Public/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-09-19 09:15:04 +02:00
Code Issues Packages Projects Releases Wiki Activity

Commit Graph

  • ef9fa1ba1f quantize: dequantize and quantize large tensors in row bands danielhanchen 2026-08-26 17:18:44 +00:00
  • 6a69a0c12c qwen4exp: drop an unused variable that breaks -Werror builds danielhanchen 2026-08-27 00:00:41 +00:00
  • c52ed2a0b0 llama: give qwen4exp a large-graph node budget danielhanchen 2026-08-26 17:37:20 +00:00
  • 0ac4b18025 qwen4exp: support a quantized KV cache in the QSA attention path danielhanchen 2026-08-26 16:13:21 +00:00
  • d4a943f9a5 qwen4exp: tidy comments and simplify image token read danielhanchen 2026-08-26 15:46:37 +00:00
  • d22d2be2b4 llama: make the qwen4exp PLE n-gram history per context and serialise it Daniel Han 2026-08-26 15:15:11 +00:00
  • cfbdc0a50d llama: save and restore the qwen4exp indexer KV cache Daniel Han 2026-08-26 15:06:10 +00:00
  • c619e22d0f Merge remote-tracking branch 'origin/master' into xsn/video_args Xuan Son Nguyen 2026-08-27 01:11:19 +02:00
  • 925e117994 llama: add token ID tracking to KV cell (#27762) b10642 Xuan-Son Nguyen 2026-08-26 23:34:28 +02:00
  • 2bd7078179 add get_prev_tokens xsn/kv_track_token_id Xuan Son Nguyen 2026-08-26 21:39:07 +02:00
  • d07a0ae352 nits Xuan Son Nguyen 2026-08-26 21:22:49 +02:00
  • c0ea2b3038 rm get_prev_tokens, move it to the main pr Xuan Son Nguyen 2026-08-26 21:17:09 +02:00
  • 980256bda6 kv: track token id Xuan Son Nguyen 2026-08-26 21:16:29 +02:00
  • a0d616e9cb context : relax on-device seq-copy chunk alignment gg/tests-save-load-all-arch-tmp Georgi Gerganov 2026-08-26 22:10:17 +03:00
  • 539f24529b ui: Move Settings and MCP Servers routes to dialog-based views (#27744) Aleksander Grygier 2026-08-26 21:07:24 +02:00
  • 6358971e51 chore: Formatting & linting allozaur/feat/chat-form-actions/settings-mcp-dialogs Aleksander Grygier 2026-08-26 15:56:40 +02:00
  • 06453e1c90 fix(ui): adjust MCP dialogs for new base sizing Aleksander Grygier 2026-08-26 15:30:02 +02:00
  • b60d6fe497 refactor(ui): remove settings route and URL-based settings navigation Aleksander Grygier 2026-08-22 09:55:31 +02:00
  • 9a0868b310 feat(ui): open settings as dialog from sidebar Aleksander Grygier 2026-08-22 09:55:24 +02:00
  • 9cd139973c refactor(ui): switch SettingsChat to in-app section navigation Aleksander Grygier 2026-08-22 09:55:15 +02:00
  • a29da144a4 feat(ui): add DialogSettingsChat dialog Aleksander Grygier 2026-08-22 09:55:08 +02:00
  • 2569828602 ui : remove unused MCP servers submenu component Aleksander Grygier 2026-08-22 06:53:45 +02:00
  • 076a415ac1 ui : remove mcp-servers route and sidebar entry Aleksander Grygier 2026-08-22 06:53:37 +02:00
  • 560d61c731 ui : browse MCP resources from the server card Aleksander Grygier 2026-08-22 06:53:29 +02:00
  • 11bed339d9 ui : open MCP servers in a dialog from the chat form Aleksander Grygier 2026-08-22 06:53:22 +02:00
  • 0379a19f09 ui: Update Dialog component styling (#27743) Aleksander Grygier 2026-08-26 20:19:19 +02:00
  • 78898478fb tests : fix dsv4 save-load n_stream mismatch Georgi Gerganov 2026-08-26 21:15:40 +03:00
  • a538337fbb tests : use 64 indexer heads for deepseek4 Georgi Gerganov 2026-08-26 21:06:43 +03:00
  • f3e1b82b11 tests : add --help and document -o in test-llama-archs Georgi Gerganov 2026-08-26 20:58:34 +03:00
  • 571bf769e2 tests : match dummy DSA indexer to fused Lightning Indexer kernel Georgi Gerganov 2026-08-26 20:58:29 +03:00
  • 46ca1855de tests : run test-save-load-state across all architectures Georgi Gerganov 2026-08-26 20:35:22 +03:00
  • 60a1474ad9 ggml-metal: FA tensor path: gate to the (256, 256) shape gg/metal-fa-tensor-save gg/metal-fa-tensor ggerganov 2026-08-26 19:39:29 +03:00
  • 5e6a37cb11 vulkan: warptiles currently assume warp sizes <= 64, clamp to work around larger warps (#27726) b10639 Ruben Ortlam 2026-08-26 18:02:06 +02:00
