webui: fix syntax highlighting lost after streaming for non-common languages (#21206)
* webui: fix syntax highlighting lost for non-common languages after streaming
rehype-highlight uses lowlight internally, which only bundles 37 "common"
languages. The streaming code path uses highlight.js directly (192 languages),
so languages like Haskell highlight correctly while streaming but lose all
color once the code block closes. Pass the full lowlight language set to
rehype-highlight so both paths support the same languages.
ggml-cuda: ds_read_b128 for q4_0 and q4_1 mmq kernels (#21168)
* ds_read_b128 for q4_0 and q4_1 mmq kernels
Current for loop generates ds_read_b32 instructions with hip compiler, the new solution generates ds_read_b128 instructions for the same operation, saving some LDS bandwidth. Tested on MI50 and RX6800XT, its faster on both.
* Vectorized lds load update: used ggml_cuda_get_max_cpy_bytes and ggml_cuda_memcpy_1 functions for generic implementation
* Explicit for loop in mmq, renamed vec into tmp
* Fixed max_cpy usage in the loading loop
* Fixed typo in q4_1 kernel
* Update ggml/src/ggml-cuda/mmq.cuh
Co-authored-by: Johannes Gäßler <redacted>
* Update ggml/src/ggml-cuda/mmq.cuh
Co-authored-by: Johannes Gäßler <redacted>
* Update ggml/src/ggml-cuda/mmq.cuh
Co-authored-by: Johannes Gäßler <redacted>
* Renoved trailing white line 500
* Update mmq.cuh removed other whitelines
* Remove trailing whitespaces
---------
Co-authored-by: iacopPBK <redacted> Co-authored-by: Johannes Gäßler <redacted> Co-authored-by: iacopPBK <redacted>
Daniel Bevenius [Tue, 7 Apr 2026 19:24:25 +0000 (21:24 +0200)]
gguf-py : fix missing comma after bad merge in tensor-mapping (#21558)
This commit adds a missing comma in the vision encoder attention qkv
block.
The motivation for this change is that without the comma there will be
a string concatenation of the Kimi-K2.5 and the Nemotron Nano v2 VL
tensor mappings which will be broken.
Son H. Nguyen [Tue, 7 Apr 2026 13:13:38 +0000 (22:13 +0900)]
unicode : add custom Qwen2 regex handler to fix segfault on long input (#21257)
* unicode : add custom Qwen2 regex handler to fix segfault on long input
std::regex uses recursive backtracking internally, which causes a stack
overflow (segfault) when tokenizing long sequences of repeated characters
(e.g. 43K 'A's). The Qwen2 tokenizer regex differs from Llama3 only in
the digit pattern (\p{N} vs \p{N}{1,3}), so it was falling through to
the std::regex fallback path instead of using a custom handler.
Add unicode_regex_split_custom_qwen2() following the established pattern
used by gpt2, llama3, kimi_k2, and afmoe custom handlers.
Closes: https://github.com/ggml-org/llama.cpp/issues/21113
* cont : remove TODO comment
* cont : update comment to reflect original regex
* use the correct regex in the comment this time... [no ci]
vulkan: add FA dequant for q4_1, q5_0, q5_1, iq4_nl (#21029)
Add dequantize4() implementations for Q4_1, Q5_0, Q5_1, and IQ4_NL
in the flash attention base shader. Register them in the shader
generator, pipeline creation, and enable in the scalar/coopmat1 FA
support check.
- Add dir='auto' attributes to markdown containers and blocks
- Implement post-processing to add dir='auto' to all text elements
- Replace directional CSS properties with logical properties for proper RTL list alignment
- Ensure bidirectional text support for mixed Arabic/English content
* Clean up commented duplicate function
Remove the commented-out duplicate transformMdastNode function
that was left over from refactoring.
* Fix Arabic RTL text rendering in web UI
- Add dir='auto' attributes to markdown containers and blocks
- Implement post-processing to add dir='auto' to all text elements
- Replace directional CSS properties with logical properties for proper RTL list alignment
- Minor code formatting improvements
This ensures bidirectional text support for mixed Arabic/English content in the llama.cpp web UI.
