ollama

Commit Graph

Author	SHA1	Message	Date
Devon Rifkin	ad3c7c9bda	strip out thinking tags in message history for qwen3 & r1 (#10490 ) * strip out thinking tags in message history for qwen3 & r1 This is in advance of "proper" support where we'll make reasoning configurable and we'll parse out thinking/reasoning tags and provide them to the caller. These models expect there to be no thinking tags in the message history, so this should improve quality * parse model names instead of hacky prefix check	11 months ago
Daniel Hiltgen	415c8fcc3d	Fix "Stopping..." scheduler hang (#10487 ) * Adjust initial scheduler refCount Ensure we only set the refCount on success * sched: fix lock order inversion deadlock Under certain race conditions, there was a scenario where the scheduler would get into a deadlock while trying to update free space information while a model was trying to unload.	11 months ago
Daniel Hiltgen	718eda1b3e	Narrow set of paths we load GGML from (#10485 ) Users may have other incompatible GGML installs on their systems. This will prevent us from trying to load them from the path.	11 months ago
Shahin R	421b7edeb4	readme: add link to lumina, a lightweight React frontend client (#10378 )	11 months ago
batuhankadioglu	7b68e254c2	all: update several golang.org/x packages (#10436 )	11 months ago
Daniel Hiltgen	7bec2724a5	integration: fix embedding tests error handling (#10478 ) The cleanup routine from InitServerconnection should run in the defer of the test case to properly detect failures and report the server logs	11 months ago
Jesse Gross	a27462b708	ollamarunner: Temporarily disable worst case graph preallocation When we later have a large batch running purely on a CPU, this results the error: GGML_ASSERT(talloc->buffer_id >= 0) Disabling this means that we will incrementally reallocate memory as the graph grows. Fixes #10410	11 months ago
crStiv	6bf0b8193a	readme: fix typos (#10399 )	11 months ago
Devon Rifkin	db428adbb8	Merge pull request #10468 from ollama/drifkin/num-parallel-1	11 months ago
Devon Rifkin	fe5b9bb21b	lower default num parallel to 2 this is in part to "pay" for #10452, which doubled the default context length. The combination isn't fully neutral though, because even though the old 4x2k limit and the new 2x4k limit are memory equivalent, the 1x fallback is larger with 4k	11 months ago
Devon Rifkin	6ec71d8fb6	Merge pull request #10452 from ollama/drifkin/4096-context-length config: update default context length to 4096	11 months ago
Devon Rifkin	44b466eeb2	config: update default context length to 4096	11 months ago
Devon Rifkin	a25f3f8260	Merge pull request #10451 from ollama/revert-10364-drifkin/context-length Revert "increase default context length to 4096"	11 months ago
Devon Rifkin	dd93e1af85	Revert "increase default context length to 4096 (#10364 )" This reverts commit `424f648632`.	11 months ago
Devon Rifkin	d2ee599dcf	load arrays with up to 1024 elements when estimating This mirrors the old behavior before #10382	11 months ago
Devon Rifkin	6ed8898590	ggml: fix crash for array head counts If it's an array, it uses the max value in the array If array values for head counts becomes more popular, we can consider a more invasive change like #10225 to calculate more accurate estimates. Fixes: #9984	11 months ago
Michael Yang	5cfc1c39f3	model: fix build (#10416 )	11 months ago
Michael Yang	f0ad49ea17	memory	11 months ago
Michael Yang	7ba9fa9c7d	fixes for maverick	11 months ago
Michael Yang	8bf11b84c1	chunked attention	12 months ago
Michael Yang	470af8ab89	connect vision to text	12 months ago
Michael Yang	178761aef3	image processing Co-authored-by: Patrick Devine <patrick@infrahq.com>	12 months ago
Michael Yang	f0c66e6dea	llama4	1 year ago
Michael Yang	54055a6dae	fix test	11 months ago
Michael Yang	340448d2d1	explicitly decode maxarraysize 1024	11 months ago
Michael Yang	ced7d0e53d	fix parameter count	11 months ago
Michael Yang	a0dba0f8ae	default slice values	11 months ago
Michael Yang	5e20b170a7	update comment	11 months ago
Michael Yang	d26c18e25c	fix token type	11 months ago
