Commit Graph

Select branches

Hide Pull Requests

fable5/prefetch-experts

fable5/vram-ondemand

fable5/vram-ondemand-rebased-20260726

gemini

Mono Color

• edd4d9bca5 vulkan: add FA dequant for q4_1, q5_0, q5_1, iq4_nl (#21029) mkoker 2026-04-07 07:41:29 -04:00
• 482192f12d webui : store reasoning_content so it is sent back in subsequent requests (#21249) Aldehir Rojas 2026-04-07 06:32:44 -05:00
• 71a81f6fcc ggml-cuda : fix CDNA2 compute capability constant for gfx90a (MI210) (#21519) Antoine Viallon 2026-04-07 12:18:55 +02:00
• ecce0087da fix: Detect streaming state in reasoning content blocks (#21549) Aleksander Grygier 2026-04-07 12:04:41 +02:00
• d1f82e382d Fix rtl text rendering (#21382) Kabir08 2026-04-07 15:07:20 +05:30
• 0988accf82 [SYCL] Add Q8_0 reorder optimization (~3x tg speedup on Intel Arc) (#21527) PMZFX 2026-04-07 04:12:49 -04:00
• 0033f53a07 docs: fix typo in build.md (emdawbwebgpu -> emdawnwebgpu) (#21518) Dmytro Romanov 2026-04-07 06:37:26 +02:00
• d0a6dfeb28 ggml-webgpu: Add the support of MUL_MAT_ID (#21147) Masashi YoshimuraandReese Levine 2026-04-07 05:08:46 +09:00
• 2e1f0a889e ggml: add Q1_0 1-bit quantization support (CPU) (#21273) Pasha KhosraviandSigbjørn Skjæret 2026-04-06 11:55:21 -07:00
• 506200cf8b cli: fix stripping of \n in multiline input (#21485) Bipin YadavandSigbjørn Skjæret 2026-04-07 00:24:06 +05:30
• 15f786e658 [CUDA ] Write an optimized flash_attn_stream_k_fixup kernel (#21159) Gaurav Garg 2026-04-07 00:04:29 +05:30
• 94ca829b60 llama-bench: add -fitc and -fitt to arguments (#21304) Aman Gupta 2026-04-06 22:26:02 +08:00
• 4aa962e2b0 vocab : add byte token handling to BPE detokenizer for Gemma4 (#21488) Aldehir Rojas 2026-04-06 09:08:37 -05:00
• 941146b3f1 convert : fix block_ff_dim retrieval for lfm2 (#21508) Sigbjørn Skjæret 2026-04-06 14:05:18 +02:00
• 482d862bcb server : handle unsuccessful sink.write in chunked stream provider (#21478) lainon1 2026-04-06 13:03:02 +01:00
• 3979f2bb08 docs: add hunyuan-ocr gguf, also add test [no ci] (#21490) Xuan-Son Nguyen 2026-04-06 14:02:37 +02:00
• 400ac8e194 convert : set "add bos" == True for Gemma 4 (#21500) Georgi Gerganov 2026-04-06 13:52:07 +03:00
• f51fd36d79 sycl : handle other FA case (#21377) Neo Zhang 2026-04-06 18:28:00 +08:00
• 25eec6f327 hexagon: slight optimization for argosrt output init (#21463) Yarden Tal 2026-04-06 04:30:25 +03:00
• 58190cc84d llama : correct platform-independent loading of BOOL metadata (#21428) anchortense 2026-04-06 09:40:38 +10:00
• af76639f72 model : add HunyuanOCR support (#21395) 3 people 2026-04-05 23:32:14 +02:00
• 761797ffdf ci : use default RISE RISC-V Runners (#21263) Ludovic Henry 2026-04-05 20:29:48 +02:00
• 5d3a4a7da5 server : fix logging of build + system info (#21460) ddh0 2026-04-05 09:14:02 -05:00
• c08d28d088 ci: lower cuda12 floor to 12.8.1 for broader host compatibility (#21438) M1DNYT3andM1DNYT3 2026-04-05 04:04:00 +03:00
