All pull requests

Pull requests

• feat: Support for training and inference on Ascend NPU

  #367·liwenmurong opened Oct 31, 2026·

  ·
  1

• data: supply ESM features during training

  #366·dajiaohuang opened Oct 27, 2026·

  ·
  1

• esm: preserve inference sample identity in cache error handling

  #365·Linxiushen opened Oct 26, 2026·

  ·

• fix: make DeepSpeed an optional extra instead of a hard install dependency

  #364·aaronkanzer opened Oct 25, 2026·

  ·

• fix: fall back to the native LayerNorm when the fused CUDA kernel cannot build

  #363·aaronkanzer opened Oct 25, 2026·

  ·

• fix: gate ESM checkpoint download on esm.enable, not the model name

  #360·aaronkanzer opened Oct 25, 2026·

  ·
  1

• metrics: support scalar NumPy arrays in aggregation

  #359·Linxiushen opened Oct 19, 2026·

  ·

• Add Intel GPU (XPU) support for inference

  #358·rajeshgangireddy opened Oct 15, 2026·

  ·
  1

• tfg: handle empty and single-chain steric inputs

  #353·Linxiushen opened Sep 27, 2026·

  ·
  1

• template: reject invalid negative JSON indices

  #345·dajiaohuang opened Aug 22, 2026·

  ·
  1

• web_service: fix covalent position validation

  #344·dajiaohuang opened Aug 22, 2026·

  ·
  1

• docs: clarify inference demo script usage

  #343·dajiaohuang opened Aug 22, 2026·

  ·
  1

• tfg: avoid NaN gradients for unbounded distances

  #339·dajiaohuang opened Aug 22, 2026·

  ·
  1

• docs: make README links portable to PyPI

  #338·dajiaohuang opened Aug 22, 2026·

  ·
  1

• web_service: fix viewer modification schema

  #337·dajiaohuang opened Aug 22, 2026·

  ·
  1

• model: fall back to torch layer norm on sm_120

  #333·dajiaohuang opened Aug 22, 2026·

  ·
  1

• Log a concise warning instead of a full traceback when MSA search falls back

  #328·ssiddhantsharma opened Jul 26, 2026·

  ·
  1

• Support job-scoped custom CCD definitions for PTM inference

  #314·taivu1998 opened May 10, 2026·

  ·

• Support direct monomer A3M inputs for MSA pairing

  #313·taivu1998 opened May 10, 2026·

  ·
  1

• Support direct CIF/mmCIF template inputs for inference

  #312·taivu1998 opened May 10, 2026·

  ·

• Add native score-only confidence CLI for existing structures

  #311·taivu1998 opened May 10, 2026·

  ·

• Add TFG contact restraints for JSON constraints

  #310·taivu1998 opened May 10, 2026·

  ·

• CUDA graph caching for diffusion: auto-mode with token-count threshold

  #285·longleo17 opened Mar 24, 2026Contributor·

  ·

• Batched multi-seed inference: 2x speedup on 5-seed runs

  #284·longleo17 opened Mar 24, 2026Contributor·

  ·