]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/history
pkg/ggml/sources/llama.cpp
2026-07-20 Aleksander Grygierui: Sidebar Conversations Bulk Action + Improved Settin...
2026-07-20 Aman Guptallama_dsv4: write only used rows in state (#25325)
2026-07-20 Pascalui: fix collapsed user bubble with markdown rendering...
2026-07-20 PascalUI: fix Settings/Display tool call content toggle ...
2026-07-20 Pascalui: enable the agentic flow when only the JS sandbox...
2026-07-20 Hongqiang Wangopencl: Support broadcast for Adreno MUL_MAT and honor...
2026-07-18 Ruixiang Wangmodel: rotate injected K/V cache for DFlash (#25823)
2026-07-18 Yash Raj Pandeyllama-quant : exclude i32 ffn_gate_tid2eid routing...
2026-07-17 lhezopencl: load and use `kernel_gemm_moe_q6_k_f32_ns`...
2026-07-17 Hongqiang Wangopencl: read/write MoE dp4a activation tiles to local...
2026-07-17 Hongqiang Wangopencl: transpose q4_K noshuffle scales for coalesced...
2026-07-17 Georgi Gerganovsync : ggml
2026-07-17 Georgi Gerganovggml : bump version to 0.17.0 (ggml/1568)
2026-07-17 fairydreamingtests : initialize all tensors in test_dsv4_hc to avoid...
2026-07-17 Georgi Gerganovcommon : auto-download dflash- and eagle3- HF sidecars...
2026-07-17 Aaron Teoggml-blas: default hadamard mul_mat to cpu routine...
2026-07-17 Jeff Bolzvulkan: Support Q2_0 (#25430)
2026-07-17 Todd Malsbarysycl: fix row calculation when K_QUANTS_PER_ITERATION...
2026-07-17 Gezahegneopencl: add ABS op (#25115)
2026-07-16 Hongqiang Wangopencl: loads quants as uint for q4_K and q5_K flat...
2026-07-16 akleinedocs: added a note about using OpenCl with Adreno 810...
2026-07-16 Aman GuptaDeepseekV4: Add fused hyper-connection ops (#25585)
2026-07-16 Max Krasnyanskyhexagon: L2 cache handling rework (dirty bit tracking...
2026-07-16 Rajendra Matchakleidiai: Add SME vs SME2 distinction in kernel dispatc...
2026-07-16 Ruben Ortlamvulkan: when using transfer queue for async copies...
2026-07-16 Khashayar Ghafouriconversion: accept BitNetForCausalLM architecture name...
2026-07-16 Johannes GäßlerTP: fix Phi3, Bert, Plamo2/3, ChatGLM (#25536)
2026-07-16 Alessandro de Oliv... vendor: update BoringSSL to 0.20260713.0 (#25624)
2026-07-16 Aman Guptatests: actually exercise `test-recurrent-state-rollback...
2026-07-16 Chipmunkserver : allow text-only slot save/restore with mtmd...
2026-07-16 Ruixiang Wangconvert : fix dflash target tokenizer mismatch during...
2026-07-16 Anav PrasadCUDA: Support CUDA Virtual Devices (#25228)
2026-07-16 Alexander HeislerEnable CUDA graphs on volta+turing (#25749)
2026-07-16 Sebastian Drögeserver: Ignore empty / non-existing `Origin` headers...
2026-07-16 liminfei-amdggml-cuda : restore prop.integrated on HIP builds ...
2026-07-16 Pranesh GonegandlaCUDA: dedup MoE gate/up activation quantization (#25441)
2026-07-16 Georgi Gerganovci : add official website link to release notes (#25728)
2026-07-16 Georgi Gerganovquant : allow using manual tensor types with --pure...
2026-07-16 Hongqiang Wangopencl: disable FA and MoE weights repack to work aroun...
2026-07-16 David Friehscuda: extract Q1_0 elements via __byte_perm (#25628)
2026-07-15 Hongqiang Wangopencl: exclude some moe kernels on Adreno a7x (#25698)
2026-07-15 Aleksander Grygierui: Agentic Content UX improvements (#25450)
2026-07-15 fairydreamingcuda : CUDA GGML_OP_LIGHTNING_INDEXER implementation...
2026-07-15 Adrien Gallouëttokenize : drop --stdin mutual-exclusion check (#25672)
2026-07-15 Hongqiang Wangopencl: fix two issues on flash attention for Adreno...
2026-07-15 leonardHONGCUDA: tighter MMQ src1 buffer size for native fp4 ...
2026-07-15 Gaurav GargFix crash with draft-simple (#25720)
2026-07-15 Pascalserver: fix read_file append_loc space breaking edit_fi...
2026-07-15 Pascalui: fix thinking menu never appearing in single-model...