• 661e9acb36 ci: fix vulkan workflow referencing non-existent action (#21442) Nicholas Sparks 2026-04-04 20:59:51 -04:00
• b8635075ff common : add gemma 4 specialized parser (#21418) Aldehir Rojas 2026-04-04 13:39:00 -05:00
• 9c699074c9 server: Fix undefined timing measurement errors in server context (#21201) Dan HoffmanandDan Hoffman 2026-04-04 07:11:19 -07:00
• d01f6274c0 common : respect specified tag, only fallback when tag is empty (#21413) Adrien Gallouët 2026-04-04 15:08:03 +02:00
• 650bf14eb9 llama-model: read final_logit_softcapping for Gemma 4 (#21390) SamareshSingh 2026-04-04 06:05:10 -05:00
• b7ad48ebda llama: add custom newline split for Gemma 4 (#21406) Aman Gupta 2026-04-04 15:06:34 +08:00
• d006858316 ggml-webgpu: move from parameter buffer pool to single buffer with offsets (#21278) Reese Levine 2026-04-03 11:40:14 -07:00
• e439700992 ci: Add Windows Vulkan backend testing on Intel (#21292) Masato Nakasaka 2026-04-04 02:16:44 +09:00
• 50e0ad08fb server: save and clear idle slots on new task (--clear-idle) (#20993) Yes You Can Have Your OwnandGeorgi Gerganov 2026-04-03 20:02:27 +03:00
• f1f793ad06 common/parser: fix call ID detection (Mistral parser mostly) + atomicity for tag-json parsers (#21230) Piotr Wilkin (ilintar)andSigbjørn Skjæret 2026-04-03 17:51:52 +02:00
• af5c13841f common : fix tool call type detection for nullable and enum schemas (#21327) Samanvya Tripathi 2026-04-03 11:51:23 -04:00
• 277ff5fff7 docker : bump cuda12 to 12.9.1 (#20920) 3 people 2026-04-03 16:06:45 +03:00
• 384c0076bc docs: Update build.md: HSA_OVERRIDE_GFX_VERSION clarification (#21331) jeromew 2026-04-03 15:05:14 +02:00
• 1f34806c44 jinja: coerce input for string-specific filters (#21370) Sigbjørn Skjæret 2026-04-03 15:03:33 +02:00
• 887535c33f ci: add more binary checks (#21349) Aaron Teo 2026-04-03 20:50:00 +08:00
• d3416a4aa9 fix: remove stale assert (#21369) Piotr Wilkin (ilintar) 2026-04-03 13:40:41 +02:00
• 43a4ee4a2c HIP: build eatch ci build test for a different architecture (#21337) uvos 2026-04-03 11:38:22 +02:00
• f851fa5ab0 fix: add openssl to nix dependencies (#21353) (#21355) Tillerino 2026-04-03 11:21:07 +02:00
• f1ac84119c ggml-zendnn : add MUL_MAT_ID op support for MoE models (#21315) Vishal SinghandAaron Teo 2026-04-03 14:49:08 +05:30
• b069b10ab4 vocab: fix Gemma4 tokenizer (#21343) 3 people 2026-04-03 10:33:03 +02:00
• 0c58ba3365 rpc : reuse compute graph buffers (#21299) Radoslav Gerganov 2026-04-03 10:28:09 +03:00
• 57ace0d612 chat : avoid including json in chat.h (#21306) Georgi Gerganov 2026-04-03 09:07:59 +03:00
• 39b27f0da0 (revert) kv-cache : do not quantize SWA KV cache (#21332) Georgi Gerganov 2026-04-03 09:07:01 +03:00
• f49e917876 ci : add AMD ZenDNN label to PR labeler (#21345) Vishal SinghandAaron Teo 2026-04-03 08:05:15 +05:30
• 7c7d6ce5c7 [HIP] Bump ROCm version to 7.2.1 (#21066) Slobodan JosicandSigbjørn Skjæret 2026-04-03 00:59:20 +02:00