* Implement rehype plugin for comprehensive RTL text support
- Add rehypeRtlSupport plugin that applies dir='auto' to all elements with children
- Replace DOMParser-based approach with efficient HAST tree processing
- Remove hardcoded element lists for better maintainability
- Ensure proper bidirectional text rendering for mixed RTL/LTR content
* Fix RTL text rendering with rehype plugin and cleanup
Extend the existing reorder optimization to Q8_0. The reorder
separates scale factors from weight data for coalesced memory
access -- was implemented for Q4_0/Q4_K/Q6_K but Q8_0 was missing.
On Arc Pro B70 (Xe2), Q8_0 tg goes from 4.88 to 15.24 t/s (3.1x)
on Qwen3.5-27B. BW utilization: 21% -> 66%.
The key fix beyond the kernels: Q8_0 was missing from the type
check in ggml_backend_sycl_buffer_init_tensor() that allocates
the extra struct carrying the reorder flag -- so the optimization
was silently skipped.
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.
[CUDA ] Write an optimized flash_attn_stream_k_fixup kernel (#21159)
* Write an optimized flash_attn_stream_k_fixup kernel
Write a specialized and more optimized kernel for cases where nblocks_stream_k is multiple of ntiles_dst.
Make nblocks_stream_k to multiple of ntiles_dst if nblocks_stream_k > 2 * ntiles_dst
* Use the new kernel only for nblocks_stream_k_raw > 4 * ntiles_dst to make sure we have enough concurrency on GPUs
server : handle unsuccessful sink.write in chunked stream provider (#21478)
Check the return value of sink.write() in the chunked content provider
and return false when the write fails, matching cpp-httplib's own
streaming contract. This prevents logging chunks as sent when the sink
rejected them and properly aborts the stream on connection failure.
server : fix logging of build + system info (#21460)
This PR changes the logging that occurs at startup of llama-server.
Currently, it is redundant (including CPU information twice) and it is
missing the build + commit info.
common : fix tool call type detection for nullable and enum schemas (#21327)
* common : fix tool call type detection for nullable and enum schemas
* common, tests : fix grammar delegation for nullable/enum schemas and add tests
Fix enum type inference to scan all enum values (not just index 0) so
schemas like {"enum": [0, "celsius"]} correctly detect string type.
Fix schema_delegates in peg-parser to handle nullable type arrays
(["string", "null"]) and typeless enum schemas in raw mode, allowing
the tagged parser to use raw text instead of JSON-formatted strings.
Add test cases for Qwen3-Coder (TAG_WITH_TAGGED format):
- nullable string ["string", "null"]
- nullable string with null first ["null", "string"]
- nullable integer ["integer", "null"]
- enum without explicit type key
The `HSA_OVERRIDE_GFX_VERSION` variable can be used in ROCm to override an unsupported target architecture with a similar but supported target architecture.
This does not and has never worked on Windows. I think the clarification could avoid driving Windows people towards this solution that does not work.
ggml-zendnn : add MUL_MAT_ID op support for MoE models (#21315)
* ggml-zendnn : add MUL_MAT_ID op support for MoE models
- Add MUL_MAT_ID op acceleration for Mixture-of-Experts models
- MUL_MAT_ID op fallback to CPU backend if total experts > 32
- Point ZenDNN lib to latest bits ZenDNN-2026-WW13
* ggml-zendnn : add braces to sgemm failure condition for consistency
Reuse the buffer for the ggml context which is used for creating the
compute graph on the server side. This partially addresses a memory leak
created by the CUDA backend due to using buffer addresses as cache
keys.
