Commit History

Автор SHA1 Съобщение Дата
  Xuan-Son Nguyen 3b24d26c22 server : update docs (#13432) преди 9 месеца
  Sigbjørn Skjæret 43dfd741a5 llguidance : set tokenizer slices to default (#13424) преди 9 месеца
  Thammachart Chinvarapon b064a51a4e ci: free_disk_space flag enabled for intel variant (#13426) преди 9 месеца
  Xuan-Son Nguyen 053367d149 mtmd : support InternVL 2.5 and 3 (#13422) преди 9 месеца
  Johannes Gäßler d8919424f1 CUDA: fix FlashAttention on Turing (#13415) преди 9 месеца
  Xuan-Son Nguyen 7fef11766c arg : add env var to control mmproj (#13416) преди 9 месеца
  Jeff Bolz dc1d2adfc0 vulkan: scalar flash attention implementation (#13324) преди 9 месеца
  Helton Reis 7c28a74e07 chore(llguidance): use tagged version that does not break the build (#13413) преди 9 месеца
  Xuan-Son Nguyen 33eff40240 server : vision support via libmtmd (#12898) преди 9 месеца
  Alberto Cabrera Pérez 17512a94d6 sycl : implementation of reordered Q4_0 MMVQ for Intel GPUs (#12858) преди 9 месеца
  Georgi Gerganov 611aa914ef metal : optimize MoE for large batches (#13388) преди 9 месеца
  Johannes Gäßler 0cf6725e9f CUDA: FA support for Deepseek (Ampere or newer) (#13306) преди 9 месеца
  Diego Devesa 27ebfcacba llama : do not crash if there is no CPU backend (#13395) преди 9 месеца
  Johannes Gäßler 5c86c9ed3e CUDA: fix crash on large batch size for MoE models (#13384) преди 9 месеца
  Bartowski efb8b47eda imatrix : Add --parse-special for enabling parsing of special tokens in imatrix calculation (#13389) преди 9 месеца
  R0CKSTAR 0527771dd8 llama-run: add support for downloading models from ModelScope (#13370) преди 9 месеца
  Xuan-Son Nguyen 2189fd3b63 mtmd : fix batch_view for m-rope (#13397) преди 9 месеца
  Xuan-Son Nguyen 3f96aeff39 llama : one-off chat template fix for Mistral-Small-2503 (#13398) преди 9 месеца
  Radoslav Gerganov b486ba05bf rpc : add rpc_msg_set_tensor_hash_req (#13353) преди 9 месеца
  Jeff Bolz 02115dcd9a vulkan: Allow up to 4096 elements for mul_mat_id row_ids (#13326) преди 9 месеца
  Xuan-Son Nguyen d9c4accaff server : (webui) rename has_multimodal --> modalities (#13393) преди 9 месеца
  Diego Devesa 15e03282bb ci : limit write permission to only the release step + fixes (#13392) преди 9 месеца
  Matt Clayton f05a6d71a0 mtmd : Expose helper_decode_image_chunk (#13366) преди 9 месеца
  Xuan-Son Nguyen ee01d71e58 server : (webui) fix a very small misalignment (#13387) преди 9 месеца
  Xuan-Son Nguyen 8c83449cb7 server : (webui) revamp the input area, plus many small UI improvements (#13365) преди 9 месеца
  Sigbjørn Skjæret 1a844be132 convert : support rope_scaling type and rope_type (#13349) преди 9 месеца
  welix 0ccc121354 mtmd : fix the calculation of n_tokens for smolvlm (#13381) преди 9 месеца
  Georgi Gerganov 6562e5a4d6 context : allow cache-less context for embeddings (#13108) преди 9 месеца
  Georgi Gerganov 51fb96b1ff context : remove logits_all flag (#13284) преди 9 месеца
  Diego Devesa 70a6991edf ci : move release workflow to a separate file (#13362) преди 9 месеца