transformers
feat: Sequential beam search(a.k.a Low-memory beam search)
#26304
Merged
Overview
Commits
18
Changes
View On GitHub
Commits
  • add sequential beam search (clean version)
    Saibo-creator committed 2 years ago
  • fix bug in key_value split, this solves the problem of output mismatch
    Saibo-creator committed 2 years ago
  • update the logger message with accordant weights_file_name (#28181)
    Saibo-creator committed 1 year ago
  • [`Llava`] Fix llava index errors (#28032)
    Saibo-creator committed 1 year ago
  • [`Awq`] Enable the possibility to skip quantization for some target modules (#27950)
    Saibo-creator committed 1 year ago
  • fix FA2 when using quantization (#28203)
    Saibo-creator committed 1 year ago
  • small typo (#28229)
    Saibo-creator committed 1 year ago
  • Update docs around mixing hf scheduler with deepspeed optimizer (#28223)
    Saibo-creator committed 1 year ago
  • Fix trainer saving safetensors: metadata is None (#28219)
    Saibo-creator committed 1 year ago
  • fix bug:divide by zero in _maybe_log_save_evaluate() (#28251)
    Saibo-creator committed 1 year ago
  • [Whisper] Fix errors with MPS backend introduced by new code on word-level timestamps computation (#28288)
    Saibo-creator committed 1 year ago
  • Remove fast tokenization warning in Data Collators (#28213)
    Saibo-creator committed 1 year ago
  • Bump tj-actions/changed-files from 22.2 to 41 in /.github/workflows (#28311)
    Saibo-creator committed 1 year ago
  • Merge branch 'main' into fix_issue_22639
    gante committed 1 year ago
  • refactor _split_model_inputs function for sequential beam search.
    Saibo-creator committed 1 year ago
  • fix error with test_modeling_whisper.py
    Saibo-creator committed 1 year ago
  • fix syntax error(feature only available in python3.10)
    Saibo-creator committed 1 year ago
  • fix typo
    Saibo-creator committed 1 year ago
FAQ Terms Privacy Refunds Impressum
Loading