server: tests: fetch random media marker via /apply-template (#21962) (#21980)
* server: tests: fetch random media marker via /apply-template (#21962 fix)
* server: allow pinning media marker via LLAMA_MEDIA_MARKER env var
get_media_marker() checks LLAMA_MEDIA_MARKER at first call and uses it
as-is if set, falling back to the random marker otherwise.
Tests no longer need to fetch the marker dynamically via /apply-template:
the fixture sets LLAMA_MEDIA_MARKER=<__media__> so the hardcoded prompts
work as before.
Address review feedback from ngxson
* server: make get_media_marker() thread-safe via magic statics
Use a C++11 static local with a lambda initializer instead of a global
static with an empty-check. The runtime guarantees initialization exactly
once without explicit locking.
The Q8_0 reorder optimization (#21527) was missing a reorder-aware
dequantizer for the GEMM code path used during prompt processing.
After token generation reordered Q8_0 weights (via DMMV/MMVQ), the
next prompt processing pass would read them with the standard
dequantizer, producing garbage output.
Add dequantize_block_q8_0_reorder() and wire it into both
ggml_get_to_fp16_sycl() and ggml_get_to_fp32_sycl(), matching the
pattern already used by Q4_0, Q4_K, and Q6_K.
Fixes #21589
AI (Claude) was used to assist with root cause investigation and
writing the kernel code. All code was human-reviewed and tested
on real hardware.
* SYCL: fix reorder crash when device memory is full
The reorder optimization allocates a temporary buffer the full size of
the weight tensor on the device. When VRAM is nearly full (large models
on a single GPU), this allocation fails and the subsequent memcpy crashes
on a NULL pointer.
Fix: try device allocation first, fall back to host memory if device
memory is full. The reorder kernel still works correctly reading from
host memory over PCIe. This is slower for the one-time reorder (~21 t/s
vs ~38 t/s on Intel Arc Pro B70), but the optimization is preserved for
all subsequent inference. If both device and host allocation fail, skip
the reorder and fall back to the unoptimized kernel path.
Also fixes a bug where opt_for_reorder() marked tensors as reordered
even when the reorder was skipped due to allocation failure. This caused
DMMV/MMVQ kernels to read the original AoS data as if it were SoA,
producing garbage output or NaN results.
Tested on Intel Arc Pro B70 (32GB) with Q8_0, Q4_K_M models. Coding was
AI-assisted (Claude), reviewed and tested on hardware by a human.
Fixes #20478
* SYCL: add RAII temp buffer class + macro guard for host fallback
Replace sycl_ext_malloc_with_fallback/sycl_ext_free_fallback free
functions with sycl_reorder_temp_buffer RAII class. The host_fallback
bool is now a private member, and cleanup happens automatically at
scope exit.
Add GGML_SYCL_HOST_MEM_FALLBACK cmake option (default ON) to guard
the host memory fallback code path. Device access to host memory
requires Linux kernel 6.8+ (Ubuntu 26.04+); users on older kernels
can set -DGGML_SYCL_HOST_MEM_FALLBACK=OFF to disable it.
Addresses arthw's review on PR #21638.
Co-Authored-By: Claude Opus 4.6 (1M context) <redacted>
* SYCL: document GGML_SYCL_HOST_MEM_FALLBACK build option in SYCL.md
Co-Authored-By: Claude Opus 4.6 (1M context) <redacted>
* SYCL: add reorder-aware DMMV dequantizers for Q4_K and Q6_K
Q4_K and Q6_K had reorder support for MMVQ and GEMM paths but not
DMMV. When the DMMV path encountered reordered data it would abort.
Add DMMV kernels that read from the SOA reorder layout for both
types. Same math as the non-reorder versions, different memory
access pattern.
Co-Authored-By: Claude Opus 4.6 (1M context) <redacted>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <redacted>
Introduce hmx-worker (dedicated thread for HMX compute) to overlap HMX
matmul with HVX dequant/DMA stages in the pipeline path, replacing the
previous synchronous HMX calls that blocked the main thread.
* hexagon: cost-based VTCM chunk search for out-stationary matmul
* hexagon: fix futex race in hmx_worker_drain
Store the boolean to local variable avoid atomic load twice
* hex-mm: hmx optimize scatter/transpose and use HMX intrinsics
* hex-vmem: drop vmem limit a touch under 3GB on v73
* hexagon: add fwd declaration of htp_context
* hex-hmx: replace hmx-worker with hmx-queue that mimics dma-queue interface
Simplifies the overall implemantion, reduces thread wakeup roundtrips.
