master
common/gemma4 : handle parsing edge cases (#21760)
tests : add unit test coverage for llama_tensor_get_type (#20112)
libs : rename libcommon -> libllama-common (#21936)
tests: allow exporting graph ops from HF file without downloading weights (#21182)
ci : add model tests + script wrapper (#4586)
chore : correct typos [no ci] (#20041)
ci, tests : use cmake to download models and remove libcurl dependency (#18791)
docs : Minor cleanups (#19252)
metal : add XIELU unary op (#20802)
tests: enable kv_unified to prevent cuda oom error on rtx 2060 (#20645)
Fix race conditions in threadpool when dealing with dynamic/frequent n_threads changes (#17748)
ggml : remove kompute backend (#14501)
common/parser: fix reasoning whitespace bugs + extra parser tests (#21085)
common/parser: add proper reasoning tag prefill reading (#20424)
chat : add Granite 4.0 chat template with correct tool_call role mapping (#20804)
common/autoparser : allow space after tool call (#22073)
ggml : minor naming changes (#8433)
cmake : do not include ./src as public for libllama (#13062)
llama: fix llama-model-saver (#20503)
common/grammar: fix grammar parsing issues to prevent stack overflow and hangs (#18604)
tool/ex/tests: consistently free ctx, then model (#18168)
jinja : support ensure_ascii=true, string repetition and int/float self-filtering (#21623)
common : handle unicode during partial json parsing (#16526)
tests : remove obsolete .mjs script (#21615)
mtmd: add Gemma 4 audio conformer encoder support (#21421)
common : use common_ prefix for common library functions (#9805)
cli: new CLI experience (#17824)
llama : update llama_model API names (#11063)
mtmd: add pos_0 to mtmd_image_tokens_get_decoder_pos (breaking change) (#22082)
tests : fix test-opt with GGML_BACKEND_DL (#15599)
Autoparser - complete refactoring of parser architecture (#18675)
ggml: add Q1_0 1-bit quantization support (CPU) (#21273)
ci: run the x64 and arm ci on the github machines instead (#16183)
common : inhibit lazy grammar sampler while reasoning is active (#20970)
common/grammar : replace problematic backtracking regex `[\s\S]*` (#18342)
ggml-cpu: templateify ggml_compute_forward_rope_f32 and _f16 (#16805)
sampling : optimize samplers by reusing bucket sort (#15665)
common : move up common_init() and fix Windows UTF-8 logs (#21176)
requirements : update transformers to 5.5.1 (#21617)
model : add Jina Embeddings v5 Nano (partial EuroBERT) support (#19826)
devops: add s390x & ppc64le CI (#15925)
common : implement new jinja template engine (#18462)