7515) 2024-11-05 14:21:45 -08:00
manifest_test.go One corrupt manifest should not wedge model operations (#7515) 2024-11-05 14:21:45 -08:00
model.go tools: refactor tool call parsing and enable streaming (#10415) 2025-05-23 14:19:31 -07:00
modelpath.go server: add hint to the error message when model path access fails (#10843) 2025-05-24 13:17:04 -07:00
modelpath_test.go lint: enable usetesting, disable tenv (#10594) 2025-05-08 11:42:14 -07:00
prompt.go fix(openai): handle reasoning_effort (#11868) 2025-08-12 11:02:01 -07:00
prompt_test.go gpt-oss (#11672) 2025-08-05 12:21:16 -07:00
quantization.go skip quantizing per_layer_token_embd (#11207) 2025-06-26 21:49:35 -07:00
quantization_test.go Reapply "feat: incremental gguf parser (#10822)" (#11114) (#11119) 2025-06-20 11:11:40 -07:00
routes.go llm: New memory management 2025-08-14 15:24:01 -07:00
routes_create_test.go Move quantization to new backend (#10363) 2025-05-06 11:20:48 -07:00
routes_delete_test.go Update the /api/create endpoint to use JSON (#7935) 2024-12-31 18:02:30 -08:00
routes_generate_test.go llm: New memory management 2025-08-14 15:24:01 -07:00
routes_harmony_streaming_test.go llm: New memory management 2025-08-14 15:24:01 -07:00
routes_list_test.go Update the /api/create endpoint to use JSON (#7935) 2024-12-31 18:02:30 -08:00
routes_test.go server: use slices.Equal to simplify code (#11502) 2025-07-23 14:25:39 -07:00
sched.go llm: New memory management 2025-08-14 15:24:01 -07:00
sched_test.go llm: New memory management 2025-08-14 15:24:01 -07:00
sparse_common.go Don't hard fail on sparse setup error 2024-08-09 12:16:19 -07:00
sparse_windows.go Don't hard fail on sparse setup error 2024-08-09 12:16:19 -07:00
upload.go server: always print upload/download part info (#8832) 2025-02-04 19:30:49 -08:00