llama.cpp
ggml-webgpu: Fix some binding alias issues to support all archs, fix recurrent-state-rollback test
#25931
Merged

Commits
  • Add overlap glu variant to support all archs, fix recurrent-state-rollback test
    reeselevine committed 124 days ago
  • format
    reeselevine committed 124 days ago
  • Fix all arch overlapped ranges
    reeselevine committed 124 days ago
  • format
    reeselevine committed 124 days ago
  • diagnose bus error on apple ci
    reeselevine committed 124 days ago
  • More testing
    reeselevine committed 124 days ago
  • more testing
    reeselevine committed 124 days ago
  • more targeted testing
    reeselevine committed 124 days ago
  • Fix bug in alignment for > 4gb buffer offsets
    reeselevine committed 124 days ago
  • Fix bug in view offsets
    reeselevine committed 124 days ago
  • Try avoiding multi_buffers
    reeselevine committed 124 days ago
  • not fixed yet, more logging :(
    reeselevine committed 124 days ago
  • Handle edge case in set_rows
    reeselevine committed 123 days ago
  • Try looking at view source
    reeselevine committed 123 days ago
  • Skip deepseek32 for now and clean up trace infrastructure
    reeselevine committed 123 days ago
  • simplify skipping
    reeselevine committed 123 days ago
  • last cleanup
    reeselevine committed 123 days ago
  • actually final cleanup
    reeselevine committed 123 days ago
  • Merge remote-tracking branch 'upstream/master' into webgpu-all-archs
    reeselevine committed 123 days ago
  • update handling of overlap
    reeselevine committed 120 days ago
  • format
    reeselevine committed 120 days ago
  • try skipping other failing model
    reeselevine committed 118 days ago
Loading