* hex-mm: add debug log to hmx work func called from hmx-queue
* Update hmx-queue.h
Co-authored-by: Max Krasnyansky <redacted>
---------
Co-authored-by: Kim-Chyan Gan <redacted> Co-authored-by: Max Krasnyansky <redacted> Co-authored-by: Max Krasnyansky <redacted>
cmake: fix CMP0194 warning on Windows with MSVC (#21630)
* cmake: fix CMP0194 warning on Windows with MSVC
Set CMP0194 policy to NEW before project() call in ggml/CMakeLists.txt to suppress the "MSVC is not an assembler for language ASM" warning introduced in CMake 4.1.
The ggml project enables ASM globally for Metal (macOS) and KleidiAI (ARM) backends. On Windows/MSVC, no assembler sources are used, but CMake 4.1+ warns because cl.exe is not a valid ASM compiler.
This follows the same pattern used in ggml-vulkan (CMP0114, CMP0147).
Berk Idem [Tue, 14 Apr 2026 10:43:06 +0000 (06:43 -0400)]
common: skip reasoning budget sampler when no budget is requested (#21870)
* common: skip reasoning budget sampler when no budget is requested
After I added thinking_start_tag / thinking_end_tag for gemma4 in #21697, the reasoning budget sampler gets unconditionally created even when no budget is configured (the default -1). The same applies to kimi_k2, lfm2, lfm2_5, and ministral_3 which also set these tags. The budget gets converted to INT_MAX, so the sampler never actually forces any tokens but still runs per-token checks (start tag matching in IDLE state, token-to-piece conversion + UTF-8 checks in COUNTING state).
More importantly, the mere existence of the sampler (non-null rbudget) disables backend sampling. Backend sampling lets the GPU select tokens directly, avoiding a full logits transfer from GPU to CPU every token. This could explain the 30% speed regression reported in #21784 (98 t/s to 70 t/s on Vulkan).
So I added a reasoning_budget_tokens >= 0 check to the sampler creation condition. When the budget is unlimited, the sampler is not created, backend sampling stays enabled, and no per-token overhead is added. When a budget is explicitly set (0, 128, 1024, etc.), the sampler is created and works as before.
* common: preserve rbudget when grammar is lazy
Following up on the review feedback on #21870: keep the reasoning budget sampler when grammar_lazy is true, so the thinking-block grammar suppression from #20970 still works when tools are in use. This way, we only skip the sampler when both no budget is set AND grammar is not lazy.
* Add test case for dispatch to DeviceSegmentedRadixSort
We currently lack a way to force graph mode in CUDA, patch callback to
invoke ggml_backend_compare_graph_backend twice to enforce each test to
run in graph mode
webui: add setting for first-line chat titles (#21797)
* webui: add setting for first-line chat titles
Add an opt-in setting (`titleGenerationUseFirstLine`) to use the first
non-empty line of a prompt as the generated conversation title.
Previously, the complete multi-line prompt was being used, which created
long titles for complex queries. Coupled with
"Ask for confirmation before changing conversation title", the dialog
would overflow.
* Update tools/server/webui/src/lib/utils/text.ts
Co-authored-by: Aleksander Grygier <redacted>
* Update tools/server/webui/src/lib/utils/text.ts
Co-authored-by: Aleksander Grygier <redacted>
* webui: Run build to update the bundle
As requested in:
https://github.com/ggml-org/llama.cpp/pull/21797#pullrequestreview-4094935065
Stephen Cox [Sun, 12 Apr 2026 12:15:26 +0000 (00:15 +1200)]
mtmd: add Gemma 4 audio conformer encoder support (#21421)
* mtmd: add Gemma 4 audio conformer encoder support
Add audio processing for Gemma 4 E2B/E4B via a USM-style Conformer.
Architecture:
- 12-layer Conformer: FFN → Self-Attention → Causal Conv1D → FFN → Norm
- Subsampling Conv Projection: 2x Conv2D(stride=2) with LayerNorm
- Full self-attention with sinusoidal RPE and sliding window mask (24)
- Logit softcapping at 50.0, ClippableLinear clamping
- Output: 1024 → 1536 → RMSNorm → multimodal embedder
Mel preprocessing (dedicated mtmd_audio_preprocessor_gemma4a):
- HTK mel scale, 128 bins, magnitude STFT, mel_floor=1e-3
- Standard periodic Hann window (320 samples), zero-padded to FFT size
- Semicausal left-padding (frame_length/2 samples)
- Frame count matched to PyTorch (unfold formula)
- No pre-emphasis, no Whisper-style normalization
- Mel cosine similarity vs PyTorch: 0.9998
Key fixes:
- Tensor loading dedup: prevent get_tensor() from creating duplicate
entries in ctx_data. Fixed with std::set guard.
- ClippableLinear clamp_info loading moved after per-layer tensors.
- Sliding window mask (24 positions) matching PyTorch context_size.
- Skip Whisper normalization for Gemma4 mel output.
Tested on E2B and E4B with CPU and Vulkan backends.
