diff options
| author | Georgi Gerganov <ggerganov@gmail.com> | 2023-03-25 19:47:21 +0200 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2023-03-25 19:47:21 +0200 |
| commit | ecbe466a364876927994e2f1ec14f4d82301d201 (patch) | |
| tree | 618c3e4e3f178f20bdc219e80b455ff45381ada5 /examples/perplexity | |
| parent | 502a400192013d3e95ed87b777e8fa3bec45713c (diff) | |
Retire the ggml_mul_mat() branch for transposed src0 (#500)
* Retire the ggml_mul_mat() for transposed src0
- It can always be made contiguous with ggml_cpy()
- The code is now simplified
- The results are deterministic in respect to num threads
* SIMD-ify dequantize_row_q4_0() for ARM_NEON (#502)
* Attempt to SIMD-ify dequantize_row_q4_0() for ARM_NEON
* Fix dequantization - forgot to interleave the quants
Diffstat (limited to 'examples/perplexity')
0 files changed, 0 insertions, 0 deletions