Commits

Branch selector

 main

User selector

All users

Datepicker

All time

Commit History

Commits on Jan 6, 2025

• vlm: fix feature size calculation for llava-next models (#1079)

  AlpinDale

  authored
  766ea79

• tests: refactor model tests (#1078)

  Show description for 7b6501b
  AlpinDale

  authored
  7b6501b

Commits on Jan 5, 2025

• fix: unexpected kwarg for the legacy API server (#1076)

  AlpinDale

  authored
  f6df92b

• fix: validate n in the sampling params (#1075)

  AlpinDale

  authored
  814c850

Commits on Jan 3, 2025

• api: support LoRA lineage and base model metadata management (#1072)

  Show description for 6212072
  AlpinDale

  authored
  6212072

• rocm: enable multi-step scheduling for rocm (#1071)

  AlpinDale

  authored
  d9d287a

• fix: Phi3.5 Mini and MoE LoRA inference (#1070)

  AlpinDale

  authored
  ec17b6c

• vlm: add support for molmo vision model (#1069)

  AlpinDale

  authored
  acc0c72

• build: guard against changes in cuda library name (#1068)

  AlpinDale

  authored
  7dd001e

• sampler: simplify logits resort in _apply_top_k_top_p (#1067)

  AlpinDale

  authored
  ca7028d

• rocm: add support for FP8 KV cache in the custom paged attention kkernels (#1066)

  AlpinDale

  authored
  61aed09

• api: enable MQAphroditeEngine for embedding models (#1065)

  AlpinDale

  authored
  12b0059

Commits on Jan 2, 2025

• fix: encoder-decoder models for beam search (#1064)

  AlpinDale

  authored
  314fa7f

Commits on Jan 1, 2025

• api: non-zero exit code if MQ engine startup fails (#1063)

  AlpinDale

  authored
  34cf9b7

• rocm: add more quants, fix _scaled_mm call (#1062)

  AlpinDale

  authored
  92cee43

• readme: add attribution to Ruliad

  AlpinDale

  committed
  b12d5c0

• distributed: bind only to 127.0.0.1 for local-only usage (#1061)

  AlpinDale

  authored
  f81e7d7

• core: support prompt logprobs in multi-step (#1060)

  AlpinDale

  authored
  58aff37

Commits on Dec 31, 2024

• feat: introduce MQAphroditeEngine (#1056)

  Show description for 9a7d551
  AlpinDale

  authored
  9a7d551

Commits on Dec 30, 2024

• fix: add missing logit index increment in sampling metadata prep (#1059)

  AlpinDale

  authored
  0b5588d

Commits on Dec 27, 2024

• build: fix compilation for causal_conv1d_fwd kernel signature (#1057)

  AlpinDale

  authored
  525edc1

• mamba: enable continuous batching for mamba kernels (#1055)

  AlpinDale

  authored
  9bdf8d5

• fix: granite logit scale in logit computation (#1054)

  AlpinDale

  authored
  11f49b5

• api: add mistral function calling format to all models loaded with "mistral" format (#1053)

  AlpinDale

  authored
  1264e0b

• quant: add tensor parallel support for bitsandbytes (#1052)

  Show description for b3f9ab3
  AlpinDale

  authored
  b3f9ab3

• core: add cuda graph support for encoder-decoder models (#1051)

  AlpinDale

  authored
  a985143

• torch.compile: register all-reduce operations as custom ops (#1050)

  AlpinDale

  authored
  239a8ca

• chore: remove dead code from triton sampling kernels (#1049)

  AlpinDale

  authored
  4593a3b

• kernel: asymmetric AQ AZP quantization kernels (#1048)

  AlpinDale

  authored
  8976805

• fix: clean shutdown issues (#1047)

  AlpinDale

  authored
  638c08d

• tpu: implement multi-step scheduling (#1046)

  AlpinDale

  authored
  4b1b658

• torch.compile: fix functionalization (#1045)

  AlpinDale

  authored
  960dee2

• model: add support for MiniCPM-3 (#1044)

  AlpinDale

  authored
  ce7b602

• rocm: add custom paged attention kernels for ROCm (#1043)

  AlpinDale

  authored
  4a7cb8f

• xpu: bump IPEX to 2.3, support GQA (#1042)

  AlpinDale

  authored
  6951928

Pagination

PreviousNext