Transcribes: "Glad to see things are going well and business is starting
to pick up" (matching ground truth).
The mmproj GGUF is generated via convert_hf_to_gguf.py --mmproj on the full
MERaLiON-2 model directory (architecture: MERaLiON2ForConditionalGeneration).
The decoder is converted separately as a standard Gemma2 model after stripping
the text_decoder. weight prefix.
server : ignore --alias when using --models-preset (#21380)
I'm not sure what the purpose of keeping `--alias` was when using
`--models-preset`, but the result is really weird, as shown in the
following logs:
$ build/bin/llama-server --models-preset preset.ini --alias "Gemma 4 E4B UD Q8_K_XL"
...
init: using 31 threads for HTTP server
srv load_models: Loaded 2 cached model presets
srv load_models: Loaded 1 custom model presets from preset.ini
main: failed to initialize router models: alias 'Gemma 4 E4B UD Q8_K_XL' for model 'angt/test-split-model-stories260K:F32' conflicts with existing model name
So I propose to simply ignore `--alias` too in this case. With this
commit, the server starts in routing mode correctly.
Berk Idem [Fri, 10 Apr 2026 09:49:14 +0000 (05:49 -0400)]
common : enable reasoning budget sampler for gemma4 (#21697)
* fix: enable reasoning budget sampler for gemma4
Add thinking_start_tag and thinking_end_tag to
common_chat_params_init_gemma4(). Without these, the reasoning
budget sampler never activates for gemma4.
Make the newline after "thought" optional in the PEG parser to
handle budget=0 (sampler forces end tag before the newline).
Add test case for empty thinking block.
Fixes #21487
* use p.space() instead of p.optional(p.literal("\n")) in gemma4 thought parser
- vendors/hip.h: Add CDNA4 preprocessor define for __gfx950__
- common.cuh: Add GGML_CUDA_CC_CDNA4 and GGML_CUDA_CC_IS_CDNA4 macros
- mma.cuh: Route CDNA4 to compatible MFMA instructions:
* f32 matmul: mfma_f32_16x16x4f32 (xf32 variant unavailable on gfx950)
* bf16 matmul: mfma_f32_16x16x16bf16_1k (same as CDNA3)
* int8 matmul: mfma_i32_16x16x32_i8/32x32x16 (same as CDNA3)
- mmq.cuh: Include CDNA4 in stream-k kernel dispatch
CDNA4 is largely compatible with CDNA3 except:
- No xf32 MFMA (mfma_f32_16x16x8_xf32) — routes to f32 path
- Different FP8 format (e4m3fn vs e4m3_fnuz) — not changed here
Tested on AMD Instinct MI355X (gfx950), ROCm 7.0.1:
- Build: compiles cleanly with -DAMDGPU_TARGETS=gfx950
- llama-bench (Qwen2.5-1.5B Q4_K_M, single GPU):
* f16+FA: 40,013 tok/s prefill, 254 tok/s decode
* q8_0+FA: functional
- Flash attention: works correctly
- MMQ: works correctly with stream-k dispatch
* Remove shfl and AllReduce from backend interface
* move allocation workaround out of ggml-alloc.c
* 2d tensor set/get support
* Fix the seg fault without NCCL
* Apply suggestion from JohannesGaessler
* support for tensor dims % n_devs != 0
* fix view_offs scaling
* arbitrary num. of GPUs/tensor split
* fix compilation
* better granularity estimate
* Support device-specific host buffer types if all underlying backends expose the same type. This allows using pinned memory instead of pageable memory for CUDA.
Fix compilation errors.
* partial Qwen 3 Next support
* Fix qwen3 30b (#8)
* Fix crash with Qwen-30B-A3B Q4_0
Qwen-30B-A3B Q4_0 has an intermediate dimension of 768. Using a granularity of 256 forces an uneven split between GPUs, which is not supported by the current implementation.
* Decide block size based on tensor quantization type
* Fix crashes due to KV cache serialization (#9)
KV cache serialization requires non-zero offsets on the tensor. Add support in the meta backend to set/get a tensor with a non-zero offset.