2026-07-15 fairydreamingcuda : relax tensor contiguity requirements for quantiz...
2026-07-15 Georgi Gerganovci : add HF_TOKEN to self-hosted workflows (#25706)
2026-07-15 Pascalmetal: fuse snake activation (mul, sin, sqr, mul, add...
2026-07-15 Michael Lamotheggml: add f16 out_prod support for CPU and out_prod...
2026-07-15 Aman GuptaDeepseekV4: reduce graph splits (#25702)
2026-07-15 Neo Zhangsycl : fix get_rows Q2_K, Q4_K, Q5_K (#25656)
2026-07-15 Neo Zhangsycl : support kernel type fp16 for conv2d_dw (#25653)
2026-07-15 Andrew Smithsycl : implement xielu op (#25550)
2026-07-15 Francois Dugastsycl: Increase minimum buffer size for USM system alloc...
2026-07-15 hmscider[SYCL] Flash Attention with XMX engine via oneDNN ...
2026-07-15 Hongqiang Wangopencl: do not use `clCreateBufferWithProperties` when...
2026-07-14 Hongqiang Wangopencl: handle OOB write in noshuffle GEMV kernels...
2026-07-14 Hongqiang Wangopencl: avoid the vec path in GEMV for unaligned row...
2026-07-14 Chyanhexagon: fix hmx-queue signal enum-narrowing problem...
2026-07-14 Georgi Gerganovserver : refactor prompt cache state ownership (#25649)
2026-07-14 Xuan-Son Nguyenserver: add --cors-* options (#25655)
2026-07-14 Bill Siderisui: Fix spacing in tool-call request (#25634)
2026-07-14 Emanuil Rusevwebui: parse effective-parameter sizes (E2B, E4B) as...
2026-07-14 Hongqiang Wangopencl: fix a dp4a bug for devices where cl_khr_integer...
2026-07-14 Pascalui: fix mcp panel for toggle + timeout + proxy + ON...
2026-07-14 Aman GuptaDeepseekV4: fix seq_rm (#25588)
2026-07-14 Jeff Bolzvulkan/cpu: Support f16 as SET_ROWS src. (#25432)
2026-07-14 Adrien Gallouëttokenize : align usage by using common args (#25516)
2026-07-14 fairydreamingggml : add a set of functions for checking contiguity...
2026-07-14 Christian Kastnertests: export-graph-ops: exit gracefully when called...
2026-07-14 JusteLeoggml: uniformize im2col dst_type for all conv ops ...
2026-07-14 Charles Xukleidiai : add SME2 f32 kernel (#24414)
2026-07-14 Pascalui: add reasoning effort control to mobile add sheet...
2026-07-14 Thiago Padilhaconvert_hf_to_gguf: support split MTP export for HY...
2026-07-14 Christian Kastnerarg: Flush log before exiting after usage() (#25504)
2026-07-14 Titaniumtownsycl: set fattn_vec_nthreads to 256 for Battlemage...
2026-07-14 Pasha Khosravimetal : add Q2_0 support (#25419)
2026-07-13 Satinder Grewalmodel: add Hy3 (hy_v3) support with MTP speculative...
2026-07-13 Johannes GäßlerCUDA: refactor MMQ kernel configuration (#24127)
2026-07-13 Jeff Bolzvulkan: Use native e2m1 and e4m3 conversions for mxfp4...
2026-07-13 Adrianspec: add Minimax2 eagle3 support
2026-07-13 Georgi Gerganovreadme : add link to maintainer PRs (#25621)
2026-07-13 Christian Kastnertests: Harmonize header use (#25616)
2026-07-13 QuintinShawgguf : add tensor shape accessor (#24405)
2026-07-13 Frosty40chat : fix reasoning leak with force-opened bare <think...
2026-07-13 Frosty40sycl: add fused top-k MoE (#25217)
2026-07-13 Todd Malsbarysycl: add Q2_K to DMMV reorder path (#25064)
2026-07-13 Aleksander Grygierui: Remove recommended MCP Servers + improve MCP Server...
2026-07-12 Bernard Ladenthinserver: honour per-request reasoning_budget_tokens...
2026-07-12 Alessandro de Oliv... vendor : update cpp-httplib to 0.50.1 (#25576)
2026-07-12 Sebastian Drögeserver: Don't consider models with --no-mmproj-auto...
2026-07-12 Pascalmtmd: fix silent prompt truncation on embedded NUL...
2026-07-12 Aldehir Rojasserver : evict checkpoints within min-step of each...
2026-07-12 queiserver : fix image blocks in tool_result being dropped...
2026-07-11 kdkdFix conditional to display 'LLAMA_SPLIT_MODE_TENSOR...
2026-07-11 Rohit Maheshgguf : reject empty metadata keys (#24917)
next