All pull requests

Pull requests

• Bump doc-builder to get the docs machine-translated

  #956·mishig25 opened Oct 29, 2026·

  ·

• Respect truncation direction when cutting long inputs by characters

  #955·ViderStar opened Oct 25, 2026·2 of 5 tasks·

  ·
  1

• Fix batched DeBERTa-v2 scores: mask padded keys in attention

  #954·cagyirey opened Oct 21, 2026·2 of 5 tasks·

  ·
  1

• Reduce peak memory of the Qwen3 causal attention bias

  #953·aalquwayfili opened Oct 18, 2026·

  ·
  1

• Do not panic when the backend channel is full

  #952·spicysharkco opened Oct 14, 2026·3 of 5 tasks·

  ·
  1

• Bump the actions group across 1 directory with 16 updates

  dependenciesgithub_actions
  #950·dependabot[bot] opened Oct 13, 2026Bot·

  ·

• Fix Gemma3 attention masks in unpadded batches

  #943·anudit opened Oct 2, 2026·

  ·

• Fix CUDA_ERROR_INVALID_VALUE in Nomic MoE router for batches > 65535 tokens

  #942·dietmar opened Sep 30, 2026·2 of 5 tasks·

  ·
  1

• Add OTLP HTTP/protobuf export with HTTPS support

  #940·beriktassuly opened Sep 30, 2026·

  ·
  1

• Make native CPU optimizations opt-in

  #939·beriktassuly opened Sep 29, 2026·

  ·
  1

• Upgrade hf-hub to 1.x

  #938·assafvayner opened Sep 29, 2026Collaborator·

  ·
  2

• Keep requests in te_queue_size until their batch is dispatched

  #936·bhaskargurram-ai opened Sep 26, 2026·3 of 5 tasks·

  ·
  1

  2

• Add OTel trace_id and span_id to JSON logs

  #935·modelpath-dev opened Sep 26, 2026·4 of 5 tasks·

  ·
  1

• fix: panic on NaN scores in predict softmax (infer.rs)

  #932·dsingal0 opened Sep 19, 2026·

  ·

• docs(rocm): build from source inside the TEI ROCm image

  #929·Abdennacer-Badaoui opened Sep 15, 2026Member·

  ·
  1

• docs: deploy TEI on Nebius Serverless AI Endpoints

  #923·SalikovAlex opened Sep 11, 2026·4 tasks done·

  ·

• Fix single embed_sparse and embed_all gRPC request metrics

  #919·shoemoney opened Aug 24, 2026·4 of 5 tasks·

  ·

• Add BF16 support for CUDA (Ampere or newer)

  #915·MyoungHaSong opened Aug 14, 2026·2 of 5 tasks·

  ·

• Add OTLP over HTTP/protobuf export support

  #910·AmreshSinha opened Aug 4, 2026·5 tasks·

  ·
  1

• Add SigLIP model - text embeddings only

  #903·m-toman opened Jul 28, 2026·4 of 5 tasks·

  ·
  5

• Add BAAI/bge-m3 sparse (lexical_weights) support via --pooling m3_sparse

  #899·SpongeRobert opened Jul 18, 2026·

  ·
  1

• Speed up ModernBERT inference by 1.15x on CUDA

  #897·hotchpotch opened Jul 11, 2026·2 of 4 tasks·

  ·

  · v1.9.4
  9

• Add --root-path to specificy HTTP prefix

  #894·alvarobartt opened Jul 9, 2026Member·4 of 5 tasks·

  ·

  · v1.9.4
  1

  1

• feat(qwen3): support reranker on candle backend

  #886·malaiwah opened Jun 26, 2026·4 of 5 tasks·

  ·
  1

  1

• feat: allow controlling startup warmup tokens

  #884·malaiwah opened Jun 25, 2026·4 of 5 tasks·

  ·
  2