llama.cpp
metal : implement q5_0 and q5_1 kernels
#3648
Merged

Commits
  • metal : implement dequantize_q5_0
    jhen0409 committed 2 years ago
  • metal : block_q_n_dot_y for block_q5_0 (broken)
    jhen0409 committed 2 years ago
  • metal : revert unnecessary change
    jhen0409 committed 2 years ago
  • metal : implement dequantize_q5_1
    jhen0409 committed 2 years ago
  • metal : block_q_n_dot_y for q5_1 (broken)
    jhen0409 committed 2 years ago
  • metal : fix block_q_n_dot_y
    jhen0409 committed 2 years ago
  • minor : spaces / formatting
    ggerganov committed 2 years ago
Loading