* metal : fix build (#7)
* static memory allocations, fix usage count
* fix tensor granularity
* more even memory distribution
* use BF16 for allreduce
* rebase fixup
* better error message for unsupported architectures
* Fix device mismatch during scatter of allReduce. (#11)
There is a mismatch between the dst buffer device and the backend device, causing the use of sync copies
* Enable the previous allreduce implementation. It is better in both perf and stability (#12)
* delay AllReduce for Moe for less I/O
* build : clean-up compile warnings
* backend : move most of the meta backend API to ggml-backend-impl.h
* cont : hide unused public API in the implementation
* llama : use llama_device + remove ggml_backend_dev_is_meta()
* ggml-backend : remove unused alloc include
* minor : remove regex include
* ggml : introduce ggml-ext.h for staging new APIs
* rebase fixup
* fix tests
* llama : more robust logic for determining Meta devices (#16)
* llama : more robust logic for determining Meta devices
* cont : fix devs size check
Co-authored-by: Johannes Gäßler <redacted>
* cont : fix log type
Co-authored-by: Johannes Gäßler <redacted>
---------
Co-authored-by: Johannes Gäßler <redacted>
* disable roundtrip for meta backend
* fix arch selection
* Qwen 3.5 support
* fix Gemma 4 MoE
* fix OpenVino, SYCL
* fix test-llama-archs for CPU-only builds
* Fix Qwen 3.5 MoE
* disable meta backend tests for WebGPU
* tests : filter CPU-based devices from the Meta backend tests (#17)
* meta : formatting, naming, indentation (#18)
* formatting : llama-model.cpp
* formatting : ggml-ext.h
* formatting : ggml-backend-meta.cpp
* meta : add TODO
* add documentation
* better error messages
* fix GPT-OSS
---------
Co-authored-by: Carl Philipp Klemm <redacted> Co-authored-by: Gaurav Garg <redacted> Co-authored-by: Georgi Gerganov <redacted>
Daniel Bevenius [Thu, 9 Apr 2026 10:36:29 +0000 (12:36 +0200)]
requirements : update transformers to 5.5.1 (#21617)
* requirements : update transformers to 5.5.0
This commit updates the transformers dependency to version 5.5.0.
The motivation for this is that transformers 5.5.0 includes support for
Gemma4 and is required to be able to convert Gemma4 models. This is also
causing issues for user of gguf-my-repo.
* convert : add ty ignore directives to convert_hf_to_gguf.py
This commit adds `ty: ignore` directives to transformers tokenizers
field/methods to avoid type check errors. There might be better ways to
handle this and perhaps this can be done in a follow up commit.
The motivation for this is that it looks like in transformers 5.5.0
AutoTokenizer.from_pretrained can return generic tokenizer types or None
and the type checker now produces an error when the conversion script
accesses field like tokenizer.vocab.
* convert : add ty ignore to suppress type check errors
* convert : remove incorrect type ignores
* convert : fix remaining python checks
I was running a newer version of ty locally but I've switched to
version 0.0.26 which is what CI uses and I was then able to reproduce
the errors. Sorry about the noise.
Kwa Jie Hao [Thu, 9 Apr 2026 09:28:33 +0000 (17:28 +0800)]
jinja : support ensure_ascii=true, string repetition and int/float self-filtering (#21623)
* feat: jinja engine improvements for reka-edge
Port three Jinja engine improvements needed for the reka-edge model:
1. Python-style string repetition ("ab" * 3 → "ababab")
2. ensure_ascii=true support for tojson filter (escapes non-ASCII to \uXXXX)
3. int() builtin on value_int_t (identity, needed for Reka Edge template)
* fix: escape invalid utf8 bytes when ensure_ascii=true
The json_ensure_ascii_preserving_format function does not correctly
handle an edge case where if UTF-8 parsing fails, it adds the non-ascii
character back to the output as a raw byte.
This commit fixes that by adding the unicode standard replacement
character \\ufffd to the output instead. This is the standard behavior
for various programming languages like Python, Rust, Go, etc.
* chore: address PR comments
1. Add todo comment for supporting string repetition for array/tuples
2. Add support for float identity operation
3. Move invalid ascii test case to test_fuzzing
* chore: accept suggestion for common/jinja/value.cpp
sycl : add flash-attn support for head size 512 (#21654)
* sycl : add flash-attn support for head size 512
This patch extends the SYCL Flash Attention implementation to support head sizes (DKQ/DV) of 512.
Changes:
- Added DKQ/DV 512 cases to both tile and vector Flash Attention kernels.
- Updated kernel selection logic to allow vector kernels for head sizes up to 512 (previously 256).
- Removed unused/redundant AMD and RDNA-specific configuration functions in `fattn-tile.hpp`.
- Refactored `ggml_backend_sycl_buffer_init_tensor` to use a switch statement for clearer tensor extra buffer initialization.
- Added necessary template instances for the new 512 head size across various quantization types.
* remove defunct mxfp4 reorder from setting buffer type
ci: drop v5 `all:` composition from labeler.yml (#21627)
actions/labeler@v6 removed the `all:` / `any:` composition keys.
The `server/webui` and `server` entries used `all:` to combine
`any-glob-to-any-file` with negated `all-globs-to-all-files`,
which now errors on every PR with:
Unknown config options were under "changed-files": all
Flatten both entries to a single `any-glob-to-any-file`. PRs
touching both webui and other server files will now receive both
labels instead of only `server/webui`.
overview / pkg / ggml / sources / llama.cpp / log