Bump ROCm version on Linux from 7.2 to 7.2.1
Add gfx1102 target
Delete LLVM workaround since ROCm 7.2.1 has fix for ROCm 7.2 perf regression https://github.com/ROCm/rocm-systems/issues/2865
tests : add unit test coverage for llama_tensor_get_type (#20112)
* Add unit test coverage for llama_tensor_get_type
* Fix merge conflicts, add more schemas
* clang formatter changes
* Trailing whitespace
* Update name
* Start rebase
* Updating files with upstream changes prior to rebase
* Changes needed from rebase
* Update attn_qkv schema, change throw behaviour
* Fix merge conflicts
* White space
* Update with latest changes to state counters
* Revert accidental personal CLAUDE.md changes
* Change quotation mark
* Reuse metadata.name since we have it
* Move test-only stuff out of llama-quant.cpp
* Hide the regex functionality back in llama-quant.cpp, use a unique pointer to a new struct 'compiled_tensor_type_patterns' which contains the patterns
* cont : inital deslop guidelines
* Cleanup based on review comments
* Continue cleanup
* Small cleanup
* Manually set proper ordering of tensors, mostly applies to gemma
Jesus Talavera [Thu, 2 Apr 2026 09:28:56 +0000 (11:28 +0200)]
chat : add Granite 4.0 chat template with correct tool_call role mapping (#20804)
* chat : add Granite 4.0 chat template with correct tool_call role mapping
Introduce `LLM_CHAT_TEMPLATE_GRANITE_4_0` alongside the existing Granite
3.x template (renamed `LLM_CHAT_TEMPLATE_GRANITE_3_X`).
The Granite 4.0 Jinja template uses `<tool_call>` XML tags and maps the
`assistant_tool_call` role to `<|start_of_role|>assistant<|end_of_role|><|tool_call|>`.
Without a matching C++ handler, the fallback path emits the literal role
`assistant_tool_call` which the model does not recognize, breaking tool
calling when `--jinja` is not used.
Changes:
- Rename `LLM_CHAT_TEMPLATE_GRANITE` to `LLM_CHAT_TEMPLATE_GRANITE_3_X`
(preserves existing 3.x behavior unchanged)
- Add `LLM_CHAT_TEMPLATE_GRANITE_4_0` enum, map entry, and handler
- Detection: `<|start_of_role|>` + (`<tool_call>` or `<tools>`) → 4.0,
otherwise → 3.x
- Add production Granite 4.0 Jinja template
- Add tests for both 3.x and 4.0 template paths (C++ and Jinja)
Co-Authored-By: Claude Opus 4.6 <redacted>
* Code review: follow standard format and use common logic in test-chat-template.cpp
* Rename custom_conversation variable for extra_conversation to give it a more meaningful name
memory: respect unified KV cache in hybrid memory for eval tasks (#21224)
The hybrid memory paths (`llama-memory-hybrid.cpp` and
`llama-memory-hybrid-iswa.cpp`) always used sequential equal split,
ignoring the unified KV cache flag. This caused hellaswag, winogrande,
and multiple-choice evaluations to fail on hybrid models (models with
both attention and recurrent/SSM layers, such as Qwen3.5-35B-A3B) with:
split_equal: sequential split is not supported when there are
coupled sequences in the input batch (you may need to use the
-kvu flag)
PR #19954 fixed this for `llama-kv-cache-iswa.cpp` by automatically
enabling unified KV mode and setting n_parallel >= 4 for multi-choice
eval tasks. However, the hybrid memory paths were not updated.
This commit mirrors the iswa fix: use non-sequential split when KV
cache is unified (n_stream == 1), which is automatically set by
llama-perplexity for hellaswag/winogrande/multiple-choice since #19954.
CUDA/HIP: Fix kernel slection for mmvq mmid kernel to align host selection with device launch bounds (#21238)
The conditions cc == GGML_CUDA_CC_VOLTA || cc >= GGML_CUDA_CC_ADA_LOVELACE and cc >= GGML_CUDA_CC_TURING match all non-nvidia devices. This causes us to attempt to launch the kernel for batch sizes with larger configurations than our launch bounds on HIP devices. This pr fixes the conditionals in get_mmvq_mmid_max_batch.
Abhijit Ramesh [Tue, 31 Mar 2026 22:38:16 +0000 (15:38 -0700)]
ggml-webgpu: port all AOT operators to JIT (#20728)
* port cpy pipeline to shader lib with JIT compilation
* port glu pipeline to shader lib with JIT compilation
* port rope pipeline to shader lib with JIT compilation
* port soft_max pipeline to shader lib with JIT compilation
* removed unused functions from embed_wgsl.py which were used for
old AOT template expansion
overview / pkg / ggml / sources / llama.cpp / log