História revízii

Autor SHA1 Správa Dátum
  AlpinDale d8c4193704 feat: Speculative Decoding using a draft model (#432) 9 mesiacov pred
  50h100a f67b5be198 chore: port sampler+metadata changes from main to dev (#427) 9 mesiacov pred
  AlpinDale fa083286e3 Speculative Decoding Part 4: Lookahead scheduling (#402) 10 mesiacov pred
  AlpinDale 3abc641d68 directly use in forward pass 10 mesiacov pred
  AlpinDale c3c374396b logprobs fixes 10 mesiacov pred
  AlpinDale 2efee6bcc6 optimize logprob ranks 10 mesiacov pred
  AlpinDale 777b6f6d51 add logprob ranks 10 mesiacov pred
  AlpinDale 0c4ead5e9f min_tokens 10 mesiacov pred
  AlpinDale d1786645a3 fix formatting 10 mesiacov pred
  AlpinDale f01c668259 clean up sampler 10 mesiacov pred
  AlpinDale 78d66f16d1 Chunked Prefill Part 1 (#384) 10 mesiacov pred
  AlpinDale 9181fa0396 feat: Triton kernels for sampling (#383) 10 mesiacov pred
  AlpinDale f8dfac6372 chore: attention refactor and upstream sync apr01 (#365) 10 mesiacov pred
  50h100a d5dbd29db4 hoist sampler internals into a single function. 10 mesiacov pred
  AlpinDale da223153c6 feat&fix: cohere support and missing GPU blocks (#333) 11 mesiacov pred
  AlpinDale e42a78381a feat: switch from pylint to ruff (#322) 11 mesiacov pred
  AlpinDale 9fa99215f8 feat: add cubic sampling (#280) 11 mesiacov pred
  AlpinDale 657aec0cbd refactor: OpenAI endpoint (#261) 11 mesiacov pred
  AlpinDale 4d04ade9ef feat: fine-grained seeds (#279) 11 mesiacov pred
  anon998 35b9033782 fix: crash in quadratic sampling when batch > 1 (#253) 1 rok pred
  50h100a f619c96c79 fix: zero token output due to temperature bias (#243) 1 rok pred
  50h100a 53a9c60442 fix: logit processor declarations and application (#242) 1 rok pred
  AlpinDale e73a92ad2f fix: remove the mask for quadratic sampling (#236) 1 rok pred
  AlpinDale 1c46fa31ad feat: add quadratic sampling (#233) 1 rok pred
  AlpinDale c3a221eb02 feat: GGUF, QuIP#, and Marlin support (#228) 1 rok pred
  AlpinDale c0aac15421 feat: S-LoRA support (#222) 1 rok pred
  AlpinDale 8fa608aeb7 feat: replace Ray with NCCL for control plane comms (#221) 1 rok pred
  Stefan Gligorijevic 9e7e108dc8 chore: clamp dynatemp_min (#214) 1 rok pred
  Stefan Gligorijevic 56446a04bb feat: dynamic temperature (#209) 1 rok pred
  AlpinDale d54791aaa8 feat: reduce sampler overhead by making it less blocking (#198) 1 rok pred