shipfeedAI news, curated daily

00:32:35 CET
21 MAY00:32:35shipfeed⋯
pull to refreshlast sync
Just in — 30 new
§ tools · storyline

llama.cpp b9075

llama.cpp b9075 releases build b9075, adding fused CUDA snake activation support for audio decoder models such as BigVGAN and Vocos across F32, F16, and BF16 precisions.

May 8 · · primary fetch1 sourceupdated May 11 ·

cuda: fuse snake activation (mul, sin, sqr, mul, add) (#22667) cuda: fuse snake activation (mul, sin, sqr, mul, add) Add ggml_cuda_op_snake_fused with F32 / F16 / BF16 templates. The matcher recognizes the naive 5 op decomposition emitted by audio decoders (BigVGAN, Vocos) for snake activation y = x + sin(ax)^2 inv_b and rewrites it to a single elementwise kernel. Add test_snake_fuse comparing CPU naive vs CUDA fused across F32 / F16 / BF16. cuda: address review feedback from @am17an Use ggml_cuda_cast for F32/F16/BF16 conversions and rename kernel_snake to snake_kernel to match upstream conventions.

cuda: snake fusion fastdiv on T_len, Suggested-by: @am17an Update tests/test-backend-ops.cpp Co-authored-by: Aman Gupta cuda: snake fusion check add->type matches x->type Address review feedback from @am17an cuda: snake fusion check add->type matches x->type Moved for readability (equivalent) Address review feedback from @am17an --------- Co-authored-by: Aman Gupta macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled) macOS Intel (x64) iOS XCFramework Linux: Ubuntu x64 (CPU) Ubuntu arm64 (CPU) Ubuntu s390x (CPU) Ubuntu x64 (Vulkan) Ubuntu arm64 (Vulkan)…

read full article on github.com ↗
§ sources23 publications · timeline below
  1. github.comllama.cpp b9075primary
  2. github.comllama.cpp b9110
  3. github.comllama.cpp b9106
  4. github.comllama.cpp b9105
  5. github.comllama.cpp b9103
  6. github.comllama.cpp b9102
  7. github.comllama.cpp b9100
  8. github.comllama.cpp b9099
  9. github.comllama.cpp b9097
  10. github.comllama.cpp b9095
  11. github.comllama.cpp b9094
  12. github.comllama.cpp b9093
  13. github.comllama.cpp b9090
  14. github.comllama.cpp b9089
  15. github.comllama.cpp b9088
  16. github.comllama.cpp b9087
  17. github.comllama.cpp b9085
  18. github.comllama.cpp b9084
  19. github.comllama.cpp b9082
  20. github.comllama.cpp b9080
  21. github.comllama.cpp b9079
  22. github.comllama.cpp b9077
  23. github.comllama.cpp b9076

§ how this story moved

  1. primary — llama.cpp — Releases publishes the launch post.
  2. llama.cpp — Releases picks up coverage.
  3. llama.cpp — Releases picks up coverage.
  4. llama.cpp — Releases picks up coverage.
  5. llama.cpp — Releases picks up coverage.
  6. llama.cpp — Releases picks up coverage.
  7. llama.cpp — Releases picks up coverage.
  8. llama.cpp — Releases picks up coverage.
  9. llama.cpp — Releases picks up coverage.
  10. llama.cpp — Releases picks up coverage.
  11. llama.cpp — Releases picks up coverage.
  12. llama.cpp — Releases picks up coverage.
  13. llama.cpp — Releases picks up coverage.
  14. llama.cpp — Releases picks up coverage.
  15. llama.cpp — Releases picks up coverage.
  16. llama.cpp — Releases picks up coverage.
  17. llama.cpp — Releases picks up coverage.
  18. llama.cpp — Releases picks up coverage.
  19. llama.cpp — Releases picks up coverage.
  20. llama.cpp — Releases picks up coverage.
  21. llama.cpp — Releases picks up coverage.
  22. llama.cpp — Releases picks up coverage.
  23. llama.cpp — Releases picks up coverage.