• 5208e2d5ba fix: gemma 4 template (#21326) Piotr Wilkin (ilintar) 2026-04-02 23:31:02 +02:00
• 7992aa7c8e tests : add unit test coverage for llama_tensor_get_type (#20112) 3 people 2026-04-02 16:53:58 -04:00
• a1cfb64530 ggml-webgpu: add vectorized flash attention (#20709) Zheyuan ChenandReese Levine 2026-04-02 10:40:42 -07:00
• 5803c8d115 tests: allow exporting graph ops from HF file without downloading weights (#21182) Ruben Ortlam 2026-04-02 18:19:20 +02:00
• 63f8fe0ef4 model, mtmd: fix gguf conversion for audio/vision mmproj (#21309) Xuan-Son Nguyen 2026-04-02 17:10:32 +02:00
• 223373742b common : add commentary rules for gpt-oss-20b (#21286) Aldehir Rojas 2026-04-02 08:59:59 -05:00
• e15efe007d Relax prefill parser to allow space. (#21240) Piotr Wilkin (ilintar) 2026-04-02 11:29:11 +02:00
• 6137c325a1 chat : add Granite 4.0 chat template with correct tool_call role mapping (#20804) Jesus TalaveraandClaude Opus 4.6 2026-04-02 11:28:56 +02:00
• 17193cce34 kv-cache : do not quantize SWA KV cache (#21277) Georgi Gerganov 2026-04-02 11:54:05 +03:00
• d6dac92bfd Ignore Transfer-Encoding header. (#20269) Roger Chen 2026-04-02 16:41:19 +08:00
• dae2bf41c9 sync : ggml Georgi Gerganov 2026-04-02 10:38:24 +03:00
• bc07d55922 ggml : bump version to 0.9.11 (ggml/1456) Georgi Gerganov 2026-04-02 10:37:26 +03:00
• 4888137b17 sycl : fix llama_kv_cache hang when kv_cache is huge: 5GB (#21283) Neo Zhang 2026-04-02 15:08:32 +08:00
• fbd441c379 hexagon : add cumsum op support (#21246) Todor BoinovskiandMax Krasnyansky 2026-04-01 17:44:02 -07:00
• c30e012253 contrib : rewrite AGENTS.md, make it more clear about project values (#21270) Xuan-Son Nguyen 2026-04-01 23:31:51 +02:00
• 95a6ebabb2 opencl: fix leak in Adreno q8_0 path (#21212) lhez 2026-04-01 12:54:58 -07:00
• 12dbf1da95 server: Bypass API Key validation for WebUI static bundle assets (#21269) Aleksander Grygier 2026-04-01 21:32:15 +02:00
• 86221cf6da CUDA: fix FA kernel selection logic (#21271) Johannes Gäßler 2026-04-01 21:28:19 +02:00
• 6de97b9d3e kleidiai: add CPU feature detection to CI run script (#20394) Martin Klacer 2026-04-01 18:02:41 +01:00
• 5a0ed5150a Update Dawn version in WebGPU CI (#20784) Nikhil Jain 2026-04-01 09:53:05 -07:00
• 8710e5f9b9 hexagon: improve RMS_NORM and DIV accuracy (#21251) Aparna M PandKrishna Sridhar 2026-04-01 21:13:08 +05:30
• 1d6d4cf7a5 fix: tool call parsing for LFM2 and LFM2.5 models (#21242) Jonathan 2026-04-01 07:22:44 -07:00
• 744c0c7310 llama : rotate activations for better quantization (#21038) Georgi Gerganov 2026-04-01 16:58:01 +03:00
• 0356e33aaf scripts: add function call test script (#21234) Xuan-Son Nguyen 2026-04-01 15:31:58 +02:00
• 6422036fcb sync : ggml Georgi Gerganov 2026-04-01 16:02:34 +03:00
• 296bc0538b ggml : bump version to 0.9.10 (ggml/1454) Georgi Gerganov 2026-04-01 16:01:45 +03:00
• 6b949d1078 sycl : support nvfp4 type in mul_mat (#21227) Neo Zhang 2026-04-01 18:54:15 +08:00
