diff options
| author | Onur Gündüz <ogunduz@gmail.com> | 2026-06-11 19:54:11 +0200 |
|---|---|---|
| committer | Onur Gündüz <ogunduz@gmail.com> | 2026-06-14 08:27:10 +0200 |
| commit | 69c612d012ff65135373c66aa5dc9f26ac0b24e3 (patch) | |
| tree | 17fef11047cdee1c620ca0ea76ae167552a87ed0 /embed.h | |
| parent | 2a4c9932cfebb597422aef5aaa836b084f6e50f4 (diff) | |
| download | ffwd-69c612d012ff65135373c66aa5dc9f26ac0b24e3.tar.gz ffwd-69c612d012ff65135373c66aa5dc9f26ac0b24e3.zip | |
chore: comments housekeeping
Diffstat (limited to 'embed.h')
| -rw-r--r-- | embed.h | 6 |
1 files changed, 2 insertions, 4 deletions
@@ -1,7 +1,5 @@ /* - * embed.h - Pure C inference for Perplexity AI's pplx-embed models. - * - * Supports all pplx-embed-v1 and pplx-embed-context-v1 variants (0.6B, 4B). + * embed.h - pplx-embed API */ #ifndef PPLX_EMBED_H @@ -116,7 +114,7 @@ PPLX_API size_t pplx_workspace_nbytes(const pplx_workspace_t *ws); * 3. Final RMSNorm * 4. Mean pooling over all positions * - * Perplexity embeddings are intentionally not L2-normalized. Use cosine + * pplx-embed-v1 embeddings are intentionally not L2-normalized. Use cosine * similarity for comparisons, or normalize explicitly before storing in vector * databases that only support inner product. * |
