llama.cpp
ggml: Add initial WebGPU backend
#14521
Merged

Commits
  • Minimal setup of webgpu backend with dawn. Just prints out the adapter and segfaults
    reeselevine committed 317 days ago
  • Initialize webgpu device
    reeselevine committed 316 days ago
  • Making progress on setting up the backend
    reeselevine committed 316 days ago
  • Finish more boilerplate/utility functions
    reeselevine committed 315 days ago
  • Organize file and work on alloc buffer
    reeselevine committed 315 days ago
  • Add webgpu_context to prepare for actually running some shaders
    reeselevine committed 315 days ago
  • Merge remote-tracking branch 'upstream/master' into webgpu
    reeselevine committed 301 days ago
  • Work on memset and add shader loading
    reeselevine committed 301 days ago
  • Work on memset polyfill
    reeselevine committed 300 days ago
  • Implement set_tensor as webgpu WriteBuffer, remove host_buffer stubs since webgpu doesn't support it
    reeselevine committed 297 days ago
  • Implement get_tensor and buffer_clear
    reeselevine committed 296 days ago
  • Finish rest of setup
    reeselevine committed 295 days ago
  • Start work on compute graph
    reeselevine committed 294 days ago
  • Merge remote-tracking branch 'upstream/master' into webgpu
    reeselevine committed 294 days ago
  • Basic mat mul working
    reeselevine committed 293 days ago
  • Work on emscripten build
    reeselevine committed 288 days ago
  • Basic WebGPU backend instructions
    reeselevine committed 288 days ago
  • Merge remote-tracking branch 'upstream/master' into webgpu
    reeselevine committed 288 days ago
  • Use EMSCRIPTEN flag
    reeselevine committed 288 days ago
  • Work on passing ci, implement 4d tensor multiplication
    reeselevine committed 282 days ago
  • Pass thread safety test
    reeselevine committed 281 days ago
  • Implement permuting for mul_mat and cpy
    reeselevine committed 273 days ago
  • minor cleanups
    reeselevine committed 272 days ago
  • Merge remote-tracking branch 'upstream/master' into webgpu
    reeselevine committed 272 days ago
  • Address feedback
    reeselevine committed 269 days ago
  • Remove division by type size in cpy op
    reeselevine committed 269 days ago
  • Fix formatting and add github action workflows for vulkan and metal (m-series) webgpu backends
    reeselevine committed 260 days ago
  • Fix name
    reeselevine committed 260 days ago
  • Fix macos dawn prefix path
    reeselevine committed 259 days ago
Loading