GitLab

* working (other than tool call is the incorrect order) for tool calls and tools

* Tests work, other than image tags (tests do not go through server) and tools (not in the correct order, but contents are the same)

* testing for qwen3vl parser - toolparser is working

* made changes to JSON tool parser, wraps the TollCallFunction with a TollCall object

* Working parser for thinking models - assumes state of thinking, emits unambiguous content in thinking, does not call tool call in thinking

* changed the parser to start with collecting content

* thinking prefill

* add hasThinkingSupport parameter to parser

* qwen3-vl -> qwen3-vl-instruct for renderer/parser

* Add hasThinkingSupport=false to QwenVLParser

---------
Co-authored-by: Devon Rifkin <drifkin@drifkin.net>

Llama cpp bump (df1b612): granite docling / mamba2 optimizations / multimodal encoding fixes (#12552)

* feat: Bump llama.cpp to df1b612

Branch: LlamaCPPBump-GraniteDocling
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix(mtmd): Correctly encode text chunks during mtmd tokenization

There can be text chunks that appear interspersed with the image embeddings
that contain template delimiter tokens for some models. These need to be
correctly translated to text tokens.

Branch: LlamaCPPBump-GraniteDocling
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* tests: Use MtmdChunk in image_test

Branch: LlamaCPPBump-GraniteDocling
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* style: Fix unnecessary conversion linting

Branch: LlamaCPPBump-GraniteDocling
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix(ggml): Revert changes to ggml_hip.cpp

These changes were done largely by our code assistant and are likely wrong

Branch: LlamaCPPBump-GraniteDocling
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix: Revert changes in mem_nvml.cpp

Branch: LlamaCPPBump-GraniteDocling
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* feat: Update sync point to 1deee0

This brings in several more optimization commits and model support for
EmbeddingGemma

Branch: LlamaCPPBump-GraniteDocling
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* feat: Update patches for 1deee0

Branch: LlamaCPPBump-GraniteDocling
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* feat: sync for bump to 1deee0

Branch: LlamaCPPBump-GraniteDocling
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix: Bad patch updates with errant `+`

Branch: LlamaCPPBump-GraniteDocling
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* feat: Bump llama.cpp/ggml to 7049736

Branch: LlamaCPPBump-GraniteDocling
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix: format-patches after latest bump

Branch: LlamaCPPBump-GraniteDocling
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

---------
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

This reverts commit 3d32249c.

deepseek's qwen3 distill uses a different rope scheme so support both

routes: fix built-in renderers for `api/generate`

Made it so when api/generate builds up a message array and generates the
prompt it now goes through the same function as `api/chat` for
consistency. This is where we hook the optional built-in renderers to
bypass templates, which was missing for `api/generate` before this
change.

Closes: #12578

hardErrCh will deadlock since forwardBatch is blocked on
computeStartedCh which never gets sent. since the response to
hardErrCh is to panic, just panic instead

* implement nvml for linux

* Improve scheduler logging when VRAM doesn't recover

Signed-off-by: yajianggroup <yajianggroup@outlook.com>

this change updates how metrics are collected. until now, performance
metrics, specifically initial input processing and subsequent generation
durations, were collected by taking the timestamp when creating a new
sequence, the first token generation, and completing generation. the
processing duration is taken as first token generation sub sequence
creation while generation is taken as completing generation sub first
token generation.

while this approach is an accurate end-to-end metric of processing and
generation, it's not comparable to other tools which only measure the
active, i.e. decode, duration.

this change updates the metrics to only capture decode duration so it
can be more directly compared to other tools

* logs: quiet down context canceled on completion

If the client closes the connection before Completion finishes, we were
logging at error level implying the runner crashed which was misleading.

time=2025-10-08T22:59:20.566-07:00 level=ERROR source=server.go:1490 msg="post predict" error="Post \"http://127.0.0.1:57736/completion\": context canceled"

* quiet down scheduler log error on expected case

Since we don't hold the lock while performing memory load calculations, other
runners can unload in parallel, so finding no runner to unload is a valid scenario
which we shouldn't log at error level.

This reverts commit 6a62b894.

Sliding windows models (e.g. gpt-oss, gemma3) remove tokens that
are out of the cache's window each time we start a new forward pass.

The cache storage needs to handle the window size for each sequence
plus the batch size, since the batch needs to attend to the full
window size. This means that we have greater than a window size
stored while processing the batch.

When the next batch comes, we are currently only looking at the
sequences in the incoming batch to slide the window forward.
However, we also need to clean up the other sequences that might
be occupying space in the batch processing buffer to ensure each
sequence is only using its window size of storage. Failure to do
this can result in "no kv cache slot found" errors.

Fixes: #10127

GGML picks the wrong kernel and these systems fail with:
Sep 28 22:25:39 xavier ollama[48999]: //ml/backend/ggml/ggml/src/ggml-cuda/fattn-wmma-f16.cu:437:
ERROR: CUDA kernel flash_attn_ext_f16 has no device code compatible with CUDA arch 720. ggml-cuda.cu
was compiled for: __CUDA_ARCH_LIST__

Fixes #12442

Remove some flaky scenarios, and switch to chat for better reliability

* Bring back escape valve for llm libraries

If the new discovery logic picks the wrong library, this gives users the
ability to force a specific one using the same pattern as before. This
can also potentially speed up bootstrap discovery if one of the libraries
takes a long time to load and ultimately bind to no devices.  For example
unsupported AMD iGPUS can sometimes take a while to discover and rule out.

* Bypass extra discovery on jetpack systems

On at least Jetpack6, cuda_v12 appears to expose the iGPU, but crashes later on in
cublasInit so if we detect a Jetpack, short-circuit and use that variant.

openai: refactor to split compat layer and middleware

This variable isn't currently documented or intended as something the user can
override, but if the user happens to set OLLAMA_LIBRARY_PATH we were doubling
this in the subprocess environment which will cause problems with the new
bootstrap discovery logic.

This makes the core openai compat layer independent of the middleware
that adapts it to our particular gin routes

Resolve subtle erroraction stickiness difference between x86 and arm builder setup