• 84f82e846c ggml-cuda: Add generic NVFP4 MMQ kernel (#21074) Michael WandandJohannes Gäßler 2026-04-01 03:04:58 -07:00
• e1cb817483 memory: respect unified KV cache in hybrid memory for eval tasks (#21224) Ettore Di Giacinto 2026-04-01 11:50:17 +02:00
• 88d5f8ffc3 CUDA/HIP: Fix kernel slection for mmvq mmid kernel to align host selection with device launch bounds (#21238) uvos 2026-04-01 10:21:20 +02:00
• d43375ff7f ggml : fix RWKV ops thread assignment (#21226) Georgi Gerganov 2026-04-01 11:10:25 +03:00
• 2b86e5cae6 ggml-cpu: fix fallback for RVV kernels without zvfh (#21157) Taimur Ahmad 2026-04-01 13:10:03 +05:00
• 88458164c7 CUDA: Add Flash Attention Support for Head Dimension 512 (#20998) Anav PrasadandJohannes Gäßler 2026-04-01 07:07:24 +00:00
• 4951250235 llama : refactor llama_model_quantize_params to expose a pure C interface (#20346) Ed AddarioandGeorgi Gerganov 2026-04-01 06:43:00 +01:00
• 82764c341a ggml webgpu: quantized buffers to u32 + wider browser/device support (#21046) Reese Levine 2026-03-31 22:38:24 -07:00
• 825eb91a66 ggml-webgpu: port all AOT operators to JIT (#20728) Abhijit Ramesh 2026-03-31 15:38:16 -07:00
• 0fcb3760b2 fix: Use lower-case proxy headers naming (#21235) Aleksander Grygier 2026-03-31 17:47:46 +02:00
• 6307ec07d3 common : cleanup logs and modernize the progress bar (#21215) Adrien Gallouët 2026-03-31 16:18:00 +02:00
• 632219af73 CANN: fix multi-thread set_tensor race conditions (#20151) hipudding 2026-03-31 22:00:51 +08:00
• 4a00bbfed6 server: (webui) no more gzip compression (#21073) Xuan-Son NguyenandAleksander Grygier 2026-03-31 15:44:26 +02:00
• 624733d631 common : gpt-oss handle builtin and unsolicited tool calls (#21213) Aldehir Rojas 2026-03-31 06:52:42 -05:00
• 0b6ff47996 fix: correct misspellings in code comments (#21217) lainon1 2026-03-31 12:50:51 +01:00
• eec6f85d7b CI: Enable CPU and Vulkan ARM64 Release (#21207) Seungmin Kim 2026-03-31 20:02:56 +09:00
• 9281dd135d sync : ggml Georgi Gerganov 2026-03-31 13:08:13 +03:00
• 0be6c7c9ce ggml : bump version to 0.9.9 (ggml/1449) Georgi Gerganov 2026-03-30 18:34:29 +03:00
• 41361c8599 common : move up common_init() and fix Windows UTF-8 logs (#21176) Adrien Gallouët 2026-03-31 12:53:41 +02:00
• 62278cedde sycl : enhance fattn perf (#21185) Neo Zhang 2026-03-31 18:31:50 +08:00
• 90aa83c6bd common: add bounds check in common_init_result::sampler to prevent segfault on failed model load (#21082) mtmcp 2026-03-31 07:04:42 -03:00
• fcc2d598c8 fix: include API key in CORS proxy requests for MCP connections (#21193) SATISH K C 2026-03-31 03:52:34 -05:00
• 4453e77561 server/webui: cleanup dual representation approach, simplify to openai-compat (#21090) Piotr Wilkin (ilintar)andAleksander Grygier 2026-03-31 10:42:06 +02:00
• 26dac845cc vendor : update BoringSSL to 0.20260327.0 (#21211) Adrien Gallouët 2026-03-31 09:21:54 +02:00