Commit History

Автор SHA1 Съобщение Дата
  Chenguang Li 67e3f6f601 CANN: add operator fusion support for ADD + RMS_NORM (#17512) преди 3 седмици
  Francisco Herrera 92ac1e016b doc: clarify that steps also apply to linux for opencl (#18002) преди 3 седмици
  Ali Tariq 8e3a761189 ci : init git lfs in every build for RISC-V (#18590) преди 3 седмици
  Daniel Bevenius d3dce4e0a5 sampling : add support for backend sampling (#17004) преди 3 седмици
  Tarek Dakhran 4974bf53cf model : mtmd : make input norm optional in LFM2-VL (#18594) преди 3 седмици
  Aman Gupta 908a9e5a1e CUDA: disable cuda graph when using n-cpu-moe (#18593) преди 3 седмици
  Aman Gupta 5126c41c1c ggml-cuda: remove unused params in ggml_cuda_graph (#18579) преди 3 седмици
  Aldehir Rojas cef1d23c5a common/grammar : replace problematic backtracking regex `[\s\S]*` (#18342) преди 3 седмици
  Georgi Gerganov c69c7ebc90 graph : fix graph reuse logic when `n_pos_per_embd > 1` (#18566) преди 3 седмици
  Aman Gupta e57f52334b ggml-cuda: fixes for concurrent streams (#18496) преди 3 седмици
  Georgi Gerganov a554a1ecc7 context : fix reserve token padding to n_seqs (#18536) преди 3 седмици
  Johannes Gäßler 0f2e42ca1d CUDA: only allocate FA tmp buffer if needed (#18564) преди 3 седмици
  pl752 9dba9f5352 (Bugfix, ggml-cuda) Pool alloc count fix + small size computation type adjustment (#18559) преди 3 седмици
  Shouyu bcfc8c3cec ggml-hexagon: optimize activation function (#18393) преди 3 седмици
  Jeff Bolz 18ddaea2ae vulkan: Optimize GGML_OP_CUMSUM (#18417) преди 3 седмици
  Jeff Bolz 706e3f93a6 vulkan: Implement mmvq for iq1_s/iq1_m (#18450) преди 3 седмици
  Prabod 5755e52d15 model : Maincoder-1B support (#18534) преди 3 седмици
  Georgi Gerganov f38de16341 metal : adjust extra size for FA buffer to avoid reallocations (#18545) преди 3 седмици
  Georgi Gerganov af1e8e1a6c graph : reduce topology branching (#18548) преди 3 седмици
  Georgi Gerganov d84a6a98be vocab : reduce debug logs about non-EOG control tokens (#18541) преди 3 седмици
  Chris Rohlf c6f0e832da rpc : use unordered_map::reserve and emplace (#18513) преди 3 седмици
  MeeMin e86f3c2221 cuda : fix copy of large tensors (ggml_nbytes <= INT_MAX assertion) (#18433) преди 3 седмици
  Sigbjørn Skjæret 169ee68ffb model : remove modern-bert iswa template (#18529) преди 3 седмици
  tt ced765be44 model: support youtu-vl model (#18479) преди 3 седмици
  Piotr Wilkin (ilintar) 3ccccc83f7 Add conversion support for IQuestCoderForCausalLM (#18524) преди 4 седмици
  o7si d0a6a31470 model : add support for JinaBertModel with non-gated ffn (#18475) преди 4 седмици
  o7si 2b2afade9f convert : fix encoding of WPM vocab for BERT models (#18500) преди 4 седмици
  HelloKS f4f5019254 model: add Solar Open model (#18511) преди 4 седмици
  Anri Lombard d5574c919c webui: fix code copy stripping XML/HTML tags (#18518) преди 4 седмици
  Aman Gupta 26831bded9 ggml-cuda: remove unneccesary prints on ggml_cuda_init (#18502) преди 4 седмици