  • a76cd9f8ee chore: Formatting & linting allozaur/feat/chat-form-actions/base-dialog Aleksander Grygier 2026-08-26 15:54:32 +02:00
  • d3e02783b0 ui: move dialog close button to the sticky header Aleksander Grygier 2026-08-22 13:12:05 +02:00
  • c53a39b669 feat(ui): make base dialog responsive and support sticky headers Aleksander Grygier 2026-08-22 09:55:00 +02:00
  • 035e22731a llama: hold the qwen4exp indexer cache in a new llama_memory_hybrid_idx Daniel Han 2026-08-26 14:43:13 +00:00
  • 7fee670a95 llama: keep the qwen4exp top-k attention mask arch-local Daniel Han 2026-08-26 14:42:46 +00:00
  • bf94216469 Implemented vulkan cross_entropy_loss and cross_entropy_loss_back (#27216) b10638 Pranav Uttarkar 2026-08-26 09:49:32 -05:00
  • 3f56636c8b ggml-metal: FA tensor kernel: register-light softmax in the per-d-block path ggerganov 2026-08-26 17:49:19 +03:00
  • d0132a680a rpc : implement event and async backend APIs (#18626) Radoslav Gerganov 2026-08-26 17:34:46 +03:00
  • c3148ebe38 ggml-metal: FA tensor kernel: 32-wide chunks, register-light global-max softmax ggerganov 2026-08-26 17:22:40 +03:00
  • bea3b12dae qwen4exp: support a non-unified KV cache in QSA danielhanchen 2026-08-26 04:09:11 +00:00
  • eb95d125ba qwen4exp: hash the image placeholder for multimodal batches danielhanchen 2026-08-26 00:58:41 +00:00
  • 25aa77a4e2 quantize: size the output buffer exactly instead of nelements * 4 danielhanchen 2026-08-26 00:31:26 +00:00
  • 4597a86f12 quantize: let --tensor-type name per_layer_token_embd danielhanchen 2026-08-25 23:28:58 +00:00
  • 570c26d3d6 quantize: fall back to F16 for 32-block types with an odd ncols danielhanchen 2026-08-25 23:18:12 +00:00
  • 53c0f624ae convert: stream the qwen4exp PLE table instead of concatenating it Daniel Han 2026-08-25 19:21:37 +00:00
  • 40e90413fd tests: cover qwen4exp in test-llama-archs Daniel Han 2026-08-25 19:17:58 +00:00
  • dd8962766f tests: record what the qwen4exp arch-test skip actually observes Daniel Han 2026-08-25 19:09:47 +00:00
  • 1b599101d2 llama: give the qwen4exp indexer cache the attention cache's slots Daniel Han 2026-08-25 21:07:43 +00:00
  • c88c916636 llama: QSA sparse attention for qwen4exp Daniel Han 2026-08-25 19:09:43 +00:00
  • ff24f3874f llama: optional indexer key cache in llama_memory_hybrid Daniel Han 2026-08-25 18:16:16 +00:00
  • ddf0980e52 llama: fix the qwen4exp PLE conv state and unblock test-llama-archs Daniel Han 2026-08-25 15:29:13 +00:00
  • c89e67b08e llama: carry the qwen4exp PLE conv state across ubatches Daniel Han 2026-08-25 15:21:05 +00:00
  • ad4fa3fa7b llama: qwen4exp PLE n-gram hash embedding Daniel Han 2026-08-25 15:15:11 +00:00
  • c9ad86f6b9 llama: qwen4exp text graph with hyper-connections, GDN and MoE Daniel Han 2026-08-25 14:51:49 +00:00
  • 058b7a1b48 qwen4exp: shorten comments Daniel Han 2026-08-26 12:37:53 +00:00
  • 331fa7e560 llama: load qwen4exp (Qwen3.8-Flash-Next) hparams and tensors danielhanchen 2026-08-25 14:09:17 +00:00
  • 6e5b8b7e4d gguf: add qwen4exp (Qwen3.8-Flash-Next) arch and converter danielhanchen 2026-08-25 14:00:29 +00:00
  • 36591b9d62 ggml-metal: hoist QK^T + softmax out of the d-block loop in the FA tensor kernel ggerganov 2026-08-26 16:16:03 +03:00
  • 59549e01b6 ci : rename backend-only release jobs to windows-backend-* (#27731) allozaur/chore/gh-release-ui-fallback Aleksander Grygier 2026-08-26 14:46:02 +02:00
  • c185acffe9 ci : remove the ui-publish job Aleksander Grygier 2026-08-25 17:21:19 +02:00
  • 25f3c82768 ui : replace HF bucket download with GH Releases assets Aleksander Grygier 2026-08-25 17:21:13 +02:00
  • 21eb1b7040 ggml-metal: flash attention kernel for the MPP tensor API ggerganov 2026-08-26 15:30:32 +03:00
  • 4d19b28769 ci: Clean up UI builds from releases (#27706) b10636 Aleksander Grygier 2026-08-26 14:12:09 +02:00