Michael Yang	8d376acc9b	zero means zero use a default of 1024 when asking for zero is confusing since most calls seem to assume 0 means do not ready any data	11 months ago
Michael Yang	dc1e81f027	convert: use -1 for read all	11 months ago
Michael Yang	5d0279164c	generic ggml.array	11 months ago
Michael Yang	214a7678ea	fix superfluous call to WriteHeader the first call to http.ResponseWriter.Write implicitly calls WriteHeader with http.StatusOK if it hasn't already been called. once WriteHeader has been called, subsequent calls has no effect. Write is called when JSON encoding progressUpdateJSON{}. calls to http.ResponseWriter.WriteHeader after the first encode is useless and produces a warning: http: superfluous response.WriteHeader call from github.com/ollama/ollama/server/internal/registry.(*statusCodeRecorder).WriteHeader (server.go:77)	11 months ago
Michael Yang	4892872c18	convert: change to colmajor	11 months ago
Michael Yang	0b9198bf47	ci: silence deprecated gpu targets warning	1 year ago
Jeffrey Morgan	e9e5f61c45	llama: update to commit 2016f07b (#10352 )	11 months ago
Parth Sareen	11dde41824	server: improve spacing for JSON grammar (#10131 )	11 months ago
Parth Sareen	a53d744b01	llama: remove model loading for grammar (#10096 )	11 months ago
Adrien Duermael	40b10eee6d	api: fix ImageData struct comment to expect raw image bytes (#10386 )	11 months ago
Devon Rifkin	424f648632	increase default context length to 4096 (#10364 ) * increase default context length to 4096 We lower the default numParallel from 4 to 2 and use these "savings" to double the default context length from 2048 to 4096. We're memory neutral in cases when we previously would've used numParallel == 4, but we add the following mitigation to handle some cases where we would have previously fallen back to 1x2048 due to low VRAM: we decide between 2048 and 4096 using a runtime check, choosing 2048 if we're on a one GPU system with total VRAM of <= 4 GB. We purposefully don't check the available VRAM because we don't want the context window size to change unexpectedly based on the available VRAM. We plan on making the default even larger, but this is a relatively low-risk change we can make to quickly double it. * fix tests add an explicit context length so they don't get truncated. The code that converts -1 from being a signal for doing a runtime check isn't running as part of these tests. * tweak small gpu message * clarify context length default also make it actually show up in `ollama serve --help`	11 months ago
Richard Shiue	2eb1fb3231	readme: add AppFlowy to community integrations (#10335 )	11 months ago
greengrass821	0806521642	cmd: add support for escaping ~ in filepath (#10339 ) Co-authored-by: tooth paste <tooth_paste91@Poorneshwars-MacBook-Pro.local>	11 months ago
Michael Yang	88738b357b	create tempdir in models directory the models directory should have plenty of storage and also ensure there's no cross-device copy	12 months ago
Blake Mizerany	4e535e6188	server/internal/registry: make pull send errors with Error field (#10326 ) Previously, the pull handler would send an error message in the Status field, this prevented the client from using the message as a signal to stop. In the case of the "run" command, it would follow the pull with a "show" which would print a nearly identical "not found" message for unresolved models. Fixes #10307	12 months ago
Michael Yang	40b8fdbdca	arange	1 year ago
Blake Mizerany	1d99451ad7	server/internal/client/ollama: handle some network errors gracefully (#10317 )	12 months ago
Jeffrey Morgan	09bb2e30f6	ml/backend/ggml: use default CUDA compression mode (#10314 )	12 months ago
Jeffrey Morgan	dc264be6ff	ml: add missing cmake property and remove additional CMakeLists.txt (#10310 )	12 months ago
Devon Rifkin	fbe7039618	Merge pull request #10290 from ollama/drifkin/template-highlighting docs: change more template blocks to have syntax highlighting	12 months ago
Jeffrey Morgan	943464ccb8	llama: update to commit 71e90e88 (#10192 )	12 months ago

... 2 3 4 5 6 ...

4366 Commits (a3f7dd3e98df803695f1ae165bc61a1b52142449) All Branches Search

4366 Commits (a3f7dd3e98df803695f1ae165bc61a1b52142449)

All Branches