Commit History

Автор SHA1 Съобщение Дата
  slaren 49006c67b4 llama : move random seed generation to the samplers (#9398) преди 1 година
  Xuan Son Nguyen bfe76d4a17 common : move arg parser code to `arg.cpp` (#9388) преди 1 година
  Georgi Gerganov f12295b8a9 llama : fix empty ring buffer push (#9358) преди 1 година
  Georgi Gerganov df270ef745 llama : refactor sampling v2 (#9294) преди 1 година
  Georgi Gerganov 938943cdbf llama : move vocab, grammar and sampling into separate files (#8508) преди 1 година
  Kevin Wang 470939d483 common : preallocate sampling token data vector (#8363) преди 1 година
  Kevin Wang ffd00797d8 common : avoid unnecessary logits fetch (#8358) преди 1 година
  Daniel Bevenius e6bf007744 llama : return nullptr from llama_grammar_init (#8093) преди 1 година
  Georgi Gerganov 6ff13987ad common : normalize naming style (#7462) преди 1 година
  Olivier Chafik e402de364b `grammars`: fix resampling logic regression (#7424) преди 1 година
  Johannes Gäßler 5ae3426b0b server: fix reported top tokens for temperature 0 (#7203) преди 1 година
  Johannes Gäßler af0a5b6163 server: fix incorrectly reported token probabilities (#7125) преди 1 година
  David Renshaw 3f167476b1 sampling : use std::random_device{}() for default random seed (#6962) преди 1 година
  Johannes Gäßler 28103f4832 Server: fix seed for multiple slots (#6835) преди 1 година
  Minsoo Cheong 586e7bc561 sampling : deduplicated code for probability distribution access (#6240) преди 1 година
  Clint Herron 463628372d grammar : handle missing "root" node (#6004) преди 1 година
  Minsoo Cheong 6d341ab6c5 speculative : implement stochastic speculative sampling (#5625) преди 1 година
  Pierrick Hymbert e3965cf35a server: tests - slow inference causes timeout on the CI (#5715) преди 1 година
  Robey Holderith 5ee99c32f5 common, server : surface min_keep as its own parameter (#5567) преди 1 година
  Georgi Gerganov 689a091bbe sampling : do not set min_keep to n_probs (#5564) преди 1 година
  Alexey Parfenov 6dcc02d244 server : add "samplers" param to control the samplers order (#5494) преди 1 година
  Alexey Parfenov a803333a4e common : use enums for sampler types (#5418) преди 1 година
  Georgi Gerganov 139b62a839 common : fix compile warning преди 1 година
  Johannes Gäßler 26d4efd11e sampling: fix top_k <= 0 (#5388) преди 1 година
  Michael Klimenko 35a2ee9143 Remove unused data and add fixes (#5154) преди 2 години
  l3utterfly 5eaf9964fc llama : dynamic temperature sampling (#4972) преди 2 години
  David Friehs 4483396751 llama : apply classifier-free guidance to logits directly (#4951) преди 2 години
  Alexey Parfenov 6123979952 server : allow to specify custom prompt for penalty calculation (#3727) преди 2 години
  kalomaze b9ec82d262 grammar : check the full vocab only if necessary (opt) (#4306) преди 2 години
  Georgi Gerganov caa9249217 common : fix compile warning преди 2 години