  • fc35562ba4 cuda: unblock mmq for MoE on sm_60 (#26264) b10635 David Friehs 2026-08-26 12:35:54 +02:00
  • da9b5d68c3 ci : make cache bucket public (#27728) Sigbjørn Skjæret 2026-08-26 12:08:23 +02:00
  • dac869b0a0 conversion : fix Nemotron 3.5 Lightning layers (#27729) Daniel Bevenius 2026-08-26 12:05:31 +02:00
  • 11cd988428 ggml-metal: add chunked SSD MMA for Mamba-2 prefill optimization (#26647) b10632 Dominik Pantaleoni 2026-08-26 01:57:07 -07:00
  • e9dab6c5fe cont : disable self-hosted gg/ci-disable-openvino Georgi Gerganov 2026-08-26 09:05:17 +03:00
  • d9e300aa50 cont : add ref Georgi Gerganov 2026-08-11 08:51:53 +03:00
  • db87106252 ci : disable OpenVINO builds Georgi Gerganov 2026-08-11 08:47:54 +03:00
  • 5d5cb4c3a4 ggml-meta: propagate buffer usage and call init on the new tensors (#27586) b10631 Max Krasnyansky 2026-08-25 22:27:51 -07:00
  • d222767c7a kleidiai: Rework KleidiAI Build System/Integration (#26077) b10630 Jonathan Clohessy 2026-08-25 22:07:29 +01:00
  • eab8ee41f8 ci : update OS used for ROCM to Ubuntu 24.04 (#27681) b10629 Mario Limonciello 2026-08-25 12:14:16 -05:00
  • b114b47397 rpc: support apple RDMA as an RPC transport (#26421) b10628 Ryan C 2026-08-25 17:12:15 +00:00
  • 0a5ac49bce devops: use GGML_NATIVE=OFF for OpenVINO (#27338) krzsztf 2026-08-25 18:37:11 +02:00
  • 1729ed5371 server : reject prefilled assistant messages with tool calls (#27626) Kyozzz 2026-08-25 22:35:22 +08:00
  • 0cc5b14959 chat : scope qwen3-coder workarounds (#27679) b10625 Aldehir Rojas 2026-08-25 09:33:00 -05:00
  • 790b5713ca ci : store ccache on HF buckets (test with cuda-ubuntu for now) (#27699) Sigbjørn Skjæret 2026-08-25 16:28:25 +02:00
  • 4d2d1c2437 chore: Format allozaur/feat/chat-form-actions-ui-ux-improvements Aleksander Grygier 2026-08-22 18:02:25 +02:00
  • 0b24f4a36c ui: move dialog close button to the sticky header Aleksander Grygier 2026-08-22 13:12:05 +02:00
  • 515b993632 ui: show modality icons instead of file submenu in chat add menu Aleksander Grygier 2026-08-22 13:11:56 +02:00
  • c390ec891b fix(ui): adjust MCP dialogs for new base sizing and close add sheet before opening settings Aleksander Grygier 2026-08-22 09:55:38 +02:00
  • 3d7ca1a82a refactor(ui): remove settings route and URL-based settings navigation Aleksander Grygier 2026-08-22 09:55:31 +02:00
  • 8573111704 feat(ui): open settings as dialog from sidebar Aleksander Grygier 2026-08-22 09:55:24 +02:00
  • 29776f6a48 refactor(ui): switch SettingsChat to in-app section navigation Aleksander Grygier 2026-08-22 09:55:15 +02:00
  • d72e57944a feat(ui): add DialogSettingsChat dialog Aleksander Grygier 2026-08-22 09:55:08 +02:00
  • dcbf990808 feat(ui): make base dialog responsive and support sticky headers Aleksander Grygier 2026-08-22 09:55:00 +02:00
  • 553f3bc944 ui: replace per-conversation MCP overrides with per-conversation tool policy Aleksander Grygier 2026-08-22 08:03:42 +02:00
  • 45b0c1756a chore: Format Aleksander Grygier 2026-08-22 07:10:06 +02:00
  • a8aef5d05e ui : remove unused MCP servers submenu component Aleksander Grygier 2026-08-22 06:53:45 +02:00
  • c2ef793ea3 ui : remove mcp-servers route and sidebar entry Aleksander Grygier 2026-08-22 06:53:37 +02:00
  • b9c3bb6ec5 ui : browse MCP resources from the server card Aleksander Grygier 2026-08-22 06:53:29 +02:00
  • 20ec1c9539 ui : open MCP servers in a dialog from the chat form Aleksander Grygier 2026-08-22 06:53:22 +02:00
  • c68470019d ui : fix eslint issues in chat form and model selector Aleksander Grygier 2026-08-17 11:04:02 +02:00
  • dfd6ed8564 feat: Enable microphone input as default for audio models Aleksander Grygier 2026-08-16 18:03:52 +02:00
  • 7af64f3a29 ui : add raw model id tooltip to model selector options Aleksander Grygier 2026-08-15 23:26:25 +02:00
  • 8961962316 ui : make model option hover and focus highlight override the active state Aleksander Grygier 2026-08-15 22:54:05 +02:00